UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A cooperative scheme for image understanding using multiple sources of information Glicksman, Jay 1982

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1983_A1 G55.pdf [ 21.94MB ]
Metadata
JSON: 831-1.0051869.json
JSON-LD: 831-1.0051869-ld.json
RDF/XML (Pretty): 831-1.0051869-rdf.xml
RDF/JSON: 831-1.0051869-rdf.json
Turtle: 831-1.0051869-turtle.txt
N-Triples: 831-1.0051869-rdf-ntriples.txt
Original Record: 831-1.0051869-source.json
Full Text
831-1.0051869-fulltext.txt
Citation
831-1.0051869.ris

Full Text

A COOPERATIVE SCHEME FOR IMAGE USING MULTIPLE SOURCES OF  UNDERSTANDING  INFORMATION  by JAY  GLICKSMAN  B.Sc. The U n i v e r s i t y o f T o r o n t o , 1975 M.E. The U n i v e r s i t y o f U t a h , 1977  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE  REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in  THE  FACULTY OF GRADUATE STUDIES  DEPARTMENT We  accept t h i s to  THE  OF COMPUTER thesis  the required  as conforming standard  UNIVERSITY OF BRITISH November  SCIENCE  COLUMBIA  1982  © Jay Glicksman,  1982  DE-6  In p r e s e n t i n g requirements  this thesis f o r an  of  British  it  freely available  agree t h a t for  Library  shall  for reference  and  study.  I  for extensive copying of  that  h i s or  her  copying or  f i n a n c i a l gain  be  shall  g r a n t e d by  not  be  Ct>fnPuT€i So(C/JC€  The U n i v e r s i t y o f B r i t i s h 1956 Main Mall V a n c o u v e r , Canada V6T  (3/81)  1Y3  of  Columbia  make  further this  thesis  head o f  this  my  It is thesis  a l l o w e d w i t h o u t my  permission.  Department o f  the  representatives. publication  the  University  the  s c h o l a r l y p u r p o s e s may  understood  the  I agree that  permission by  f u l f i l m e n t of  advanced degree at  Columbia,  department or for  in partial  written  Abstract  One method of r e s o l v i n g t h e a m b i g u i t y inherent in interpreting images i s t o add d i f f e r e n t s o u r c e s of i n f o r m a t i o n . The m u l t i p l e i n f o r m a t i o n source paradigm emphasizes the ability to u t i l i z e knowledge g a i n e d from one s o u r c e t h a t may not be p r e s e n t in another. However, u t i l i z i n g d i s p a r a t e i n f o r m a t i o n may create situations in which data from different sources are inconsistent . A s c h e m a t a - b a s e d s y s t e m has been d e v e l o p e d that can take a d v a n t a g e of m u l t i p l e s o u r c e s of i n f o r m a t i o n . Schemata are.comb i n e d i n t o a s e m a n t i c network v i a t h e relations decomposition, specialization, i n s t a n c e o f , and n e i g h b o u r . C o n t r o l depends on t h e s t r u c t u r e of the e v o l v i n g network and a c y c l e of p e r c e p t i o n . Schemata cooperate by message p a s s i n g so t h a t a t t e n t i o n can be d i r e c t e d where i t w i l l be most a d v a n t a g e o u s . T h i s s y s t e m has been i m p l e m e n t e d t o i n t e r p r e t a e r i a l p h o t o graphs of s m a l l urban scenes. Geographic f e a t u r e s are i d e n t i f i e d u s i n g up t o t h r e e i n f o r m a t i o n s o u r c e s : t h e i n t e n s i t y image, a s k e t c h map, and i n f o r m a t i o n p r o v i d e d by the u s e r . The product i s a r o b u s t s y s t e m where t h e a c c u r a c y of the results reflects the q u a l i t y and amount of d a t a p r o v i d e d . Images of s e v e r a l geog r a p h i c l o c a l e s a r e a n a l y z e d , and p o s i t i v e r e s u l t s a r e r e p o r t e d . -  T a b l e of C o n t e n t s  1 Introduction  1  1 . 1 The Domain  1  1.2 S t e p s Toward a S o l u t i o n  3  1.3 The Scope o f t h e T h e s i s  7  1.4 R e a d i n g  8  2 Multiple  Information Sources  2.1 A New 2.1.1  Guide  10  Paradigm  10  Paradigms  10  2.1.2 The C u r r e n t S t a t e o f C o m p u t a t i o n a l V i s i o n  11  2.1.3 A New  13  Focus  2.2 I n f o r m a t i o n S o u r c e s 2.2.1 What  14  i s an I n f o r m a t i o n S o u r c e ?  14  2.2.2 The C o n j e c t u r e  16  2.2.3 I l l u s t r a t i v e  17  2.3 R e s e a r c h 2.3.1  that  Examples  h a s Combined  In A r t i f i c i a l  Information Sources  Intelligence  20  2.3.2 In C o m p u t a t i o n a l V i s i o n 2.4 Summary  23  and L o o k i n g Ahead  26  3 A Framework o f t h e F i e l d 3.1 M o d e l - B a s e d  Visual  3.1.1 D i f f e r e n t 3.1.2 G e o g r a p h i c 3.2 P o s s i b i l i t i e s  Types  20  29  Perception  29  of Models  30  Models i n Matching  34 Data  t o Models  3.3 Knowledge R e p r e s e n t a t i o n ( D e s c r i p t i v e  Adequacy)  41 44  iii  iv  3.4 Knowledge R e p r e s e n t a t i o n 3.5 G r a c e f u l D e g r a d a t i o n 4 How  t o Combine  Multiple  ( P r o c e d u r a l Adequacy)  49  (Robustness) Information  51 Sources  53  4.1 I n t r o d u c t i o n  53  4.2 A Schemata M a n i p u l a t i o n 4.2.1  Schemata Have F o u r  4.2.2 M a n i p u l a t i n g 4.3 I n t e r a c t i o n 4.3.1  54  Parts  55  Schemata  as a Source  Accepting  System  60  of I n f o r m a t i o n  and A c c o m o d a t i n g  61  Information  from  Users  4.4 A C y c l e o f P e r c e p t i o n 4.4.1  Interaction  4.5 M i x i n g 4.5.1  64  i n the C y c l e of P e r c e p t i o n  Top-Down and Bottom-Up  Top-Down a n d Bottom-Up  63  67  Control  72  Control  72  4.5.2 R e l a x i n g T h r e s h o l d s  75  4.5.3 M i x e d C o n t r o l S t r a t e g i e s  77  4.5.4 D i s t r i b u t e d 4.6 U s i n g  A I , VLSI, and P a r a l l e l  Processing  G l o b a l s t o Remove T h r e s h o l d s  4.7 C l u s t e r A n a l y s i s a s an I n f e r e n c e Mechanism 4.7.1  A Method  4.7.2 G a u s s i a n  —  80 81 84  for Clustering  88  Smoothing  90  4.7.3 Peak and T r o u g h D e t e c t i o n  91  4.7.4 The S t a b i l i t y  94  4.7.5 E x t e n s i o n s  Heuristic  to Multivariate  4.8 Summary 5 The I m p l e m e n t a t i o n  Data  97 104 106  5.1 I n t r o d u c t i o n  106  5.2 MAIDS--MISSEE A i d s  108  V  5.3 The I n p u t 5.3.1  Information  Sources  111  The D i g i t i z e d Image  5.3.1.1  111  Edge D e t e c t i o n  113  5.3.1.2 R e g i o n M e r g i n g  115  5.3.1.2.1 A C a t e g o r i z a t i o n  f o r the Regions  120  5.3.2 The S k e t c h Map  123  5.3.3 The U s e r  129  5.4 I n s t a n t i a t i n g 5.4.1  Schemata  Instantiation  130  Directly  from t h e I n f o r m a t i o n  Sources 5.4.1.1  132  Instantiating  %bridge-1  with  the A i d of a  S k e t c h Map 5.4.1.2  Instantiating  132 %bridge-2  from t h e I n t e n s i t y  Image  139  5.4.2 B u i l d i n g  up t h e H i e r a r c h i e s  144  5.4.3 Top-Down  Attached  147  5.5 C l u s t e r 5.5.1  Analysis  Determining  5.6 C o n t r o l - - T h e 6 Results  Procedures  Using Gaussian-Smoothed Histograms  .  Rectangularity  Priority  153  Queue, M e s s a g e s , and Demons ...  and E v a l u a t i o n  Instantiating  O b j e c t s W i t h t h e A i d o f a S k e t c h Map  6.2  Instantiating  Objects  from t h e I n t e n s i t y  ..  Image A l o n e  Up H i e r a r c h i e s  User  Interaction  6.6 D e s c r i p t i v e 6.7 P r o c e d u r a l  Adequacy Adequacy  Can I n f l u e n c e  163 177 190  6.4 M o v i n g Up and Down t h e H i e r a r c h i e s 6.5 How  158 163  6.1  6.3 B u i l d i n g  151  193 Interpretation  ...  200 203 205  vi  6.7.1 S o l u t i o n s  t o Problems  i n Matching Data  6.8 R o b u s t n e s s  207 209  7 Summary and C o n c l u s i o n s 7.1 A Summary  t o Models  o f What  i s New  211 and I n t e r e s t i n g  211  7.2 Open I s s u e s  214  7.3 F u t u r e  215  Directions  7.4 C o n c l u s i o n s  217  References  218  Appendix  A  230  Appendix  B  240  Appendix  C  243  Appendix  D  246  Appendix  E  247  List  Table  4.1  Modification  and  of  Tables  Retrieval  f o r Schemata  Attributes Relevant  60  Table  4.2  The  Statistics  Table  5.1  Interpretations  Table  5.2  O r i e n t a t i o n s and  Confidence  Table  6.1  A s h c r o f t : Sketch  Map  Table  6.2  Houston: Sketch  Table  6.3  Spences B r i d g e : Sketch  Table  6.4  S p e n c e s B r i d g e West: S k e t c h  Map  t o Image  176  Table  6.5  Spences B r i d g e  Map  t o Image  176  Table  6.6  Cranbrook: Sketch  Table  6.7  A l l Images: I n s t a n t i a t i o n  Table  6.8  The  of  f o r F i g u r e s 4.12-17  the A s h c r o f t Chains  Map  Values  t o Image  Map  155  175  t o Image  East: Sketch  126  174  t o Image Map  99  t o Image Results  Number of R e g i o n s S e a r c h e d  175  176 189 198  vii  List  Aerial  of  Figures  Figure  1.1  An  Photograph  2  Figure  1.2  A S k e t c h Map  Figure  2.1  An  Image of N o i s e  18  Figure  2.2  An  Object  18  Figure  2.3  A  Figure  2.4  Some c o n t e x t  19  Figure  3.1  A M o d e l of a R i v e r  35  Figure  3.2  A  37  .Figure  3.3  Mapsee2 D e c o m p o s i t i o n  Figure  3.4  Mapsee2 S p e c i a l i z a t i o n  Figure  3.5  Possibilities  Figure  3.6  A Maya I n s t a n c e :  Figure  3.7  An  Figure  4.1  A Schema: % r i v e r - 6  57  Figure  4.2  A Cycle  66  Figure  4.3  Interaction  in a Cycle  Figure  4.4  Hierarchies  in a Geographic  Figure  4.5  Edge Segments S u p e r i m p o s e d  Figure  4.6  A Histogram  Figure  4.7  An  Figure  4.8  The  Figure  4.9  A G a u s s i a n - S m o o t h e d H i s t o g r a m and  Figure  4.10  The  Figure  4.11  Finding  5  of  Interest  letter  19  S k e t c h Map  FRL  Hierarchy  39  Hierarchy  i n Matching Data  f o r Geosystems  40  to Models  42  *BRIDGE-1  45  Frame: SUPPLY  of  46  Perception  of  Idealized,  of  Perception  68  Domain on  73  Cranbrook  86  Edge Segment O r i e n t a t i o n s  87  Continuous Histogram  88  Gaussian D i s t r i b u t i o n  Effects the  of  o on  Largest  92 Derivatives  ..  Smoothing Stable  R e g i o n of  93 95  a Values  .  96  vi i i  ix Figure  4.12  A  2-D  H i s t o g r a m of a B i v a r i a t e D i s t r i b u t i o n  Figure  4.13  The  Histogram Convolved  with  the  Gaussian  Figure  4.14  The  Histogram Convolved with  the  Laplacian  Figure  4.15  Zero C r o s s i n g s  Figure  4.16  Determining  Figure  4.17  Using  Figure  5.1  The  Figure  5.2  Ashcroft,  Figure  5.3  A s h c r o f t : Edge D e t e c t i o n  Figure  5.4  Ordering  Edges  in Ashcroft,  a =  Figure  5.5  Ordering  Edges  in Ashcroft,  a = 2.2  Figure  5.6  A s h c r o f t : Region Merging  119  Figure  5.7  Ashcroft:  121  Figure  5.8  A S k e t c h Map  Figure  5.9  Ashcroft  Figure  5.10  A Mapsee2 D e c o m p o s i t i o n  Figure  5.11  The  Generic  Figure  5.12  The  Stereotype  Figure  5.13  Three %curb  Figure  5.14  Instance  Figure  5.15  Ashcroft:  %bridge-1  Regions  Figure  5.16  Ashcroft:  %bridge-1  Edge Segments  Figure  5.17  Instance  %bridge-2  144  Figure  5.18  Building  up  146  Figure  5.19  An  Figure  5.20  Clustering Orientations  Figure  5.21  Ashcroft:  of  the  a Using  ...  100 ....  101  Second D e r i v a t i v e  the  Stability  C h a n n e l s as D e c i s i o n  Heuristic  102 ...  Boundaries  107 Columbia  112 114 1.1  116 . .'  117  Region C l a s s i f i c a t i o n of A s h c r o f t  with  124  a Superimposed  Objects  S k e t c h Map  Hierarchy  125  of A s h c r o f t  .  i n MISSEE  %bridge  Schema  133 137  %bridge-1  Ashcroft  139 140 (Curbs)  141  S e m a n t i c Network  Instance  127 131  Instances  the  103 105  MISSEE System British  98  Hierarchy  %road-4 Edges  in Ashcroft  148 %road-4  ...  154 156  X  Figure  5.22  Clustering  O r i e n t a t i o n s along  Figure  5.23  The  Figure  6.1  Houston, B r i t i s h  Figure  6.2  Spences B r i d g e ,  Figure  6.3  Cranbrook, B r i t i s h  Figure  6.4  Ashcroft:  Figure  6.5  Houston: Information  Figure  6.6  Spences B r i d g e :  Figure  6.7  Spences B r i d g e  West: I n f o r m a t i o n  Sources  170  Figure  6.8  Spences B r i d g e  East:  Sources  171  Figure  6.9  Cranbrook:  Information  Sources  172  Figure  6.10  Ashcroft:  S k e t c h Map.  %town-1  178  Figure  6.11  A s h c r o f t : S k e t c h Map.  River  179  Figure  6.12  Ashcroft:  S k e t c h Map.  Mountain Edges  180  Figure  6.13  Ashcroft:  S k e t c h Map.  Roads  1-3,  Edges  181  Figure  6.14  Ashcroft:  S k e t c h Map.  Roads 4-6,  Edges  182  Figure  6.15  Ashcroft:  S k e t c h Map.  Roads  Regions  183  Figure  6.16  Ashcroft:  S k e t c h Map.  %road-2  Figure  6.17  Ashcroft:  Image A l o n e . F o u r Roads  186  Figure  6.18  Ashcroft:  Image A l o n e . % r o a d - 1 2  187  Figure  6.19  Ashcroft:  Image A l o n e . S e v e r a l  Figure  6.20  Ashcroft:  Instance  Hierarchy  2  192  Figure  6.21  Ashcroft:  Instance  Hierarchy  3  194  Figure  6.22  H o u s t o n : An  Figure  6.23  Spences B r i d g e :  Execution  the  Boundary  ....  Cycle  161  Columbia British  164  Columbia  165  Columbia  Information  166  Sources  167  Sources  Information  168 Sources  Information  Instance An  157  Regions  1-3,  185  Urban R e g i o n s  Hierarchy  Instance  169  Hierarchy  ..  188  195 196  Acknowledgments  F i r s t and foremost, I wish t o thank my supervisor, Alan Mackworth, f o r h i s support (both i n t e l l e c t u a l and f i n a n c i a l ) , guidance, and advice. I am a l s o i n d e b t e d to him for writing Mapsee a n d e x p o s i n g me t o t h e p o t e n t i a l f o r e x p l o i t i n g t h e c o n s t r a i n t s i n h e r e n t i n models of geographic o b j e c t s . I would a l s o like t o acknowledge h i s p e r m i s s i o n t o use the diagrams that a r e f o u n d i n F i g u r e s 3.3 a n d 3 . 4 . Thanks a r e a l s o extended t o t h e r e s t o f my committee, Bob Woodham, Richard R o s e n b e r g , D a v i d K i r k p a t r i c k , a n d Anne T r e i s man, who made h e l p f u l c o m m e n t s a l o n g t h e way t o t h i s f i n a l d o c u ment . I am g r a t e f u l t o B i l l H a v e n s , Jan Mulder, and Alan f o r explanations o f t h e i n n e r w o r k i n g s o f Maya a n d M a p s e e 2 , a n d f o r f i x i n g t h e bugs t h a t I found. J i m L i t t l e g a v e me a great deal of assistance with algorithms f o r l i n e generalization, finding channels and passes, and edge detection using a Sobel edge detector. The latter program was superceded by the MarrH i l d r e t h e d g e d e t e c t o r ( s o w i l l n o t be described herein) but t h a t was n o t J i m ' s f a u l t . I w o u l d l i k e t o t h a n k my o f f i c e m a t e s f o r b e i n g interested in what I was d o i n g a n d f o r n o t b e i n g v i s i o n a r i e s w h i c h f o r c e d me t o e x p l a i n my w o r k i n E n g l i s h . I l e a r n e d a l o t a b o u t what I had done from those discussions with Randy G o e b e l a n d Bob Mercer. Many members o f t h e L a b o r a t o r y f o r Computational Vision provided i n s p i r a t i o n and feedback. E s p e c i a l l y h e l p f u l were J i m L i t t l e , J a n Mulder, Roger Browse, A l a n Mackworth, Bill Havens, B o b Woodham, a n d M a r c M a j k a . And  i n a class  a l l by h e r s e l f ,  Susan  Suh.  Thanks.  xi  1  CHAPTER  Introduction  J_.J_.  The Domain Remote s e n s i n g  vision,  it  is  representation remote  of  values.  Unlike  dimensional lation  and  cannot  cues.  of which  tricted  to  vision  vision  narrow often  world.  range  utilizes  Computer-based  array  problem,  of  computer-based  additional  of  as the v i s i o n  three-  ( t h e simu-  problem)  the electromagnetic  several different  usual  intensity  whereas human v i s i o n  typified  computer  b o t h have a s t h e i r  t a k e a d v a n t a g e o f many  Furthermore,  Like  t h e i d e n t i f i c a t i o n and  two-dimensional  general  i s often  a  remote s e n s i n g  a  the  with  i n the e x t e r n a l  computer  information  remote s e n s i n g  at a distance.  p r i m a r i l y concerned  of objects  sensing  source  i s perception  i s res-  spectrum,  portions  of  the  spectrum.  Figure  1.1 was t a k e n  British  Columbia.  entities  such  tracks. tasks  the  Aerial  as land  erosion,  as  and  from an a i r p l a n e  flying  over  P e o p l e have no p r o b l e m d i s c e r n i n g the  roads,  rivers,  photo-interpretation  management,  i s quite  cartography,  urban growth, e s t i m a t i n g  movement o f s h i p s ,  buildings,  geographic or  railway  useful  f o r such  monitoring crop  Ashcroft,  yield,  pollution, and t r a c k i n g  c a r s , and t h e l i k e .  1  2  (scale Chapter  1.  F i g u r e 1.1 An A e r i a l P h o t o g r a p h r e d u c e d from o r i g i n a l 23 cm. x 23 cm. Introduction  photograph)  3  Computer-based restriction tially  erally  do  is  ure  true  number of  However, as  in A r t i f i c i a l like  problem because objects  has  been  I n t e l l i g e n c e , tasks  l a n g u a g e and  vision,  What  salient  features  of  an  1.1  enable  a person  to  identify  change  are  erally,  we  their  of  the  must use  are  objects  i t i s essen-  that  are  con-  f o u n d t o be  that  the  useful  seem  gen-  "easy"  most d i f f i c u l t  to  modelled  dependent  used  where  s u c h as  Fig-  Certainly  the  and  the the  the  intensities  is rarely sufficient.  knowledge of  dissertation discusses  t o be  i t s parts?  and  that  sizes,  photograph  objects  Gen-  themselves—  configurations  of many  of  knowledge  of  I t aims t o  combine  together.  information  data  our  aerial  regions But  relative  taken  This  various  important.  shapes,  objects  be  the  is also a  machine.  intensities  the  machine v i s i o n  manageable.  people, by  the  photo-interpretation  t w o - d i m e n s i o n a l and  sidered  for  of  aerial  that  the  in geographic  can  be  extracted  on  the  domain of  to organize  the  data  and  the  use  scenes. directly task.  to  from  Model  control  images  the  with  knowledge  the  can  process  of  interpretation. _.2.  S t e p s Toward a The  as  information  Figure  1.1  bright  to a dark  bridge  and  Chapter  1.  Solution found  i s often region  water, a road  Introduction  in a very  might and  s i n g l e , monochromatic ambiguous.  represent  The  the  a b u i l d i n g , the  image  transition  boundary sunlit  such  from a  between and  a  shadowed  4 parts  of a r i d g e , or many o t h e r  aerial  p h o t o g r a p h s can  ers are  so ambiguous t h a t  discovering  One people (e.g.  the  way  and  t o a i d the  a great 1.2).  it  the  can  be  is  the  nal  drawn,  two  utilized, who  ways  can  be  used  in this  an  oth-  difficulty  image,  in  for  both  more i n f o r m a t i o n .  Maps  many  stylized  them.  sparse  A  symbols  sketch  area.  picture  t o a i d the  that  be  document  of c o m b i n i n g so.  besides  map  is a  I t encodes (see  Figure  i n t e r p r e t a t i o n of  general  i n the be  They a r e used  to  of  or  very of  the  information  (e.g.  "The  information  intensity  emphasize d i f f e r e n t  map, nomi-  image").  extracted  that  a sketch  specific  from an  generally  of  concerned  information  image and  information  centre  derived  types  source  intensity  i s 1:10000")  can  is primarily  different  A third  the  provide  image  regions.  and  some  a novice,  have  f e a t u r e s of an  in a very  i t can  i s a bridge  features  e d g e s and  can  the  "There  understand  crucial  mechanisms t o do  user  While  image.  usefulness  s c a l e of  (e.g.  the  i s t o add  maps) c o n t a i n  to  work d e s c r i b e d  the  and  road  information  a corresponding The  may  i n t e r p r e t a t i o n of  easier  of  Easily  with  and  d r a w i n g of deal  i n t e r p r e t e d by  even e x p e r t s  a computer program,  make  freehand  readily  objects.  correct interpretation.  topographic  which  be  p a i r s of  in  Also,  image  are  different  aspects  of  the  image.  All  these  models  of  Chapter  1.  the  different objects  Introduction  s o r t s of t o be  data  must be  identified  i n the  related scene.  to  the  However,  5  Figure  while  i t i s hoped t h a t  conclusion, models. robust the  to  q u a l i t y of  the  way  relate  frames, are  Chapter  may  1.  of  that  must be  the  accomodating  them a data  all  to  Introduction  handled of  the  the  the  data  same  and  in a consistent results will  the and  reflect  data.  disparate  the  structure  support  between  accuracy  available  Map  evidence w i l l occur  These d i f f i c u l t i e s so  A Sketch  a l l the  mismatches  manner  One  1.2  objects  that  can  t y p e s of  information  themselves. be  used  to  Schemata,  store  all  is or the  6 facts pertinent contain  a declarative part  interpretation this  road  part  where  c a n be s t o r e d  contains  instantiate  of  the  t h e image, u s i n g  Schemata c a n be c o n j o i n e d  13")  organization  into  cohesive  all  entities  the  body o f w a t e r . instances ple,  that  there  tion  of  under  it.  others steps  existence  will a  found  Using  model  strategy nearby.  that  is crucial  group  objects  attached object  that  in this  objects  without  the system  should  jump t o c o n c l u s i o n s .  C h a p t e r - 1.  Introduction  This  F o r exam-  of a r i v e r  parts  in  leads to  subsequent  of t h e scene.  be p r o c e d u r a l l y instantiated.  is  flowing  i s an i s l a n d - d r i v e n  expands  support  Furth-  Identifica-  o f one o b j e c t  not presuppose  f i n d i n g the proper  by a  neighbouring  nearby.  way  the discovery of  For  r a i s e s the expectation  t o the p o s s i b i l i t y  a r e not i n c o r r e c t l y  t o schemata  scene.  to b u i l d i n g s , not s h i p s .  t o encompass a l l o f t h e r e l e v a n t  It  the  reflects  of the e n t i r e group.  knowledge  wherein  This  in  that  o f a landmass d e l i n e a t e d  r e l a t e d instances  points  the  c a n be r e l a t e d a s c o u l d  o f some i n s t a n c e s  be o t h e r  bridge  next  sources to using  i n the i n t e r p r e t a t i o n of  r a i s e s the confidence  the  control  are part  Agreement  roads a r e o f t e n  ermore,  they  that  procedural  alone).  units  a network of r o a d s and b r i d g e s  a  as  network  instance,  an  correspond to  and  (such  t h e image  a  from  information  object  into  Schemata  derived  ( e . g . "The edges t h a t  methods f o r u t i l i z i n g  the existence  map w i t h  their  information  i n t h e image a r e 11, 12, a n d  that  sketch  t o a p a r t i c u l a r e n t i t y such as a road.  adequate  The p r o c e d u r e s  the e x i s t e n c e i n the data  intimately  so  of  nor  related  an  should  to  the  7 descriptive to c o n f i r m able  adequacy  o r deny t h e e x i s t e n c e  sources  of i n f o r m a t i o n ?  midrange, n e i t h e r tion,  the a v a i l -  The d e s c r i p t i o n s must  f i t into a addi-  t h e d e s c r i p t i o n s must be s u i t a b l e f o r any p o s s i b l e  appli-  t h a t a r e of i n t e r e s t  nor t o o  encompassing.  ( e . g . "How  long are the bridges  interpret  disparate  aerial  information  from  photographs. sketch  interpretation information  information  sources  Along  maps and t h e  results  with  are  combined  the d i g i t i z e d  user  is  made  system  ferent  uniform  depend on t h e amount and q u a l i t y o f  them.  The  repository for  information  sources.  schemata  information Schemata  be  created  growing  i n t h e image t h e y  and l i n k e d i n t o a s e m a n t i c  network  enables  bottom-up, w h i c h e v e r  The  The  i s most  messages  t i o n s and a r e p l a c e d  Chapter  control  coming  contain  on a  1. I n t r o d u c t i o n  instances  network.  to  be  a con-  from  dif-  attached As  a  pro-  entities  o f schemata t o  T r a v e r s a l of the  either  top-down  or  appropriate.  nodes o f t h e s e m a n t i c  messages.  cause  with  provide  cedures which c o n t r o l the i n t e r p r e t a t i o n p r o c e s s . discovered  image,  provided.  f o r manipulating  sistent,  to  available.  A new v a r i e t y o f schemata h a s been d e v e l o p e d a l o n g  are  in  The Scope o f t h e T h e s i s Several  the  using  image").  j_.3_.  The  too limited  of the o b j e c t s  sufficient  In  cations an  of the m o d e l s — a r e d e f i n i t i o n s  network  communicate  are suggestions priority  queue.  for future  by  passing  instantia-  Priorities  depend  8 upon t h e and  are  the  confidence  values  determined during  models.  Processing  of  proceeds  of p e r c e p t i o n .  This  where  interaction  the  Several computational instantiation  of  Guassian-smoothed determining method of  gle  values  can  take  t o o l s are  is  an  w e l l the  data f i t  convenient  i s based  points  place. to  aid  in  the  C l u s t e r a n a l y s i s using  inference  elements best  knowledge  messages  manner f o l l o w i n g a  mechanism  for  f i t model d e f i n i t i o n s . on  the  In a d d i t i o n , many t h r e s h o l d s c a n when model  the  developed  histograms  relaxing thresholds  how  provides  entities.  data  sending  in a c y c l i c  cycle user  by  geographic  which  knowledge.  instances  instantiation  cycle  with  the  i s combined  used  be  of  reduced  with  A  model to  global  sinsche-  mata . In that  summary, t h i s  can  accomodate  object-oriented network It  J_.4.  and  based around  geographic the  Reading next  title  of  should  of  three the  chapters  of  instantiated  correspond  thesis in reverse  Chapter  this  work.  image u n d e r s t a n d i n g : 1.  instances  form  a  It i s  semantic  those  objects.  entities  to  con-  Guide  hypothesis  Chapter  system  be d i r e c t e d .  sources  of  vision  from s e v e r a l s o u r c e s .  o b j e c t s and  current context  a  schemata t h a t  the m u l t i p l e i n f o r m a t i o n of  describes  information  where f u t u r e e f f o r t  The the  of  utilizes  trol  dissertation  Introduction  to the  order. paradigm  Chapter as  3 i s concerned  model-based v i s u a l  three 2  parts  of  introduces  the  underlying  with  one  perception.  aspect Previous  9 relevant criteria are  research  for judging  described.  schema-based sources. the  the  field  is outlined.  success  Chapter  system  Each  of  a model-based v i s i o n  4 discusses  that  i s used  s e c t i o n of  this  In a d d i t i o n ,  the  cooperative  t o combine m u l t i p l e chapter  describes  three system  scheme, a information  a component  of  system.  Chapter tem  5 discusses  described  tem  are  the  work and  the  proceed  tions  Results 6.  implementation and  Finally,  of  evaluation Chapter  to p o s s i b l e d i r e c t i o n s f u t u r e  the  of  7  the  syssys-  summarizes  research  might  in. who  are  and  dissertation are  Appendix  Chapter  4.  of C h a p t e r  points  representation the  MISSEE, t h e  in Chapter  topic  Readers  of  i n the  3.3,  not  with  issues  of  vision  wish to read  only  may  concerned with  3.4,  4.2,  A.  1.  concerned  Introduction  schemata.  4.5.3, 5.2,  The  5.4.2, 5.6,  knowledge those  parts  relevant  sec-  6.6,  6.7,  and  CHAPTER Multiple  2.]_.  A New  2.]_.]_.  Sources  Paradigm  his influential  lutions"  (Kuhn,  "Normal  book,  science"  science  generally  "paradigm". Paradigms  consist defines  exist  revolves  around  of s c i e n t i f i c  astronomy)  field.  notion  spring  research"  of a  particu-  (p. 1 0 ) . Where  advances  i n the  field  by t h e work  I t i s important t o note t h a t  that  the paradigm  Thus s c i e n t i s t s  t h e p a r a d i g m a r e not c o n s i d e r e d  problems  the  Revo-  to account  in a particular  which problems a r e s i g n i f i c a n t .  working w i t h i n  of S c i e n t i f i c  out t h e p r o b l e m s n o t s o l v e d  the paradigm.  meaningful  progresses  (e.g. Ptolemaic  of s o r t i n g  influences  Structure  " p r o v i d e m o d e l s from w h i c h  coherent t r a d i t i o n s  paradigms  "The  1970), Thomas Kuhn p r e s e n t s a t h e o r y  t h e means by w h i c h  larly  Information  Paradigms  In  for  .2  t o be w o r k i n g  by t h e a d h e r e n t s o f t h e c u r r e n t l y  not on  dominant  paradigm.  Paradigms can  be  which the  c a n be a d o p t e d i n two  supplanted  redefines  by a new  legitimate  Copernican "revolution"  theory,  ways.  the " s o - c a l l e d  p r o b l e m s and which  An e x i s t i n g  research  replaced  paradigm  revolution",  methods  Ptolemaic  (e.g.  astronomy).  [The new s c i e n t i f i c ] a c h i e v e m e n t [must be] sufficiently unprecedented t o a t t r a c t an e n d u r i n g g r o u p of a d h e r e n t s away from c o m p e t i n g modes o f s c i e n t i f i c a c t i v i t y . S i m u l -  10  11 t a n e o u s l y , i t [must be] s u f f i c i e n t l y o p e n - e n d e d t o l e a v e a l l s o r t s of problems f o r the r e d e f i n e d group of p r a c t i t i o n e r s t o r e s o l v e , ( p . 10)  P a r a d i g m s c a n a l s o emerge  i n a young  science  where none a r e  currently  e s t a b l i s h e d . In p r e - p a r a d i g m a t i c  science,  generally  accepted  focus  Rather,  there  views which a c t  a r e many  small  own v i e w s and m e t h o d o l o g i e s . more  as  a  there  for  groups of p r a c t i t i o n e r s In t h i s  state,  a r e no  research. with  science  their  proceeds  slowly:  . . . when t h e i n d i v i d u a l s c i e n t i s t c a n t a k e a p a r a d i g m for g r a n t e d , he need no l o n g e r , i n h i s m a j o r works, a t tempt t o b u i l d h i s f i e l d anew, s t a r t i n g f r o m f i r s t p r i n ciples and justifying t h e use of each c o n c e p t introd u c e d . T h a t c a n be l e f t t o t h e w r i t e r of t e x t b o o k s , ( p p . 19-20) O n l y when a p a r a d i g m h a s been e s t a b l i s h e d cumulative  so t h a t one c a n s t a n d  can  progress  on t h e s h o u l d e r s  Newton w r o t e ) r a t h e r t h a n  on t h e i r  2.J_.2.  of C o m p u t a t i o n a l V i s i o n  The C u r r e n t  Artificial Vision,  central  question  physics,  Intelligence,  in particular,  recognized ple  State  the  are very  lack  in  general,  young  of  a r e even  unifying  of g i a n t s (as  toes.  and  Computational  sciences' without  paradigm. F o r e x a c t l y t h i s  whether t h e y  become  reason,  sciences at a l l . principles  is  a clearly some  peo-  Compared t o noteworthy[1].  [ 1 ] U n f o r t u n a t e l y t h e r e a r e many o l d e r d i s c i p l i n e s that have the same d i f f i c u l t y . T h i s l e a d s some t o q u e s t i o n whether K u h n s theory i s a p p l i c a b l e to everything that i s l a b e l l e d science. I t leads others t o q u e s t i o n whether the l a b e l " s c i e n c e " i s always appropriately applied. 1  Chapter  2. M u l t i p l e I n f o r m a t i o n  Sources  12  While  several  written  t e x t b o o k s on  (Nilsson,  1971  Artificial  and  1980;  has a t e x t b o o k on c o m p u t a t i o n a l Brown,  Winston,  vision  computer  t h e r e has n e v e r been vision,  there  e x t e n d e d beyond  labelling history  of  been  several  the l a b o r a t o r y  line  (Huffman,  have  a generally  drawings  1971;  of  Clowes,  use  further  which s u r f a c e s  1971;  in early  vision  and  research.  The  primal  representation  in  vision  biological In  attention  discontinuities  sketch and  ( e d g e s ) i n an  1973;  Kanade,  as  primal  1982)  The  a  has  long means  objects.  sketch  and  inspired  much  a  rich,  1981;  1975;  texture:  1978), t h e t h r e e - d i m e n s i o n a l the m o d e l l i n g  detection  an  the  of  1975;  image  s t e r e o : Grimson,  Render,  1980;  Sources  that  intensity  Davis,  of a  1975), and  surface  1981;  con-  motion: Ullman,  of o b j e c t s  of what m i g h t a p p e a r  1977).  equally  (Yakimovsky  orientation  representation  Information  seem  some of t h e p r o b l e m s  (Shirai,  in  1980),  ("shape f r o m " s h a d i n g : H o r n ,  2. M u l t i p l e  1981)  i s d e s i g n e d t o be  the  image  Feldman,  and B a j c s y ,  originator.  a l l problems  include  of homogeneous r e g i o n s  Witkin,  that  i s m o d e l l e d on what t a k e s p l a c e  field,  detection  and  and  In c o m p u t a t i o n a l v i s i o n ,  received  models  s y s t e m s — y e t another paradigm.  a pre-paradigmatic  interesting.  unifying  b e l o n g t o which d i s t i n c t  ( M a r r , 1976  intermediate  Chapter  (Ballard  paradigm i n  Kanade,  work of D a v i d Marr and h i s g r o u p on t h e  1981),  recently  i n t h e b l o c k s w o r l d has had a  The  tour:  been  accepted  their  determining  have  1977), o n l y  appeared  of  its  have  1982).  While  have  Intelligence  (Brooks,  i n scenes  (Sloan  1 3 2'1«3.  A New  The  Focus  major  hypothesis  information  sources.  research  utilized  a  has  secondary  proposal  As  f a c e t of  and  to  the  will  multiple  dissertation be  seen  notion  relate  the  the  aspects  multiple  sections,  sources,  but  problem at  of m u l t i p l e  other  concerns  in later  information  i t s s o l u t i o n to  i s t o make t h e  primary,  of  only  hand.  information of p r o b l e m  other as The  sources  s o l u t i o n to  it.  In being is  this  sense the  presented  a major  as  a new  paradigm  field[2].  Rather,  to  best  that  can  provide  a  The  rest  this  for using  tion.  cert,  with  information  in l i g h t  lems w i l l  be  theoretical  of  the  the  addressed thrust  hypothesis--one  of  presumed  coherence  coordinate  for  research.  will  for  sources,  sources  that the  One  as  is the  a  a  is this  whole  the  s o l u t i o n to overall as problem  sources)  convincing  focus  individually  Thus t h e  thesis  u s e f u l s o l u t i o n to  information  discuss  both  paradigm.  in d e t a i l .  the  t r y to provide  thesis will  the  sources  g e n e r a t e s a p r o b l e m domain ( i . e .  information  the  i s not  and  chapter  multiple  information  It  provides  paradigm  focus  Subsequently,  dealing  paradigm.  manipulate  of  of m u l t i p l e  that  the  how  case  use  of  problems  of  and .in c o n those  prob-  (simplified)  follows: of  atten-  major  understanding  [ 2 ] I t i s c l e a r t h a t a c r u c i a l a s p e c t of p e r c e p t i o n would be m i s s i n g from t h i s p a r a d i g m b e c a u s e i t i s p o s s i b l e , f o r p e o p l e a t l e a s t , t o f o r m a p e r c e p t from a s i n g l e i n f o r m a t i o n source such as a p h o t o g r a p h .  Chapter  2.  M u l t i p l e Information  Sources  14 images i s t o  use  multiple  information  secondary h y p o t h e s i s — u s e f u l tage  of  multiple  languages, control  user  information  interaction,  (Chapter  4).  5)  domain o f  understanding  2_.2_.j_.  What  An either ate  demonstrates  Information  result  Standard IS)  i s an  what  array  amount of  and  line  some i d e a l  scale. is  a  input  sources  of  of  t o the  those  tools  geographic  IS  of  i s the  handled.  range  A digitized  geometric  of  the  abbreviated  captures  M u l t i p l e Information  Sources  one  scene  surfaces  two-  of  to  i n view)  them  forming  objects  s y s t e m and  the  in  out-  scene.  symbolic  of  (a  some r e l a t i o n  between  the  mapping between them.  i t only  images  i n the  image i s a t  representation  intermedi-  sometimes  lines  of  can  processing.  to a v i s i o n  representation  iconic  because  the  It  of  bearing  edges of  input  a  stages  digitized  values  the  data.  s y s t e m o r an  to l a t e r  are  intensity  i s , of c o u r s e ,  direct  of  bottom-up  i n the  input  (henceforth  vision  representation  I t i s an  2.  set  and  more a b s t r a c t  Chapter  is a  (vertices  i s some s y m b o l i c  be  usefulness  drawings  The  schema-based  implementation  from o b j e c t s  put  can  top-down and  reflected  scene).  that  combined  light  the  There  include  advan-  Source?  becomes a v a i l a b l e  computational  dimensional the  source  information  in  to take  2);  photographs.  I nf ormat i o n  is initially  that  one  sources  their  aerial  (Chapter  Sources  information be  and  that allow  A particular  (Chapter  2.2.  tools  sources  representations  the  low  scene A  aspect  since  line of  end  of  there  drawing the  the  is  image,  15 namely, scale  the  to a  edges of  representation  objects  s u c h as  standing  system  range and  mation  t h a t can  illustrated intrinsic model  images  tance,  (Barrow  up  the  from  the  and  cooperate  Several vision  of  This  with  of  include  arrays  Each  the  each  intensity  images--for  images[3],  range d a t a  on  the  iterative  in the  type  infor-  This  is on  Their  basic  which are  iconic  array  contains  a  illumination, reflecsystem  other  there  i s only  representations  relaxation process.  example of  is,  Tenenbaum  1978).  their  both  that scale.  of  source.  Barrow and  intensity,  under-  representation  individual  In  image  IS's  and  the  how  information  are Thus  final  output sources  other.  other  colour  based  image; t h e  are  i s an  a  the  specific  somewhere  information  distance.  intensity  of an  registered arrays  image.  to  f u r t h e r up  Tenenbaum,  information:  non-intensity  task  abstract  work done by  f u r t h e r up  finally  represented  an  i t t h r o u g h an  representation. can  in  i n the  o r i e n t a t i o n , and  input,  built the  found  k i n d of  The  dimension  c o n s i s t s of a s t a c k  different  IS  progress  and  tables. an  orthogonal  of  can  objects  r a n g e of how  be  One  "more a b s t r a c t " o u t p u t  clearly  representations  one  i s to take  this  e x i s t s an  of  c h a i r s and  produce a  Besides there  surfaces.  types images  red, which  of  information  (usually  green, gives  and  used  in  represented blue),  distance  computer as  three  multispectral  directly,  digital  [ 3 ] T h i s i s a g e n e r a l i z a t i o n of c o l o u r images where certain bands i n t h e e l e c t r o m a g n e t i c s p e c t r u m a r e s a m p l e d . The most common e x a m p l e s of t h i s a r e i n f r a - r e d p h o t o g r a p h y and s a t e l l i t e ima g e r y s u c h as L a n d s a t w h i c h has 4 MSS bands.  Chapter  2.  M u l t i p l e Information  Sources  16 terrain  models  sketch  maps  bolic tion  from  goal  interest.  1981).  actually  scene),  between sym-  (containing the  loca-  landmarks).  sensed,  Unfortunately, The  i s t o d e t e r m i n e t h e mapping  the  the problem  factors  image,  o r from  resulting  direction,  factors arising  path  phenomena,  to  i s vastly from  shape and s u r f a c e m a t e r i a l , c a n n o t  other  viewing  from or  the scene of  underconstrained  properties be  separated  illumination,  sensor  of the  noise  Images a r e ambiguous and c a n be t h e r e s u l t  from  shadows, (Woodham,  o f an  infin-  number o f s c e n e s [ 4 ] . The  hypothesis  i s that,  Multiple information biguities .  sources  are useful i n resolving  am-  simply, More  i s better.  This viewpoint will  the  relationships  bases  o f image u n d e r s t a n d i n g  is  general.  objects,  Or,  of  data  in  The C o n j e c t u r e  what  each  of p o i n t s  the t o p o l o g i c a l  and c h a r a c t e r i s t i c s  The  ite  (giving  the heights  o b j e c t s ) , and g e o g r a p h i c  2.2_.2.  in  (yielding  be seen  i s widely  h e l d by many v i s i o n  i n the d i s c u s s i o n of systems t h a t  r e s e a r c h e r s as  take  advantage of  [ 4 ] T h i s i s t h e o p p o s i t e o f human v i s i o n i n w h i c h images a l l y seem t o be o v e r c o n s t r a i n e d and n o t ambiguous a t a l l .  Chapter  2. M u l t i p l e I n f o r m a t i o n  Sources  usu-  1 7 multiple  IS's.  however,  that  the  notion  There has  i s a stronger  conjecture  n o t been a d d r e s s e d  of d i f f e r e n t  types  of  t h a t c a n be made,  before.  This  relates to  information.  The more d i s p a r a t e t h a t t h e i n f o r m a t i o n s o u r c e s combined a r e , the b e t t e r a b l e they a r e t o r e s o l v e a m b i g u i t i e s . Or,  restated, The  more d i f f e r e n t ,  This  the b e t t e r .  i s b a s e d on t h e i n t u i t i o n  scene  will  mation  sources.  important  be e n c o d e d  in different  one s t a r t s  additional image  of  ways by t h e d i f f e r e n t  these  ideas.  the  infor-  an i n t e n s i t y  image,  Examples  with  information  source  (more i s b e t t e r ) .  would  like  cess.  Changing  which  results  to  vary  at least  the viewpoint  the p o s i t i o n  allows  one  take  to  tometric  stereo  altered,  one  allows  is  able  one t o  (Grimson, of  advantage  (Woodham,  something  one p a r a m e t e r  i n a d e p t h map  Changing  then  the  simplest  t o a d d would be a n o t h e r  To o b t a i n  1981).  Chapter  aspects  i n the i n t e r p r e t a t i o n .  Illustrative If  similar  One t h u s h a s more c h a n c e o f d i s c o v e r i n g what i s  Some e x a m p l e s may c l a r i f y 2.2.3.  that  the  If  to determine  2. M u l t i p l e I n f o r m a t i o n  interesting,  one  of the imaging  pro-  perform  stereopsis  1981,'.Baker and B i n f o r d , source  of a type  1980a).  intensity  of  illumination  of s t e r e o c a l l e d  the  time  parameter  shape from m o t i o n  Sources  phois  (Ullman,  18 1 978) .  A s i m p l e example w i l l advantage. Clearly, sider  C o n s i d e r the there  below t o be  t o p of F i g u r e 2.1.  spring  i f one  into  The  moved  view.  tinguishable  from  square  the  overlay  Figure  2. M u l t i p l e  good  Figure  2.1.  image.  and  An  An  Information  resting  on  noise. would  the square  becomes  indis-  Image o f  Sources  con-  square  one  had  two  subtraction  1970).  Object of  Now  the  If  (Anstis,  2.1  2.2  in  to  s t a n d out. of t h e  around,  then a simple  example  dots in this  would n o t  the background.  in this  motion  on a t r a n s p a r e n t o v e r l a y  Stop the motion  Figure  Chapter  interest  square  images s e p a r a t e d i n t i m e , the  p e o p l e use  image of random  i s n o t h i n g of  F i g u r e 2.2  However,  show how  Noise  Interest  intensity  would  reveal  19 If have can  the  been  s q u a r e had  been a d i f f e r e n t  colour,  apparent,  to people,  away.  i m a g i n e two  noise,  which  is  square, which  Now  a  letter  the  above, the  Figure  drawn  half  is deliberately  be  any  mation  larly,  help.  The  c a u s e us if  the  for  and  IS's  contains way  would  t h i s case, colour  f o r the  have been t h e  one  of  colour  the  of  the  was  created  that as  the  A  LETTER_BEFORE = T LETTER AFTER = E  was in  an  similar  in Figure  LETTER_BEFORE LETTER AFTER  b  2.4  2. M u l t i p l e I n f o r m a t i o n  Some  context  Sources  Now  an  this  an  infor-  2.4a  would  "H".  Simi-  2.4b  = C = T  of  image would  added as Figure  kinds  represents  "H".  letter  a Figure  that  l e t t e r was  provided  2.3  and  other  i f context  expectations  to conclude  no  "A"  same--intensity  different  a l i n e drawing  ambiguous and  context  using  between an  Figure  Chapter  the  one  a d v a n t a g e s of  However, what  source?  probably  the 2.3  image of  one  In  it  interesting.  consider  information.  arrays,  uninteresting,  is  In a l l of images.  intensity  right  then  then  the  20 letter  is likely  The  Stroop  of  information  had  subjects  an  "A"  effect can  read  colour  than  the  ink),  then  slower  r e a c t i o n time,  Here,  then,  Research  greatly  represents diagram  the  to  p r u n e 995  node  i n the  as  [5]Much  Chapter  was  they  i n the  "brown"  1976).  Stroop  colours  (e.g.  in a  drawn  and  names and  a  different in  green  significantly  colours  compelled  coincided. t o use  p e r f o r m a n c e of a  I nf ormat i o n  the  a  second  recognized  work on  proving  information  mul-  task.  Sources  out  of  source i n the  on  the  information  e a r l y d a y s of theorems  i n a diagram  1963).  hypotheses, 1000,  of  geometry  provided  (Gelertner,  impossible  successors  Information sources  Combined  problem[5]  search  (e.g.  where p e o p l e a r e  solver  r u l e out  of  more d i f f i c u l t y ,  of c o m b i n i n g  by  names  types  Intelligence  Gelertner's aided  different  (Neisser,  i t degrades t h e i r  t h a t has  problem  example of how  t h a n when t h e  even t h o u g h  usefulness  research.  to  had  I_n A r t i f i c i a l  a  .  of a word were drawn  word r e p r e s e n t e d  i s a c a s e of  2.3_.  with  letters  subjects  IS's  The  i s another  words w h i c h were t h e  When the  tiple  1967).  influence perception  "brown").  2^3_.J_.  (Neisser,  When  was which  using  h i s p r o g r a m was average,  at  Al  a  able each  tree.  sources  are  directly  were  defined  i n the  same way  that  2. M u l t i p l e I n f o r m a t i o n  high  related HEARSAY I I  school  Sources  to  knowledge  speech  students  are.  under-  21 standing  project  source its  is  area  (Erman  viewed  a s "an a g e n t  and  Lesser, (Hewitt  seem b e s t  suited  3.4  4),  and  based  1975, p . 4 8 3 ) . T h i s  on  that  i s similar  e t a l . , 1973) o f a community  cooperatively  to  solve problems.  sources  knowledge" t o t h e ACTOR  of e x p e r t s  work-  However, whereas ACTORs  t o an o b j e c t - o r i e n t e d  knowledge  knowledge  w h i c h embodies t h e knowledge o f  and w h i c h c a n t a k e a c t i o n s  philosophy ing  (Erman e t a l . , 1980). I n HEARSAY e a c h  solution  (see  Chapters  were i m p l e m e n t e d a s p r o d u c t i o n  rules.  Information that  each  IS  sources  are related  requires  at least  I S ' s and KS's c a n be seen The  major  former  concentrates  multiple speech that  difference  IS's.  reside  subsequent information  Despite several tem  on  of  retrieved  is  from  blackboard;  used  in  Chapter  both  systems  2. M u l t i p l e  coin.  deals  with  as  intermediate  output  and p r o v i d e c o n t e x t f o r with  the  relevant  t h e Long Term Memory.  4 and 5.  ( i nproduction  the  Thus  work i s t h a t t h e  the l a t t e r  together  similarities  They b o t h  i t .  in  i n p u t IS i n HEARSAY--the  IS's exist  philosophical  d e s c r i b e d i n Chapters  (the  one  processing,  methodological  the o t h e r ) .  only  on t h e b l a c k b o a r d  t h a t , major  knowledge m o d u l a r  one KS t o d e a l w i t h  on m u l t i p l e KS's w h i l e  together  sources  t o be t h e two s i d e s o f t h e same  The o t h e r  stages  knowledge  between HEARSAY and t h i s  There  waveform.  to  exploit  semantic  there  are  between HEARSAY and t h e s y s They b o t h  rules  attempt  network).  Sources  to  on t h e one hand;  a hierarchical  (hypothesize  Information  difference,  and  keep  schemata  representation  Feedback c o n t r o l i s test;  a  cycle  of  22 perception). when t h e  They  both  strive  to achieve g r a c e f u l  degradation  i n f o r m a t i o n i s poor.  Knowledge " s o u r c e s have been u s e d  e x t e n s i v e l y in Al  systems,  c usually  when  that  be  can  sources that  applied  can  they  i t is useful  be  to  the  encode t h e i r  from  new  problem  integrated  w h i c h a l l communicate acquired  to d i s t i n g u i s h at  the  t y p e s of  hand.  New  i n t o DENDRAL b e c a u s e  knowledge  i n the  form  of  instruments  knowledge  i t is required  of p r o d u c t i o n  i n t e r m s of a common g r a p h  types  knowledge  turned  rules  language.  out  to  Data  be  very  useful. . . . t h e impact of b r i n g i n g j u s t one a d d i t i o n a l s o u r c e of knowledge t o b e a r on a p r o b l e m can be s t a r t l i n g . In one d i f f i c u l t ( b u t n o t u n u s u a l l y d i f f i c u l t ) mass spectrum a n a l y s i s p r o b l e m , t h e p r o g r a m u s i n g i t s mass s p e c t r o m e t r y knowledge a l o n e would have g e n e r a t e d an impossibly l a r g e s e t of p l a u s i b l e c a n d i d a t e s ( o v e r 1.25 m i l lion!). Our e n g i n e e r i n g r e s p o n s e t o this was to add another source of d a t a and knowledge, p r o t o n NMR. The a d d i t i o n of a s i m p l e i n t e r p r e t i v e theory of this NMR d a t a , from w h i c h t h e p r o g r a m c o u l d i n f e r a few a d d i t i o n a l c o n s t r a i n t s , r e d u c e d t h e s e t of p l a u s i b l e candidates to one, the r i g h t s t r u c t u r e ! T h i s was n o t ,an i s o l a t e d r e s u l t but showed up d o z e n s of t i m e s i n s u b s e q u e n t a n a l y s e s . ( F e i g e n b a u m , 1977, p. 1020)  Several types  of  obtained  knowledge. by  emphasized to a  Brown  and  (Burton  their  and  Burton  talk  importance  Brown,  1975,  (Davis,  1980a) and  p.  use  of  disparate  of  the  "synergism  of  the  312). even  procedural Davis raised  2. M u l t i p l e  Information  redundancy — i s  Sources  nicely  has  their  principle:  A f o u r t h pr inc i p l e - - e x p l o i t i n g  Chapter  f o r the  f o c u s i n g the d i v e r s e c a p a b i l i t i e s  specialists"  use  r e s e a r c h e r s have a r g u e d  il-  23 lustrated by work on HEARSAY (Erman e t a l . , 1980) t h a t i l l u s t r a t e d how r e d u n d a n c y c a n be a remedy f o r incomp l e t e and i n e x a c t knowledge. The t r i c k i s t o f i n d m u l t i ple overlapping sources of knowledge with different areas of s t r e n g t h and d i f f e r e n t s h o r t c o m i n g s . P r o p e r l y u s e d , t h e e n t i r e c o l l e c t i o n o f knowledge s o u r c e s c a n be a good deal more robust than any one of them t a k e n a l o n e . ( D a v i s , 1982, pp. 6-7)  2.3.2.  In C o m p u t a t i o n a l  As  i n the l a r g e r  argued  for  Chandrasekaran, has c l a i m e d range  field  bringing  (Hanson and Riseman,  that  of i n p u t s  tioned. found  AI,  "diverse 1978, p.  several  sources 316;  h i s s y s t e m "must  see  of  Another c l a s s i c  in pattern  use  recognition  several  where  Each  represents a d i f f e r e n t  dimension  derived  by s a m p l i n g  ferent  spectral  because  some p r o p e r t i e s For  example,  quite  clear  This  methodology  Chapter  analysis  to  find  stand the  2. M u l t i p l e  when  over  1979, p.  intensity  He  terrain  been menimages  Chapter  is by 8).  image w h i c h i s  band.  Using  dif-  features  works  bands more t h a n  b o u n d a r y between 7 of Landsat  image  c a n be done  1979,  intensity  spectral  Information Sources  179).  intensity  distinguished  the  wide  belief.  out i n c e r t a i n  i n i m a g e r y from band works  a  (see H a l l ,  from a d i s t i n c t  bands  t o work  classification  cluster  and  Russell  o r m o t i o n has a l r e a d y  of  together"  Flinchbaugh  combines more t h a n one stereo  have  1981, p. 4 4 ) .  (Russell,  multidimensional  others.  also  the consequences of t h a t  advantage  researchers  o f knowledge  be a b l e  f r o m many s o u r c e s "  P r e v i o u s work t h a t take  of  1981, p. 391; and W i t k i n ,  does n o t e x p l o r e  to  Vision  in  land  and w a t e r i s  (near  infra-red).  is flat,  b u t i s not  24 effective  when t h e  surface slope causes  F o r more c o m p l i c a t e d combined 1978; the  effectively  with  Woodham, 1980b). points  in  the  scenes,  digital  intensity  A DTM scene.  If  t h e DTM,  making  remotely and  a  sensed low  illumination  between t h e teristics nique  two of  either  of  the  At  but  digitized  image  imag-  s y n t h e t i c image t o a f o r the e f f e c t s this  p o i n t , the  correspond  work  process  been done a t  concerning  of  difference  to the  SRI  continues  may  be  real slope  charac-  In m o u n t a i n o u s t e r r a i n ' t h i s  more knowledge or more I S ' s  Work has  As  imaging  the  ideally  surface cover. results  some a s s u m p t i o n s  account  angle.  images s h o u l d  p r o d u c e s good  aspects  can  illumination  o b j e c t s p l u s some o t h e r  registering  image, one  of  a  image from  By  elevation  Bachman,  i t i s p o s s i b l e to generate  synthetic  parameters.  and  is  then  ing  (Horn  be  direction  the  scene  m o d e l s can  of  imagery,  the  shadows.  all  i n Landsat  s u r f a c e m a t e r i a l of  terrain  images  r e c o r d s the  known, as  the  o c c l u s i o n s and  so  tech-  that  understood.  more  That i s ,  are r e q u i r e d .  t o combine a map  (Tenenbaum e t a l . , 1978;  data  Bolles  base w i t h  a  e t a l . , 1979).  they e x p l a i n ,  Map knowledge can p r o v i d e i m p o r t a n t c o n s t r a i n t s on where t o l o o k i n an image, what t o l o o k f o r , and how t o i n t e r p r e t what i s s e e n . Such c o n s t r a i n t s , p r o p e r l y e x p l o i t e d , p e r m i t t h e e x t r a c t i o n of complex i n f o r m a t i o n w i t h o u t ext e n s i v e computation. (Tenenbaum e t a l . , 1978, p. 1)  The  map  shapes by  data of  contains  l a n d m a r k s and  location,  Chapter  base  entity  2. M u l t i p l e  three-dimensional  monitoring  name, and  entity  Information  stations. type.  Sources  SRI  locations  I t can  be  and  indexed  r e s e a r c h e r s have  25 produced using the  a  this  of  ships  other  intensity  water  of  This  A method of  The  road  use  suitably has  that  defined,  Chapter  image.  railway  different  t o good  road  used  This  measuring boxcars,  In c o n t r a s t  i s a very  abstract  in kind  from  an  i t is possible  to  operators  has  also  I t u s e s one  input  to produce d i f f e r e n t that are  each p i x e l  internal  i s part  then o p t i m i z e d  program  to  advantage.  e t a l . , 1981).  likelihood  of  reasoning  done  a model of I t has  in  is discussed  the  other  described be  can  i f the  Hanson,  the  belief  i s imprecise,  i t can  information  ( W e s l e y and  been  There are be  the  be  base  shows t h a t  different  can  evidential  plausibility.  will  very  information  (Fischler  sources  encoded  g r a p h and  tion  and  data  of  to  of  a  produce  further in  3.1.2.  Disparate  This  of  utilizing SRI  map  g l o b a l f i g u r e s of m e r i t  "best"  Sect ion  that  types  containing  road. the  at  feasibility  t r a c k i n g roads.  work c l e a r l y  many o p e r a t o r s  images  r e s e r v o i r , counting  information  combine d i s p a r a t e  but  in a  i n a h a r b o u r , and  image.  originated  demonstrates the  i n s e v e r a l a p p l i c a t i o n s , s u c h as  systems examined, the  representation  IS,  system t h a t  technique  volume  counting the  working  be  model  1982;  combined  knowledge has  manipulates  ability  a  both  to e f f e c t i v e l y  u n c e r t a i n , and  considered  3.1. an  2. M u l t i p l e I n f o r m a t i o n  handle  and  informa-  inexact.  Where t h e information  Sources  dependency  support  e x a m p l e s of m o d e l - b a s e d v i s i o n  in Section  been  G a r v e y e t a l . , 1981).  VISIONS s y s t e m u s i n g  that  i n systems  model  systems is  source.  that  explicitly In  other  26 cases  the  carry line  out  model the  knowledge  interpretation.  between t h e s e  2._.  Summary and  This notion  influence current  for  utilize  aspects.  the  pursuits  the  First,  available  framework  to  later  fields this  be  described  input  data  be  an  processes  or  iterations. t o how  from  iconic  a l l the  way  representations  t h a t can  distinction,  information  information.  In a c c o r d a n c e w i t h  will of  be  on  using  information  Evidence presented Chapter  are  several  input  varying  f o r the in  this  be  sources  different  and  will  strongly the and  provides  the to  can  better)  greatly  efficacy  to  containing in their  Several  Sources  is  will are:  "chair".  the  conjectures  emphasis  that  they  "names" l i k e  of c o m b i n i n g  chapter.  2. M u l t i p l e I n f o r m a t i o n  abstract  sys-  IS's  encode d i f f e r e n t  two  three  result  Second,  orthogonal  the  IS's  to  having  i s f e d t o the  intermediate  respect  (more and  a  The  p r o b l e m of how  as  that  i t can  in a dimension  thesis.  outlining  notion  o v e r a wide r a n g e w i t h  2.2.2  draw  of computer v i s i o n  s o l u t i o n s t o the  have been  IS can or  chapter  vary  Finally,  to  that  sources.  sources  outside  the  sources  next  i n the  proposed  an  for  information  Moreover,  information  from t h e  It i s often d i f f i c u l t  d e s c r i p t i o n s i n the  Information  tem  laid  representation.  rationale  procedures  Ahead  multiple  relevant  knowledge  best  has  using the  i n t o the  cases. Looking  chapter  of  i s compiled  previous types  of  of  Section  in l a t e r  chapters  different  symbolic  multiple  types  nature.  IS's  advantages  was were  27 d e m o n s t r a t e d , and  claimed,  for their  ity  to r e s o l v e a m b i g u i t i e s .  one  IS or  in  another.  better  i s not  deal  "experts" they  can  be  with that  inexact  sufficiency  be  a map  i n the  ried  out  disparate  over  may  base or  an  result  types  the  derive a  these  For  t o be  map,  local  Such  the fact  problem  a  savings  as  2. M u l t i p l e I n f o r m a t i o n  Sources  not.  problem  reasons  (the  there  are  on  intensity IS,  "interesting"  need not  be  car-  of a t t e n t i o n  in computation.  there  to  combine  i s the  correspon-  to r e g i s t e r  represen-  in s t e r e o p s i s , before generally,  one  any must  i n f o r m a t i o n where one  c o n t r a d i c t o r y t o what was IS.  the  I f a second  focus  to  together  (the  in trying  More  of c o n f l i c t i n g  out  isolate  operations  First,  derived.  could  one If  then  IS's  carried  disadvantages  image t o e a c h o t h e r ,  allows  example,  can  image.  knowledge.  be  they  in  retrievable,  also efficiency  I t i s sometimes n e c e s s a r y  from a d i f f e r e n t  Chapter  are  in considerable  of  cooperate,  f o r combining  sketch  entire  i n f o r m a t i o n can with  can  f o r edge d e t e c t i o n .  also potential  dence p r o b l e m . of  be  abil-  exist  knowledge.  individuals  operations  image, t h e n  There are  deal  IS's  there  convolution  data  mechanism can  useful  the  i s the  does not  found, or  incomplete  reasons  them),  local  images, s u c h a s  areas  be  s o l v e d more q u i c k l y ) .  many " e x p e n s i v e "  tations  and  know a b o u t  solved with  s u c h as  information  i t may  s o l v e p r o b l e m s t h a t as  p r o b l e m can  Primary  Furthermore, combining m u l t i p l e IS's  Besides can  conclusive  Where  use.  discovered  by  a  KS KS  28 The sistency  Chapter  concepts will  of  ambiguity,  be made more s p e c i f i c  2. M u l t i p l e  Information  incompleteness, i n the next  Sources  and  chapter.  incon-  CHAPTER  3  A Framework o f t h e  3.J_.  Model-Based V i s u a l Computer  types.  Domain  low l e v e l , mation of  vision  as  possible  the u n d e r l y i n g  model-based  Percept ion  s y s t e m s can r o u g h l y  independent  or e a r l y )  or  f r o m an  properties regarding  T h i s work  These  without  applies  a s much  vision  (Zucker et a l . ,  to  1975).  into  two  image-based,  recourse  semantics t o a i d the  systems  Model-based v i s i o n  and  and  to  infor-  knowledge  (also  known a s  objects  i n the  interpretation.  general, Certain  intrinsic assumptions  Only  deal  to influence  such  or  image  model-based  to scenes c o n t a i n i n g  chairs,  those  dependent  tables,  of  t h e p r o c e s s of  certain  roads,  o b j e c t s c a n be  s y s t e m s can t a k e a d v a n t a g e  scene  with  more  reflectance.  applies  such as cubes,  houses.  Domain d e p e n d e n t the  generally  i s c o n c e r n e d w i t h domain  types of o b j e c t s  cerning  known a s  l e v e l ) attempts to i d e n t i f y  f e a t u r e s as edges, s u r f a c e s ,  trees,  divided  t h e images a r e u s u a l l y " made t o make t h e p r o b l e m s  tractable[1 ].  vision.  image  independent v i s i o n o f images  (also  Domain d e p e n d e n t  s c e n e and make use of t h e i r Domain  vision  be  i s concerned with capturing  scene. high  Field  cars,  "recognized".  knowledge  con-  interpretation.  [ 1 ] S o m e t i m e s t h e s e a s s u m p t i o n s r e l a t e t o what c a n appear the s c e n e , making them m o d e l - d e p e n d e n t i f not m o d e l - b a s e d .  in  29  30 For  example, h o r s e s n e v e r  less,  model-based  image-based  have more t h a n  systems  component  must  four  contain  legs.  Nonethe-  some i n t e r f a c e  to p r o v i d e the a p p r o p r i a t e  data  t o an  from  the  image. 3.J_.J_.  Different  The by  first  Roberts  vision  lines,  tic  search.  was  1976,  model,  modelled  et  for a review).  al.,  1979)  relationships  ning  system  the  other prominent  the  (Kelly,  extended features  1971)  i n terms  of  i t s vertices  image v i a  heuris-  for  more  with 9 objects;  see  three-dimensional  cylinders  generalized  a  thereof.  allowed  1972,  the  graph  lets  used  take  ( A g i n and  cones  Duda,  (Marr  idea  ( e y e s , nose, to include  (Brooks et  and  generated  C h a p t e r .3. A Framework of t h e  from  Field  of  heuristic  faces to h e l p mouth).  and  1979;  locate  Fischler  Elschlager, Brooks,  the o b j e c t  the plan-  the r e l a t i o n s h i p s  (Fischler al.,  advantage  Kelly's  knowledge o f  of t h e f a c e  is  one  among them.  features  In t h e ACRONYM s y s t e m prediction  of a c u b e ,  combinations  More complex  or  developed  1978).  known s p a t i a l  all  encoded  (e.g. Falk,  H a v i n g m o d e l s of o b j e c t s  Elschlager  p r i s m , and  be m o d e l l e d a s g e n e r a l i z e d  Brooks  Nishihara,  "world" c o n s i s t e d  matched t o the t w o - d i m e n s i o n a l  be  o b j e c t s can 1973;  The  s i m p l e m o d e l s was  work i n t h e b l o c k s w o r l d domain  to  Mackworth,  t o use  wedge, a h e x a g o n a l  Further objects  Models  system  three-dimensional  and  of  ( R o b e r t s , 1965).  rectangular The  Types  and  among 1973).  1981)  models o f  a  such  31 t h i n g s a s m a c h i n e p a r t s and a i r c r a f t . in  the  graph  restrict  the  The c o n s t r a i n t s  possibilities  embodied  when a n a l y z i n g  the  image.  In B o l l e s ' s work on models system  of  the o b j e c t s  t o run almost  models  of  dictate  what  Model  the  totally  top-down.  as  knowledge c a n be c a p t u r e d between  knowledge  general about  structure  reflecting  the c o n t r o l  tuples  in a relational  (Bajcsy  1978).  complete  t h e models c a n  ways.  Freuder  a b o u t t h e domain  (Freuder,  1976).  were p r o c e d u r e s embedded  data  and His  in~a tree  r e g i m e u s e d f o r image  base  et a l . ,  The  model  in a procedural (Jain  The  explicitly  interpre-  an  the  form  1976),  in  a  knowledge  can  also  1975)  1981), i n  semantic rules  network  (Bajcsy  be  Kuipers  and  implicitly  or u s i n g  fuzzy set  source image.  utilized  by t h e above  r e p r e s e n t e d p e o p l e ' s map  3. A Framework o f t h e  systems  O t h e r I S ' s have been u s e d s u c h  d r a w i n g s ( o f h o u s e s : M u l d e r , 1979; of human body 1982).  of  1982).  information intensity  in  ( L e v i n e and Shaheen,  system ( S h i r a i ,  and Haynes,  input  generally  Chapter  has  tasks,  and L i e b e r m a n , 1974), o r i n p r o d u c t i o n  encoded  Browse,  image  c a n be r e p r e s e n t e d  terms of frames (Sakai  as l i n e  1975),  o f hammers.  The model  is  one  in various  knowledge  the  modules  theory  (Bolles,  be l o o k e d f o r .  knowledge  Joshi,  When  in inspection  "active"  tation  vision  (machine p a r t s ) and t h e s c e n e e n a b l e t h e  objects,  should  distinguishes particular  verification  Field  parts:  knowledge  and  32  how  t h e y use Two  stages  i t t o move a b o u t  tasks are prevalent of  visual  segmentation—also 1973;  Few of  1976;  1973;  of  houses,  input  IS  the  1981, system  (Hanson  for  cars,  Barrow,  applied  1979)  (Yakimovsky  encompassed  and  and  1976;  FeldBarrow  1978a; Weymouth,  and Riseman,  some  and  intensity  sources.  s h o r t - t e r m memory  memory  has been  several  more  1978a  recent  1981).  aspects  Two  other  (particular  (general  knowledge),  b a s e d on  levels  of a b s t r a c t i o n .  through  segments,  regions,  scenes  objects.  There  image) and a p l e t h o r a major  and  repositories  large,  composed  is a of  single  internal  of knowledge  knowledge) and  the  levels  surfaces,  go  volumes,  are  long-term  both of which a r e h i e r a r c h i c a l The  of  1978b;  work) i s a  for understanding natural  trees,  (a c o l o u r  information  Yachida et a l . ,  first  The VISIONS s y s t e m , d e v e l o p e d a t t h e U n i v e r s i t y  Weymouth,  comprehensive  1975;  Tenenbaum and  1977).  the  knowledge  s y s t e m s have  M a s s a c h u s e t t s a t Amherst see  with  Model  Hanson and Riseman,  comprehensive  vision.  (Shirai,  (Kuipers,  dealing  known a s r e g i o n m e r g i n g  Tenenbaum,  and Tenenbaum,  settings  i n systems  processing.  t o b o t h edge d e t e c t i o n  man,  i n urban  from and  and  vertices objects  to  schemata.  The used  schemata  to represent  mation of what  required  t o t h e VISIONS s y s t e m .  both o b j e c t s  to b u i l d  i s represented  instantiation  Chapter  are c e n t r a l  and  scenes.  up t h e volume  i n an  image.  of hypotheses d u r i n g  3. A Framework o f t h e  Field  They  provide  (or s u r f a c e )  Finally,  They  infor-  description  schemata  interpretation.  are  guide the  The  system  33 follows  a  verify) could  hypothesize  which,  exhibit  The  depending either  (Brooks  et  al.,  modelled  in three  triction" ships  matches.  From  models  generalized  on  1981b). cones,  the  input  and  Stanford Objects,  and  their  graph.  The " r e s -  spatial  relationa  "predic-  t o image  feature  "interpretation"  graphs a r e b u i l t  of an image.  IS i s t a k e n  from t h e r e s u l t s  two-dimensional  m a n i p u l a t i o n system, i n the r e s t r i c t i o n  ribbons  are  a  line  found.  these ribbons are  graph.  more s p e c i f i c  of  finder.  Using a con-  matched  against  As t h e i n t e r p r e t a t i o n  graph i s  m o d e l s c a n be d e d u c e d  ( e . g . an  L1011  a type of wide-bodied j e t ) .  An of  important aspect of these v i s i o n  program  operate  control.  working  towards  dent  systems  generally  but  most c o n t a i n some s o r t .  Chapter  Most  i n a bottom-up,  and  of  and  at  From t h e s e r e p r e s e n t a t i o n s ,  "Observation"  constructed, is  developed  1981a  involved,  control.  r e p r e s e n t e d i n an " o b j e c t "  objects.  (focus-expand-  sources  i s generated to hypothesize object  this,  straint  Brooks,  dimensions as  the a n a l y s i s  The  knowledge  graph c o n t a i n s c o n s t r a i n t s  graph  during  the  s y s t e m h a s been  1979;  are  between  tion"  on  mode o f c o n t r o l  b o t t o m - u p o r top-down  ACRONYM v i s i o n  relationships  and t e s t  linear  independent  fashion  the p o s s i b l e have a w i d e r  either  Such  domain  systems  objects  systems  range  i n i t . Domain of c o n t r o l  have t h e p r o b l e m  3. A Framework o f t h e F i e l d  the  matter tend t o  beginning with the  a top-down component  systems  is  image depen-  strategies,  or a feedback of  deciding  loop which  34 object  to  hypothesize  Mackworth, once  object-specific expectations  clues  derived  try  to plan  their  be s e e n  the  bridges,  dissertation  i n remotely  maps  (Thorndyke,  p r o j e c t s a t the  utilized  models  Tavakoli,  1973), m o d e l s e x i s t  bridges,  of  (Bajcsy three  a  of  list  texture);  relationship  Figure  and T a v a k o l i ,  o f images.  images.  Objects  imagery  on  University  of  entities.  f o r water, 3.1 shows  one might  include  the  Work h a s been  to bridge:  below).  computational knowledge.  3. A Framework o f t h e F i e l d  have  rivers,  lakes,  description This  of  simple  (e.g.  aspect  a  model model:  a r i v e r has  (e.g. texture  between o b j e c t s  The l a s t  infor-  In one ( B a j c s y and  t h e model  properties  relationships  roads,  Pennsylvania  land,  1973, p . 2A-57).  those  objects  o f how p e o p l e a c q u i r e  p r o p e r t i e s which d e f i n e  and  priori  important  o f t h e f o u r major p r o p e r t i e s o f a g e n e r i c  restrictions  homogeneous);  use of  A  provide  1979) and s e v e r a l  geographic  and i s l a n d s .  contains  Chapter  both  s y s t e m s have been d e v e l o p e d w h i c h use s u c h  Several  but  as w e l l as the context  h o u s e s , m o u n t a i n s , and f i e l d s .  mation  river  efficient  strategies.  or s a t e l l i t e  the c o g n i t i v e aspects  vision  1975),  c o n c e r n s m o d e l s of  sensed  done t o s t u d y from  make  the understanding  to d i s t i n g u i s h i n a e r i a l  rivers,  Palmer,  models  interpretation  and egg p r o b l e m :  Models  domain o f t h i s  might  can  from  from p r e v i o u s  problem:  programs  heuristics  Geographic  that  these  (and c u e s ) t o g u i d e  The  as the chicken  1977b; o r t h e p a r s i n g  determined,  resulting  (known  (e.g.  is  spatial  o f a model  isa  35  R i v e r s : G r a y v a l u e of t h e w a t e r T e x t u r e : homogeneous B o u n d a r i e s : open Contrast: large S p a t i a l r e l a t i o n s h i p s t o b r i d g e : below T o p o l o g i c a l r e l a t i o n s h i p s : continuous S p a t i a l r e l a t i o n s h i p s to land: surrounded Figure  mechanism match  for determining  those  derived used  (Bajcsy  feld,  1980). Several  several  i n the  and  other  of  i f the  a  1976;  researchers at  River  f e a t u r e s of  In a s i m i l a r  Tavakoli,  Scientists types  A M o d e l of  model d e s c r i p t i o n .  procedurally.  roads.  3.1  have  information  about  a  specific  In t h i s  system  image  these  s y s t e m , m o d e l s of  roads  see  and  also Tavakoli  have d e v e l o p e d  SRI  by  developed  systems  are were  Rosen-  to  find  a method o f  using  roads.  The a p p r o a c h i s b a s e d on a new paradigm for combining local information from m u l t i p l e and p o s s i b l y incommens u r a t e , s o u r c e s , i n c l u d i n g v a r i o u s l i n e and edge detection operators, map knowledge a b o u t t h e l i k e l y p a t h of r o a d s t h r o u g h an image, and generic knowledge about roads (e.g., connectivity, and width c o n s t r a i n t s ) . ( F i s c h l e r e t a l . , 1981, p. 201) They d i v i d e a l l t h e i r  information  I,  where  almost  false  II,  where  possible  false  elements  parameters  are  accurately  relevant  no  applied  in d i f f e r e n t  maximum  likelihood  Chapter  road  ways and scores  3. A Framework of  sources  i n t o two  elements are  a l l the  are  Field  accepted,  accepted  measured. results  (using assumptions  the  c l a s s e s : Type  are  and  but The  Type local,  IS's  optimized  of c o n t i n u i t y  are for and  36 width) t o d e r i v e is  that  the  the "best"  operator  track.  that  numerical  likelihood  estimate  multiple  IS's  combined  representation  are plus  a d hoc  A requirement  produces  the IS's  them must a l s o s u p p l y  f o r each p o i n t by  for  reducing  modifications  i n the array.  a  The  them a l l t o a common in  the  optimization  procedure.  Work was a l s o done a t SRI on map-guided was d e s c r i b e d  in Section  Selfridge  uses  buildings,  roads,  creditable  feature  tors  which allow  having  2.3.2.  "appearance" models and  shadows  of h i s system  contrast,  mechanism t h a t a c c o m p l i s h e s  for  i s t h e use o f  etc.  similar  et  aerial  board  and t h e knowledge  models  f o r such  roads,  and  procedurally tion  more s p e c i f i c loop ous  allows  i s encoded  regions  (cf. Section  fields,  as p a r t  of "model-driven  ones  Chapter  bare  system  on  regions)  rules.  Object  grass  land,  but are a v a i l a b l e Interpreta-  segmentation) t o and  to a i d i n the c l a s s i f i c a t i o n  3. A Framework o f t h e F i e l d  (Nagao black-  soil,  (region  for  a  subsystems".  properties  regions.  for a  a f t e r HEARSAY-II  explicitly  ( s e l e c t i o n of cue  the context  4.5.2  a vision  i n production  are not s t o r e d  from g e n e r a l  A  when t r y i n g t o f i n d  i s available  rivers  proceeds  1981).  tasks.)  Common d a t a  items as crop  as  opera-  photographs modelled  a l . , 1978 and 1979).  such  adaptive  Nagao and h i s c o w o r k e r s have d e s c r i b e d interpreting  entities  ( S e l f r i d g e and Sloan,  parameters to vary  c e r t a i n areas,  interpretation. It  a  feedback o f ambigu-  37 Many maps u s e a b s t r a c t as  • for a building, — j — | — | — | — | —  containing only on  only  a little  such s t y l i z e d  the information  called  Mapsee  different found line  in  for a railway.  symbols  is  a  A line  sketch  found  map.  map  (see F i g u r e 3.2).  a  to  interpret  (Mackworth,  use o f s k e t c h a t a b l e which  program  sketch  With only  maps  1976c; s e e B a l l a r d e t a l . , 1978 f o r a  maps i n  vision).  i s indexed  Model  END).  (Mackworth,  3.2 A S k e t c h Map  3. A Framework o f t h e F i e l d  knowledge  by p i c t u r e c u e s  ambiguous but network c o n s i s t e n c y  Figure  such  drawing  in a sketch  d r a w i n g s s u c h a s T E E , OBTUSE L, a n d FREE  Chapter  objects,  p r a c t i c e , one c a n f o r m a c o g n i t i v e map b a s e d  Mackworth d e v e l o p e d  are  symbols t o r e p r e s e n t  is  ( f e a t u r e s of The  cues  1977a) i s used  38 to reduce the  the  label  relationships  bridges)  been  Havens, language  river  are on  This  Maya the  is  domain  for  that  be  find  a schema  the  instances  Chapter  3.  1980;  into  hierarchy that  the  f o r a l l the  ele-  to roads,  Associated  a  with  to  coherent  3.3) carry in  hierar-  network.  in Figure  under  the  addition  specialization  i s shown  exists  rivers,  in Section  Another  and  and  in  them p l u s a mechanism  i s completed.  schemata  Mackworth  written  i n more d e t a i l  by  3.3  geosystems  creating hypothetical  correspond by  closed  the  to each c h a i n .  s h a p e s of  curves,  the  The  and  the  can  be  instances.  i n the must  example,  The  be  A Framework of  hierarchy  c o n s i s t e n t with  r o a d s must  the  be  Field  next  for  These are  res-  Thus,  blobs,  ambiguous and  newly c r e a t e d  decomposition  instances  chains.  towns must be  However, many s h a p e s a r e  multiple  For  "world".  of d e c o m p o s i t i o n  might  inverted V s .  schema.  discussed  begins  i n some c a s e s must  erate  the  problem--Mapsee2--has  through chains  instantiating  use  same  Mackworth,  to  under  table.  Schemata e x i s t  up  flow  of  3.4.  Recognition entities  on  hierarchy  in Figure  lines  be  decomposition  specialization  tricted  so  advantage  rivers  i s schema-based and  1978).  instantiation the  the  (Havens and  taking  (e.g.  the  from p o i n t s  and  to organize  complete  found  objects  system  (which w i l l  when t h a t  chies  i s done by  solve  (Havens,  procedures  Mapsee2  to  developed  systems,  schemata  between  system  1981).  ments of  This  t h a t were u s e d t o b u i l d  A second also  sets.  shore-  mountains can  instances  to complete  gentry  to  s p e c i a l i z a t i o n s of to regions  labelled  to but the as  Figure Chapter  3.3 Mapsee2 D e c o m p o s i t i o n  3. A Framework  of t h e F i e l d  Hierarchy  40  Figure  land.  are  instances,  pruned.  separate  creation This  Mapsee2 S p e c i a l i z a t i o n H i e r a r c h y  Inconsistent  bridge, form  3.4  of  continues  such as  Hypothetical,  a  Unconnected  new  instances,  higher  level  river  mutually  specializations.  f o r a l l the  possible  this  extended?  f o r Geosystems  going  exclusive instances  s u c h as  over  instances cause  road  interpretations  might  work be  Mackworth  for  states,  . . . i n the l o n g run, . . . u n d e r s t a n d i n g [of LANDSAT images] would p r o c e e d more s u c c e s s f u l l y i f p r o g r a m s were a b l e t o a c c e p t a d v i c e , i n t h e f o r m of s k e t c h maps, a b o u t  Chapter  3.  A Framework of  the  Field  the  systems.  chain.  How  a  each  41 the geography 55)  u n d e r l y i n g the  This dissertation 3.2.  i s concerned  Possibilities  models form  and  data.  t h e domain and  incorrect  interpretation,  Figure  bottom-up  occur  3.5  objects.  image  include  regions that  (where  might  ship  illustrates  edges  intensities  between  one a  region,  (arbitrarily  An  like  from say  t o one  the  roads,  problem  of  onto.  input  Some scene  and  objects  and b r i d g e s . t o one  i f rivers  relation-  expect best  identified  to  if  as a  be one  river  t o keep i t d i s t i n g u i s h a b l e ) .  exists  where more t h a n one  of a m o d e l .  situation  where t h e r e  evidence  Field  can  from an  i t would be be  that  discontinuities)  a one  So,  image,  river-2,  3. A Framework o f t h e  is  (data) to models of  rivers,  the requirements  Chapter  the  IS d e r i v e d  intensity  satisfies  is  and  image  i n an  a model.  situation  models  situation  possibilities  r e g i o n - 1 2 , can  numbered, as  inconsistent  Besides  to d i s c o v e r  the  between  then the  The  a r e homogeneous).  datum and  matched w i t h a r e g i o n particular  from  include  would  range.  five  (actually  mapping  t h e wrong model w i t h a datum,  one  Example e l e m e n t s  a  i s top-down  the  the  elements  be m o d e l l e d  Ideally,  be  p.  system.  considered  relating  1978,  to Models  evaluation.  not always  when m a t c h i n g  scene  be  data elements  for  r e l a t i o n s may  Data  I f the e v a l u a t i o n  reversed  the  can  (Mackworth,  w i t h such a  i n Matching  Image i n t e r p r e t a t i o n  image.  Alternatively,  both  for  and  datum  this  against  is a an  Figure  Chapter  3.  3.5  Possibilities  A Framework o f t h e  i n Matching  Field  Data  to Models  43  interpretation. sources, can  With  regard t o combining  the data elements  in fact  cause  IS's c o n f l i c t  rather  than  t h e model  find  the data  Ambiguity vision  i f no d a t a  dent be  linked  with  organization the o n l y f a c t s  Finally, scene  either tem,  fulfill  the requirements  whether  than  features  this  "special  ble  of  computer  i s domain  i n f o r m a t i o n such  so t h a t  facing  Model-based  single  t h e r e c a n be d a t a e l e m e n t s c a n be f o u n d .  model e x i s t s case".  Until  i n model-based  One c r i t e r i o n  as  depen-  they can  systems can the  spatial  image e l e m e n t s  a r e not  This  they handle  ticular,  the problems  f o r which  can occur  vision  systems this  no mapping t o  f o r two r e a s o n s :  represented i n the  but i s t o o r i g i d l y  defined  to accept  i n c l u d e modules  will  remain  a  sys-  capa-  difficulty  vision.  for evaluating  how w e l l  Chapter  possibilities  a r e ambiguous when  one m o d e l .  a b s t r a c t i o n and l e a r n i n g ,  encountered  of a  available.  element  the  problem  the system  t h e " c o r r e c t " model h a s n o t been or  different  t h e model.  Image f e a t u r e s  more  from  This  Here m u l t i p l e i n f o r m a t i o n  the g r e a t e s t  o f more g l o b a l of  data  t h e r e a r e more  satisfy  It exists  or independent.  take advantage  a  will  since  i s generally  systems.  t h e same I S .  i f the  i s unsatisfied.  that  from  information  i n agree.  s o u r c e s c a n be an a d v a n t a g e , to  n o t come  great d i f f i c u l t y  On t h e o t h e r hand, model, t h e n  need  multiple  model-based v i s i o n  the four problems of i n c o n s i s t e n c y  3. A Framework o f t h e F i e l d  described  systems  above.  is  In p a r -  and u n s a t i s f i a b i l i t y  are  44  germane. the  These  solutions  3^.3.  that  Knowledge Knowledge  totally tures  Representation  between  Schema-based  (Bartlett,  scenarios,  associated  utilized.  can Also,  semantic  1932)  or  frames  are  from struc-  modification. reasoning  a  compromise  (Minsky,  1975) ( o r  i n a convenient methods  schemata  that  are related  network,  (methods  part  advantage  allow  (the  are  developed. knowledge  many  Their  been d e s i g n e d .  how e x p l i c i t  Since  explicit  t o be drawn  way,  from t h e  they c o n t a i n  facts),  of  schemata  both  a  l i n k s ) and a p r o schemata c a n t a k e  that  c a n be j u d g e d on how w e l l  to carry  out the  tasks  for  have they  a s new i n f o r m a t i o n  3. A Framework o f t h e F i e l d  a r e they  i s discovered?  been encode  which  Q u e s t i o n s of d e s c r i p t i v e adequacy  a r e t h e f a c t s ? how u n i f o r m l y  be m o d i f i e d  Further-  of each.  varieties  merits  required  i n an  inferences  f o r manipulating  way.  by w h i c h t h e f a c t s may be  f a c t s a n d t h e network  of the s t r e n g t h s  There  e t c . ) a r e a means by w h i c h  the  cedural  Chapter  continuum  for  representations  i n which they a r e c o n n e c t e d .  facts,  a  Declarative  available  encode  part  have  l i e in  u n i t s , templates,  declarative  the  6.7.1).  t o c o n t r o l a n d c a n make  f a c t s c a n be r e l a t e d  they  manner  regard to  Adequacy)  procedural.  and  with  t h e two e x t r e m e s .  scripts,  a  explicit  systems a r e e a s i e r  Schemata  more,  schemes  to t o t a l l y  make knowledge  (Section  (Descriptive  representation  more d i r e c t .  be examined a g a i n  have been d e v i s e d  declarative  Procedural  in  problems w i l l  they  include:  encoded?  can  how m o d u l a r  45 are  individual  schemata?  representation baum,  capture  This  distinguishes  particular  ing  i s the only to  the  name/value pointers the  invoked,  link  pairs.  Maya  schemata  that  schemata  schema-specific  are  values  The  instance  encoded can  tuplebases  procedures  as  be  in  mean-  attribute  which  of  pattern-  may be s t o r e d .  ( ( 5 7 . 33) (0.886914  side2:  *CHAIN-5  side2-desc  ((69  regions:  (*REGION-1  *REGION-2  C/labels:  ((*CHAIN-3  . *BRIDGE)  Q/models:  ((*RIVER-SYSTEM *RIVER-SYSTEM-1) (*ROAD-SYSTEM *ROAD~SYSTEM-1))  instance->  *BRIDGE  . 0.461934) 54.1202)  . -0.396436) 47.918062) *REGION-3 (*CHAIN-5  3.6 A Maya I n s t a n c e :  3. A Framework o f t h e F i e l d  relation-  combinations  sidel-desc:  Chapter  t h e use  S-expressions,  *CHAIN-3  . 62) (-0.918062  Maya  t h a t has a s p e c i a l  ( e . g . *CHAIN-5), o r  supports  bridge.  sidel:  Figure  the  bridge  knowledge t h r o u g h  respectively.  between  also  1978) schema f o r a  from g e n e r a l  Note  does  See (Barrow and Tenen-  of a p a r t i c u l a r  interpreter. - Facts  to other  two.  (Havens,  i s an i n s t a n c e  i n s t a n c e s and o b j e c t s  ship  and c o r r e c t l y  considerations.  3.6 shows a Maya  i n Mapsee2.  completely  our i n t u i t i o n s ?  1975) f o r some o t h e r  Figure  of  how  *REGION"4  *REGION~5)  . *BRIDGE))  *BRIDGE-1  4 6  A sample is  displayed  have  FRL frame in Figure  more p a r t s  Besides  (Roberts 3.7  (from  t o an a t t r i b u t e ,  t h e name and v a l u e ,  REQUIRE  predicate  which  there  and G o l d s t e i n , Rosenberg, called  1977a a n d 1977b)  1977). . FRL  a slot,  t h a n Maya  c a n be a d e f a u l t v a l u e  restricts  the values  frames  t h a t may  does. and  f i l l the  SUPPLY AKO  $VALUE  thing  domain  $DEFAULT  carryover,  old-carryover  $1F-ADDED $1F-REMOVED $DEFAULT  total-supply, total-supply, 0  carryover carryover  product ion  $IF-ADDED $IF-REMOVED $DEFAULT  total-supply, total-supply, 0  carryover carryover  imports  $IF-ADDED $IF-REMOVED $DEFAULT  total-supply, •total-supply, 0  carryover carryover  total-supply  $DEFAULT  total-supply  use  $DEFAULT  domestic,  domestic  $IF-ADDED $IF-REMOVED $DEFAULT  carryover, carryover, 0  total-use total-use  exports  $IF-ADDED $IF-REMOVED $DEFAULT  carryover, carryover, 0  total-use total-use  total-use  $DEFAULT  total-use  carryover  $DEFAULT  carryover  date  $DEFAULT  nowc  Figure  Chapter  production,  exports  3.7 An FRL Frame: SUPPLY  3. A Framework o f t h e F i e l d  a  imports  47  slot.  There are  a l s o IF-ADDED and  IF-REMOVED  evaluated  whenever a v a l u e  i s a d d e d or  question.  T h e s e demons can  be  changes to the The kind  one  or  inheritance. how  t h e AKO  AKO  general  links  The also are  of  slots  value  by  ture  t o the  which  are  slot  in  effect  of  i n FRL  link  (a  can  links  be  form a  used  no  particular  (see  Brachman,  for  clear  knowledge and 1979;  i n t e g e r or  that are As  with  values  like FRL  t h a t can  be  i s one  between  knowledge.  (frames).  Slots  t o be  (the type confused in  of with  FRL),  and  inheritance hierar-  inherited  This provides  represented  1980)  require clauses there  an  Friedland,  in units  not  on  for  distinction  S m i t h and  string;  property  1982,  particular  basic attribute  special-  restrictions  n o t i o n s have been  i s no  (Stefik,  i s t h e AKO  are  another  type  determined of  struc-  knowledge.  properties, slots form o t h e r  part-of role,  Chapter  the  as  roles.  in effect  decomposition  used  there  inheritance role.  Besides  the  such  However, t h e  their  the  names, v a l u e s , d e f a u l t s , d a t a t y p e s  restrictions  inheritance chy.  t o be  Also,  as  from t h e  spread  These  which  there are  stereotypical  composed of  value  type  many d i f f e r e n t  UNIT p a c k a g e  has  the  are  how  links). or  hierarchy  However,  explanation with  to  i t s i n v e r s e , INSTANCE.  subset  removed  which  base.  distinguished slot  o f ) and  ization  data  used  forms  and  can  hierarchies.  i t s inverse,  h i e r a r c h y , and,  3. A Framework of  have o t h e r  the  a  In p a r t i c u l a r ,  super-unit,  "relation"  Field  definitional  role  which  roles  there  is  forms  a  which p r o v i d e s  a  48 general  mechanism  for describing  c e d u r e s c a n be a t t a c h e d can  to units,  be p a s s e d between u n i t s  been  used  (Stefik,  class  A class  properties. Employing  A  slots.  the  These  building  the notion  classes--that instances.  relations  Pro-  messages  mechanisms  of  expert  have  systems  relation  is  a  mapping  basic  operations  destroy,  fetch, and  have  F o r example, that  share  between classes  PART-OF  the Student not  enti-  and  relations  t h e PSN d e v e l o p e r s  hierarchies—both  of p r o p e r t i e s .  not  common  two c l a s s e s .  possessed  of  There i s  c l a s s whose i n s t a n c e s  properties  would  the basic  which  and t e s t ) ,  of a m e t a c l a s s - - a  Average-Age  on  i n the inheritance  can  from  ( L e v e s q u e and M y l o p o u l o s ,  i s a c o l l e c t i o n of o b j e c t s  can take part  perty  and d a t a t y p e s ;  language b u i l t  binary  IS-A ( s p e c i a l i z a t i o n )  which also  and  simple  (create/assert, built  in  i s a representation  of  1979).  successfully  or  slots,  relationships.  1980).  PSN ties  binary  are also by  those  m e t a c l a s s may have t h e p r o make  sense  i n the Person  instance.  Numerous o t h e r been KRL  developed.  frame-based Some  (Bobrow and W i n o g r a d ,  tualizations  knowledge  Chapter  languages  representation  1977) have d e v e l o p e d  knowledge more  to  directly.  3. A Framework o f t h e F i e l d  Lisp  languages  elaborate  a n d i m p l y a c e r t a i n model o f memory.  Maya, c a n be v i e w e d a s e x t e n s i o n s access  representation  that  like  concep-  Others, allow  have  one  like to  49 3.4.  Knowledge  Representation  (Procedural  Adequacy)  . . . a d a t a s t r u c t u r e i s n o t k n o w l e d g e , any more than an e n c y c l o p e d i a i s k n o w l e d g e . We c a n s a y , m e t a p h o r i c a l l y , t h a t a book i s a s o u r c e o f knowledge, b u t w i t h o u t a reader, t h e book i s j u s t i n k on p a p e r . S i m i l a r l y , we o f t e n t a l k o f t h e 1 i s t - a n d - p o i n t e r d a t a s t r u c t u r e s i n an Al d a t a b a s e a s knowledge p e r s e , when we r e a l l y mean t h a t t h e y r e p r e s e n t f a c t s o r r u l e s when u s e d by a cert a i n p r o g r a m t o behave i n a k n o w l e d g e a b l e way. ( B a r r and F e i g e n b a u m , 1981, p . 143)  Schemata  i n a semantic  a  base  data  which  network c a n be c o n s i d e r e d i s manipulated  versely,  procedures attached  process  of  interpretation.  object-oriented  The  in  (Robson  an  and G o l d b e r g ,  of d e f i n i n g  sent  by  based  l a n g u a g e s a l l o w messages  other  objects.  I n an image u n d e r s t a n d i n g  the  within  existing  schema,  oriented  river-system  of  systems  for effecting  Chapter  control  the  i s an example o f  of  which  respond  to  messages  a r e o b j e c t s and most between  ( e . g . UNITs a n d  message  "A r i v e r  h a s been  might found  response might  checks i f r i v e r - 6  schema-  schemata ( e . g .  system, a  The a p p r o p r i a t e  should  come i n the  be  that  be added t o an  instance.  procedural  adequacy  i n c l u d e : how e f f e c t i v e interpretations?  3. A Framework  language,  ( e . g . F R L ) , or both  saying,  s y s t e m schema  Questions  tures  schemata  i t i s river-6".  river  case  will  t o be s e n t  KRL).  image,  can  Con-  1981) i s a p r i m e example, c o n -  Schemata  slots  river  latter  o b j e c t s and how t h e y  Maya),  a  schemata  object-oriented  sists  from  by s e p a r a t e p r o g r a m s .  as  programming.  Programming Smalltalk  to the  collectively  of the F i e l d  relevant  to  are the c o n t r o l  how c o n v e n i e n t  is  objectstruci t to  50 "program" ferent sages  with  how  how  procedures attached  schema-specific  must  to  the  difmes-  be?  (Havens  butes  is  and  written a  a  schema-based  Mackworth,  i n Maya schema  may  have  of methods.  variables  are  matching.  Hence e a c h  from  which  chosen. access  the  bridge  or  others.  entity.  instance,  I f so,  associated  that  with  s y s t e m s , or  decomposition  Chapter  3.  chain  schemata The  current  one  (line) roads,  appropriate  and  determine  whether  higher  i s suspended  and  objects,  s u c h as  level  geosystems, are  hierarchy.  A Framework of  the  to  Field  initiated  to  and  functions will  that  be  one  can  succeed.  in  an  almost  is a possibly rivers,  towns,  schema-specific  possibly hypothetical,  procedure  so  a  pattern  context  maps  chain  s u c h as  by  does not  sketch  Each  attri-  retrieved  local  mechanism  first  which  the  context  several  i n the  i f the  fashion.  c e d u r e s examine the an  one  1981)  of  are  current  have  system  to a database, c a l l e d  (functions)  interprets  for several  sides,  creating  is a pointer  schema can  functions  Havens,  In Maya, one  i s also a backtracking  bottom-up  ambiguous cue  1978).  appropriate  other  understanding  Mackworth and  Methods  Mapsee2 c u r r e n t l y strictly  image  bound t o e s t a b l i s h t h e  There the  1980;  (Havens,  tuplebase,  river  w e l l can  schemata c o o p e r a t e ?  Mapsee2  is  them?  it  supports  represent the road build  pro-  that  functions systems, up  the  51 3_.5_.  G r a c e f u l Degradation  A computer degradation conditions  program  if  (Robustness)  i s said  i t s performance  f o r which  conditions  f o r a robust  will  not prevent  one  input  (Marr,  a r e n o t met.  program  t o be t h a t  delivering  1976, p . 4 8 6 ) .  at  still  be a b l e  to analyze  s y s t e m t h a t does n o t d e g r a d e  infinite  l o o p because  graceful  Marr  least  some  gracefully  i t "expected"  the  might  of  data the  sentence i s  program, then  p a r t s of  defines  "degrading  I f an u n g r a m m a t i c a l  t o a n a t u r a l language u n d e r s t a n d i n g  system w i l l A  from  or d i s p l a y  d i m i n i s h e s g r a d u a l l y when t h e  i t was d e s i g n e d  the  answer"  t o be r o b u s t  a robust  sentence[2].  just  go i n t o an  a p a r t of the sentence  that  was n o t t h e r e .  Graceful tems.  degradation  Both the user  when c o m m u n i c a t i o n "graceful  important  not  interaction"  by  complete.  (keeping  1979; Hayes e t a l . , 1981).  t r a c k of  what  the  from  internally  external descriptions),  (what ing  the system  description  i s doing  has  been  They have  interaction,  identification from  This  sys-  called  w o r k e r s on t h e S p i c e p r o j e c t  s e v e r a l components o f g r a c e f u l  tracking  in interactive  and t h e s y s t e m must have ways o f r e c o v e r i n g is  (Hayes and Reddy, fied  i s very  identi-  including  conversation  (matching  a t CMU  objects  is  focus  about),  represented  an e x p l a n a t i o n  and why), and p e r s o n a l i z a t i o n  facility (adjust-  t o the p r e f e r e n c e s of the u s e r - - c f . S e c t i o n 4.3).  [2]People  Chapter  are very  robust  understanders  3. A Framework o f t h e F i e l d  of n a t u r a l  language.  52 Graceful tion  sources.  remaining As  an  one  degradation I f an  source(s)  IS  is also  r e l e v a n t to combining  i s removed, can  to advantage,  the  or does  system u t i l i z e  it  fail  e x t r e m e example, c o n s i d e r s t e r e o p s i s and  IS  i s removed, t h e n  directly IS's,  from  i t is  a monocular  s t e r e o s y s t e m s can  impossible  image.  not  IS's.  calculate  i n t e r m s of  degrade g r a c e f u l l y  the  completely?  i t s two  to  Hence,  informa-  If  depth  i t s input  when o n l y one  IS  remains.  In  systems combining  sources,  more  using  an  data  base,  able  the  DTM,  results  correct  nothing  desirable  3.  can  then  an  t o be  as  Thus, w h i l e  f o r a l l computer  sources.  A Framework of  the  Field  is  (a  map  still  be  image  One  would  the a d d i t i o n a l  almost  always  those  i t can  be  not  I S ; how-  better  i s a generally  programs,  s y s t e m s and  I f one  system w i l l  image.  robustness  information  developed.  a robust  are  input  intensity  good w i t h o u t  results  of  be  o n l y the o r i g i n a l  in interactive  information  types  interpret  s k e t c h map)  partial  at a l l .  geous p r o p e r t y  Chapter  solutions  " a i d " IS t o h e l p  to f u n c t i o n with  expect ever,  robust  different  than  advanta-  particularly  attempting  to  combine  CHAPTER How  4_.J_.  t o Combine M u l t i p l e  Information  Sources  I ntroduct ion Earlier  of  4  in t h i s  combining  This  point, there  IS's  chapter The  so  was  is s t i l l  t h a t t h e y may  i s concerned  with  major h y p o t h e s i s  information  a case  begun  multiple information sources.  as a s t a r t i n g combine  thesis  sources  of  be  i n image  solution  this  is useful";  Taking  a r e a way  goal.  also  thought  which can that  be  may  assembled  exploit  different  The  tools  described  categories  corresponding,  Descriptive  adequacy  of  and  schema  (Section can  be  encoded  4.2). very  i s described  of  The  to  the  of  best  to  problem.  is  "Combining i s that  realizing  the that  of as d e s c r i b i n g t o o l s a  vision  system  inputs. can  be  last  divided  into  three  s e c t i o n s of C h a p t e r  for creating  f u n c t i o n s f o r schemata t h a t comes d i r e c t l y  means by  paradigm  which user  a  new  3. type  manipulation  from  the  user  i n f o r m a t i o n can  be  (4.3).  adequacy  perception  of  below  Information  Procedural cycle  types  of  5 ) t o form  i s the m o t i v a t i o n  a library  useful.  be  (see C h a p t e r  can  to that  the minor h y p o t h e s i s  chapter  chapter  that  interpretation.  dissertation  methods d i s c u s s e d i n t h i s This  usefulness  t h e q u e s t i o n of how used  one  f o r the  as  is primarily the  the  result  basic control  regime  of  using  (4.4).  a The  53  54 cycle  contains  (4.4.1)  and  convenient can  be  used  mixed c o n t r o l s t r a t e g y  Several  that  expected  (4.5.2) . the  context  permits  be  scene  based  on  connects  Chapters that in  w h i c h can  (4.6).  5 are  be  sources  them u s i n g  representation  languages d e s c r i b e d language  perspective, suffice  for  domains t h a t Chapter  relaxed  weaker  4.  one  can  have an  to to  more  mechan-  sources  means (4.7).  4  introduces  multiple  information  describes  three  implementation of  with  incremental  Chapter  so  those  specific  of  a  system  or  sche-  tools.  System s y s t e m s have u s e d medium. 3.3  would hope t h a t domains;-  5  given  thresholds  information  instances  in Section  Combining  a  evidence  inference  flexible,  actual  understanding  all  an  to u t i l i z i n g  the  image u n d e r s t a n d i n g a  user  graceful  knowledge  linked.  Chapter  and  a  from  closely  applied  perception.  analyze  as  natural  t o be  Relating  Finally,  provides  A Schemata M a n i p u l a t i o n Few  mata  4 and can  information  j4.2.  promote  with  model-specific  information clustering  (4.5.3).  categorizing features derived  sources  to  thresholds  identified  hypotheses  schemata  tools  the  top-down, b o t t o m - u p , or  have been d e v i s e d  o b j e c t s can  reliable  relevant ism  to generate  with  I s l a n d - d r i v e n c o n t r o l c a u s e s a t t e n t i o n t o be  most  global  of  Using  for interaction  (4.5).  techniques  degradation.  points  Most  Information  of  the  and  expert  one  representation  on  schema-based  have been a p p l i e d t o t a s k s  however,  effect  frames  how  there  systems.  are  information  Sources  From an  in AI  scheme would  d i f f e r e n c e s in can  best  be  55  represented. In p a r t i c u l a r , knowledge d e r i v e d tive  and h y p o t h e t i c a l ,  This  knowledge must  unlike  must  be  that  t o be  incorrect.  prove  visual  made  tinguished cient.  relational  Finally,  great  deal  usually  met  Maya  structure  of p r o c e d u r a l  provides  instantiations  hierarchy  languages,  how  but  out t o  Thus, a s i n g l e  dis-  of r e a l  imagery  requires  Scenes can vary  syntactic  only  methods  a  so much  have  has been used  a simple data  Maya c a n be e x t e n d e d more complex  often  differ  in  result  data  more  turns  with  schemata i s i n s u f f i -  as  a  that,  between  r e q u i r e m e n t s o f image u n d e r s t a n d i n g is  often  name/value p a i r s .  sometimes s i m i l a r t o t h e o t h e r tation  hypothetical  is  and  not  success.  Maya  shows  program.  knowledge  flexibility.  uniform,  knowledge--attribute follows  t o an NLU  tenta-  from g e n e r i c  i s a schema-based l a n g u a g e t h a t  understanding. encode  input  Another d i f f e r e n c e  the u n d e r s t a n d i n g  w i t h much  images i s v e r y  as s p e c i a l i z a t i o n .  image t o image t h a t  that  t o remove  knowledge, t h e d e c o m p o s i t i o n  be a t l e a s t a s i m p o r t a n t  from  t h e words  be d i s t i n g u i s h e d  provision  from  differentiated  i n image  structure  The  to  description  i n ways t h a t a r e  knowledge  represen-  accordance  with the  described  above.  The  structure  (Glicksman,  types of  attributes.  1982) .  4_._2.j_.  Schemata Have  Four  Parts  Schemata a r e made up o f f o u r The  VALUE  Chapter  type  is  used  to  4. C o m b i n i n g I n f o r m a t i o n  distinct  store  the f a c t s that  Sources  define  the  56 represented  object  It  extension  of Maya's a t t r i b u t e  and  Although  is  slot  an  facets  expression, a  (e.g. the  more.  " a r e a " of  generally i t will  a  be  river-6  value  is  type  76  to  can  pixels).  i n c l u d e FRL's  be  any  a number, a l i s t  Lisp  S-  of numbers,  or  string.  VALUEd t y p e s value. it,  As  such  provide a slot  i n FRL,  the  as a d e f a u l t  slot or  can  the  that  required  Furthermore,  f u n c t i o n s can  is  removed, or  needed.  exists  which  bility  i s used  and  i s invoked  to spread The  type  the e f f e c t s  final  factor  with  culate  the  overall  nothing  review  of  relaxation  the  issue  less  confidence  (Jain  and  4.  where t h e (see  be  a  these  value  function  also  This  the  schema.  and  when  with  be  a  VALUEd  reliability  used  to  presented  types.  Combining I n f o r m a t i o n  is  Sources  cal-  It is useful  Rosenfeld,  s i n c e the  F i g u r e 4.1  flexi-  interpretation  to associate a  Confidence  degradation  attribute  when t h e  slots.  associated  Davis  methods).  1982).  a  slot  e x i s t e n c e of o b j e c t s i s not  in i t s results  Haynes,  with  the  i s modified.  of c h a n g e s t o the  CONFIDENCE of  of g r a c e f u l  schema h a v i n g  Chapter  FRL,  t h e c o n s i s t e n c y of t h e  T h i s a l l o w s one  proposition  of  initiated  t h e components of a schema t h a t can  image u n d e r s t a n d i n g or  be  Unlike  m o d i f i e r t h a t can  i s confidence.  filled  properties  when t h e v a l u e  to maintain  be  have e x p r e s s i o n s a s s o c i a t e d w i t h  filler. added,  can  also system with  c o n t a i n s an  in  an a l l  1981,  for a  related  to  should  have  poor  data  example o f  a  57  length  VALUE: REQUIRED: IF-ADDED:  width  VALUE: REQUIRED: IF-ADDED: CONFIDENCE:  6 DEFAULT: ( g r e a t e r p % v a l 0) . . . similar to length 48  VALUE: REQUIRED: IF-NEEDED:  276 DEFAULT: n i l ( g r e a t e r p % v a l 0) (and ( s g e t v %name ' l e n g t h ) ( s g e t v %name ' w i d t h ) (times length width)) 42  46 DEFAULT: n i l ( g r e a t e r p % v a l 0) ( a n d ( s g e t v %name ' w i d t h 'n) ( s p u t v %name ' a r e a (times % v a l width))) IF-MODIFIED: ( a n d ( s g e t v %name ' w i d t h 'n) ( s p u t v %name ' a r e a (times % v a l width))) IF-REMOVED ( s p u t v %name ' a r e a n i l ) IF-NEEDED: (and ( s g e t v %name ' w i d t h ) ( s g e t v %name ' a r e a ) (quotient area width)) 37 CONFIDENCE:  area  CONFIDENCE: intensity  n i l . . .  VALUE: REQUIRED:  102 DEFAULT: nil ( a n d ( g r e a t e r p % v a l -1) ( l e s s p % v a l 257)) IF-ADDED: (modify-interpretation-range %val) IF-MODIFIED: ( m o d i f y - i n t e r p r e t a t i o n - r a n g e % v a l ) IF-REMOVED: (modify-interpretation-range %val) IF-NEEDED: ( q u o t i e n t ( s u m - a l l - p i x e l - v a l u e s %name) ( s g e t v %name ' a r e a ) ) CONFIDENCE: 26  aio—>  %river  apo—>  %river-system-2  ako—>  ( % t r a n s p o r t a t ioh-system-1  %waterbody-3)  n e i g h b o u r s - -> ( % b r i d g e - 7 % r i v e r - 8 ) CONFIDENCE  43  CONF-ALG  (cluster  PROCEDURES  BOTTOM-UP—>  (sgetv  TOP-DOWN—> Figure  Chapter  4. C o m b i n i n g  %name ' i n t e n s i t y ) r i v - i n t e n s i t i e s )  river-bottom-up river-top-down  4.1 A Schema:  Information  %river-6  Sources  58 The  schemata  knowledge  are  base),  contained  with  and  can  their  be  traversal  of  and  the  hierarchy. for  One in  i s used  possible  ultimate used  part  and  The overall  this  (see  separate  scene.  Other  or  relations  VALUEd  often  compensate  hierarchies  Figure  global  can any  that  might  a  be  used t o  commitment would  number  be  as  from  explore to  their  generally or  subset  or  sub-  representing  the  i n a schema a t a g i v e n  time  available information.  Combining  (see  4.1).  have  4.  which  stereotypic objects  scheduler  and  The  CONFIDENCE v a l u e  can  be  used  to  [ 1 ] L i n k s a r e r e f e r r e d t o as l i n k / i n v e r s e . AIO s t a n d s I n s t a n c e Of; AKO f o r A K i n d Of; a n d , APO f o r A P a r t Of.  Chapter  types  specialization  form a d e c o m p o s i t i o n  one  a  pointers automatic  form a s p e c i a l i z a t i o n  confidence  to  the  binary  control strategies  Instances  is  available  of  i s INSTANCE/AIO[1] w h i c h  general  attribute  the  LINKS—the  of  used with  k i n d can  for  CONFIDENCE  considering  forms  induce  relation  DECOMPOSES-TO/APO t o  hierarchy  number  (or  1978).  SPECIALIZES-TO/AKO t o  hierarchy  be  i n t e r p r e t a t i o n s without  existence.  are  of  possibilities  r e a l i z a t i o n s in a  several  LINKs can  built-in  to  several  Some r e l a t i o n s  Bajcsy,  prominent  Maya  their  and  Any  by  system m a i n t a i n s  for  Knowledge s h a r i n g  interesting  Rosenthal  the  network  indicated  i n h e r i t a n c e , u s u a l l y down a  incomplete data.  create  allows  network.  to generate property  semantic  of a schema.  f o r m e d , and  inverses  a  relationships  second d i s t i n g u i s h e d p a r t relations  in  Information  Sources  is  encourage  for  An  59 evaluation  of t h e most p r o m i s i n g  Associated rithm  to  with  modify  found  for this  rely  on  the  MYCIN  uses  the  (Shortliffe  for  of o t h e r  values). sider  the  tation It  Using  allows  attributes  tiate  and  importance  the  VALUEd  (VALUEd  the  and  them  (i.e. f i l l  procedures  CERTAINTY  These would  a model  (top-down) or  cedures  would a l s o  in their  send  and  Information  average to  con-  interpre-  i t s existence.  used  to weight  be  spe-  to  attached  associated used  control invoke  with  to  the the  to a  instan-  interpreschema  (bottom-up).  other  Sources  the  i s i t s procedural  The  o u t messages w h i c h w o u l d  evaluated.  Combining  of  which a r e  f o r data  schemata t o become " c u r r e n t " and  4.  or  the  t h a t can  include procedures  typically  1977,  confidence.  types),  slots)  to account  see G o e b e l ,  as w e l l as components and  of a schema when c a l c u l a t i n g  Besides  its  flexibility  knowledge t o be  For  of  t h e maximum  the p r o b a b i l i t y  types  systems"  factor  of c o m p l e t i n g  there are g e n e r a l l y procedures  Chapter  1975;  a l g o r i t h m p r o v i d e s the  tation.  other  Buchanan,  are  reliability.  certainty  f o u r t h d i s t i n g u i s h e d p a r t o f a schema  attachments.  schema  and  algo-  VALUES  Many " b e l i e f  calculating  minimum  o b j e c t as w e l l as  of  cializations The  for  for schema-specific  importance  schema-specific  v a l u e whenever new  s y s t e m s w h i c h use  an  difficulty  of an  is a  i t s components.  functions  antecedents a review  attribute  confidence  o b j e c t or  uniform  example,  this  interpretations.  procedures  as  procause  to  be  60  4.2.2.  Manipulating  Schemata  Numerous f u n c t i o n s a r e a v a i l a b l e t o m a n i p u l a t e that  the  describes 5.2)  information  they  i s a particular  scheme d e s c r i b e d  Essential  i n the l a s t  stereotype  i n s t a n c e s ) and d e s t r o y i n g for  merging  Also  there  are basic  the a t t r i b u t e  used  can  Chapter  (Section  snewi — f o r  hypothetical  and s e r a s e i ) ,  f o r modifying  into and  schemata  plus  one  one ( s m e r g e ) . accessing  the  the f u n c t i o n s a v a i l a b l e f o r each  types.  be u s e d  any a t t r i b u t e VALUEd LINKs CONFIDENCE  system  for creating  (seraseo  4.1 l i s t s  standard  between two  sputa sputv saddl sputc sputcl  functions, there of the schemata.  values  for property  ADD  Table  those  add t o the u s e f u l n e s s  s h o r t e s t path  Appendix A  of the representation  o f two i n s t a n c e s  t o f e t c h d e f a u l t s and/or  which the  them  functions  In a d d i t i o n t o t h e s e that  so  section.  objects;  the a t t r i b u t e s  knowledge b a s e . T a b l e  more  implementation  functions include  (screate—for  of  e n c o d e c a n be u t i l i z e d .  t h e f u n c t i o n s a v a i l a b l e i n t h e MAIDS  which  schemata  along  inheritance.  schemata  MODIFY sputa sputv saddl sputc  along  a  are  several  S g e t v c a n be  specified  LINK  One c a n d e t e r m i n e a  certain  REMOVE  FETCH  sremovev sremovel  sgeta sgetv sgetl sgetc  link  4.1 M o d i f i c a t i o n and R e t r i e v a l f o r Schemata A t t r i b u t e s  4. C o m b i n i n g I n f o r m a t i o n  Sources  61 (slink?),  or  discover  (sanylink?). adding  The  missing  whether  semantic  back  a path  exists  network  pointers  and  can  between  two  schemata  b e made c o n s i s t e n t  removing  superfluous  by  LINKs  (sconsist) . Given the  a  type,  attributes  versely,  if  can  found  be  tions  branches  schema  knows t h e  show t h e  the  data  a  structures functions  tem t h a t  combine  modularity, tion  4.3_.  them t o  advice  will  truly machine  enable  symbiotic then  solved  Chapter  as  computer  human  4.  it  are  of  for  of  that  several of  schemata  type  (sattr).  then  Contype  pretty printing  func-  and  (sprint)  or  the  plus  the  a vision  sys-  assets  are  ssprintn).  blocks  Their  a l l  its  representation  building  to  to determine  schemata  (sprintn  multiple IS's.  the  possible  an a t t r i b u t e ,  Also,  network  is  for  greatest  encapsulate  interrelationships  diverse among  interpreta-  schemata  which  cooperate.  ultimate goal  that  name  provide  a Source  vision  f r o m a human u s e r .  minimal  be  and  Interaction Most  that  flexibility  strategies,  enables  the  the  it  components  semantic  manipulation can  LINK,  (sattrtype).  to  of  The  in  one  exist  such as  in  this  of  I nf ormat i o n  systems  have  Whereas field,  totally it  is  i n t e r v e n t i o n may p r o v i d e i n t e r p r e t a t i o n to relationship would  for  be  often the  cooperation  than  could  Combining  Information  Sources  to  use  systems  the  crucial  case  are that  information  Furthermore,  formed  more  through  attempted  automated  succeed.  can  be p o s s i b l e  not  if  b e t w e e n man  difficult  problems  be h a n d l e d  b y man  a and to or  62 machine  alone.  Even though the detail,  several  researchers  human  interaction  1975;  Agin  1978)  and  was  and  in both  Duda,  expert  s u b j e c t has  been d e a l t  have  vision  1975;  systems  not  advocated  systems  Barrow and  with the  in  u s e f u l n e s s of  (Barrow and  Tenenbaum,  ( N i i e t a l . , 1982).  Tenenbaum,  1976;  One  great  of  Stockman,  the  reasons  w e l l d e s c r i b e d as f o l l o w s :  The s u b s t a n t i a l amount of ad hoc world knowledge required t o p l a n p e r c e p t u a l s t r a t e g i e s i s most r e a s o n a b l y a c q u i r e d i n an i n c r e m e n t a l f a s h i o n . The system should thus be d e s i g n e d t o r e q u e s t a d d i t i o n a l i n f o r m a t i o n from a u s e r a t t i m e s of f a i l u r e , i n d e c i s i o n , or on e n c o u n t e r i n g a new o b j e c t and t o i n c o r p o r a t e t h i s i n f o r m a t i o n imm e d i a t e l y i n a r e v i s e d s t r a t e g y . (Tenenbaum, 1973, pp. 8-9)  Human i n t e r a c t i o n  w o u l d be most u s e f u l  systems  t o r e s o l v e p r o b l e m s due  a vision  p r o g r a m has  interpretation mation"  reduced  to ambiguity  the  interaction  user  to  can  be  p r o b l e m s of m o d e l - b a s e d v i s i o n  (as  inconsistency,  One system  is  Hence, tion  of  unsatisfiability,  the c r i t e r i a that  i f the  user  process,  the  without  the  tinue  Chapter  4.  f o r the  i t display  Combining  or  and  additional  Information  the  in  for  If an  "infor-  difficulty. the  other  Section  3.2):  incompleteness.  success  of  a  computer  degradation  to p a r t i c i p a t e  should  little  i n combating  outlined  vision  indecision.  very  resolve  useful  graceful  c h o o s e s not system  current  t h e number o f p o s s i b i l i t i e s  down t o a s m a l l number, t h e n  i s r e q u i r e d from  Furthermore,  in  still  be  guidance.  Sources  (Section  i n the  robust The  vision 3.5).  interpreta-  enough t o more  con-  useful  63 information one  the user  would e x p e c t .  would be u s e f u l  An of  ever,  how  cal?  I f the user  for a vision  interactive  the user  chooses t o p r o v i d e , the b e t t e r the r e s u l t s  vision  a t times  wishes t o input  system  system  o n l y be making  way  requests  d e c i s i o n s must be made.  i s t h e s y s t e m t o know what t h e u s e r  One  i s f o r the system  considers  t o have a u s e r  Howcriti-  model t h a t h a s  knowledge o f t h e p r e f e r e n c e s and h a b i t s o f t h e i n d i v i d u a l s use  the  refer  system  (Rich,  t o t h e model  determining 4.3.J_.  1979a and 1979b).  i n the context  i f the user  should  puter  is  language In  able  to  relate  tant,  of the c u r r e n t  graphic  i s best  in  usually  consists  of  either  from  achieved  when  obscure  Users  when  a n a t u r a l manner.  where p i c t o r i a l  i n p u t and o u t p u t  decision  be q u e r i e d and how.  i s t o be p r e f e r r e d o v e r  image u n d e r s t a n d i n g  that  Then t h e s y s t e m c a n  A c c e p t i n g and Accomodat i n g I n f o r m a t i o n  Man-machine c o m m u n i c a t i o n  i t  to accept i t .  should  when c r i t i c a l  information,  the  com-  Thus n a t u r a l  programming  languages.  i n f o r m a t i o n i s so i m p o r -  becomes c r i t i c a l . pointing  Graphic  input  a t p a r t s o f an image o r  drawing.  Sketch the  user  maps c a n be c o n s i d e r e d a t y p e and  the v i s i o n  system.  of i n t e r a c t i o n  They d e p i c t c e r t a i n  symbols w h i c h a r e a b s t r a c t i o n s o f what m i g h t a p p e a r (cf. is  Section  i n a scene  3.1.2).  They c o n t a i n  as w e l l as the t o p o l o g i c a l  4. C o m b i n i n g  Information  standard a  information concerning  entities. Chapter  in  between  Sources  relationships  scene what  among t h e  64 More d i r e c t via  text.  of  the  It  is  Whereas a s k e t c h  intensity  To  or  be  requires  able  The  a  input  input  vision  is  generally t h a n an  system  properties  not  iconic.  intensity  existence  of  disparate  types  representation  schemata d e s c r i b e d  utilize  provide  the  filled  become  to a  r e t a i n s some a n a l o g i c  accomodate  flexible,  s t o r e and  be  to  a method o f  provide  can  image, d i r e c t  be  an  image.  object,  its  its characteristics.  images.  can  map  i n c l u d e s p e c i f y i n g the  location,  to  might  a l s o u s u a l l y much more a b s t r a c t  Examples  the  information  part  advice  i n the  from a  user.  the  values  directly.  context  which  of  last  representation  i n t e r f a c e between man  with of  modular  independent  of  knowledge  the  nature  section  scheme t h a t Attached  and  (4.2) i s able  procedures  m a c h i n e so  These  of  that  slots  can  then  values  i n f l u e n c e s subsequent  process-  ing .  In  interactive  information display  to  places  4.4.  A Cycle  of  collections  ceptual ties  4.  a  user,  what and  s e c t i o n on  feedback  i s a l s o the  of  what  when i t i s a p p r o p r i a t e when i n f o r m a t i o n  a c y c l e of  loop  question  where  perception  interaction  should  to be  indicates  can  occur.  of  schemata  Percept ion  p s y c h o l o g i s t s have examined  the  of  that  s t r u c t u r e s and  information  (Neisser,  Chapter  next  i n the  Cognitive  to  conversely,  The  three  as  provide  i t , and,  received.  systems t h e r e  and  1976).  Combining  direct  processes  movements and  Neisser  Information  describes Sources  how  role  both accept  exploratory they  might  per-  activibe  used  65 as  follows: In my v i e w , the c o g n i t i v e s t r u c t u r e s c r u c i a l f o r vision a r e t h e a n t i c i p a t o r y schemata t h a t p r e p a r e t h e p e r c e i v e r t o a c c e p t c e r t a i n k i n d s of i n f o r m a t i o n r a t h e r t h a n others and t h u s c o n t r o l t h e a c t i v i t y of l o o k i n g . Because we can see o n l y what we know how to look for, it is these schemata ( t o g e t h e r w i t h the i n f o r m a t i o n a c t u a l l y a v a i l a b l e ) t h a t d e t e r m i n e what w i l l be p e r c e i v e d . Perception is indeed a c o n s t r u c t i v e process . . . At e a c h moment t h e p e r c e i v e r i s c o n s t r u c t i n g a n t i c i p a t i o n s of certain k i n d s of i n f o r m a t i o n , t h a t ' e n a b l e him t o a c c e p t i t as i t becomes a v a i l a b l e . O f t e n he must a c t i v e l y exp l o r e . . . These e x p l o r a t i o n s a r e d i r e c t e d by the a n t i c i p a t o r y schemata, which a r e p l a n s f o r p e r c e p t u a l a c t i o n as well as readiness f o r p a r t i c u l a r k i n d s of o p t i c a l structure. The outcome o f t h e e x p l o r a t i o n s - - t h e information picked up--modifies the o r i g i n a l schema. Thus modified i t directs further exploration and becomes r e a d y f o r more i n f o r m a t i o n . ( N e i s s e r , 1976, pp. 20-21)  Neisser  calls  slightly  this  modified  A similar model-based are  cue  tion),  continual process v e r s i o n can  feedback  computer  discovery  loop  vision  (object),  model v e r i f i c a t i o n  boration  (exploration)[2].  cycle  perception.  of  D e p e n d i n g on different methods of  be has  modes of o p e r a t i o n  seen  been p r o p o s e d  model  c o n t r o l i n computer  I t s four (schema  instantiation), "paradigm"  which correspond science.  one  and  for stages  invoca-  model  is called  will  t o the  If the  A  4.2.  1978).  i s entered,  cycle.  specifically  invocation  Mackworth's  cycle  perceptual  in Figure  (Mackworth,  (schema  where t h e  the  elathe  observe  traditional  c y c l e begins  at  [2]What has been l a b e l l e d "schema i n v o c a t i o n and instantiation" i n F i g u r e 4.2 i s s i m p l y c a l l e d "schema" i n N e i s s e r ' s p e r ceptual .cycle.  Chapter  4.  Combining  Information  Sources  66  SCHEMA INVOCATION AND INSTANTIATION  F i g u r e 4.2  A C y c l e of P e r c e p t i o n  the " o b j e c t " stage then bottom-up, or a l i n e a r result.  Cues  from the image d e r i v e d from o b j e c t s i n the  cause schemata  (models) to be  linear  model,  stage  invoked and  processing  would  exploration  i n the image and  i n the schema w i l l  of . o b j e c t s Chapter  in  a  pro-  further  the c y c l e would c o n t i n u e .  would r e s u l t . be used  the image.  scene In  iterative  i s e n t e r e d at the stage of schema  top-down behaviour  tained  However, an  use the i n f o r m a t i o n j u s t o b t a i n e d t o d i r e c t  If the c y c l e then  instantiated.  can  would stop at t h i s p o i n t : the  invoked model i s a c l a s s i f i c a t i o n . cess  stage model,  The model knowledge con-  to h y p o t h e s i z e  There w i l l a l s o be  4. Combining I n f o r m a t i o n  Sources  invocation,  the  existence  i n f o r m a t i o n con-  67 cerning that  how  and  will  will  verify  cause the  schema w i l l  use  be  other  suitably  more  information So,  the  be  to  find  sampled  on  what  modified.  information  knowledge g a i n e d  is  to help  the i n the  image, schema  i t s existence  hypothesize  image  appropriate  instantiated  to confirm  the  e x p l o r a t i o n module  found  The  in  the  or  the can  i t can  whereabouts  of  objects.  In among  this the  way,  the  c y c l e of p e r c e p t i o n  schemata.  As  cate  to other  schemata  tic  network.  This  cycle  as  a new  individuals  schema (via  o b j e c t s are  to give in  advice  turn  the  promotes  instantiated and/or  initiates  is instantiated.  attached  sages) t o b r i n g about tion  the  hypotheses.  Depending  request the  the  image t o  information.  then  where t o o b t a i n  procedures)  interpretation  cooperation they  to b u i l d  another The  communi-  the  loop around  schemata  and  seman-  work  the as  together  ( v i a mes-  of p a r t i c u l a r  informa-  sources.  The digms  c y c l e of  such  as  perception hypothesize  1975).  It i s particularly  because  i t identifies  into  and  out  this  are  discussed  4.4_.J_.  The  the  the  focus i n the  I n t e r a c t i o n i n the  c y c l e of  munication  Chapter  of  4.  with  is similar  well  Combining  user.  test  feedback  para-  ( e . g . Erman and  Lesser,  s u i t e d t o a schema-based  system  n a t u r a l p l a c e s where schemata can of  attention.  The  following section  Cycle  perception the  and  to other  of  Perception  e a c h node  Information  of  (4.5).  provides- convenient At  consequences  move  Sources  i n the  points  f o r com-  c y c l e there  is a  68  p a r t i c u l a r , u s e f u l k i n d of i n f o r m a t i o n t h a t  can  These  Figure  inputs  described  and  outputs  are  shown  in  be  exchanged. 4 . 3 and a r e  below.  OUTPUTS GLOBAL: When a schema i s i n v o k e d ,  t h e model  INPUT: GLOBAL what t h e user knows about t h e image  type  (e.g.  river)  OUTPUT: GLOBAL what schema has been i n v o k e d and why  SCHEMA INVOCATION AND INSTANTIATION  OUTPUT: LOCAL what f e a t u r e s have been d i scovered  INPUT: LOCAL what t h e user sees i n the c\ image t N  samples  EXPLORATION  v  INPUT: PRIORITIES where t o s e a r c h and what f o r  OUTPUT: PLANNING what t h e schema or a d v i s o r want to look f o r  F i g u r e 4 . 3 I n t e r a c t i o n i n a C y c l e of P e r c e p t i o n "i (  C h a p t e r 4 . Combining I n f o r m a t i o n  Sources  69 can  be  described  current  focus  f o r the of  user  so  attention  that  is.  he/she  A curious  like  t o know t h e  reasons behind  the  schema  i n s t a n t i a t e d - - b y having  is  values—the' information river  i s 276  LOCAL: The the of  so  i s sampled.  values:  succeed),  relevant units), of  the  can  be  or d a t a  As  output  I t can  schema  they  (e.g.  t h a t can  that  is a  then  the  would  the often  Furthermore,  i t s slots  (e.g.  the  are  the the  filled  "area"  as with  of  the  (e.g.  shift  other  the  the  user's  is  102:  that  may  be  the  average  region (e.g.  46 one  intensity; i t  further  models t h a t  input  is  under).  for  the  low  region  schema-specific  s e n d a message t o a l e r t  T h o s e messages p l u s  this  when  have  in attention  i s flowing  executing,  rivers  of  a high  of p o s s i b i l i t i e s  output  verifications  image f e a t u r e s  has  river  be  form of  l o n g a x i s of  cause a  the  i n the  intensity  regions  list  be  image can  schema  concerning  make i n f e r e n c e s a b o u t  They w i l l  choice.  from the  average  neighbouring a bridge  ing.  the  facts  t o the  PLANNING: T h i s  can  derived  h y p o t h e s e s d i r e c t e d by  intensity  be  that  user  what  pixels).  information  image  can  knows  should  processprocedures  be  appropriate  make up  the  invoked. schema.  possibilities  list. The one  of  current might  three  4.  reasons.  schema w i l l  fail,  troyed.  Chapter  schema-specific  Or  procedures w i l l The  i t might  of  the  wait  Sources  for  control  s u c c e e d and  semantic  hypothetical  suspend t o  Combining Information  up  p r o c e d u r e might  become p a r t  i n w h i c h c a s e the  give  network.  schema w i l l more  be  for the It des-  information.  70 Attention  would t h e n  possibilities  The are  by  fidence  caused  the  user can  which  list  the  t o the  schema w i l l  the  is a priority  schema t h a t  (cf. Section  that  and  by  the  shift  first  entry  in  the  list.  possibilities  ranked  given  normally  sent  g e n e r a l l y be  4.2.1) and  message t o be  the  see  assert his/her p r i o r i t i e s  i s done v i a p r i o r i t i e s  a  factor  The  how  Its  t h e message.  certainty  sent.  so t h a t h e / s h e can  queue.  of of  entries The  i t s own the  rankings  are  rearranging  con-  inference shown  processing w i l l  by  value  to  continue  the  queue  input.  INPUTS GLOBAL: The the  user  scene.  levels  of  that  enter  General  or  the  is  a  c a u s e e n t r i e s t o be  or  f o r example, of  objects. road  the  the For  i n the  added  specific  information  i n c l u d e s parameters  location  to s p e c i f i c  there  general  data  processing,  photograph pertains  can  scale  sun.  t o the  that a f f e c t a l l of  i n s t a n c e , one This  possibilities  c a u s e g l o b a l schemata or v a r i a b l e s t o be  the  Specific  picture.  about  aerial  information  can  indicate  information list  modified  or  (see  can  might Section  4.6) . LOCAL: More s p e c i f i c ing the  the  interpretation  form o f an  ple, red)  System:  Chapter  4.  Or  that  interaction "Is the  or edge 26  choice.  information  the  user  Combining  with a  The  might  a l s o be  introduced  is taking place.  bank of  (blue)?"  can  the user  just  Information  specific river  would  T h i s might  routine.  edge  12  For  take exam-  (displayed in  point  i n p u t what h e / s h e  Sources  concern-  at sees  his/her in  the  71 image: r e g i o n would the  be  72  is a  used  to  an  entry  latter,  river.  In  influence on  the  the  how  former  case,  a schema was  possibilities  the  data  i n s t a n t i a t e d ; in  list  may  be  added  or  modified. PRIORITIES: The modifying the  the  search  can  priority  and  i f one  relating  to b r i d g e s  be  of  no  removed  the  object  from the  will  The  for  the that  can  clearly  types  of  describes  the  perception vision  altering  the  4.  objects.  As  move a l l t h e  to  the  l o c a t i o n in  the  an  entries  Objects  be  by  changes  that  fallacious  parameters  image  can  within  i n t e r p r e t a t i o n context.  distinguished  a complex can  is well-suited  can  control  I t allows  to provide Also,  This  where  the  i n the  exchanged.  strategy  take advantage  Combining  Information  a  schema-based,  feedback  from p r e v i -  informed  context  interaction is facilitated  points be  to  a more  of  cycle The  both general  Sources  by  where d i f f e r e n t next  w h i c h , b e c a u s e of  knowledge.  Chapter  entries  queue.  perceived  system.  processing.  information  perception,  By  modify  i n s t a n t i a t e d schemata  having  are  the  the  of  in bridges,  processing  sought.  c y c l e of  subsequent  subsequent  Rearranging  head o f  queue.  changing  be  object-oriented ously  to  i n t e r e s t or  include  queue.  is interested  e n t r i e s , one  would  influence  instantiation priorities  example,  are  of  user  and  section the  cycle  particular  72 4.5.  Mixing  Top-Down and  Besides e n c e d by  the  the  cate.  c y c l e of p e r c e p t i o n ,  hierarchical  object-oriented determine  system,  which  of  the  goal  be of  the  tion  between n e i g h b o u r i n g  important and  intertwined  i s found  in  general  or  this  of an  In a g r a p h i c  the  Top-down,  below,  s i n c e the  road,  river)  Chapter  4.  however,  i s to j o i n  An 4.4  Kiefer,  (modified  1979,  and  knowledge. graph  the  be  data.  Combining  from a  the  the  sche-  communica-  would  spebeen  classifica-  represents  separate be  have  decomposition  This A  standard  but  created  control  behaviour  r o u t i n e s respond  and only  related  during  the  paradigms  are  the  " b o t t o m " of  considered The  data  Information  to  (the  will  t o messages  Bottom-up, d a t a - d r i v e n  o b j e c t s at  can  t o the  a l l of  example where t h e s e  model-driven  hierarchies.  "closest"  schema t o a n o t h e r ;  Bottom-Up C o n t r o l  form,  when s c h e m a - s p e c i f i c in  communi-  image.  Top-Down and  apparent.  most p r o f i t a b l y  h i e r a r c h i e s in Mapsee2).  i n s t a n c e v e r s i o n of  4.5.j_.  in Figure and  an  objects  h i e r a r c h i e s i n terms of c o n t r o l a r e  stereotypic  interpretation  between  In  nodes.  decomposition.  Lillesand  specialization  schemata.  network n e c e s s i t a t i n g s i g n i f i c a n t  cialization  tion  from any  influ-  the  interpretation  into a unified  two  connections  schemata can  sent  mata  The  control is strongly  r e l a t i o n s h i p s between  p a i r s of  M e s s a g e s can  part  Bottom-Up C o n t r o l  be  observed  from  "above"  c o n t r o l comes  the  graph  the  most  intensity  Sources  be  image  from  (runway, primitive, and  the  Figure  Chapter  4.  4.4  Combining  Hierarchies  Information  i n a Geographic  Sources  Domain  74 s k e t c h map b u t n o t n e c e s s a r i l y thought The  of as b e i n g another  relationship  type  The d i r e c t i o n s  h a s been d i s p l a y e d  b o t t o m by i n c r e a s i n g  top-down  parsing:  then  either  find  knowledge  reached. hierarchies  can  control  branch  must be t a k e n .  part  results ing. used  already  location  to  schema  t h e model  that  i n the road  look  from  4. C o m b i n i n g  like  geosystem  | W).  Model  can  symbol)  in place.  is  Other-  and a d i f f e r e n t  take  effect  place.  only  Previous  of  process-  knowledge o f t h e schema a r e F o r example,  schema, over  a  search f o r evidence  Sources  then  i f i t has i f control  good  the b r i d g e .  o f t h e b r i d g e a r e known,  Information  is  the I S ' s then the  a bridge exists,  system  a  (terminal  has taken  search.  f o r a road going  the r e s u l t i n g  the  how t o move down t h e  the c o n t e x t of t h e c u r r e n t stage  established  from  first  (G —> L  up c a n remain  interpretation  and o r i e n t a t i o n  constrains  The  in deciding  o f top-down c o n t r o l  plus  because  and c o m p o s i t i o n .  up t o a c h o i c e p o i n t  a more e f f i c i e n t  resides  be  back  the  context  been  currently  Chapter  type  of  t o produce  would  must  h e l p form The  t h e schemata o r d e r e d  instantiated  have been b u i l t  wise,  after  be  level  b u t some  make s e n s e  one wants t o e s t a b l i s h  stage  a bottom  i t  that  other  if  a t each  until If  or d e c o m p o s i t i o n  c a n t a k e two f o r m s .  e.g.  t h e u s e r ) c a n be  "below" t h e d e p i c t e d s c h e m a t a .  a landmass o r a w a t e r b o d y  i s used  hierarchies  from  i n the graph  with  generality  Top-down c o n t r o l  The  layer  coming  i s not s p e c i a l i z a t i o n  of mapping.  the graph  that  this  of a r o a d .  strategy Since the greatly  75 4_.5.2.  Relaxing  This trol  can  cutoff  Thresholds  s e c o n d t y p e of a l s o be  points  top-down, or  used to  that  relax thresholds.  schema-specific  mine whether  the  features  i n an  the  For  example,  the  object.  water  are  threshold  usually  less  f o r water  i s an  water  a l l p i x e l s having  erally  true  divides will  that  all  tricted pixel fied will  having  not  then  an  have  schema-specific lake  node,  the  to  find  are  water.  It  can  While t h i s  routine  that  can  implies in that are  be  found  that  land),  there  a that  such  reliably  there  res-  that  any  be  will  i s expected  classiobjects  be  But  i s executing,  single a  non-water  ( i . e . the  as  value  case  misclassified.  relaxed  A  incorrectly.  the  that  land.  i t i s gen-  considers  range,  where water  to  However,  r a n g e can  f o r a waterbody  in a context  threshold  be  deter-  classifies  (water and  i f one  those  representing  intensity  i s often  i n t e n s i t y i n the  t o water  that  classified  desperate  intensities  an  categories  i n t e n s i t y values  water[3].  corresponding  the  s u c h as  r a n g e df  as  tries  value  are  to i n s t a n t i a t e  of p i x e l s  intensities.  some p i x e l s t h a t  classification,  sufficient  con-  to  those corresponding  intensity lower  s i t u a t i o n i s not  IS a r e  than  Thresholds  r o u t i n e s must use  intensities  p i x e l s i n t o two  a l w a y s be  The  i f one  expectation-driven,  pixels if  perhaps  t o be  the at  found,  range extended)  so  [ 3 ] T h i s w i l l not a l w a y s be the case in complicated images containing shadows, clouds, l i g h t s o u r c e s , e t c . but t h e n , t h e r e l i a b i l i t y w o u l d be low. Also, single p i x e l values are prone to several types of errors, so that groups of p i x e l s a r e p r e f e r a b l e — c f . S e c t i o n 5.3.1.2.  Chapter  4.  Combining Information  Sources  76 that  more of  the  In t h i s  way,  will the  only  of  feature  values  should  be  will  intensities  might  water, be  a  still  t o the  road  will  road,  The  of  the  the  reliable  current  between t h e s e  con-  thresholds  that a l l p i x e l s  in t h i s  pixels  boundaries. water  extremes,  .  i s s o u g h t , and  Combining  test  can  the  10 and  i s the  s e t of  a l l the  then other  Information  range.  then  there  values  Another as  use  20  i s no and  expectation  of v a l u e s  as  but a 50  that  i t i s between 0 and use  in  range  water  Thus i f t h e r e  i s a high  pixels  classified  considered,  whether  prompt  be  pixel  for  having  classified  s u c h as  . ., Cm), Cn  being  If there  then  expectations  C1,  t o be  to  c a t e g o r i e s as w e l l .  the  possible  mountain,  values  More f o r m a l l y , i f C (CO,  and  intensity  other  of  classifications  urban,  include a l l pixels  represents  Intermediate  the  pixels  knowledge of  decision  and  range  i s , no  priori  4.  less  much  That  include  Chapter  i t s presence.  i f the  [20,50] s u c h  might  that  that  how  commitment--  for existence  so  that  i n d i c a t o r s of  least  rise  determine  position  reliable of  relaxed  verify  field,  water.  have an  [ 0 , 7 5 ] would  tions  be  example,  [0,255]  range  range a r e  pixel  that are  expectations  knowledge  is  this  as  from a c o n s e r v a t i v e  relaxed.  Returning  include  the  classified.  (the p r i n c i p l e  t h r e s h o l d s can  p l u s model  correctly  starts  object  As  the  be  feature values  an  1976).  object,  text  one  accept  presence  Marr,  image can  a  75.  somewhere  62.  a l l possible  l e t Cy  be  the  categories  Sources  classifica-  classification  (C - C y ) .  Then,  77 one  starts  with  [a,b]—no  two  ranges:  feature corresponding in  this  [a',b*]—all  to elements  i n Cn  has  a  value  range.  features corresponding  t o Cy  have v a l u e s  in  this  range. a' Call  interval  b'  of a c c e p t a n c e ,  1(0)  =  [a,b]  1(1)  =  [a',b']  I(e)  = [f(e,a,a'), f(e,b,b')]  and is  the  < a & b <  e a  For  i s the  expectation  function example,  f(e,x,y)  that  returns  i t can =  that  be  the  I,  f e a t u r e e x i s t s (0 < e <  a number  a step  where  that  i s u s e d as  a  1).  f  threshold.  function, ~  i f 0.00  < e < 0.33  ->  x  0.33  < e < 0.67  ->  (x +  0.67  < e <  ->  y  1.00  y)/2  linear, f(e,x,y)  = x + e *  exponential, 4.5.3_.  or  (y - x ) ) ,  s o m e t h i n g more  Mixed C o n t r o l  Since  both  complex.  Strategies  top-down and  b o t t o m - u p c o n t r o l can  what d e t e r m i n e s w h i c h w i l l  be  tion  c y c l e of  pertains  Initially  Chapter  4.  this  t o where the largely  Combining  a p p l i e d and  depends on  Information  when?  perception  the  amount of  Sources  A  take  place,  similar  i s t o be user  ques-  entered.  direction.  78 The for  user  can s p e c i f y  interpretation.  instantiated  If  directly  like  a location,  that  point.  Otherwise,  be i n i t i a t e d .  he/she  from  tion  will  any s c h e m a ( t a ) a s t h e  the data  then  data. of  the  routine might  without  If work  filling  will  case  i n the  default  I t would above  but i f i t  filled  node  send  then  passes  message  is  used.  Chapter  will  bottom-up from t h e create  the data  Because  out model-based selects  i s instantiated two t y p e s  instances  to  rou-  another  structure  to  itself  i s the  instance to j o i n .  4. C o m b i n i n g  most  the s t a r t i n g  applica-  from  the data  point. by  having  looks  request f i t into road  to  Sources  higher-  the semantic will  f o r a compatible  t h e schema may send  its  F i r s t , the  I f one does n o t e x i s t ,  Information  general  top-down messages a s i n t h e  instantiated  system" which  Second,  i n the net-  i t represents  t o d i s c o v e r where i t s h o u l d  "road  place  o f messages c a n be s e n t .  F o r example, a newly  road-system created.  which  control  g e n e r a l l y t r a n s m i t a bottom-up  schemata  network.  routine  i s , a schema-specific  in i t s slots,  i n which the user  schema w i l l level  hierarchy  does n o t n e c e s s a r i l y That  from  o u t an a p p r o p r i a t e  does n o t p r o v i d e any s t a r t i n g  Once a schema slots  top-down  send  knowledge, one h i g h up i n t h e h i e r a r c h i e s ble.  informa-  remain.  the user  a  down  schemata.  be e v a l u a t e d ,  not  schema-specific  a schema c a n be i n s t a n t i a t e d  higher-level  might  a schema t h a t c a n be  p l u s some r e l e v a n t  T h i s procedure  T h i s communication  the  tine  until  point  bottom-up p r o c e s s i n g c a n s t a r t  message t o some schema l o w e r propagate  specifies  starting  send  existing  i twill  out s u g g e s t i o n s  a  be  of other  79 possible  interpretations.  instantiated, for  a car,  schema.  a  so  dark,  the  These  hierarchy--up,  user  of  both  up  and  be  down t h e  type  takes  organizes Bolles,  passed  more  of  laterally  across  heterarchical island-driven determine  few  for  the  order  direction  car  in  further  the  there  will  processing.  be  modified  include a  Control  between  by  mixture  will  move will  it  (top-down),  laterally  rankings  on  search.  the  going  up.  closest  strategy  the  priority Since  be  and  The  which  Loftus,  laterally  across  modes;  bottom-up.  schemata  and  from a schema may  1972).  of  stages,  a t e a c h node more messages  relations  three  (Winston,  the  first  (see a l s o C o l l i n s  down, o r  latter  with  i n any  candidate  to the  i s n e i t h e r top-down nor  A message s e n t  the  sent  queue w h i c h c a n  the  ( b o t t o m - u p ) , down  name f o r  a  being  processing.  hierarchy  no  be  These p o s s i b i l i t i e s  i n t o a graph  and  was  w o u l d be  bottom-up m e s s a g e s [ 4 ] .  advantage  1979).  go  possibilities  of p r o c e s s i n g  them  schema  r e g i o n can  the  h i e r a r c h i e s and  induce  region  could  in a p r i o r i t y  top-down and  This It  has  ( S e c t i o n 4.4.1).  spawned t o  the  road  laterally.  several  These are p l a c e d the  of  suggestions down, or  be  the  rectangular  location  Once p r o c e s s i n g generally  While  1975; up  across, There might  is best-first queue  hypotheses  the  is be or  used  to  can  be  [ 4 ] A s w i l l be seen i n C h a p t e r 5, there are two types of bottom-up entries, one d e a l i n g w i t h the i n t e n s i t y image a l o n e and a n o t h e r t h a t u s e s t h e r e s u l t s of s k e t c h map interpretation i n c o n j u n c t i o n w i t h t h e image.  Chapter  4.  Combining  Information  Sources  80 retracted,  i t i s more a k i n  to non-monotonic  vision  systems which u s u a l l y  mitment  (Marr,  1976).  motivated  by  dently 4.5.4.  group  of  1980b, p.  Al  the  this field).  model.  Each  schema can  to  parallel  with  instance  t h e n much of  a  being  in VLSI,  If can  the be  least  com-  is  indepen-  various  to  into  solver  this that  I f each to  can  possibilities schema, or  i t s own proceed  would be  in  sent  determine  in  better  processor,  number of p r o c e s s o r s  used  (Davis,  attitudes  fits  problem  with  modelled.  access  they  that deals solvers"  exciting  imagined.  queued,  of A l  described  i n t e r p r e t a t i o n process  rankings  parallel.  to was  which  active, limited, schemata  receive attention.  Parallel theoretical is,  object  be  most  Processing  the  considered  work  of  w e l l and  problem  being  schema, had  routines.  priority  would  a  of messages b e i n g  executing the  the  system  the  can  of  subset  summarizes  be  recent  processing  each  Instead  The  recognize  Coupled  Parallel  cooperating  report  the  knows how  principle  than  perception.  i s a q u i t e new  toward  yet  c y c l e of  "distinct,  42;  the  It i s also described  D i s t r i b u t e d A l , V L S I , and  Distributed a  follow  reasoning  processing power of  i t would not  problematic. venient  4.  able  to  would  system only  interpret  such  a  might  Information  not  increase  its efficiency.  images  that  were  s y s t e m would make i t more  t o communicate  cooperation  Combining  described  a vision  However,  f o r schemata  increased  Chapter  be  as  while  create  Sources  a  t h e y are  active.  situation  where  the That more conThis the  81 interpretation 4.6.  Using  potential  Globals  Thresholds evil  that  number  This  of  increased.  t o Remove  Thresholds  were d e s c r i b e d  will  routines.  is  often  be  in Section  found  (ad  hoc)  as  a  in schema-specific  section describes  magic  4.5.2  a  method  instantiation  for  numbers t h a t must  necessary  reducing  be  u s e d as  the  thres-  holds. Take a c h a i n preted the  by  of  sketch  sketch  intensity the  sketch a  to  search  map  create the  The  length  be  defined  width  a n a l y s i s and  same s c e n e ,  chain  to search  has  no  river's  an  ellipse  the as  or  end  rivers tated. between  chosen  that are Similar features  to  4.  Combining  a l w a y s be  in  a  river one  of  the  be  an  Sources  the  chain. ad  location image. chain  chain  and  image.  ellipse)  However,  the  number  or  hoc the  widths  be  interpre-  i n almost sources  an  probably  the  the  If  to  the  intensity  images t o  information  a river.  would  i s known a b o u t  required  inter-  does; the  around  i n the  approximate.  Information  river  So,  i n the  are  from d i f f e r e n t  correspondence w i l l  Chapter  what  to appear  thresholds  use  (or m a j o r a x i s of  points  reflect  likely  and  been  i t as  can  rectangle  ( o r m i n o r a x i s ) would have t o  threshold  a  region  rectangle the  for  has  registered  one  location.  corresponding  of  that  consider  then  thickness  the  map  approximately  the  chain  within  from a s k e t c h  i s known t o be  approximates  want  can  map  map  image of  However, only  (line)  of  a l l matches because  the  82 All are If  of  the  thresholds  r e l a t e d to the one  then  knows t h e  feet  For  per  example,  the  and  of  size,  s u c h as the  each p i x e l ,  rivers  i f the  rivers  rectangle  on  r e s o l u t i o n of  knowledge of  pixel  wide, then  s c a l e or  actual size  real-world  threshold.  based  can  be  question.  a r a n g e of  the  sizes,  image  g e n e r a l l y between  should  in  width,  used to generate  r e s o l u t i o n of  are  image  or  be  river  between  .25  5 and and  the  is 200  10  20 feet  pixels  wide.  In  the  (10)  to  might  be.  of  was that and  be  least  tations  situation sure  commitment allow  one  described  and  to r e l a x the  ranges that  The  image i s t h a t schemata  of  the  global  thresholds  Chapter  scale  i n the  4.  In the  might  be  into a are  to  the  the  value. to the  value river  principle  Higher  expec-  l a r g e r value  terminology  ranges are  relaxed are  described  [a,b]  =  as in  [0,.25]  Hence, routines.  f a c t o r of  a l l reduced  such the  that  i t s value For  the  the holds  can  s c a l e , then a value  Sources  plus  scale  as of  f o r a l l of out  above example,  the  where  be  such  factored  resolution  size-related  t o one  Information  routine,  t y i n g them t o  i s a g l o b a l value  all  a p p l i c a b l e even i f  schema-specific  "200/resolution",  If  Combining  be  to a  image.  variable.  translated  thresholds  larger  image where t h e  would use  smaller  a d v a n t a g e of  schema-specific  reference  can  known o n l y  used.  the  one  the  4.5.2.  example  the  the  [0,10].  values are  choose  in Section  [a',b'] =  river, the  i n c l u d e enough of  section, for this  hoc  would c h o o s e  .However, i n many c a s e s ,  Threshold ad  to  above, one  is  thresholds large  the  number  a are of  f a c t o r s , which  83 are  based  on model  This gle is  as  i s not a case  basket", applied.  the  i f the p r i n c i p l e  The s c h e m a - s p e c i f i c of the value  in generating  use  o f " p u t t i n g a l l o f one's e g g s  however,  certainty  value  knowledge.  schemata  calculating cedures  have a s l o t  o r removed.  used  no  cedures  value  exist.  "200/(sgetv MAIDS  Thus  is  were.  a  schemata  i s also a default  schema,  have  value  to calculate  'feetperpixel  pro-  i s added,  t h a t c a n be  Also, there are attached t h e v a l u e when  'yes  I t c a n be e n t e r e d by t h e u s e r  4.4.1) s i n c e s c a l e  imagery.  Or,  knowledge. tiate  an  it  Before,  can  'one  pro-  i t does n o t refer to  'default)"  i n the  object.  to limit size  Chapter  as g l o b a l  i s often provided with be  inferred  from  However, test  once  input  or other  by t h e r e c i p r o c a l  Information  Sources  as  ( c f . Sec-  remotely  to help  sensed  size  values—by This  instan-  i s instantiated,  I S , i t s known  of the f a c t o r s .  such  t h e image p l u s model  the object  t h e p o s s i b l e range o f s c a l e  4. C o m b i n i n g  of a g l o b a l  t h e t h r e s h o l d r a n g e was u s e d  p e r h a p s by some o t h e r  the  as  i n s t e a d o f " 2 0 0 / r e s o l u t i o n " one m i g h t  'scale  that  system.  scale.  used  of  when t h e v a l u e  T h e r e a r e two ways t o o b t a i n t h e v a l u e  tion  i t i s used,  Such a v a l u e ' s  represented  In a d d i t i o n ,  is available.  t h a t can attempt  account  Then, t h e q u a l i t y  them t o r e s p o n d There  degradation  into  whenever  sin-  f o r CONFIDENCE p l u s a mechanism f o r  i t (Section 4.2).  modified,  yet  i f scale  a s s o c i a t e d with  if  for resolution  how good t h e r e s u l t s  c a n be f a c i l i t a t e d  since  r o u t i n e s can take  threshold factors.  would d e t e r m i n e  of g r a c e f u l  in a  can  be  multiplying  value  can  be  84 modified  and  improved,  proceeds.  A mechanism  i.e.  restricted,  for carrying  i t out  as  interpretation  i s found  i n the  next  section.  These g l o b a l v a l u e s they  are  not  similar  to  differ  from  part  of  global  the  holds. the  the  local  often  semantic  three-dimensional  schemata  objects.  programming  v a r i a b l e s that  are  since  They  languages  lexically  are  which  ordered  by  i n s t a t i c a l l y - s c o p e d languages. in  thresholds  reducing can  be  e s p e c i a l l y when shadows a r e  available  unusual  network of  in  is a useful global  Many a n g u l a r  sun,  somewhat  variables  procedure d e c l a r a t i o n s Scale  are  with  satellite  models,  the  size-related  r e l a t e d to  involved.  imagery.  e l e v a t i o n of  the  azimuth  of  value  is  This When  the  sun  thres-  dealing  with  i s also  use-  ful. 4.2-  Cluster Analysis In  was  Section  used  entities scene  as  4.5.2, the  t o make an  Inference  similarity  inference  i n t o water and  have  an  similar  land:  feature  about  used  to  in  the  c l u s t e r s of image  image p o i n t s features  Chapter  of  4.  are  are  values.  a l l part  compound  Combining  of  the  feature  Information  meaningful  Sources  values  the  the  values  of  those  in  the  feature can  they w i l l  individual  ( c l u s t e r e d ) , then  s u c h as  i f the  because  If s i m i l a r ,  objects,  objects  Further,  objects  some  intensity  classification  similar  these  different  grouped  pixel  values.  is selected appropriately,  different  of  namely,  space  discriminate  Mechanism  be have  features  corresponding  class.  Derived  o r i e n t a t i o n of  edge  85 segments, a r e even more u s e f u l Histograms apparent. of  British  F i g u r e 4.6  number o f  however,  sions  between  A smoother  used  make  shows t h e edge segments  found  Columbia,  these  diagram  in  sections would be  space  that  i s known as group  can  intercepts  (y  (Duda and  Luk,  applied  of  histogram  showing  the  the  (Ballard, dimensional  1977).  used  Ballard  flow, e t c .  Chapter  4.  Combining  line  rough  two to  roads.  and  that  to  determine.  divi-  of  recently,  and  Information  Clowes,  of  and  slopes  1981).  parameters,  Dudani  image  They used to  and  = xcosfl +  h i s group  of a p p l i c a t i o n s  Sources  successively  1973;  intrinsic  a  originally  d i s t a n c e s (p  Ballard  Sabbah,  in a variety  either  into  significant  I t was  segments by  O'Gorman and  transformation  optical  are  corresponds  in discriminating  t o the a n a l y s i s  histograms  model-to-image  A  difficult  or a n g l e s and  1972;  More  technique 1981;  i s very  histograms  + b)  Hart,  There  preferable.  be  one-dimensional  image  number  northwest-southeast  are o f t e n  points into  mx  i n an  a  t h e Hough t r a n s f o r m a t i o n .  edge  =  B to  clustering  image components a r e t r a n s l a t e d  clustering  and  contains  histogram:  the h i s t o g r a m  method whereby  to  ysin0)  which  this  r o a d s , and  that  values.  space  sections  Note  objects  feature  intensity  i s the one-dimensional  northeast-southwest  feature  pixel  segments h a v i n g a g i v e n o r i e n t a t i o n .  significant  The  proper  F i g u r e 4.5  Cranbrook,  roads.  i n the  than  have data  multi-  determine  surface orientation,  86  Figure Chapter  4.5 Edge Segments S u p e r i m p o s e d  4. C o m b i n i n g  Information  Sources  on  Cranbrook  35  30  25  Number of Edge Segments (a =  20  3.0) 1 5  1 0  * * A * ** ** ** ** ** ** ** ** ** ** ** *** *** *** *** *** *** **** ****** ****** ** * * * * * * * * ** * * * * * * * ** ** * * * * * * * * * * * *************** **************** ***************** ******************  * * B * * * * * * * * * * * * * * ** ** ** ** * * ** * * ** * * ** * * ** * * * ** * * * *** * * * *** * * * * ***** * * * ***** * * * ***** * * * * ***** ** ** ** ***** ** ** ** ***** *** ** ** ***** *** * * * * * * * * * * * *** * * * * * * * * * * * * *** **************** *****************  511223344556677889911111111111111111 05050 5050505050 50 50011223344 5566778 05050505050505050 Orientation  Figure  Chapter  4.6  A Histogram  4. C o m b i n i n g  i n Degrees  o f Edge Segment  Information  Sources  Orientations  88 Clustering tern  i s a method o f u n s u p e r v i s e d ,  recognition.  When u s i n g h i s t o g r a m s ,  nonparametric  the general s t r a t e g y  u s e s peaks f o r t h e modes o f c a t e g o r i e s and t r o u g h s boundaries features, 1974;  to  classify  segmentation  Ohlander, 1981).  work.  I t s major  4.7.K  c a n be done  1975; S c h a c t e r  Tomita,  facilitate  advance  A Method  as  image  i n t e n s i t i e s as  in this  manner  (Eigen e t a l . ,  e t a l . , 1976; O h t a e t a l . , 1978; below  follows  i s the use of G a u s s i a n  from  smoothing t o  for Clustering  tinuous histogram.  Intuitively,  this  idealized,  represents  w i t h modes a t p e a k s A and C and a d i v i s i o n  two  Number of Data Points  Feature 4.7 An I d e a l i z e d ,  4. C o m b i n i n g  Information  Continuous  Sources  Histogram  con-  clusters  at the trough,  A  Chapter  this  selection.  t h e f u n c t i o n i n F i g u r e 4.7 a s an  Figure  decision  Using  The method d e s c r i b e d  peak a n d t r o u g h  Consider  values.  pat-  B.  If  89 we  r e g a r d the h i s t o g r a m  A,  B,  and  s l o p e of first  C  given  boundaries  formed  analysis holds  a  in will  be  equivalently,  the  both  t h e modes  and  crossings  in  of the h i s t o g r a m  be  multivariate  of  be p u t  of a  space  to  be  by  finite  Chapter  4.  too  their  be q u i t e  the  case,  it  two-dimensional  finite,  or  large.  However,  i f a l l the  infinite  buckets,  I f t h e r e a r e many empty  or  indirect  nature, are d i s c r e t e ,  bucket  wide, t h e n  Combining  would  then  buckets  reduce  the  a d d r e s s i n g can  be  with buckets  or  implementation.  real-valued,  are  one-dimensional  number o f d i s t i n c t  derived.  c o n t a i n i n g a l l the v a l u e s are  histograms  for c l a r i t y ,  t r a n s f o r m a t i o n ( e . g . l o g a r i t h m i c ) can  f o r a computer  derived  feature  example o f t h e  a more manageable s i z e  Histograms,  be  4.7.5.  the v a l u e s are into a  Each  Although,  f e a t u r e v a l u e s can  can  suitable  An  can  n-dimensional  samples.  dimension.  in Section  a l l  then a  that  e x p l a i n e d u s i n g the  given  as  a histogram  i t i s clear  different  histogram  they  the  find  derivative  presumes t h a t  First,  range  v a l u e s can  data  first  points  t h e e n d s ) where  (or,  can  that  f o r z e r o s , or zero  higher dimensions.  long  bins  function  looking  notice  (ignoring  Hence, we  i n the  for  will  The  used  by  discussion  represent  a  case,  the data. be  as  the  then  feature.  This  case  to is 0.  derivative)  the d i s c r e t e  can  t h e o n l y ones  the tangent  decision  from  are  as a f u n c t i o n ,  in a c e r t a i n  boundaries  must  range. be  Thus  selected:  t o o many d a t a p o i n t s w i l l  Information  Sources  i f the  be  put  if into  90 each too  bucket  and the h i s t o g r a m w i l l  narrow, t h e h i s t o g r a m w i l l  o u t , d e s t r o y i n g t h e modes. bias  and  problem sure  large  that  that  no b u c k e t  r e c e i v e s more t h a n  number o f s a m p l e s .  there  i s some s m o o t h i n g  Gaussian  For  this  niques form  leading  The l a t t e r or grouping  of  The f i r s t the data  can  data  have  peaks.  smoothing  tech-  In most c a s e s , t h i s  I t has  been  ( E i g e n e t a l . , 1974; L e b o u c h e r  1977).  to  averaging  variance  Chapter  filter  show t h a t  t h e smoothing  tradeoff  i s , in effect,  Marr and H i l d r e t h  this  between  filter  a Gaussian, that  i n the frequency  a low p a s s  (Marr  is  and H i l d r e t h ,  localization domain.  Information  Sources  with  by  and L o w i t z , (Scott,  filter.  i . e . normal,  optimal  either  window  i s not b a n d - l i m i t e d i n  spatial  4. C o m b i n i n g  has taken the  accomplished  Dudani and Luk,  They  if  local  1979;  domain.  avoided  t o t h e g e n e r a t i o n o f many t i n y  a v e r a g i n g under a moving  1980),  of the  noisy  o r by a r i t h m e t i c  dreth,  making  very  averaging.  following  isa  of the b u c k e t s .  1978)  ever,  case  a  the b i n width  Uniform  to large  and  be  are  flatten  a small percentage case  real-world  b e f o r e peak s e l e c t i o n .  varying  o r c a n even  r e a s o n , many r e s e a r c h e r s have employed  of l o c a l  they  Smoothing  Many h i s t o g r a m s appearance,  respectively.  a v o i d e d by s c a n n i n g  total  4.2-2.  be t o o r o u g h ,  if  These c o n d i t i o n s c o r r e s p o n d  variance,  i s easily  be t o o smooth;  How-  1980; H i l -  the  frequency  function  respect  to  leads the  and m a i n t a i n i n g a s m a l l  91 The  one-dimensional  G(x) is  =  Gaussian  1/ /2~r7a  r  d e p i c t e d i n F i g u r e 4.8a.  function,  exp(-x /2a ) 2  Since  the  irrelevant  t o the  following,  There  i s one  r e q u i r e d parameter,  before vided  the  f u n c t i o n can  i n S e c t i o n 4.7.4  4.7.3_.  Peak and  be  the  z e r o c r o s s i n g s of t h e troughs  one  Figure shows  the  = 5. and  The  facter  1/  \Z2~TFO  is  T  omitted.  a,  calculated.  w h i c h must  be  A heuristic  good v a l u e s  supplied  will  be  pro-  o.  for  4.9a  of  and the  the h i s t o g r a m first  the  example  of c o n v o l v i n g  and  Specifically,  will  derivative  can  separated  indicate be  used  to  the dis-  other.  shows an  result  be  function.  derivative  second  can  second  histogram  and  i t with a Gaussian  derivatives  are  Figure filter  4.9b with  shown  in Figure  show  the  a  4.9c  d.  The  zero crossings  interest. and  from  first  be  i s smoothed, c l u s t e r s  taking derivatives  tinguish  scale  Trough D e t e c t i o n  by  and  it will  for divining  Once t h e h i s t o g r a m  peaks  2  C5  ship  are in  C2 the those  and  C4  in are  decision  Figure  t h e modes of  boundaries  clusters.  The  show t h a t p e a k s have  a  second  troughs  Chapter  derivative  4.  and  Combining  4.9c  the c l u s t e r s  or c u t o f f  second  negative  have a p o s i t i v e  Information  Sources  and  points for  derivatives  corresponding  points C1,  value.  C3,  member-  in Figure value  of  in  4.9d the  92  F i g u r e 4.8 The G a u s s i a n D i s t r i b u t i o n a . G(x) The G a u s s i a n f u n c t i o n b. G'(x) The f i r s t d e r i v a t i v e o f t h e G a u s s i a n f u n c t i o n c . G''(x) The s e c o n d d e r i v a t i v e of t h e G a u s s i a n f u n c t i o n  Chapter  4.  Combining  Information  Sources  F i g u r e 4.9 A G a u s s i a n - S m o o t h e d H i s t o g r a m and D e r i v a t i v e s a. H i s t o g r a m b. H i s t o g r a m c o n v o l v e d w i t h t h e G a u s s i a n f u n c t i o n c. H i s t o g r a m c o n v o l v e d w i t h the f i r s t d e r i v a t i v e of the Gaussian function d. H i s t o g r a m c o n v o l v e d w i t h t h e s e c o n d d e r i v a t i v e o f t h e Gaussian function  Chapter  4.  Combining  Information  Sources  94 Thus,  finding f(x)  the zero  crossings in  = d[G(x) * H ( x ) ] dx  is  the equivalent  o f peak and t r o u g h  s e l e c t i o n where  Gaussian,  H i s the histogram  operator.  By t h e d e r i v a t i v e r u l e f o r c o n v o l u t i o n s , f(x)  G  i s the  f u n c t i o n , and * i s t h e c o n v o l u t i o n  = dG(x) * H(x) dx  where, d G ( x ) = -x e x p ( - x / 2 a ) 2  2  dx and Similarly,  exp stands  f o r e r a i s e d t o a power.  the second d e r i v a t i v e of the Gaussian d G(x) = 2  function i s  (x -a )exp(-x /2a ). 2  2  2  2  dx These a r e p l o t t e d i n F i g u r e s 4.7.4.  The S t a b i l i t y H e u r i s t i c  The  method d e s c r i b e d  amount o f s m o o t h i n g . a.  A large  hence there  4.8b and 4.8c.  fewer  a will peaks  4. C o m b i n i n g  i s dependent  i s d e t e r m i n e d by t h e v a l u e  c a u s e many  are p o t e n t i a l l y  Chapter  This  f o r edge d e t e c t i o n  points  to  and fewer c l u s t e r s .  be  a s many c l u s t e r s a s d a t a  Information  Sources  chosen f o r  merged  Conversely  on t h e  together, f o r small  points.  o  95  These c o n d i t i o n s are data  used  leads  for Figure  to  results  only  in 4  Eigen describe (their for  a  single  et a l .  and  the  M  By  b.  4.  a  4.10,  =  10  for in  whereas a =  respectively).  plotting  a  P o s t a i r e and  methods t h a t depend on  =  the  Figure  1 in Figure  factor  As  i s the  determines  the  number  The  Effects  of  Vasseur  Combining  Information  of  case the  with  Sources  a,  number of  components  a on  (1981)  a quantization  1  F i g u r e 4.10 a. a = 10 b. a = 1  Chapter  value  peak,  (1974) and  quantization  clusters.  The  in Figure  same 4.10a 4.10b  peaks.  clustering  8  4.9.  shown  Smoothing  or  the  both factor value  resulting clusters  96 versus  6  ( o r M),  largest  range  they  of  8  remained c o n s t a n t . o f M was  very  shows  for  the  region  ( o r M)  a  Vasseur  c l o s e to the o p t i m a l  the  works v e r y  result  of F i g u r e  extends  equal  region  (a = 5)  Figure  4.8  There heuristic  and  i s the v a l u e  stability—the  number of  found  that  on  the  4.9.  t o 2 t o 8.  used  clusters  this ' value  Bayes minimum e r r o r  o f v a r i o u s a's  o  of  w e l l f o r the parameter  example h i s t o g r a m from  region  v a l u e s where t h e  P o s t a i r e and  This heuristic 4.11  discovered  i n the-  rate. a.  number of  clusters  The  longest  The  midpoint  Gaussian  Figure  stable of  this  smoothing  of  4.9.  i s one  immediate d i f f i c u l t y  that concerns  involved with  l a r g e v a l u e s of  o.  As  one  using  this  increases  CO  CO ZD  LJ f\l  11  O LU CD  6  7  SJGMfl  Figure  Chapter  4.  4.11  F i n d i n g the L a r g e s t  Combining  Information  8  10  S t a b l e Region  Sources  11  of  a  12  ]3  Values  a,  97 eventually will  there w i l l  also  result  always  be an  there  is  a, one  must  be  in  infinite  no  o n l y one one  cluster.  range  a priori  way  i g n o r e one  cluster  of  a's  A l l larger  Consequently,  leading  t o f i x an  cluster  left.  t o one  upper  as a v a l i d  limit  a's  there  will  cluster.  If  t o t h e v a l u e of  result  of the  heuris-  tic .  _4_.2.5-  Extensions to M u l t i v a r i a t e  As basic are  one  goes from  univariate  t e c h n i q u e remains  accomplished  boundaries  are  found,  the s t a b i l i t y  ple  show how  will  Figure ters  created  shows  the  4.12  to m u l t i v a r i a t e  t h e same: s m o o t h i n g  by c o n v o l u t i o n ,  Finally,  Data  then  although heuristic  the  this  takes  means,  number  covariances,  and  spaces  the  differentiation  becomes  shows a t w o - d i m e n s i o n a l  by a pseudo-random  and modes  is applied.  the g e n e r a l i z a t i o n  data  and  decision  more  complex.  A bivariate  exam-  place.  example w i t h  five  generator.  Table  number o f  clus4.2  samples  i n each  i s to convolve t h i s  histo-  cluster.  The  first  step i n the procedure  gram w i t h a s u i t a b l e tion  filter.  (without the s c a l e  directional  directional derivative;  Chapter  4. C o m b i n i n g  2  Gaussian  func-  + y )/2a ). 2  effects,  derivatives. however,  two-dimensional  factor) i s  G(x,y) = e x p ( - ( x To a v o i d  The  2  one  There  is  f o r the second  Information  would no  like  use  non-  non-directional  first  derivative,  Sources  we  to  can  use  the  98  Figure  Chapter  4.  4.12  A 2-D  Combining  Histogram  Information  of a B i v a r i a t e  Sources  Distribution  99 CLUSTER  MEAN  [20 [60 [20 [60 [40  2 3 4 5 Table  4.2  COVARIANCE MATRIX  60] 60] 40] 40] 20]  9 0 for  The R e l e v a n t  SAMPLE  SIZE  20 20 1 0 30 20  0 9  a l l clusters  S t a t i s t i c s for Figures  4.12-17  Laplac ian: 9 G(x,y) + 3 G(x,y) = 2  3x  2  9y  2  V G(x,y) = ( x 2  The  results  with  2  + y  2  2  - 2a )exp(-(x 2  of c o n v o l v i n g the h i s t o g r a m  a = 4.5 c a n be seen Although  in Figure  t h e r e a r e no f i r s t  number  of  image.  A small, l o c a l operator  maxima  i n t h e image.  ure  clusters  can  be  + y )/2a ). 2  with  2  these  along  As on  the  with  with  derivative  found  amount  from  the s t a b i l i t y  case,  smoothing. heuristic  to  4. C o m b i n i n g  image  derivative  Sources  a value  local Figfunc-  (the +'s).  o f modes  t o f i n d a good c h o i c e  Information  the  depends  The number o f p e a k s c a n be  5 clusters.  Chapter  find  to a c l u s t e r .  t h e number  done i n F i g u r e 4.16, w h i c h shows how  the  the Gaussian-smoothed  c a n be u s e d  t h e maxima o f t h e smoothed  of  filters  zero c r o s s i n g s ,  E a c h peak c o r r e s p o n d s  i n the one-dimensional  two  4.13 and 4.14.  4.15 shows t h e z e r o c r o s s i n g s o f t h e s e c o n d  tion  is  2  f o r a.  o f o = 4.5  used This gives  Figure  Chapter  4.13  The H i s t o g r a m C o n v o l v e d w i t h t h e f u n c t i o n , where a = 4.5  4. C o m b i n i n g  Information  Sources  Gaussian  Figure  Chapter  4.14  The H i s t o g r a m C o n v o l v e d w i t h t h e of t h e G a u s s i a n f u n c t i o n  4. C o m b i n i n g  Information  Sources  Laplacian  1 02  Figure  4.15  Finding  Zero  The  inflection  p o i n t s on  distributions  are  second d e r i v a t i v e the  boundaries  Chapter  4.  original  n o r m a l and  a l l the p o i n t s i n the  where t h e  of  the d e c i s i o n boundaries  dimensions.  of  Crossings  will  be  Combining  Second D e r i v a t i v e  i s not  is  as  simple  in  zero crossings.correspond  smoothed c o n v e x i t i e s .  rotationally  cluster  second d e r i v a t i v e  the  will  fall  negative.  somewhere  between  Information  Sources  symmetric, into Hence these  If  67  to the  percent  the the  two  portion decision  negative  second  103  .00  CO  !*•' CO  ID . _,<o CJ  u_io' O  03 21  "1 1  Figure  derivative crossing use  " 2  I 3  4  height  Chapter  1  ilGHfl  1  8  o Using  9  technique  derivative. Similarly,  1 0  1 1  1  until  for  is  to  points  a maximum  Heuristic  grow  the  t h e y meet,  use  some  form  zero then  regions having  f o r each  cluster  of  nearest  a  positive  would be i t s  l i k e l i h o o d t e c h n i q u e might  Information  1  boundaries.  in  The p r o t o t y p e  1  1 3  the S t a b i l i t y  f o r a l l the c l u s t e r s  scheme  1  1 2  can  o f t h e peak a s t h e a p r i o r i  4. C o m b i n i n g  1  One  l i n e s as d e c i s i o n  One a l t e r n a t i v e  peak.  1  r e g i o n s (see F i g u r e 4.15).  the r e s u l t i n g  second  5  4.16 D e t e r m i n i n g  boundaries  neighbour  1  I  probability  Sources  use t h e  for a cluster.  1 04 Another  alternative  three-dimensional thought along  topographic  of as peaks, the  which  (Fowler  and L i t t l e ,  found  then  channels.  ridges,  is  i s to treat  cross  at  surface.  the  Using  decision  suitable  passes,  1980).  The c h a n n e l  the nature  of the problem  dimensions,  sion  t o t h r e e and h i g h e r d i m e n s i o n s  cost  of g r e a t e r c o m p l e x i t y  clusters are  boundaries  will  c h a n n e l s and  complementary network  fall  networks  for this  example  the b a s i c  becomes more  technique  and  still  difficult  applies.  The  in  exten-  c a n a l s o be made, b u t a t t h e  computation.  Summary This chapter  put  together  SEE,  is  the  described  How  the  definitions,  form  two  of  If  as a  i n F i g u r e 4.17.  While  .4.8.  t h e smoothed h i s t o g r a m  into a v i s i o n subject  of  system.  the  i n the i n t r o d u c t i o n  descriptive well  has i n t r o d u c e d a number o f t o o l s  next  which can  Such a s y s t e m , chapter.  are designed  The  criteria includes  have  Chapter  6 which  the  several  examples.  Chapter  4. C o m b i n i n g I n f o r m a t i o n  been results  Sources  met w i l l of  MIS-  techniques  to achieve  and p r o c e d u r a l a d e q u a c y and g r a c e f u l  those  called  be  the g o a l s  degradation.  be d i s c u s s e d i n  running  MISSEE  on  x  xxxxxxx  xxxxxxxxxPxxxxxxxxxxxxxxxxxxxx  xxxxxxxxxxxxx xxxxxx  P xxxxx xxxxxx xxxxxxx  x x  X  +  #  HHHHHHHH  X  x  HHHHHHHHHH  HHHHH HHHHHH  X  X  HHHHHHH  #### X  X  HHHHH  HHHHHHHHH  X  X  HHHHHHHHHHH  X  HHHHHHHHHHHHH  HHHHHHHHHHHHH  X  HHHHHHHHHHHHHH  HHHHHHHHHHHHHHH  HHHHHHHHHHHHHHH  HHHHHHHHHHHHHHH  HHHHHHHHHHHHHHH  HHHHHHHHUHHHHHHH  HHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHH*  HHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHHHH HHHHHHHHHHHHHH  HHHHHHHH  HHHHHHHHHHHHHHHHHHHHHH HHHHHHHHHHHHHHH  HHHHHHHHHHHHHHHHHHHHHHH  Figure  + Peak -Pit P Pass x Channel # Zero v a l u e of the S u r f a c e  4. C o m b i n i n g  H  HHHHHHH  HHHHHHHHHHHHHHHHHHHHHHHH  4.17 U s i n g C h a n n e l s a s D e c i s i o n  Legend  Chapter  HH  ### X  X  *##  '  x  X  Information Sources  Boundaries  CHAPTER The  5_._1_.  MISSEE  (Multiple Information  implementation chapter. SEE.  of  Figure  The  many  5.1  core  resulting the  of  the  The  are  was  written  an  Amdahl  was  test  model-based not  subset  sketch  network of and  created  sources and  and  utilized  feed  i n t o MIS-  user. Originally  on  F r a n z L i s p and  designed ideas  yet  of  rich  t o be the  proposed  sections will  power of  describe  that  arise  in Chapter the  end  previous  can  Pascal  on. a  C.  an  enough so  and  the  MAIDS--MISSEE  that the  schemata  builds  incarnations. 1980)  previous  components of MIS-  (called  map,  (Koomen,  i n the  an  I t c u r r e n t l y e x i s t s i n a Unix environment  demonstrating  following  subsystem  gone t h r o u g h two  interpretation  everything  schemata a r e  is  ran  the  restricted,  the  i s a semantic  information  image, a  11/780 u t i l i z i n g  to  described  system  and  MISSEE i s not ten  The  in M u l t i l i s p  470.  ideas  SEE)  i n t e r p r e t a t i o n process  input  s y s t e m has  it  VAX  three  Source  overview of  manipulation  a digitized  The  the  system  representation.  schemata  Aids).  of  shows an  w h i c h b o t h c o n t r o l s the  SEE  Implementation  Introduction The  by  5  4  product. chapter a l l of  I t was  i n a domain the  the  been  key  what has  3.2).  has  in  While  implemented,  ideas  been  that  problems  (cf. Section has  writ-  a  been.  The  implemented  and  1 06  ZERO CROSSING EDGE DETECTION  MAPSEE2  / 1 1 i  \ \ \ \  MISSEE ( M u l t i p l e I n f o r m a t i o n Source SEE)  MAIDS A Schemata Manipulation System  sources of information  REGION MERGING  AN INFERENCE MECHANISM  -Schemata f o r -Geographic F e a t u r e s -Roads, R i v e r s , e t c . -Global V a r i a b l e s -Procedures f o r Schemata Instantiation -Top Down -Bottom Up -From the Image alone -From a Sketch Map -Control Strategy  S t a t i st i c a l Cluster Analysis  A V  - data  data  —  o  programs and schemata  F i g u r e 5.1 The MISSEE S y s t e m  Chapter  5. The I m p l e m e n t a t i o n  control other systems  108 point  out  what has  5.2.  MAIDS--MISSEE  MAIDS i s an system  Aids  implementation  described  extension Lisp  not.  of Maya  in  of  S e c t i o n 4.2.  (Havens,  1978),  f u n c t i o n s f o r the c r e a t i o n ,  output  of  schemata.  functions  i s found  While  A user  are  such  objects  ranges, eral  and  as  i n Appendix  the  cedures  roads,  geosystems. in  o b j e c t and  which  fill  The  L i s p as  an  a number  of and  of  these  for handling  sche-  A. system  schemata  in the  b r i d g e s , road  system c o n t a i n s form  slots  geographic  photographs.  of  slots,  possible default  those  i n Franz  modification, manipulation,  rivers,  the  manipulation  i t makes a v a i l a b l e  in analyzing a e r i a l  information  define  Written  MAIDS i s a g e n e r a l p u r p o s e  useful  schemata  manual d e s c r i b i n g t h e use  mata, MISSEE c o n t a i n s s p e c i f i c that  the  during  domain  These  include  systems,  mountain  stereotypic,  gen-  or a t t r i b u t e s ,  which  v a l u e s , and  attached  instantiation--see  pro-  section  5.4.  General ticular and  o b j e c t s , such  instantiations  i t s i n v e r s e , AIO,  during ical,  instantiation may  found  be  knowledge, a r e  Chapter  5.  The  of  an  road, them  the  rarely,  a r e d i s t i n g u i s h e d from  (road-4)  instance of.  of a p a r t i c u l a r  destroyed  regarding  as  The  i f ever,  Implementation  being  or c o n t r a r y  Objects, destroyed.  INSTANCE  instances are  image, and,  i f insufficient entity.  v i a the  being  parlink  created  hypothet-  evidence  p a r t of  is  general  109 Two  of  objects  the  and  major  instances  Specialization  The MAIDS  previous  system  It  scene. then  of  r e l a t i o n s are of  A  relation,  fourth  higher  been a d d e d  to denote  include  that  both  decomposition. decomposition  are  a natural  instances  specifically  way  to  river-4  in  is,  system  in  schemata  MISSEE  a  vice  schemata w i l l  s p e c i a l i z a t i o n and  instance  (for  sys-  bridge-2,  b r i d g e - 2 and  r i v e r - 4 and  is  particular  f l o w s under  to determine which  That  standard  ( i t s inverse  f o r the  congruency  i n the  the  usefulness  a NEIGHBOUR LINK t o  level  of  "neighbours",  that  i s used  same r i v e r  a l l part  general  spatial  fact  hierarchies. the  their  i f i t i s found  This  position part  three  river-4 will  of  and  and  hierarchies  because  i s used Thus  versa. part  form  n e i g h b o u r s ) has  tem.  specialization  within  knowledge.  organization. also  are  relationships  (SPECIALIZES-TO/AKO)  (DECOMPOSES-TO/APO) organize  organizing  be  decom-  bridge-2 w i l l  be  example,  river-  r e l a t i o n s form  hierar-  system- 1 ) .  The chies  s p e c i a l i z a t i o n and  that  are  Hierarchies instance only  define  and  relate  thought  of  present an  i n both general up-down  neighbour instances  as  Thus the  defining  initial  decomposition  (particular left-right  MISSEE  system  formed  i n t o s p e c i a l i z a t i o n and  Chapter  5.  The  Implementation  p a r t i c u l a r knowledge.  positioning  relations  the  and  do  not  of  entities.  form h i e r a r c h i e s  knowledge).  They  p o s i t i o n i n g of  contains  decomposition  The and  can  be  schemata.  generic  objects  hierarchies.  Dur-  110 ing  instantiation,  interpreted and  are  related  this  and  instances  schemata  represented  by AIO  semantic  position  those  should  general  i s organized  uses the be  to  in a p a r a l l e l  LINKS t o t h e  network  specific  by  the  image  network of  objects.  relation  grouped as  p a r t s or  to  instances  Additionally,  specialization  neighbour  being  and  decom-  determine  which  kinds  of h i g h e r  level  entities.  CONFIDENCE v a l u e s MISSEE, ties. in  again The  the  results  i s as  sketch  map  in  chies)  are  Information confidence tem  will  are  of  the  of  a d v a n t a g e of  interpretation  of  either given  instantiated  the  provided value.  by  the  Chapter  the  graceful  are attached  procedures:  intensity  5.  The  level or  capabilivariations  the  basic  the  a i d of  instantiated schemata  specialization schemata  i s done t o e n s u r e  prethe from  (that i s , hierar-  lower  down.  i s g e n e r a l l y accompanied  to s l o t s ,  t h a t the  CONFIDENCE  with  alone  interpret to respond  an  for handling  Implementation  the  by  a  sys-  values,  In MISSEE, schemata have two  top-down--to  directly  image  those  than  bottom-up—either  interact  with  in  degradation.  schemata as a whole.  to  higher  user  This ordering  some m i n o r  instantiated than  manner  t h e MAIDS  s c h e m a t a , but  confidence  bottom-up p r o c e d u r e s  with  And,  simplistic  cause  decomposition  higher  a b o v e ; and,  below o r  full  g i v e n more c r e d e n c e  exhibit  attached  from  taking  image a l o n e .  Procedures and  i n a very  f o l l o w s : schemata  intensity  higher  used  of  reliability  cedence  the  not  are  and  IS.  messages r e c e i v e d t o a message  There are  input data: those  types  that  those can  two  from types  concerned use  the  111 results  of  sketch  possibility  t h a t was  interactive interact in  5.3_«  The The  aerial is  an  5.2 of  a n a l y s i s to guide not  graphics  directly  Sections  map  5.3  Input goal  capabilities)  with  and  implemented  the  5.6,  the  Information  intensity  of  the  Two  interpretation  direct  guidance  of  from  information  specific  5_._3._1_.  Digitized  The  Most work on edge dual  d e t e c t i o n or  whereas  ties),  while  to  regions  tivity.  Chapter  are  of  that  be  seen  monochromatic  Thus t h e  primary  s u c h as  is a portion  image: a s k e t c h section  IS  Figure  a v a i l a b l e to help  This  three  with  map  describes  and the  IS's.  Image images has  (edges a r e are  location  and  b a s e d on  they  the  on  be  of  the  hypothesis  discontinui-  types  shape  of  intensi-  image.  surface  of  either  considered  h o m o g e n e i t y of  surface  both  may  intensity  o r i e n t a t i o n of  better),  Implementation  the  aspects  a r e more r e l e v a n t t o  the  concentrated  While  b a s e d on  In a c c o r d a n c e w i t h  5. The  lack  elsewhere.  T h i s photograph  region merging.  ( t h e more d i f f e r e n t ,  to a  will  digitized,  intensity  user.  to the  regions  the  As  occurs  scenes.  IS's  each emphasizes d i f f e r e n t  pertain  level.  from a p h o t o g r a p h ,  the  the  (due  third  s e t of p r o c e d u r e s  interface  urban  input  segmenting  representations  ties  this  Columbia.  other  is a  interpret  image d e r i v e d  1.1.  i n MISSEE  A  Sources  small  of A s h c r o f t , B r i t i s h  at  user  of MISSEE i s t o  photographs  Figure  user  interpretation.  this  and  Edges  boundaries connec-  dissertation  information  from  112  Figure  Chapter  5. The  5.2  Ashcroft,  Implementation  British  Columbia  11 3 the  digitized  5 . 3_._1_._l_.  image a r e made a v a i I a b l e [ 1 ] .  Edge  Detection  A Marr-Hildreth zero-crossings image lars  edge d e t e c t o r  of the L a p l a c i a n  was  implemented  of the Gaussian  The end r e s u l t  intensity  The  particu-  i n A p p e n d i x B.  of t h e edge d e t e c t i o n  image"  that  instead  of i n t e n s i t i e s ,  whether  are described  corresponds  to  the  each p i x e l  process  intensity contains  o r n o t i t i s an edge p o i n t .  a  Figures  value 5.3b  edge segments o v e r l a i d on t h e image of A s h c r o f t  of  1.1  detail the  2.2  recovered  greater  1980).  processing  is carried  location,  length,  shall  contrast,  derivative  that  specifying  and 5.3c for o  the s m a l l e r  the a  on  and  sketch  information  orientation.  i n d i c a t e s how  either  taken  o u t on e a c h edge  of t h e raw p r i m a l  F o r e a c h edge segment,  intensities  "line  show  values  o i s r e l a t e d t o t h e amount o f  by t h e edge d e t e c t o r :  a version  call  a  value,  the r e s o l u t i o n .  Further generate  respectively.  is  image e x c e p t  the  and  finds  of the  (Marr and H i l d r e t h , 1980; H i l d r e t h , 1980). of the implementation  that  side  of  from t h e o r i g i n a l  A  large the  segment  (Marr and H i l d r e t h ,  i s stored final the  edge  to  about  value,  w h i c h we  discontinuity is.  its  in  It is a first  image.  L1]See ( N e v a t i a and P r i c e , 1978 and 1982) and (Nazif and L e v i n e , 1982) f o r o t h e r s y s t e m s t h a t combine edge and r e g i o n i n formation.  Chapter  5. The  Implementation  114  Figure a. b. c.  Chapter  5. The  5.3 A s h c r o f t : Edge  Detection  (upper l e f t ) A s h c r o f t ( l o w e r l e f t ) e d g e s f r o m o = 1.1 ( l o w e r r i g h t ) e d g e s f r o m a = 2.2  Implementation  1 15 The image  zero-crossing  regardless  values.(along edge  shows  at  w i t h model  the  that  the c o l o u r  no  single  the top.  Model  the proper c r i t e r i a  5_._3.Region Region Freuder  found  produces fault mon  boundaries  unreliable  Brice  Chapter  C.  by  5. The  The  order i s  b l u e to dark red.  blue  It i s e v i -  however, t o  a  a r e most  Although quite i n appearance.  i s paid  method  edges  determine  Fennema,  t e c h n i q u e s was  techniques render  used  less  of a s i m p l e  implement.  Implementation  similar.  by  i t merges  Details  can  a simple algorithm, i t Its  leads to  1970).  described  Basically  most  t o t h e shape o r  cooperative interpretation  to  used,  subfigure  objects.  r e g i o n s which  and  come t h e d e f i c i e n c i e s i s easy  Each  the  segments  "significant"  ( F r e u d e r , 1976).  intensities  between  simple,  it  be  i s accomplished  no c o n c e r n  (see  shows how  to dark  length  ordering  criterion.  r a n k s a l l the  for individual  reasonable  merges  segmentation  a values.  bright  knowledge can  in  a g a i n show edge  of t h e segment: b r i g h t  criterion  i n Appendix  i s that  5.5  by a d i f f e r e n t  affinity  results  and  f o l l o w e d by  r e g i o n s whose a v e r a g e be  C o n t r a s t and  i n an  Merging  merging  called  5.4  using d i f f e r e n t  edges r a n k e d by  significance.  t o a l l edges  knowledge) a r e u s e f u l  Figures  the h i g h e s t rank,  dent  their  from A s h c r o f t  indicated is  of  segments.  derived  edge d e t e c t o r r e s p o n d s  prominent  l e n g t h o f com-  some  "incorrect"  Segmentation  because  1) even  than p e r f e c t  based  on  the  best  results,  2) i t  i s p o w e r f u l enough t o o v e r -  segmentation  t e c h n i q u e , and  3)  1 16  Figure a. b. c. Chapter  5.  5.4  (upper (lower (lower  The  Ordering  Edges  in Ashcroft, a =  1.1  r i g h t ) o r d e r e d by l e n g t h l e f t ) o r d e r e d by c o n t r a s t r i g h t ) o r d e r e d by c o n t r a s t * l e n g t h  Implementation  117  HILL  EC s  7 HST  LENGTH  Figure a. b. c.  Chapter  5.5  (upper (lower (lower  5. The  Ordering  Edges i n A s h c r o f t , a =  r i g h t ) o r d e r e d by l e n g t h l e f t ) o r d e r e d by c o n t r a s t r i g h t ) o r d e r e d by c o n t r a s t *  Implementation  2.2  length  1 18 Figure croft  5.6 shows t h e r e s u l t s o f r e g i o n  image s t o p p e d when t h e r e  process when  stopped  The r e s u l t  image" where e a c h p i x e l  As primal  with  eter the  and  edges, p o s t - p r o c e s s i n g  point  i s stored,  the area  is  Finally,  In t h i s  process  along  with  the l i s t  i t was  a  "region  i t belongs.  out  to  derive  The l o c a t i o n o f  the length  The a v e r a g e  i s c a l c u l a t e d as p a r t  The  algorithm  case,  is  carried  Ash-  left.  the  f o r each r e g i o n .  of the r e g i o n .  p i x e l s i n the region  measure.  of t h i s  by  i n d i c a t e s t o which r e g i o n  sketch-like information  a perimeter  or  a r e no p o t e n t i a l merges l e f t .  by t h e u s e r .  on t h e  a r e 172 and 75 r e g i o n s  c a n e i t h e r be s t o p p e d by t h e u s e r  there  merging  of the perim-  i n t e n s i t y of a l l of  of a l l n e i g h b o u r i n g  the  affinity  regions  i s made  explic i t .  B o t h edge d e t e c t i o n that  can  fairly tion  be  I f model  possible  to  and/or  region  known  or  contains small  could lem  out  limit  iterations:  the parts  discovered  of merg-  F o r example,  be  detection if i t is  a l a r g e p o r t i o n o f a p a r t i c u l a r image then  o f i n t e r e s t and when  strategy  convolu-  i t may  o f t h e image where edge  of i n t e r e s t ,  The o n l y  Both a r e  stage  effectively,  i t c a n be  o f t h e image may be s e a r c h e d  f o r an o b j e c t terminate.  that  operations  involves  one f o r e a c h  m e r g i n g must be p e r f o r m e d .  no f e a t u r e s  this  The f o r m e r  knowledge c a n be u t i l i z e d  sections  Chapter  many  merging a r e l o c a l  on s u b i m a g e s o f any s i z e .  expensive computationally.  and the l a t t e r  ing.  hood  carried  and r e g i o n  new  entails  5. The I m p l e m e n t a t i o n  ignored.  i n order  i t i s found  Or,  of  likeli-  the  search  (but not n e c e s s a r i l y major) is  how  to  "put  together"  probthe  119  Figure a. b. c.  Chapter  5. The  5.6  A s h c r o f t : Region  Merging  (upper l e f t ) A s h c r o f t ( l o w e r l e f t ) 172 r e g i o n s ( l o w e r r i g h t ) 75 r e g i o n s  Implementation  120 information  of  two  contiguous  subimages at  their  mutual  boun-  dary.  T h i s m e t h o d o l o g y can interpretation  greatly  image needs t o be increase  the  ficulty  in  i n the  device  (so  the  was  connected  5.  2  On  current  merging  user  be  preferable  in the  considered  Section  can  t o the  75  regions  This  decide  and  road/mountain). incorrectly  not of  two  of  labelled  the  f o r the i s very them  5.  The  user  t o run  not  I t has  not  the  dif-  data.  For  output  merging)  be  This  input  made  that  although  s u c h as  pro-  it  was  would  described  categorization  for  image.  because  ambiguous  i t has  some u r b a n the  "INTERPRET" box  only  (urban/hills  inaccuracies  Thus,  to  classification  due  regions  four and  to  some  have  been  c a t e g o r i z a t i o n does  i n t e r p r e t a t i o n ; i t i s only  Implementation  would  specific  the  whole  MISSEE.  can  such  crude  vice versa).  to a  overall  the  b e c a u s e of  stop  Ashcroft  (e.g.  it  system.  system.  i t has  of  p r o g r a m s and  regions  are  the  Regions  shows one  regions  p r o b l e m of  the  a u t o m a t i c method  available information—the  Chapter  the  to the  Furthermore,  " w a t e r " and  solve  of  hand,  when t o used  of  portion  is tied  f o r the  t y p e of  5.7  discovered  classification  categories  labelled  an  Figure  various  computer  another  t o use  4.7.  other  process  a d d i t i o n a l information be  small  implementation  A Categorization  can  the  the  A crude c l a s s i f i c a t i o n vide  a  efficiency  s o l v i n g power of  coordinating region  the  only  examined.  example, the  not  if  problem  been u t i l i z e d  increase  another  in Figure  5.1.  type  121  Figure a. b. c.  Chapter  5.  The  5.7  (upper (lower (lower  A s h c r o f t : Region l e f t ) Ashcroft l e f t ) 75 r e g i o n s right) classified  Implementation  Classification  regions  1 22 The ties  classification  of  the r e g i o n s .  b o u n d a r i e s can a different  image c a n be  image d i s p l a y  a  reasonable  (with  used,  2)  point  tion (for  though  4.5.2, each  mismatch regions the  from i n Cn  require  used,  priori with  information  larger  Chapter  in Figure  as w a t e r .  5.7  in  it  i n the range,  that  be  expectation  some as  Nonetheless, since  the  into  of o v e r l a p p i n g ensures that  range,  The  such  step function,  account,  interval  such t h a t is  ranges i s  a l l regions  The  [a,b],  i t is  of  w i t h no a  used,  while  f o r the c a t e g o r i z a t i o n ,  the  of  i s employed.  the c l a s s i f i c a t i o n represents  a  automatic c l a s s i f i e r . 5. The  series  Sec-  [a,b].  there w i l l  interpretation  [a*,b'],  a two  used  corresponds  range  because  intensity  the r e s t r i c t e d  [a',b'],  though  inaccurate, from any  on  reasonable  range,  Even  actually  c a t e g o r i e s c a n be  i n the proper c a t e g o r y range.  relies  some  to  on  " s e e s " where  f i t the d e f i n i t i o n  shown  In MISSEE, a s e c o n d  included  acceptance  one  other supportive evidence for i n s t a n t i a t i o n ,  corresponding be  their  arises  h a v i n g an a v e r a g e  decision  o r , 3) as was  conservative  the d e f i n i t i o n  the  regions.  t a k e t h e c r u d e n e s s of t h e  acceptable.  will  the  urban ' r e g i o n s c l a s s i f i e d  schemata and  the c l a s s i f i c a t i o n  intensi-  were f o u n d t o work f o r  s h o u l d be,  i t does n o t e x a c t l y  to  ways t h a t  mapping) u n t i l  s e t of r e g i o n s and  category)  the average  t h e b o u n d a r i e s c a n be moved  f i x the b o u n d a r i e s f o r a l l the  Even  on  1) v a l u e s t h a t  function  division  done, a t r a i n i n g  solely  There are s e v e r a l  be d e r i v e d :  an  to  i s based  Implementation  realistic  regions  is  crude  and  result  that  might  come  Statistical  pattern  recognition  123 techniques are i n s u f f i c i e n t More knowledge 5.3.2.  for interpreting  The S k e t c h Map  1980),  which  of moves/draws sketch.  like  MAIDS i s w r i t t e n  is  a  mountains,  tems, a n d t h e w o r l d . Figure a simple  Specifics  t a s k t o draw t h i s  ( i nthis  lines  case The  over  and made a v a i l a b l e  curves are joined  that  (Appendix  into  road  c a n be f o u n d  and  series of  I).  to  geosys-  i n Appendix  D. It is  on a  suitable  F i g u r e 5.9 shows be  this  grouped  t o Mapsee2.  a r e made a v a i l a b l e  to  MISSEE  are  a  ( c f . S e c t i o n 3.3). P o i n t s i n the s k e t c h l i n k s and then  into  Chains are g e n e r a l l y  chains.  of l i n e s  linear  are  gen-  representing a binary  segments.  ambiguous c u e s  for  of the  such  as  [ 2 ] T h e e x c e p t i o n a r e c l o s e d c h a i n s w h i c h a r e unambiguous s h o r e s and b l o b s w h i c h map t o towns.  cues  Implementation  The shapes  objects  and  5. The  and m o u n t a i n s [ 2 ] .  Chains  chains  Chapter  rivers,  the  chains  systems,  an image d i s p l a y e d  B) i n t o a t r e e  breakdown o f t h e c u r v e s  for  lines,  ranges,  a Comtal V i s i o n  number o f Maya s c h e m a t a  roads,  curves  p o i n t s i n t h e s k e t c h map c a n t h e n  Mapsee2 r e s u l t s  eralized  the  isa  h i e r a r c h y of i n s t a n c e s of  points,  mountain  Input  Mackworth,  5.8 shows a s k e t c h map o f t h e A s h c r o f t a r e a .  combination. into  i n Maya.  decomposition  o b j e c t s r a n g i n g from  rivers,  (Havens a n d  (or p l o t s / g o t o s ) which t r a c e  Output  geographic  device  images.  i s required.  S k e t c h maps a r e a n a l y z e d by Mapsee2  roads,  complicated  A'  2  A  Figure  their  5.8  s u c h as  the  like  all  the  will  remain  Chapter  and  a r e used  road systems,  /\,  3  A S k e t c h Map  spatial relationships  tures,  ,  of  chains  have been  aggregate  r i v e r systems, mountain  examined,  ambiguous, w h i c h  5. The  Ashcroft  to b u i l d  t o remove some o f t h e l a b e l s  Implementation  results  4  ranges,  f o r the c h a i n s .  though,  some o f t h e  in a possibly  struc-  large  and  After labels number  125  Figure Chapter  5. The  5.9  Ashcroft  with a Superimposed  Implementation  Sketch  Map  1 26 of  hypothetical Table  croft  5.1  sketch  Figure  store  their  mation  The c h a i n  The  interpretation with  shows a l l t h e l a b e l s f o r t h e c h a i n s  map.  5.8.  along  instances.  "interpretation"  first.  the  numbers c o r r e s p o n d  These  column  instances  corresponding  chains,  l o c a t i o n s a r e the major  of  the  Ash-  t o the values i n  lists  the  intended  geographic  lines,  source  in  objects  and p o i n t s  of s k e t c h  map  which infor-  used by MISSEE.  The  instances  a r e embedded  that  groups  built  from t h e " i n t e n d e d "  model  knowledge a b o u t  SEE.  Unfortunately,  version HAIN 1 2 3 4 5 6 7 8 9 10 1 1 12 1 3 14  related  of  -  Table  Chapter  5. The  5.1  decomposition 5.10  interpretations  spatial  1 74 147 240 245 340 649 702 753 900 909 992 1 061 1 126  -  73 1 46 239 244 339 648 701 752 899 908 991 1 060 1 1 25 1 168  only.  Higher  connectivity could  level  be u s e d by MISthe then  enough o f t h i s  current  information.  INTERPRETATIONS *road-6 * r i v e r - 6 *road-4 * r i v e r - 4 *bridge-1 ( f i r s t side) *town-1 *bridge-1 (second side) *road-5 * r i v e r - 5 *road-7 * r i v e r - 7 *road-8 * r i v e r - 8 * r i v e r - 1 *road-1 *road-3 * r i v e r - 3 * r i v e r - 2 *road-2 *mountain-1 * r o a d - 9 * r i v e r - 9 *mountain-2 * r o a d - l 0 * r i v e r - l 0 * m o u n t a i n - 3 *road-11 * r i v e r - 1 1  I n t e r p r e t a t i o n s of the A s h c r o f t  Implementation  hierarchy  shows t h e h i e r a r c h y  when MISSEE was d e v e l o p e d ,  LINES 31 56 101 1 04 1 52 292 313 332 406 41 1 439 475 508 530  a  Figure  Mapsee2 d i d n o t c o l l e c t  LINKS 1 32 57 102 105 153 293 314 333 407 412 440 476 509  objects.  in  Chains  127  Figure 5.10  Chapter  5 . The  A Mapsee2 Decomposition Sketch Map Instances Implementation  H i e r a r c h y of A s h c r o f t  128 In p a r t i c u l a r , town the  were other  went  of  relax  connection  discovered chain.  under  type  the  a  Also, bridge,  information the  only  between  i f the  end  of  r o a d s or a r o a d the  road  chain  went o v e r a b r i d g e  or a  that  was  Use  fact  that  the  not  recorded.  important  sketch  map  i f one  be  and  a  i s near  i f a road  w o u l d become v e r y  condition  two  river of  this  were  to  r e g i s t e r e d to  the  image.  Maya schemata a r e However, by  MAIDS  Mapsee2.  from  structured  schemata a r e  The  sketch  less  use  only  MAIDS  upwards c o m p a t i b l e  schema-specific  maps  than  routines  the  that  schemata.  with  those  use  p r i m i t i v e f o r m of  used  information the  retrieval  funct ions. Digitized kind  from  images c o n t a i n  that  images can  be  a  curve  closed  shore. (water  useful in a  i n s i d e ) or an  s h o r e l i n e can  in  the  scheme, as suffice to  land.  at  this  was  sketch gives island  guide  intensity  the  image,  described  W h i l e Mapsee2 and the  knowledge s o u r c e  no  be  whether  f o r the a  pixel  section,  not  to provide other  Chapter  Implementation  useful  source.  from  labelled a  a  lake  l o c a t i o n of features  classification will  generally  t o water and  cooperate  p o t e n t i a l e x i s t s f o r the  in  instance,  i t is  I f the  corresponds  MISSEE do  For  appropriate  simple  last  derived  unambiguously  search  region  different  maps.  outside).  of  The  can  indication  then  is  Some i n f o r m a t i o n  (water  interpretation  5.  the  map  i n the  to determine which  time,  maps.  that  in i n t e r p r e t i n g sketch  However, t h i s  the  each  in sketch  information  in t h i s  which way  i n t e r p r e t a t i o n of  information  to  aid  the  129 5.3.3.  The  The tion  user  s y s t e m and  can  perception.  rather  in  exact  Section  5.6.  is  due  to the  types  how  t o be  while  must  and  input--other  l a c k of  used. to  influencing i n the  the  cycle is  interface  the  the  is  format  i s used,  than  interactive  be  i s accomplished  know  output  information  environment  queue u s e d  which t h i s  graphic  of  displayed.  information  user  interpreta-  i t i s to  Unfortunately,  i n t h a t the  messages and,  the  interactive  information  means f o r r e c e i v i n g g r a p h i c This  the  means by  of  certain  i s v i a the p r i o r i t y  The  "unnatural"  internal  modify  means of p r o v i d i n g  interpretation  described  can  amounts of  user's  provide  d e t e r m i n e when and  he/she  differing  The the  influence several aspects He/She c a n  addition,  allow  of  can  process.  t o the In  User  there  is  sketch  h a r d w a r e on  our  of no  map.  current  system.  Section to  a  user.  includes image  the  illustrate  setting  the  the  feature  Also, col  Global  hardware  This  d e s c r i b e d the  (cf. Section  to determine what  4.4.1  of  output  global values  5.  The  image a s  accessible,  the  the user  process by  the  user. can  scale  Depending  direct  the  images  desired  "images" c l o u d  statements that  Implementation  to  the  a whole  and  used  a v a i l a b l e to  s u c h as  available  In a d d i t i o n , g l o b a l v a r i a b l e s a r e  c o n t a i n a number w h i c h must be  Chapter  input  the  instantiation  output  of  of  i s represented  a l l the  types  i n p u t p e r t a i n s t o the  4.6).  is  three  less  than  a global  that  device.  in Figure  are d i r e c t e d t o the  on  5.1.  proto-  printlevel  130 variable  before  can  be  very  the  user.  they  selective  In MISSEE, the image" X  image".  as  a  change  sage,  one  sees  (e.g. map  can  in  the  image, but s u c h as  5.4.  to  image or  specify  still  context to the  that,  the for  useful be  the  made by  in  the  "find  an  objects, adding  the  In a d d i t i o n , one  can  rearranging  specific  the  the mes-  intro-  Along item  specified  as  While  that  of  system,  features with in  information.  MISSEE t o do  in a robust  of  i n t e r p r e t a t i o n and  are.  a position.  length  needs  X  p a r a m e t e r s of  what  where t h e y  can  output  system.  everything  allowing  i s an  specific  be  of a c o r r e s p o n d i n g  object  calculating  reiterated  altering  the  statement,  find  queue.  the  on  i n t e r p r e t a t i o n by  i s concerned with  the  intensity  ability  By  *road~7) p r o v i d e s  l o c a t i o n of  the  the  context  input  s y s t e m , can  of  image and  "there  to  priority  information  i n the  road),  the  river  Hence,  depending  priorities  requests  change t h e  input  (e.g.  the  a  priorities  duce more l o c a l  user  or  complete  information,  t o the  i n t h e queue.  Local  very  message t o t h e  the  elements  or  General  road  appropriate  actually printed.  global  i s equivalent  i n the  such  are  the  sketch  a  gives  region the  user  in  the  important  u s e f u l bookkeeping a  river,  i t i s not  object  Additionally,  either  this  is  an  the  it  chores  should  be  necessary.  I n s t a n t i a t i n g Schemata  MISSEE shows Chapter  the 5.  i n t e r p r e t s images of models used The  in their  Implementation  small  urban a r e a s .  decomposition  and  Figure  5.11  specialization  131 hierarchies.  While  this  i s only  exhibited  in Figure  4.4,  i t is sufficient  that  be u s e d .  This  section describes  will  tiates  t h e schemata  particular  in this  graph  images and s k e t c h  maps.  Figure  Chapter  a small  5. The  5.11 The G e n e r i c  Implementation  from  p o r t i o n of to c l a s s i f y  domain  the scenes  MISSEE  instan-  the i n f o r m a t i o n  found i n  Objects  how  the  i n MISSEE  1 32 5.4_.j_.  Instantiation  The bridge,  "bottom  and  river--must d i r e c t l y  are  the  particular user  tain by  instantiated  ranges.  the user  road  interact  selected  knowledge.  The l o c a t i o n  finding  of  knowledge.  this  For  schema  procedures  s k e t c h map  registered  the  digitized  (in  pseudo-Lisp)  _5._4.J_._j_.  schema  section  (preceded  by  Responses  from  to  these  5. The  largest  IS's.  a  bridge.  i n i t s s l o t s by There  and  work  with  (Glicksman,  are  excerpts  explanations  the program a r e i n d e n t e d  Implementation  two  of  o f what and  with  illustrate 1982).  the A i d of a Sketch  Map  "pseudo-code"  i s accomplished.  preceded  [ 3 ] R i d g e s and c u r b s a r e c r e a t e d a s p a r t o f t h i s not i n s t a n t i a t e d independently.  Chapter  moun-  specified  the  two s u b s e c t i o n s  procedures  of b r i e f  the  b r i d g e : one t h a t r e q u i r e s a  The f o l l o w i n g  consists  "|")  be  for  involves f i l l i n g  I n s t a n t i a t i n g % b r idge-1  This  than  t o t h e image a n d one t h a t d e a l s o n l y  image. how  by  t h e landmass  to find  example,  The  first.  MISSEE  attached  starts.  chosen  F o r example,  the a p p r o p r i a t e f e a t u r e s i n the  bottom-up  i s generally  i n t h e image may a g a i n  5.12 i s t h e g e n e r i c  Instantiation  be  road,  one o r two I S ' s  This  systems a r e e a s i e r  o r by model  with  may  Sources  mountain,  when p r o c e s s i n g f i r s t  i n t h e image c a n be e x a m i n e d  Figure  are  schemata—town,  is first  g e n e r a l model  the Information  bottom-up[3].  occurs  object that  "knows" t h a t  region  level"  interface  o r from  schema  from  five  when t h e y where  Directly  by  process,  ">".  they  1 33  s k e t c h m a p i tem:  value: n i l  %confidence: n i l  %if-added: (prog n i l ( p r i n t l b " v a l u e added t o s k e t c h map i t e m = " % v a l ) ) %if-removed: (prog n i l ( p r i n t l b " v a l u e removed f r o m s k e t c h map i t e m = " % v a l ) ) %if-modified: (prog n i l ( p r i n t l b "value modified i n sketch map i t e m = " % v a l ) ) orderlist: neighbourregions: shadowregions: roadregions:  value: value: value: value:  a-part-of—> decomposes-to—> instances—> neighbours—>  (%road-system (%curb) nil nil  n n n n  i i i i  l l l l  %confidence: %confidence: %confidence: %confidence:  n n n n  i i i i  l l l l  ;r i v e r - s y s t e m )  confidence: n i l conf-alg: (prog ( v a l 1 s t ) ( s e t q v a l (cond ( ( s g e t v %name ' s k e t c h m a p i t e m 'n) 50.0) (t 2 5 . 0 ) ) ) ( s e t q 1 s t ( s g e t a %name ' d e c o m p o s e s - t o ) ) (cond ( ( n u l l 1 s t ) ( r e t u r n v a l ) ) ((atom 1 s t ) ( r e t u r n ( p l u s v a l ( t i m e s 0.2 ( s g e t c 1 s t ) ) ) ) ) (t ( r e t u r n (plus v a l ( t i m e s 0.2 (quotient (apply 'plus (mapear '(lambda (n) ( s g e t c n ) ) 1st)) (length 1 s t ) ) ) ) ) ) ) ) Figure  5.12 The S t e r e o t y p e  manipulation  *bridge-1  h a s been s e l e c t e d from t h e s k e t c h map a n a l y s i s r e s u l t s  Figure  Chapter  will  be  Schema  Schemata  (cf.  functions  %bridge  3.6) t o a i d i n t h e i n s t a n t i a t i o n  5. The I m p l e m e n t a t i o n  underlined.  of a  bridge,  If  then  134 the  variable  cedure  SKETCHMAPITEM  basis  searches  for  interpreted same  possibly 1.  be bound t o i t when t h e p r o -  i s entered.  The  the  will  of the r o u t i n e regions  as area  road  is  in or  the  the  appropriate  shadow.  which can d e f i n e  following:  sides  i s a check  ined  before.  LINKs  that  are  of  routine c a n be  found  in  t h e b r i d g e and  a shadow.  i f (SKETCHMAPSCHEMA = (sgetv '%bridge 'sketchmapitem (return 'alreadyexists) )  This  area  Then edges  the  The  Sgetv  looking  MAPSCHEMA 2. | I n s t  t o d e t e r m i n e whether  for  searches one  as i t s sketch = (snewi > create  this  '(instances))  schema h a s  from % b r i d g e  instance map  'yes 'one  schema  been exam-  down, t h e " i n s t a n c e s " containing  SKETCH-  item.  '%bridge) a new b r i d g e  instance:  p1 = ( c a r ( s g e t a SKETCHMAPSCHEMA p2 = ( c a r ( s g e t a SKETCHMAPSCHEMA > p i = (57 . 33)  %bridge-1  'side1-desc)) 'side2-desc))  > p2 = (69 . 62) Since  Mapsee2  schemata a r e u n d i f f e r e n t i a t e d , s g e t a  value  of  points  of the bridge  slots.  p1  and  p2 a r e  the  returns the  l o c a t i o n s of the mid-  sides.  3. | r e g l i s t  >  = ( p o i n t s t r i p s p1 p2 % r e g m a t f i l e ( q u o t i e n t 15.0 ( s g e t v ' % s c a l e ' f e e t p e r p i x e l 'no ' v a l u e o r d e f a u l t 'one))) r e g i o n s l i s t = (130 1186 1629 9 1750 2018)  Pointstrips from  a  enclosed  Chapter  searches  for  region-merging in  a  a l l  the  algorithm)  rectangular  5. The I m p l e m e n t a t i o n  strip  whose  regions  in  %regmatfile  corners  are  (generated that are 15  feet  1 35 from p1 4.  and  p2.  orderlist  =  (ordinterp reglist  > order Orderinterp  list  "%bridge-1"  of  HILLS.  determines  with  Other) if  ROAD,  WATER,  URBAN,  appropriate  regions  in  the  "shadowregions",  slots  t o p2.  In  The this  their  r e g i o n s a r o u n d and  The  "neighbourregions". from p1  and  Bridge  order  case  are  list  they  over  a  SHADOW,  added  to  the  "roadre-  shows t h e  are Other,  order  Shadow,  Bridge.  5. | ( s p u t v Put in  regions  consistent  schema  and  regions  and  are  Shadow B r i d g e  Interpretations include  MOUNTAIN, and  gions",  (Other  i n t e r p r e t s the  interpretations bridge.  =  Inst)  the  Inst  'orderlist  value  orderlist)  of o r d e r l i s t  i n t h e VALUEd  type  of  the  same name  Inst.  6. | ( s p u t c Put  the  Inst  'orderlist  confidence  7. | e d g e l i s t  =  of  a l l the  was  searched  orderlist  ( p o i n t s t r i p s p1 p2 (quotient 15.0 (sgetv  > edges l i s t Find  100)  =  (88  edge segments for  z e r o - c r o s s i n g s of  11  the  100  ( t h e maximum).  %edgematfile ' % s c a l e ' f e e t p e r p i x e l 'no 'valueordefault 'one)))  205  i n the  regions.  at  206  241  242)  same r e c t a n g u l a r  strip  that  Edge segments come from g r o u p s  L a p l a c i a n of  the Gaussian  applied  to  of the  image. 8.  i f (null  edglist)  then If  there  correct.  (sfail-model Inst)  are  no  edge  So,  remove  segments it  and  graph. Chapter  5.  The  Implementation  then any  of  this  model  can  not  be  i t s descendants  from  the  136 9.  if  (notwithinrange edgeangle b r i d g e o r i e n t a t i o n ) t h e n (remove edge e d g l i s t ) > e d g e s l i s t = (88 11 242)  Remove a l l edge segments t h a t a r e n o t tion 10.  of  i n the sketch  map, w h i c h  to the  orienta-  i s 67 d e g r e e s .  forall i in edgelist ( a d d t o e d g e l i s t ( l i n e n e i g h b o u r s i -1.0 35 % e d g e f i l e ) ) > e d g e s l i s t = ( ( 8 8 8 9 ) ( 1 0 11 12 13 14 15) (242 2 4 3 ) )  Expand  a l l the  ments in  the bridge  similar  out  segments  of the r e c t a n g u l a r  a similar  11.  edge  in  strip  that are  o r i e n t a t i o n as e x i s t i n g  ( f o r e a c h g r o u p o f edges  edgelist  edge  to include connected  segt o and  segments.  i in edgelist  (linestats i ) ) Linestats the  calculates  average  the  contrast  length  across  of  the  edge  segments,  t h e edge, and t h e maximum  con-  trast . 12. | ( m a t c h r e g i o n s t o e d g e s o r d e r l i s t Match the r e g i o n s One r e s u l t schema. the  from  of t h i s  A  special  of  this  shown  Figure  5.13  executed  sputc  with  after  routine will  the l a s t  the appropriate  13. | r i v e r r e g  type  has  exist. value  Inst) edges  for  %curb  slot  one  the  confidence is filled  flag set.  ' % r i v e r ' r e g i o n s 'yes 'valueonly ' a l l '(instances)) > r i v e r r e g = ( ( % r i v e r - 1 130 1058 943 874) ( % r i v e r - 2 1750 2018 2 0 9 5 ) ) f o r a l l regions r in neighbourregions if r i s in riverreg then (saddl Inst 'neighbours R i v e r ) )  Chapter  = (sgetv  5. The I m p l e m e n t a t i o n  edgelist.  edge  i s reserved  executed,  The  in  i n s t a n c e s of the % c u r b  the data  After  is  the  schema c o n t a i n s  shadows. in  with  i s t h e c r e a t i o n o f new  Each %curb  bridge.  regionlist  edgelist  of for  schemata algorithm  by a c a l l t o  1 37  avgstrength: maxstrength: length: angles: edgesegs: type: a-part-of-> conf-alg:  value: value: value: value: value: value:  3.5 5 20 (248 214) (88 89) shadow  %conf %conf %conf %conf %conf %conf  idence idence idence idence idence idence  75 75 1 00 (80) 1 00 1 00  %bridge-1  (prog n i l ( r e t u r n (quot i e n t ( a p p l y 'add ( c o n s ( s g e t c %name ' l e n g t h ) ( c o n s ( s g e t c %name ' m a x s t r e n g t h ) ( s g e t c %name ' a n g l e s ) ) ) ) (add 2 ( l e n g t h ( s g e t c %name ' a n g l e s ) ) ) ) ) )  confidence:  85  5.13a % c u r b - 1 ###################################### avgstrength: maxstrength: length: angles:  v a l u e : 96.33 v a l u e : 134 v a l u e : 54 v a l u e : (90 57 79  edgesegs:  value:  type:  value:  %conf idence %conf idence %conf idence 45 90 27) %conf idence (10 11 12 13 14 15) %conf idence road %conf idence  a-part-of—>  % b r idge-1  conf idence:  93 5.13b  100 1 00 1 00 (80) 1 00 1 00  %curb-2  ######################################  avgstrength: maxstrength: length: angles: edgesegs: type:  v a l u e : 18.0 %conf v a l u e : 22 %conf v a l u e : 20 %conf v a l u e : (228 270)% c o n f v a l u e : (242 243)% c o n f v a l u e : road %conf  a-part-of—>  %bridge-1  confidence:  93 5.13c  Figure Chapter  idence idence idence idence idence idence  1 00 1 00 1 00 (100) 1 00 1 00  %curb-3  5.13 T h r e e % c u r b  5. The I m p l e m e n t a t i o n  Instances  1 38 else  (addtoqueue (adddemon  ' t d ' % r i v e r (sgetc Inst) ( l i s t 'region r ) ) ) ( l i s t 'demonaddlink ' % b r i d g e r Inst 'neighbours))  > neighbours = (%river-1 Search they  through a l l the e x i s t i n g contain.  bridge,  add  versa.  If  links not,  instantiate schema  If  they  send  i t with  link  14. | r o a d r e g  river  schemata  f o r the  rivers  from  the  the  appropriate instantiated,  between  regions  regions  bridge  a message t o t h e % r i v e r  i s subsequently  neighbours  %river-2)  match t h e n e i g h b o u r i n g  t o the  'regions  of the  and  vice  schema t o t r y t o  regions.  If  a demon w i l l  the  %river  e s t a b l i s h the  i t and % b r i d g e - 1 .  ' % r o a d ' r e g i o n s 'yes 'valueonly ' a l l '(instances)) > roadreg = n i l f o r a l l regions r i n roadregions i f r i s i n roadreg t h e n ( s a d d l I n s t ' n e i g h b o u r s Road)) e l s e (addtoqueue ' t d '%road ( s g e t c I n s t ) ( l i s t 'region r ) ) ) (adddemon ( l i s t 'demonaddlink ' % b r i d g e ' r e g i o n s r Inst 'neighbours)) > message added t o QUEUE: ( t d % r o a d 68 ( r e g i o n 1629)) > DEMON i n i t i a t e d : d e m o n a d d l i n k  Repeat  = (sgetv  t h e a c t i o n s o f 13 f o r  the  road  that  passes  over  the  as  the  bridge. 15.  (sputv (sputc  Inst Inst  'sketchmapitem 'sketchmapitem  > value Instantiation sketch  has  map i t e m  %bridge-1  (addtoqueue Chapter  that  at this  16. | ( a d d t o q u e u e  t o sketchmapitem  succeeded  for this  imum a n d p r o p a g a t e for  added  SKETCHMAPSCHEMA) 100 t )  so  schema. fact.  add  =  *bridge-1  SKETCHMAPSCHEMA  Set i t s confidence  Figure  5.14 d i s p l a y s  t o t h e maxthe  stage.  'bu ' % r i v e r - s y s t e m 'bu ' % r o a d - s y s t e m  5. The I m p l e m e n t a t i o n  (plus 5 (sgetc Inst)) (list Inst)) (plus 5 (sgetc Inst))  schema  139  sketchmapitem:  value: *bridge-1 %confidence:  orderlist:  v a l u e : ( O t h e r Shadow %confidence: v a l u e : (2018 1750 9 %confidence: v a l u e : (1186) %confidence: v a l u e : (1629) %confidence:  neighbourregions: shadowregions: roadregions: a-part-of—> decomposes-to-> neighbours—>  100 Bridge 100 130) 100  Bridge  Other)  100 100  nil (%curb-1 % c u r b - 2 % c u r b - 3 ) (%river-1 % r i v e r - 2 )  confidence:  68 Figure  5.14  Instance  %bridge-l  ( l i s t Inst)) > messages added t o QUEUE: (bu % r i v e r - s y s t e m 73 ( % b r i d g e - l ) ) (bu % r o a d - s y s t e m 73 ( % b r i d g e - l ) ) Send messages t o t h e part tem  of. %bridge-1  two  will  higher either  i n s t a n c e s t o become p a r t  level find  of,  schemata  existing or  will  that bridges  road  and  cause  are  river  them  sys-  to  be  created. Figures edges  and  5.16  show  graphically  ( r e s p e c t i v e l y ) t h a t were u s e d  _5._4._1_._2. Like and  5.15  Instantiating the  last  explanations  intensity  Chapter  section, this of  image a l o n e .  5. The  %bridge-2  how The  one  can  to i n s t a n t i a t e  regions  I n t e n s i t y Image  section  contains  instantiate  and  %bridge-1.  from t h e  argument  Implementation  the  pseudo-code  a bridge  from  the p r o c e d u r e e n t e r s  the  with,  140  Figure  Chapter  5. The  5.15  Ashcroft: %bridge-1  Implementation  Regions  Figure  Chapter  5.16  5. The  Ashcroft:  %bridge-1  Implementation  Edge Segments  (Curbs)  1 42 if  i texists,  the  i s a number d e n o t i n g  a r e g i o n t o be  searched  for  bridge.  The WATER  routine finds  surrounding  regions  a  t h a t c a n have t h e  region with  p o s i t i o n s midway a l o n g  interpretation  the interpretation  t h e common b o u n d a r i e s  a r e used  ROAD.  The  to f i x the  edges of t h e p o s s i b l e b r i d g e . 1.I  i f #args = 1 then r e g = ( a r g 1 ) i f ( o r ( n o t (memq 'ROAD ( r e g i o n i n t e r p r e t r e g ) ) ) (regioncheck reg '%bridge)) then ( r e t u r n n i l ) e l s e go t o 3.  Whenever t h e r e  i s an argument,  image t o be u s e d . tion  used  minsize  t o attempt  bridge  If  there  image  = (quotient ( t i m e s 50 50) (square (sgetv  not  n o t have been  previ-  ' % s c a l e ' f e e t p e r p i x e l 'no 'valueordefault 'one))) 'ROAD)  the l a r g e s t  region  i n the  c a n have ROAD a s one o f i t s i n t e r p r e t a t i o n s  and has  r e g i o n = 18  been t r i e d  region  ( c f . Sec-  minsize  i s no r e g i o n  that  in the  instantiation.  reg = ( n e x t l a r g e s t r e g i o n ' % b r i d g e a r e a i f (regioncheck reg '%bridge)) t h e n go t o 2. i f (null reg) then ( r e t u r n n i l ) >  region  One o f i t s p o s s i b l e i n t e r p r e t a t i o n s  5.3.1.2.1) must be ROAD a n d i t s h o u l d  ously 2.I  i t i s a specific  i s less  before. than  area  specified,  = 1695 find  Stop s e a r c h i n g  2500 s q u a r e  3.  when  the  area  of t h e  feet.  neighb = (neighbourregions reg) > n e i g h b o u r s = (19 9 1359 738) F i n d a l l t h e r e g i o n s t h a t s h a r e a common b o u n d a r y w i t h r e g .  Chapter  5. The I m p l e m e n t a t i o n  1 43 f o r a l l regions n l i n neighb i f (memq 'WATER ( r e g i o n i n t e r p r e t n l ) ) t h e n n2 = ( m o v e a c r o s s (commonboundary r e g n l ) ) i f (memq 'WATER ( r e g i o n i n t e r p r e t n 2 ) ) t h e n go t o 5. i f (#args = 1) then ( r e t u r n n i l ) e l s e go t o 2. > n e i g h b o u r 1 = 1359 > no p r o p e r n e i g h b o u r p a i r s f o r r e g i o n 18 2. > r e g i o n = 1629 a r e a = 1330 3. > n e i g h b o u r s = (1186 130 9 2131 1750 2201 2293 and 13 o t h e r s ) 4. > n e i g h b o u r l = 130 > n e i g h b o u r l = 1750  from from from  > neighbour2 Check e a c h WATER.  region  two r e g i o n s .  in  the  If  the f i r s t  store  bounding  neighbours ment,  give  = 1186  t o s e e i f i t c a n be  the midpoint  interpreted  o f t h e common b o u n d a r y  p e r p e n d i c u l a r t o the boundary  new r e g i o n this  t h e new r e g i o n  of  shadow r e g i o n  Then move back a c r o s s r e g (1 p i x e l  direction  SHADOW, If  i n neighb  I f so, f i n d  the  = 130  fact  i s r e a c h e d can  and keep g o i n g  c a n be i n t e r p r e t e d  regions  i s found up,  that  has  been  i n neighb,  else  between  at a  time)  o f n1 and r e g . interpreted  i n t h e same  as  direction.  a s WATER, t h e n a good  found.  then  be  as  I f no s u i t a b l e  pair  p a i r of  i f r e g was g i v e n a s an a r g u -  go back and r e s e t  r e g t o t h e next  largest  region. 5.  Inst  = (snewi > create  Only  '%bridge) o f new b r i d g e i n s t a n c e :  %bridge-2  now do we have enough c o n f i d e n c e and i n f o r m a t i o n t o  an  i n s t a n c e f o r the b r i d g e .  in  the l a s t  orderlist  section are  d a r y " and " m o v e a c r o s s " Chapter  5. The  P r o c e s s i n g c o n t i n u e s much a s i t d i d  when t h e s k e t c h map was u s e d .  determined  create  from t h e r e s u l t s  functions.  Implementation  p1,  p2,  and  o f t h e "commonboun-  From s t e p 5 on, i n s t a n t i a t i o n  1 44 continues For are  almost this  identically.  particular  similar.  Note,  more e x p e n s i v e features Figure  In S e c t i o n  there  is  existence  establish  the  schema  to those  from both  second prone  sufficient  shown how formed. to  object.  value: n i l %conf idence:  orderlist:  shadowregions: roadregions:  two  of  the  Instances support New  the  the  bridge.  The  curb  nil (%curb-4 % c u r b - 5 % c u r b - 6 ) (%river-1 % r i v e r - 2 )  conf i d e n c e :  43  Figure  5.17  Implementation  Instance  four  rela-  a r e c r e a t e d when  the  hypothetical  evidence,  or a  lack  nil  a-part-of—> decomposes-to—> neighbours—>  The  missing  both  5.13.  v a l u e : ( O t h e r Shadow B r i d g e %conf idence: 100 v a l u e : (1750 130) %confidence: 100 v a l u e : (1186) %confidence: 100 v a l u e : (1629) %confidence: 100  neighbourregions  5.  to  %bridge-2.  in Figure  evidence  geographic  s k e t c h m a p i tern:  Chapter  routine is  e x i s t e n c e of  for  procedures  the H i e r a r c h i e s  5.4.1, i t was  a  t h a t the more  among schemata a r e  of  results  and  final  identical  B u i l d i n g up  tionships  to  shows t h e  schemata a r e 5.4_.__.  however,  computationally  necessary  5.17  image, t h e  %bridge-2  Bridge  Other)  1 45 thereof,  may c a u s e t h e i n s t a n c e  generic  objects  instance. determine tiated, at  the neighbour  the region  5.4.1.1  two  neighbour  the  road  two  over  regions  shared  regions (step  road  i f and when t h e r o a d  and  In S e c t i o n  been  i n common w i t h was  to  are instan-  to rivers.  It  used  instan-  %bridge-1,  assumed  associated  that  be f o r m e d with  that  was i n s t a n t i a t e d , a demon would add t h e LINKs between t h e  schemata.  cause  built  system  up ( s t e p and r i v e r  additions schemata chies, possible mata  in Section  5.4.1.1  the s p e c i a l i z a t i o n  new i n s t a n c e s to  the  network  i f demons  f o r that  the  messages  itself.  schemata  such as  T h i s communication  were u s e d  (see Section  knowledge t o e x i s t  road  t r yto f i t  network o r b u i l d  i s above them  that  hierarchies to  from below, t h e y  semantic  t o have knowledge o f what but  level  system a r e e n t e r e d  into the e x i s t i n g  were  and d e c o m p o s i t i o n  1 6 ) . When h i g h e r  on new  requires  i n the  hierar-  5.6) i t would be  i n the higher  level  sche-  alone.  A straightforward algorithm  1)  is  belongs to a  had a l r e a d y  13).  to the with the  h a d n o t y e t been e s t a b l i s h e d so no LINK c o u l d  1 4 ) . However,  would  is  that  When b r i d g e s  i t belong  two r i v e r s  LINKs  and d e s t r o y e d  information  the bridge  beside  LINK was formed  Also present  be  going  and because they  region  created  relationships.  i t was assumed t h a t  the  (step  implicitly  Instantiation provides  least  tiated  are  t o be d e s t r o y e d .  shown i n F i g u r e If  Chapter  the  new  5.18. instance  f o r b u i l d i n g up t h e h i e r a r c h i e s  There a r e three is  5. The I m p l e m e n t a t i o n  not  basic  spatially  possibilities.  n e a r any e x i s t i n g  1 46  H I G H — t h e type of h i g h l e v e l schema being j o i n e d NEW — t h e new i n s t a n c e t o be put i n t o the h i e r a r c h i e s  INSTANCES = a l l c u r r e n t  i n s t a n c e s of HIGH  -> done  ^Foreach  I i n INSTANCES  done  Cny Neighbours i n " Common Between Components of J and NEW?  N >  Figure  Chapter  -> done  -Create new Instance of HIGH: G -LINK NEW t o G -Send Message Up the A p p r o p r i a t e Hierarchy(s) to Continue B u i l d i n g  -> done  N = N +1  F = I LINK NEW t o F  0"  MERGE I Into F  5.18 B u i l d i n g up t h e S e m a n t i c  5. The I m p l e m e n t a t i o n  Network  147 elements belonging instance  of  appropriate ment  of  that  node.  then  the  schema  hierarchies.  one  that  to a higher  higher  3)  If  and 2)  level  node  higher  is  level  establish  An Figure  with  the and  generic  type  a l l merged the  a l l indicated. sketch  intensity  the  last  to  would This  map  new  LINKS  i n the  instances the  The  AIO  each  wishes  to  graph.  is links  shown  in  have  not  instances  three  result  in Ashcroft  new This  that  for  other  network would  entities  i n t o one  it  of  node  instance.  names u s e d  be.  one  requires only  of  ele-  descendant  more t h a n  descendants  i t i s c l e a r from t h e  correspondence  network..  types  of  all  of  if  were i n s t a n t i a t e d  image.  two  used  Procedures  subsections to  i t was  shown how  i n s t a n t i a t e schemata and  Top-down p r o c e d u r e s a r e  schemata  whose  instantiation  used would  bottom-up to b u i l d  to d i r e c t most  pro-  up  the  atten-  profitably  the i n t e r p r e t a t i o n .  Section  Chapter  i s n e a r an  and  semantic  first  instance  objects  The  are  advance  the  5.19.  cedures  tion  up  are  f a t h e r of  provide  Top-Down A t t a c h e d  In  message  of  "intended"  5.4.3.  another  network  the  f r o m the  new  example of a s e m a n t i c  LINKs a r e the  a  a p p l i c a b l e and  i t s ancestors  been drawn s i n c e what  creates  i t becomes a new  instances  generally instance  new  it  neighbours with  schema w h i c h a l s o becomes t h e algorithm  schema  sends  I f the  level  i t shares  higher  level  type 5.  4.5.1  described  applies The  model  Implementation  two  types  knowledge  of in  top-down c o n t r o l . the  a b s e n c e of  The other  148  Figure Chapter  5. The  5.19  An  Ashcroft  Implementation  Instance  Hierarchy  149 information. tem", to  then  search  tions)  of  that  possible  geosystem  a  landmass  higher  to  not  been  only  have  and  top-down  procedures  that  Note  the  The  that  In  these  type  of  Thus,  the  again  is  user  an  (its which  a  geosys-  decide  whether  two is  down  5.  the  The  the  of  about  specializa-  easier  they  those  try  and  more  of  this  of  find a  bottom-up routines  their  hierarchies,  So, but  Implementation  mountain,  bottom  of  control  the  to  information,  can  the the  item or  be  a  applied  redirect control  and  schemata.  to  takes  combine  place  w i t h model  " f i n d a geosystem",  the do  (road,  s k e t c h map  only  control  requests,  efficient,  connects  give  procedure  be  it  the  other  would  Consequently,  schemata  at  it  be m o r e  procedures  to  While  schemata  are  absence  available  knowledge  modularity.  level  top-down  landmass.  contain  which would  i n s t a n t i a t e d waterbody,  instantiate a  Chapter  "Find  will  hierarchies.  top-down  top-down  is  if  that  instantiate  second  the  bottom  also  context  trol  of  schemata  know  five  current  to  procedure  because  appropriate  actually  already  requests,  waterbody  levels,  to  their  routines.  not  user  knowledge  done  knowledge  bottom-up  do  a  level  needs The  the  hierarchies  to.  or  jump s e v e r a l  directly.  region  the  top-down  on a p r i o r i  the  schema  etc.)  if  moves c o n t r o l down  has  each to  for  example,  established.  All  this  the  based  reliably  sort  For  it  the  geosystem  schemata in  still  a more  when  the  knowledge. and  there  might  decide  direct  informed  con-  manner.  1 50 It  is also possible  f o r top-down  schemata  directly.  instances  in river-systems  procedure  example,  in  were two  rivers,  search  i n the  area  were a s u i t a b l e d i s t a n c e  apart,  for a  If  the  bridge  might  For  supporting directly  system. similar  This  routines the  merge t h e  type  efficiency  results  been  f r o m the  use  only  system  rivers,  i t could  not  river  if  separating  objects  of p r o c e d u r e has gain  the  bridge  three  instantiate  c a s e where t h e  between t h e  f e a t u r e s were f o u n d ,  and  to  they them.  i n s t a n t i a t e the  into  one  river-  implemented, of  but  lateral  a  mes-  sages .  Section  5.4.1.1  indicates that  tiated,  a  schema.  When schemata a r e  they  gain  this  case,  should  be  lateral  can the a  bottom  used  bridge road  message t o t h a t the  be  message  For  example,  sketch  map  While tem,  efficiency  instantiate Chapter  5.  proper i f the  he/she  send a  an  questions region  case  to the  information  not are  in  Implementation  its  specific  routines  of making region  has  come  that  an  from  In  regions  for  sure  the  number)  is  bottom-up p r o c e d u r e . the  instance  This user.  from  the  position.  designed still  the  top-down  the  specify  exact  s y s t e m was  every The  form  might  a n a l y s i s use  this  can  becomes one  this  schemata.  r i v e r s ) and the  road  knowledge  of  then  (in  the  one  others,  instan-  to the  "knows" t h a t  schemata  in  14  i n s t a n t i a t e other  of  presented  a l s o necessary  step  role  information the  in  The  contextual  is  procedure  (and  sent  was  i n s t a n t i a t e d , some of  to help  schema.  level  was  when % b r i d g e ~ 1  t o be  important.  image,  or  a production One worse  sys-  could  t r y to  yet,  every  151 point, trol  and every  curve  w o u l d be v e r y  ever,  for  i n the sketch  simple  anything  map w i t h  and s i m i l a r  other  results  than t r i v i a l  whereas t h e methods d e s c r i b e d  flexibility  and p r o c e d u r a l  one-dimensional  described several  in  Section  places  determine  instantiation  would ensue.  How-  herein  Gaussian-Smoothed  version 4.7  i n MISSEE.  the features  Con-  would  provide  prove  greater  adequacy.  C l u s t e r A n a l y s i s Using A  model.  images t h i s  unfeasible,  5.5.  every  has  of  the  been  and t o r e j e c t  should  clustering  implemented  I t has been  that  Histograms method  and u s e d i n  s u c c e s s f u l l y employed  be u s e d  the e x i s t e n c e  to f i l l  to  slots  during  of h y p o t h e t i c a l  sche-  mata.  The static road  method has been a p p l i e d b o t h fashion.  The in  ridges  i n a mountain  t h e same d i r e c t i o n .  used over  in  mountains  are  the r e c t a n g u l a r i t y of a region  ridges  t o check  the consistency  180 d e g r e e s  that  same  Several  range a t l e a s t  of t h a t v a l u e .  i s , 0 and 180  cases.  locally  tend  t o run  edge  direction  across the the  segments a r e d i s c o v e r e d  at a  example, and t h e i r  o r i e n t a t i o n s a r e added i s executed a f t e r  5. The I m p l e m e n t a t i o n  Orientation varies  represent  are close to a chain  Chapter  the s i d e s of a  incremental  degrees  time, because they  The a l g o r i t h m  a  is static.  (ignoring the bright/dark  orientation.  and  The o r i e n t a t i o n s o f edge segments c a n be  edge) a n d i s c y c l i c ,  ets.  incremental  The o r i e n t a t i o n o f e d g e s a l o n g  and o f t h e  Determining  i n an  i n the sketch  map,  to the histogram  each  new  for buck-  collection  of  152 orientations  is  provided  sistent  with  the others  mation  may  alter  w h i c h may be  result  to determine  found  so f a r .  the confidence  i n the confidence  recalculated.  This  i f t h e new Of c o u r s e ,  ridge t h e new  of the o l d v a l u e s value  recalculation  infor-  ( s e e below)  o f an e x i s t i n g  affects  i s con-  ridge to  the c o n f i d e n c e of  m o u n t a i n s , and so on up t h e h i e r a r c h y .  After the  the o r i e n t a t i o n  following  value:  is  of samples  (m), and t h e v a r i a n c e  orientation tion  information  t h e number  cluster  histogram  used  value,  is  broken  returned  for  in i t s cluster  level  each  clusters, orientation  ( n ) , t h e mean o f i t s  of i t s c l u s t e r  a confidence  into  ( v ) . Then  is calculated.  f o r each  The  func-  i n MISSEE i s 1 OOn  if  v. = 0  if  Io - mI  N  C(o) =  50n  > 3v  N 50n  * (2 -  o - m| )  N  3v  where o i s t h e o r i e n t a t i o n This  function  ences. n/N  to inter  C l u s t e r s w i t h many members  clusters,  will  values  5. The  from  o f e l e m e n t s whose  have a l a r g e r confidence  .5 t o 1.0  values.  value  the  i s similar  i s scaled  differ-  value  value.  raises  feature value  Each c o n f i d e n c e  Implementation  number o f  and i n t r a - c l u s t e r  will  have a h i g h e r  a function varying  mean o f t h e c l u s t e r .  Chapter  and N i s t h e t o t a l  responds both  so e a c h e l e m e n t  dence  o t h e r w i se  to  of  Within confito the range  153 from  0  for  to  the  well  angles  as  those  The is  This value  handled  case).  ridges previously  in  a  Since  one  manner  can  roads  clustering  for  Table  in  these  %road-4  5.2  results.  shows t h e  _5.5.__.  one  grid,  then  roads  should  will  (such as are  the not  a l l the  they  be  are  5;. The  5.20  is  roads  139,  385,  have a much lower form  the  chain  cluster-  together  392,  patusing  results  1 i n the  that are 383,  by  the  in grid  shows t h e  values  the  usually  found  to chain  roads  T h i s makes  represented  often  straight  urban  that  t o be  roads  regions  are  (city Also,  rectangular  [4]Curvature i s e a s i l y determined r e p r e s e n t a t i o n - - s e e A p p e n d i x B.  Chapter  p a r t of  included in  Figure  edges  rectangular.  a l s o appear  as  of  sketch  calculated and  confidence.  68  are  Figure  %road-4.  Rectangularity  i s assuming the  confidence  inclusion  be  (as  road  (corresponding  Since  for  ridges.  straight  edge segments t h a t  Determining  If  to  shows the c o n f i d e n c e  t h e minor c l u s t e r  5.21  are  degrees.  the  instantiated.  i n urban a r e a s  modulo 90  map).  considered  similar  also cluster  r e t u r n e d as  segments t h a t can  s k e t c h map)  orientation  from  line  that the,roads  i n the A s h c r o f t  terns  of  Roads t h a t c u r v e  ing!^].  i s then  in a l l ridges being  orientation  assumptions  6  100.  Implementation  laid  out  blocks) small  i f the  in a  that are sections  rectangular bounded of  by  rivers  scale is large.  from t h e  generalized  line  1 54  12.0 2 *  a  #clusters best a cluster  rightend  1 2 a =  11.0 2  10.0 9.0 3 3  8.0 7.0 6.0 5.0 4.0 4 6 7 7 9  numsamples  92 151  mean  22 5  12; number  3.0 10  2.0 11  variance  19.455 126.400  639.066 98.640  of c l u s t e r s = 2 1 st  * ****** 01 0  0  2 0  * * * * *  *  *  * 1 0 0  3 4 5 6 7 8 9 . 0 0 0 0 0 0  **  * *  * * * * * *  * *  1 1 1 1 1 1 2 3 4 5 0 0 0 0 0  1 6 0  1 7 0  1 8 0  orientation F i g u r e 5.20 C l u s t e r i n g O r i e n t a t i o n s i n A s h c r o f t % r o a d - 4 1st. the r e s u l t of c o n v o l u t i o n with the f i r s t d e r i v a t i v e ( p o s i t i v e or n e g a t i v e ) . 2nd. t h e r e s u l t o f c o n v o l u t i o n w i t h t h e second d e r i v a t i v e .  Hough rectangles space", whereas  transforms (Sloan,  the extent we  will  have p r e v i o u s l y been u s e d 1982).  However,  of the p r o j e c t i o n use  orientation  moves a l o n g  the boundary p o i n t s  orientation  of the l i n e  distance  (d) a p a r t  varying  both  o  joining  5. The  of  the  space.  every  Implementation  bounding  "dot  product  boundary  points,  method, one  and c a l c u l a t e s t h e  p a i r of p o i n t s a  specific  of the o r i e n t a t i o n s .  and d, a t w o - d i m e n s i o n a l number  used  In t h i s  of the r e g i o n  t o form a h i s t o g r a m  whose e l e m e n t s a r e t h e r e s u l t i n g  Chapter  Sloan  to f i n d  By  t a b l e c a n be p r o d u c e d  of c l u s t e r s .  The  centre  155  ANGLE  EDGE  1 1 1 4 1 4 18 22 23 27 37 45 45 45 45 63 72 109 1 24 127 1 36 1 36 1 54 1 58 1 75 180 180 180 180 180 Table  71 394 69 387 359 59 1 36 384 382 367 1 38 70 393 72 139 385 383 392 68 361 364 1 35 386 366 1 37 360 58  If  i s constant  the f i g u r e  clusters in  will  5.22  for result  Chapter  7 2 7 1 3 8 6 1 1 5 6 7 2 7 6 1 1 2 7 4 5 6 1 5 6 3 8  81.301 81.365 81.365 81.450 81.427 81.406 81.321 81.108 80.938 80.938 80.938 80.938 80.556 80.364 17.974 18.443 18.499 18.218 18.218 78.622 78.537 78.176 78.069 78.069 78.069 78.069 78.069  region  i n the c h a r t  i n d i c a t e s a "good"  i s rectangular,  for  other  the h e u r i s t i c  diamond-shaped  polygonal  i n three  5. The  a  Implementation  where t h e of  should This  rectangle.  shapes t o o .  clusters.  Values  choice  whose means a r e 90 d e g r e e s a p a r t .  Figure  works  CONFIDENCE  5.2 O r i e n t a t i o n s a n d C o n f i d e n c e  of mass o f t h e l a r g e s t clusters  CURB  number  of  parameters. result  is  i n two  illustrated This  F o r example, a  method triangle  156  Figure Chapter  5. The  5.21  Ashcroft:  Implementation  %road-4  Edges  (20,20)  (10,10)  (10,30)  (0,20)  a 15 dist 3 4 5 6 7 8 9  1 4 1 3 1 2 1 1 10  2 2 2 2 2 2 2  12;  a =  2 2 2 2 2 2 2  2 2 2 2 2 2 2  1  2 0  3 0  4 0  6  2 4 4 4 2 2 3  6 4 6 4 4 2 3  6 4 6 8 6 4 3  6 4 6 8 6 8 5  number  of  20 20  * * * * * * * *  * *  7  numsamples  91 0  * *  8  2 4 2 2 2 2 3  = 7;  rightend  2  1  2 4 2 2 2 2 2  2 2  distance  cluster  0  2 3 2 2  9  * * 5 0  6 0  8 0  * *  2  3  clusters  mean  1 0 0  4  6 6 6 8 8 8 6 10 10 8 8 12 1 0 10 1 2 8 1 21 2 10 14 1 4  6 8 10 1 2 1 4 1 6 18  = 2  variance  45.0 136.0  * * 7 0  5  * * 1 1 0  386.8 386.8  * * * * * * * *  * *  * *  * *  * *  1  1  1  1  1  1  3 0  4 0  5 0  6 0  7 0  8 0  or i e n t a t ion  Figure  Chapter  5.  5.22  The  Clustering Orientations of a R e c t a n g l e  Implementation  along  the  Boundary  158 This  t e c h n i q u e has been u s e d  determine longest and in  i f a region  stable  section  i n t h e image  i s roughly  i n the c h a r t  with  t h e r a n g e of (80,110) If  no  stable  means i s n o t c l o s e  This  predicate  are  alone  must have t h e  predicate  rectangular.  two c l u s t e r s  And, then  places  false  property  of  i f the region  the confidence  within  used  MISSEE,  blocks"  is  as  possible  context  cedures.  This  directed  Chapter  i s raised  i s rec-  slightly.  mechanism  in  not  Queue,  through  a schema, how  such as s c a l e  from t h e r e s u l t s  i n Chapter  4.  M e s s a g e s , and Demons the c y c l e of  perception  queue.  i t i s t o be  (cf.  Each e n t r y  entered,  and  of one o f i t s a t t a c h e d  i s a n a l o g o u s t o , b u t more l i m i t e d i n Maya and a c h i e v e s  Implementation  are  a t t e m p t s have  f o r the e v a l u a t i o n  invocation  5. The  to a river  inference  by means of a g l o b a l p r i o r i t y  queue d e s i g n a t e s  they  In p a r t i c u l a r ,  image i n t e r p r e t a t i o n , a s d i s c u s s e d  4.4)  or  that  global values  i s exercised  i n towns from t h e image  possibilities  made t o i n f e r  Priority  Regions  are other  been  Section  returns  i n two ways.  river  an  there  explored.  Control  found  i s returned.  rectangularity  of t h a t  be f r u i t f u l l y  Control—The  The  If i t i s  predicate  corresponding  could  5.6.  the  i s used by t h e s y s t e m  While c l u s t e r i n g i s several  then  to  s e c t i o n c a n be f o u n d o r t h e d i f f e r e n c e i n  i n s t a n t i a t e d as " c i t y  rejected. tangular,  degrees  enough t o 90 t h e n  that  the  a  t h e d i f f e r e n c e of t h e c l u s t e r means i s c a l c u l a t e d .  true.  of  i n MISSEE a s  than,  a similar  on a  pro-  pattern-  modularity.  159 Messages a r e sent other  schemata  able.  The u s e r  information  as  to,  to  attached of  2)  as  Lateral  reflect already  i t sits  to  rank  the importance the  when t h e a p p r o p r i a t e Consequently,  of  o r minus  importance  o f t h e message.  some  small  Finally,  such as l o c a t i o n ,  may  modify  5.4.1.1 parts  Section  for  an e x a m p l e .  described  how  were  spatially  whose p a r a m e t e r  Chapter  5. The  list  Implementation  is  a value  to  to  those  to  reflect  the  u s e f u l to  provided. to  Sometimes  the  queue;  see  i t i s necessary  queue.  For  to  example,  s c h e m a t a i n t h e h i e r a r c h i e s were  contiguous. included  value  relative  sometimes merged when a newly i n s t a n t i a t e d schema they  of  to the sender.  some i n f o r m a t i o n  of messages t h a t a r e on t h e  5.4.2  sender  g e n e r a l l y use i t s CONFI-  A f u n c t i o n , a d d t o q u e u e , a d d s messages Section  are  be known a s  can p r o v i d e  number  be  is  procedure  A priority  message  A schema w i l l  plus  procedure,  top-down.  his/her  that  "know" two a s p e c t s  relative  The u s e r  it  the  schema must  i n the h i e r a r c h y  new  (top-down o r  parameters  o r t h e u s e r ) must  queue.  entered  context  DENCE v a l u e  the  4)  schema  and  t h e message.  avail-  A message i s  number,  messages a r e a l w a y s s e n t  in  i s being  The name of t h e r e c i p i e n t  where  included  requirements.  i s evaluated.  (a schema  to  messages e i t h e r t o i m p a r t  t h e schema  the  t o t h e schema  attention  1) t h e name o f t h e g e n e r i c  how  establish  receiver.  well  initiate  3) a p r i o r i t y  t h e message  the  parts:  redirect  o f i n t e r p r e t a t i o n become  o r t o make known h i s / h e r  bottom-up), used  the r e s u l t s  can a l s o  made up o f f o u r directed  from s c h e m a t a t o  If  t h e name  there of  the  revealed  that  were any messages instance  whose  160 identity  was  submerged,  t h e queue a s w e l l . hierarchies  tions  t h e new name w o u l d  This situation  are i n i t i a l l y  Another  then  being  global f a c i l i t y  established  by  will  is  executed,  i s t h e demon l i s t .  attached  procedures,  e a c h demon  conditions are true. remove i t s e l f  While  suggest time, be  a r e g i o n t h a t might a  demon w i l l  link  between  saves  the r i v e r  bouring Section Appendix  The  user  i s evaluated  after  pro-  run  it  will  for  correspond  from h a v i n g one  to a r i v e r .  At the  f o r the r i v e r create  a  corresponding  c y c l e i s shown  with  the user  in  The demon's a c t i o n i t s neigh-  to a bridge  interaction  should  take  f o r any i n i t i a l  ( t d geosystem  5. The I m p l e m e n t a t i o n  initiated;  5.23.  controlled  place.  instance.  executed.  Figure  c a n be e a s i l y  same  neighbour  i n c l u d e s an example o f a demon b e i n g  execution  mes-  to a c t u a l l y  t o s e a r c h a l l of  E c o n t a i n s an example o f one b e i n g  message  lateral  a b r i d g e , a message may be s e n t t o  be s e t up t o w a i t  schema  i s not q u e r i e d  Chapter  func-  i f its initiation  g l o b a l v a r i a b l e % t e r s e i s u s e d a s t h e major  default  attached  a demon h a s  t h e two i n s t a n c e s and v a n i s h .  5.4.1.1  whether  the  a r e a c t i v a t e d when  every  I f i t i s , t h e demon w i l l  regions  interaction  mine  After  Demons,  i n MISSEE i n c o n j u n c t i o n w i t h  instantiating  instantiated.  The  Normally,  when  from t h e l i s t .  Demons a r e u s e d sages.  arise  i t on  formed.  some f u t u r e c o n d i t i o n becomes t r u e . cedure  often  replace  Note  how  and v a r i e d .  switch  to deter-  In t e r s e mode, t h e  information.  100 n i l ) i s s e n t .  Rather,  the  Additionally,  161  I n i t i a t e QUEUE -By U s e r -By D e f a u l t  done  Remove F i r s t E n t r y F r o m QUEUE a n d Execute Appropriate Procedure  Instant iate Schema  Executing Procedure Sends o u t Messages  Add Possibilities  Figure Chapter  5. The  -5) E x e c u t e DEMONS  5.23 The E x e c u t i o n  Implementation  1) S a t i s f i e d ? 2) E s c a p e t o Lisp -Change Global Values -Examine t h e Schemata 3) M o d i f y QUEUE -Add I t e m -Delete -Rearrange  Cycle  -> d o n e  162 the  user  is  not  proceeds.  On  each c y c l e detailed display  the  prompted  the other user  and  cution  cycle  control  Lisp  anything,  i s broken. be  of  found  perception  interpretation of  perception  i n the can  Both p r o v i d e  f o r the  and  a  maintain  then  E x a m p l e s of  This execution cycle cycle  queue.  i n Appendix  appears (Figure  found  timely  f o c u s of  for the  processing  in  that permits  the  values the  such  is  the  finished,  execution  interaction  as  environment,  When h e / s h e  either  system  i s resumed  w i t h i n the  exe-  E.  t o be q u i t e d i f f e r e n t  from  4.2).  controls  same manner and be  with  during  i s p r o v i d e d on  i s entered  interact  the p r i o r i t y  doing  can  loop  the  influence  ( t o change g l o b a l  examine s c h e m a t a ,  modify  the c y c l e  query  to  perhaps without or  to  and  to  or  chance  i n the p r o t o c o l )  escape  interpretation  i f t e r s e mode i s o f f , t h e n  above what  user  etc.)  as  and  A new  %printlevel,  a  information  (over  v a r i o u s ways[5]. to  hand,  has  information  for  However,  each aspect  somewhere i n t h e  instantiation  of  it of  the  execution  relevant  the  cycle loop.  schemata  attention.  [ 5 ] 0 n e m i g h t wonder how t h e u s e r can g e t out of terse mode when no i n t e r a c t i o n i s t a k i n g p l a c e . In a L i s p e n v i r o n m e n t , i t i s p o s s i b l e t o break the e x e c u t i o n , execute a command such as ( s e t q % t e r s e n i l ) , and t h e n c o n t i n u e t h e e x e c u t i o n where i t l e f t off.  Chapter  5. The  Implementation  6  CHAPTER R e s u l t s and  MISSEE has The  results  described. how  have  indicate  the  This chapter  Figures been are  Ashcroft  been  used  6.1,  images of  usefulness  summarizes of  image and  success  and  6.3  photograph  Columbia.  have  been  Two  of  which  sources  for  East  a l l the  to  called  distinguish  images can  the  Spences  Bridge  t o show t h a t MISSEE i s r o b u s t  extracted Bridge  purposes.  C r a n b r o o k , a l l of  The  Spences  interpretation  have  s u b i m a g e s of  be  3.  images t h a t  respect to scale d i f f e r e n c e s - - f o u r to will  its  three  with  images  indicates  for i l l u s t r a t i v e  other  extracted  and  d e s c r i b e d i n Chapter  H o u s t o n , S p e n c e s B r i d g e , and  in B r i t i s h  areas.  ideas p r e v i o u s l y  results  some r e s u l t s  show t h e  s m a l l urban  of t h e  the  in previous chapters  6.2,  used:  6  been t e s t e d w i t h  t h e y meet t h e c r i t e r i a The  Evaluation  be  one  in  this  case.  S p e n c e s B r i d g e West them.  found  The  and  information  i n F i g u r e s 6.4  through  6.9. 6.J_.  Instantiating Except  of  the  i n the case  instantiation  when a s k e t c h map Of  course,  receive  O b j e c t s W i t h t h e A i d of a S k e t c h user  process,  the  i s used  only those  guidance,  where t h e  but  in addition  is controlling best to  objects represented s i n c e the  user  results the  every  should  step  appear  intensity  i n the  probably  Map  image.  s k e t c h map  included  all  will of  1 63  164  Figure Chapter  6.  Results  6.1  and  Houston,  Evaluation  British  Columbia  Figure  Chapter  6.2  6. R e s u l t s  Spences  and  Bridge,  Evaluation  British  Columbia  166  Figure  Chapter  6. R e s u l t s  6.3  and  Cranbrook,  Evaluation  British  Columbia  167  Chapter  6.  Figure  6.4  Ashcroft:  Results  and  Evaluation  Information  Sources  Figure  Chapter  6. R e s u l t s  6.5 Houston:  and  Information  Evaluation  Sources  169  Figure  Chapter  6.  6.6  Results  Spences  and  Bridge: Informat ion Sources  Evaluation  170  Figure  Chapter  6.  6.7  S p e n c e s B r i d g e West: I n f o r m a t i o n  R e s u l t s and  Evaluation  Sources  171  Figure  Chapter  6.  6.8  Spences B r i d g e  Results  and  East: Information  Evaluation  Sources  Figure  Chapter  6.  Results  6.9  and  Cranbrook:  Evaluation  Information  Sources  173 the  items  Two ated map  tion  a  can  to f i n d  Recall  interpreted  several  as  1)  image, o r  Section  3.2,  upper the  of  Thus  one  5.1. the  s k e t c h map  their  Table  6.  could.  digi-  may  ambi-  of  be  interpreta-  and  (possibly) situations  instantiated  in  t e r m s used  in  In t h e  cases  unsatisfiability  and  s h a p e s of  the  Ashcroft  6.1  d i v i d e s the intended  R e s u l t s and  map  instances  into  by  the  chains.  The  table  i n the  image.  case. left  Thus,  ideally,  column  interpretations  Evaluation  person  because  reveals  the  instan-  They a r e p r i n t e d to  find  all  would would be  be  of  also  i n s t a n c e s were a b l e t o g u i d e  the  unintended  sketch  t h a t were p r o d u c e d  i n s t a n c e s t h a t were u n a b l e  i n lower in  the  those  counterparts  while  of  interpretation  case.  Chapter  i n the  erroneous be  sketch  respectively.  interpretations and  two  i n t e n d e d model c o u l d not  s k e t c h map  case  i t intended)  would g e n e r a l l y be  i n the  image a r e  letters  drew  these  that represent  tiation  "intended"  in Table  the  element  i s one  who  associ-  a d v a n t a g e of  so t h a t t h e r e  unintended  those  w h i c h of  to take  s k e t c h map  the  shown  ambiguities  procedure  i n the  possible interpretations  drew t h e  limitation.  a corresponding  "unintended".  were  who  a  when t h e  2) an  ambiguity,  The  result  be  that chains  the person  that are  occur:  not  schema t h a t knows how  image.  (i.e.  that should  of e r r o r s can  instances t r i e s  guously  and  interest,  types  with  tized  the  of  support  the  in  intended  i n upper  printed  in  in  case lower  174 Two whose This  errors  road  the  surprising  regions  mistaken  will  very  and  tain",  so  tended"  road  Table is  an  to f i n d  straight  edges  s k e t c h map  a heuristic  can  6.2  summarizes t h e of  situation CHAIN  the  first  i n the  image.  If  they  that correspond  interpretation used  13  search for  d i s t a n c e , then  type.  to  of  the will  roads "moun-  t o remove t h o s e  because  i s the  i n the  the  *BRIDGE-1 *TOWN-1 *RIVER-1 *RIVER-2 *ROAD-3 *ROAD-4 *ROAD-5 *ROAD-6 *ROAD-7 *ROAD-8 *MOUNTAIN-1 *M0UNTAIN-2 *M0UNTAIN-3 6.1  "unin-  R e s u l t s and  Evaluation  image  as  Here  there  to  find  a  road[l]. joined  the  the  UNINTENDED INTERPRETATION(S)  *road-1 *road-2 *r iver-3 *r iver-4 *r iver-5 *r iver-6 *river-7 *river-8 *R0AD-9, *river*R0AD-10, * r i v e r *road-11, * r i v e r -  i n the very  map.  failed  r e g i o n merger has  A s h c r o f t : Sketch one  f o r Houston.  * r o a d - 5 has  INTENDED INTERPRETATION  Table [1]Chain-5  situation  instantiated  arises  3,5 4 9 1 1 1 0 2 6 1 7 8 1 2 1 3 1 4  6.  be  and  ( r i d g e s or c u r b s ) .  However, c h a i n s  have t h e  agreement  12  mountains both  o f f at a c e r t a i n  f e a t u r e s t o be  Chapter  and  chains  interpretations.  error  proper This  that  come from  s i n c e roads  for roads.  rarely  type  were a b l e  r i d g e s happen t o p a i r be  second  schemata  i s not  bright  of  Map  top  to  Image  right  of  the  sketch  175 region corresponding resulting  average  t o the road with  intensity  t h e one n e x t  has d r o p p e d  t o i t and t h e  below t h e t h r e s h o l d f o r  acceptance.  Similar Bridge  West,  through  6.6.  were  3  results Spences  errors  of  shown  the  first  INTENDED INTERPRETATION  2,4 8 9 1 6 5 3 7  *BRIDGE-1 *RIVER-1 *RIVER-2 *ROAD-3 *ROAD-4 *road-5 *ROAD-6 *ROAD-7  1 3 1 2 9 1 1 10 5,6 2 3 7,8 4 1 4 1 5 1 6  Table  Spences  Bridge,  type  *road-1 *road-2 *river-3 *r iver-4 *r iver-5 *river-6 *RIVER-7,  *town-1 *ROAD-1 *ROAD-2 *ROAD-3 *ROAD-4 *BRIDGE-1 *RIVER-5 *RIVER-6 *BRIDGE-2 *RIVER-7 *MOUNTAIN-1 *MOUNTAIN-2 *MOUNTAIN-3  6. R e s u l t s and E v a l u a t i o n  *bridge  Map t o Image  UNINTENDED INTERPRETATION(S)  *ROAD-8, *RIVER-8 *road-9, * r i v e r - 9 * r o a d - l 0 , *RIVER-  6.3 S p e n c e s B r i d g e : S k e t c h  6.3  there  i n the second  UNINTENDED INTERPRETATION(S)  6.2 H o u s t o n : S k e t c h INTENDED INTERPRETATION  and 6 e r r o r s  Spences  i n Tables  o u t o f 100 s k e t c h map i n s t a n c e s  CHAIN  CHAIN  for  B r i d g e E a s t , and Cranbrook  In summary,  Table  Chapter  are  Map t o Image  176  CHAIN  INTENDED INTERPRETATION  4,5 2 3 10  *BRIDGE-1 *RIVER-1 *RIVER-2 *TOWN-1 *ROAD-3 *ROAD-4 *ROAD-5 *ROAD-6  9  6 7  8 Table  INTENDED INTERPRETATION  4,5 2 3 6 7  *BRIDGE-1 *RIVER-1 *RIVER-2 *ROAD-3 *ROAD-4 *MOUNTAIN-1 *mountain-2  8 9  Table  INTENDED INTERPRETATION  5 4  *ROAD-1 *ROAD-2 *ROAD-3 *ROAD-4 *ROAD-5 *ROAD-6 *ROAD-7 *ROAD-8  8  6  9  3 2 7 Table  *river-4 *river-5, *river-6,  this  Map t o Image  *river-1 *river-2 *r iver-3 *r iver-4 *r iver-5 *r iver-6 *r iver-7 *r iver-8  i s a 9% e r r o r  6. R e s u l t s a n d E v a l u a t i o n  *road-5 *road-6  UNINTENDED INTERPRETATION(S)  6.6 C r a n b r o o k : S k e t c h  Overall,  Map t o Image  UNINTENDED INTERPRETATION(S)  6.5 S p e n c e s B r i d g e E a s t : S k e t c h  CHAIN  Chapter  *river-5 *r iver-6  6.4 S p e n c e s B r i d g e West: S k e t c h  CHAIN  category.  UNINTENDED INTERPRETATION(S)  Map t o Image  rate.  1 77 Visual features 5.16  were  image. and  same as  are  (6.11).  Roads,  like  (6.15).  f o r roads  from  just  often  show  the simi-  f e a t u r e s f o r towns ( F i g u r e associated  to both  from be  the  edges 4,  5,  Intensity  instantiated  image.  with  (6.13 and  edges  and 6  Image  with  However, t h e  t h e a i d of a s k e t c h map to  figures  for  6.14)  is  the  2.  the d i g i t i z e d  discovered  found  and  o b j e c t s i n the A s h c r o f t  region f o r roads  o b j e c t s can  ambiguous w i t h o u t  image  F i g u r e s 5.15  were  following  prominent  Instant iat ing Objects Geographic  The  Mountains are  The  appropriate  instances.  edges t h a t  b r i d g e s , map  1 and  that  "bottom l e v e l "  the  rivers  regions  6.2.  r e g i o n s and  f o r the o t h e r  Regions  (6.12). and  indicates  i n t h e A s h c r o f t image.  results  6.10)  also  matched w i t h MISSEE  d i s p l a y e d the  bridge lar  evidence  instantiate  and  Alone  some  success  situation  i s more  enough e v i d e n c e  many e n t i t i e s  with  low  is  confi-  dence .  Roads and using  images  b r i d g e s a r e c o m p u t a t i o n a l l y much h a r d e r alone because the r e g i o n s that  correspond  have g e n e r a l l y merged t o g e t h e r many  individual  found  searches  a suitable  regions  for  ably  spaced  find  city  r e g i o n , the program  bridges parallel  (an example edges  b l o c k s , from  one  continuous  streets.  r e s u l t s may  be d i f f e r e n t  Chapter  6.  intersection besides  With  even when t h e y a r e  R e s u l t s and E v a l u a t i o n  roads.  Having  f o r bounding  water  roads,  this  to another, being  find  t o them  i s i n S e c t i o n 5.4.1.2) o r  for roads.  Thus,  to  less  suit-  tends  rather  efficient,  correct.  to  than the  178  Figure Chapter  6.  Results  6.10 and  Ashcroft: Evaluation  S k e t c h Map.  %town-1  179  Figure  Chapter  6.11  6. R e s u l t s  Ashcroft:  and  S k e t c h Map.  Evaluation  River  Regions  180  Figure  Chapter  6.12  6. R e s u l t s  Ashcroft:  and  S k e t c h Map.  Evaluation  Mountain  Edges  Figure  Chapter  6.13  6. R e s u l t s  Ashcroft:  and  S k e t c h Map.  Evaluation  Roads  1-3,  Edges  Figure  Chapter  6.14  6. R e s u l t s  Ashcroft:  and  S k e t c h Map.  Evaluation  Roads 4-6,  Edges  183  Figure  Chapter  6.  6.15  Results  Ashcroft:  and  S k e t c h Map.  Evaluation  Roads  1-3,  Regions  184 There found  is a  in  interurban one  Thus  the  i s just  sketch the  denotation,  to areas  ings).  Thus  sketch  The tion  %road-2 croft  one  next  few  the  sketch  niques  are  %road-l2  illustrated  one  i s shown  the  aid  to  downtown That  because  in Figure  6.19  objects  from t h e  Chapter  6.  the  or  all  examined  rivers,  etc. "down-  arbitrary  two  of an  town  buildtowns i n  it  the  a i d of  roads  from  5.21,  sketch  the  is  not  interpretation  images  alone  of  the  simi-  two  and The  6.18,  tech-  when  region  was the  s e v e r a l urban  5  where  one  rectangular.  summarized  Ash-  instantiated  s k e t c h map  the  map.  where % r o a d - 4 i s  map,  acceptable  which c o n t a i n s  are  interpreta-  a i d of a s k e t c h  image a l o n e .  not  Evaluation  one  results  f o r the  R e s u l t s and  being  r e q u i r e d to obtain  in Figures  r e g i o n was  are  images a l o n e ,  Four  i n the A s h c r o f t  examined  results  the with  are  from t h e  To  containing  only  19 e d g e s .  of  MISSEE.  rather  results  with  ( F i g u r e 6.17)  as  by  image.  However, sometimes t h e  6.10.  was  The  contains  a  roads,  t o see  i n the  and  is  instantiated  is instantiated  Figure  alone  was  and  with  corresponding  (near  as  to represent  considering  f i g u r e s compare t h e  similar  instantiated  in  several  image a l o n e  results.  urban  "town"  roads,  been u s e d That  when  image a l o n e  map  has  at  of  roads  images t h a t a r e town  would e x p e c t  but  images a n a l y z e d  city.  and  ( F i g u r e 6.16)  from the lar  the  that are  map,  using  of  semantics  whole t o w n — a n d  the  "blob"  however,  refers  the  In t h e  inside  map  core  i n the  that—the  thoroughfares. i s looking  town", or  the  mismatch  s k e t c h maps v e r s u s  Mapsee, a town  here,  slight  image This  is  regions.  bottom in  shown  Table  level 6.7.  185  Figure  Chapter  6.  Results  6.16  and  Ashcroft:  Evaluation  S k e t c h Map.  %road~2  186  Figure  Chapter  6.  6.17  Results  Ashcroft:  and  Image A l o n e . F o u r  Evaluation  Roads  187  Figure  Chapter  6.18  6. R e s u l t s  and  Ashcroft:  Image A l o n e .  Evaluation  %road-12  188  Figure  Chapter  6.  6.19  Ashcroft:  Results  and  Image A l o n e . S e v e r a l  Evaluation  Urban  Regions  189 BRIDGE  MOUNTAIN  RIVER  ROAD  Ashcroft CORRECT INCORRECT OMITTED  1 0 0  3 2 2  2 0 0  30 1 1 1  Houston CORRECT INCORRECT OMITTED  1 1 0  0 0 0  2 0 0  20 1 5  26 1 5  Spences Br i d g e CORRECT INCORRECT OMITTED  2 10 0  3 0 1  2 3 0  51 14 0  4 0 1  Spences B r i d g e West CORRECT INCORRECT OMITTED  0 7 1  4 0 0  2 2 0  81 0 0  2 1 0  Spences Bridge East CORRECT INCORRECT OMITTED  0 1 1  0 0 2  2 0 0  1 2 2 0  4 0 0  Cranbrook CORRECT INCORRECT OMITTED  0 0 0  0 0 0  0 0 0  7 0 4  6 0 1  Table  There a r e again instantiated objects cause  6.7 A l l Images:  two t y p e s  of e r r o r s :  i n t h e image t h a t d i d n o t have (omitted).  match between model and d a t a  Chapter  the  domain  that  (called the  and t h e r a n g e ) ;  6. R e s u l t s and E v a l u a t i o n  have  been  i n c o r r e c t ) ; and,  right  The f o r m e r e r r o r  (as opposed t o a  7 5 1  Results  instances  t h a t s h o u l d n o t have been  instantiation  between  Instantiation  TOWN  features  to  i s an  incorrect  numeric  mismatch  the l a t t e r  error  i s one o f  190 incompleteness.  Overall, alone this  is  the e r r o r rate  f o r i n t e r p r e t a t i o n from  (#INCORRECT + #OMITTED) / #TOTAL = 24%.  i s higher  that  t h e e r r o r r a t e when a  sketch  the  image  As e x p e c t e d , map  is  also  used ( 9 % ) . 6.3_.  B u i l d i n q Up H i e r a r c h i e s When t h e f i r s t  cause  ( v i a messages)  created. results the  bottom l e v e l  of  earlier  This  relation to each  entire  processing),  network, or cause entry  neighbour  into  subgraph  that  i s m e d i a t e d by t h e n e i g h b o u r  scene elements a r e l o c a t e d  relationships  Generally, bouring  next  regions  relation  routines after Each o b j e c t  finding  the inverse  relation  i s reflexive) i s also  instances  node up a t l e a s t  6. R e s u l t s  either  by  the existence of has  common edges o r r e g i o n s  i n the a p p r o p r i a t e  Neighbouring  established  information  n e i g h b o u r s a n d ways t o e s t a b l i s h t h e l i n k .  i t involves  established,  are  has been c o n f i r m e d .  i t s possible  Chapter  either f i t directly  the  other.  instance  parent  they  i n t o t h e network  demons o r by t h e s c h e m a - s p e c i f i c  about  a b o v e them t o be  the c r e a t i o n of a  which determines which  The  a new  hierarchy  As more schemata a r e i n s t a n t i a t e d ( p o s s i b l y u s i n g  existing  can.  the  schemata a r e i n s t a n t i a t e d , t h e y  instances.  or  neigh-  When t h e l i n k i s  ( a l s o "neighbour",  since  the  created.  o f t h e same t y p e  will  one o f t h e h i e r a r c h i e s .  and E v a l u a t i o n  have a  common  Sometimes a new  191 instance that  will  case,  indicate the  that  parent  i t relates  nodes a r e  two  or more  a l l collapsed  groups.  i n t o one  In  of  the  nodes.  The the  r e s u l t s of  neighbour  formed, the  this hierarchy  relation  hierarchies  is will  a l w a y s be  r e l a t i o n s h i p s are  are  the  However, chies  if  will The  of  hierarchies  split  formation  errors:  1)  into of  an  neighbour  2)  MISSEE,  the  former  has  rarely.  F u r t h e r m o r e , as  Hence, (and on  problematic  i t depends on  the  the  user's  w o u l d be  5.19  to  Chapter  showed t h e  aid  t h i s graph.  and  6.  can  link  the  correctly  be  might  then  If links  formed.  the  hierar-  and  just explained,  from  types  found.  In  the  latter  only  errors  of be  actually which  non-  be  l i n k s may  their slots),  two  made between  not  never o c c u r r e d  relations  i f a l l of  the  in i n t e r p r e t i n g Figure  image was  included  some of  suffer  schemata were  fill  correctly.  missing,  because neighbour  which  be  properly  omission redundant.  instantiated  in turn  depends  strategy.  established  were u s e d  Ashcroft  was  v a l u e s used t o  Figure  in  error  whether  (sub)graphs.  link  the  up  f o u n d but  still  b a s e d on  When i t i s  built  l i n k s can  incorrect  s c h e m a t a , or  less  may  separate  neighbouring  are  not  enough c r u c i a l l i n k s a r e  be  are  established.  some n e i g h b o u r redundant,  building  used.  several  Results  6.20  and  Ashcroft  the  image.  instance  s k e t c h map  the  the  correct  intervening  Evaluation  of)  that  instances  There are  shows what would happen  Only of  (except  no  errors  i f only  instantiations features  required  the are to  192  Figure Chapter  6.  Results  6.20 and  Ashcroft: Evaluation  Instance  Hierarchy  2  193 establish Thus t h e Figure  neighbour graph  l i n k s are  i s fragmented  Ashcroft  All  into  13  (especially  for  p i e c e s more t h a n  roads).  the  one  in  5.19.  Appendix E c o n t a i n s the  missing  types  of  image  a protocol that  where  schema-specific  ing  hierarchy  this  graph which c o n t a i n s  because  i s shown  the  user  the  user  shows  fewer  employed.  6.21.  There are  instances  terminated  session  i n t e r a c t s with  r o u t i n e s are  in Figure  a  the  system.  The  result-  no  than F i g u r e  processing  with  with  errors in 5.19,  only  possibilities  remaining.  Similar Figures  results  6.22  and  Spences Bridge map  goal  randomly  Down t h e  the  the  other  images.  r e l a t i o n s h i p s f o r Houston  automatically  using  the  and  sketch  searching  to  the  attention shifts newly  necessary first  Chapter  guide  6.  i n an  further  image f o r  to search  only  f o r the  and  Evaluation  manner.  determined In an  spatially  item.  Results  processing  interesting  information.  o b j e c t s are  i s t o use  intelligent  in a direction  acquired the  Hierarchies  c y c l e of p e r c e p t i o n  i s accumulated  where a l l of  the  of  and  interpretation  knowledge  and  show t h e  t h a t were d e r i v e d  M o v i n g Up The  of  6.23  for  to guide i n t e r p r e t a t i o n .  6.4.  of  have been o b t a i n e d  the so  results  that  Instead  f e a t u r e s , the by  scene  model  of focus  knowledge  i s l a n d - d r i v e n approach connected,  features  that  it  becomes  will instantiate  194  Figure Chapter  6.  Results  6.21 and  Ashcroft: Evaluation  Instance  Hierarchy  3  195  Figure  Chapter  6. R e s u l t s  6.22 H o u s t o n :  An I n s t a n c e  and E v a l u a t i o n  Hierarchy  196  Figure  Chapter  6.  6.23  Results  Spences B r i d g e :  and  Evaluation  An  Instance  Hierarchy  1 97 When a s k e t c h map to  search  (spatially)  organization network for  the  in  the  user  of  might  the  No  sketch not  map  interested i n the  former w i l l  to s t a r t  information,  of  are  both  for  That  model that  the  cycle  of  river  search  only  the  the the  should  Chapter  6.  has  s p a c e of  be  be  searched  heuristic  latter  has  R e s u l t s and  to  using  rivers,  can  i f the in  their  region  After  map  Bridges  and  mountains  for a chain  will  s k e t c h map  to r e s t r i c t the  be  useful  attached  pro-  candidates  with i n the  roads  (and  sketch  map.  In t h i s  used.  This  case, reduces  instances.  i s not  available,  the  regions  initial  object.  been u s e d : c a l c u l a t e  Evaluation  being  a s u i t a b l e message i s  associated  instance  be  sketch  much  likely  the  often  and  provide  particular,  river.  find  chains  and  will  i s chosen.  information source used  i t again  i f the  several interpretations.  road  t o do  a l s o have t h e a d v a n t a g e of  procedure  i n the  s k e t c h map  knowledge can  simple  the  the  semantic  t h a t might appear  image re'gions t h a t a r e and  fact,  own  instantiated  instance  In  only  perception  s y s t e m s and  each.  top-down  through  If  While  be.  Bridges  and  chain probably  search  of h a v i n g  necessary  In  their  necessary  entities  seldom  not.  determine  will  however,  a  i n any  the a s s o c i a t e d road  rivers)  would be  even  object.  in  requirement  s k e t c h map.  the  about  can  received,  schemata  unambiguous  road  information cedures  initial  u s u a l l y u n i q u e whereas r o a d s ,  generally part  an  i t i s not  were u n a m b i g u o u s l y  true,  towns a r e  the  search  not  So,  the  s k e t c h map  image but the  is available, for  preclude  image.  was  IS  i n the  then image  In  MISSEE,  a r a n g e of  suitable  198 areas  that a p a r t i c u l a r  value,  s c a l e ) and  Instantiation  mation  i s not  image-based  then  will  interpretations  object search  again  routines w i l l  them  regions.  the be  less  have  (using  in order  generate  for other  available,  might  the  accurate  (and  d i r e c t e d to the  global  of d e c r e a s i n g  messages Since  the  about sketch  map  less  size. likely infor-  efficient)  appropriate  regions  or p o s i t i o n s .  Table  6.8  instantiate  shows t h e  the  five  regions  level  Houston  75  1 39  With Sketchmap Intended #Relevant #Extra Unintended  t h a t were s e a r c h e d  schemata.  The  figures  Without Sketchmap Correct #Relevant #Extra Incorrect Total  Spences Spences Spences Cranbrook Br i d g e Bridge Bridge West East 161  1 25  119  285  7 6 10  9 12 1  23 4 8  31 1 0  42 1 2  24 20 0  23  22  35  32  45  44  15 17 7  36 39 7  31 15 51  36 27 9  44 13 1  10 9 3  39  82  97  72  58  22  Table  Results  6.8  and  The  to for  •  Total  6.  bottom  Ashcroft  T o t a l Number of R e g i o n s  Chapter  number of  Number of  Evaluation  Regions  Searched  199 search the  with  tests  regions the  map  the  was  i n e d when t h e  On  number  the  sketch  advice  search,  given  was  have  procedures  associated  6.1  t o 6.7.  the  The  other  requirement  rectangular.  If  searched  instantiation  to  the  anyway.  i s that  not.  is  For  the  exam-  useful in  region  reducwould  one  that  example,  that  As  without  were  the  neighbouring  URBAN can  be  i t s shape be  a suitable size,  towns. roughly  i t would  i t s intensity.  Preserving  about  regions  and  objects prevents  needless  knowledge  as  was  than  features that  indicate  f o r town  when  when i t was  the  are  slots.  regions  interpretation  region  anyway due  positive)  roads  possible  the  a l l of  searched  with  the  41%  they  searched smaller  to  Relevant  in that  is invariably  present,  been  with  of  correspond  in instance  regions  22%  if  not  regions  fill  u s e d and  even  are  would  (and  map  considerably  i n messages  because,  generally  been  sketch  to  of  average,  map  instantiation  only  of a  t o compose T a b l e s  employed was  s k e t c h map.  enable  use  e v e n t u a l l y used  expected,  The  the  d i s t i n g u i s h e d from e x t r a n e o u s ones  ones t h a t a r e  sketch  ing  without  t h a t were u s e d  are  would be  the  and  have  negative  their  attempted  subsequent  process-  ing .  Search 1) tion  i n the  Positional i n the  registered  regions  average  intensity  (the  Chapter  6.  i s a u t o m a t i c a l l y reduced  information  filters  text  image  for of  intensity further  the  information  Results  from t h e  and  image.  2)  processing  region. gained  s k e t c h map  in three  guides  instantia-  Model  knowledge  b a s e d on  3) M o d e l knowledge from the  Evaluation  ways.  the  size  plus  interpretation  and con-  process)  200 suggest focus  likely of  interpretations  attention. giving  6.5.  Interaction  User  In MISSEE, t h e he/she  desires  taking will  an be  system  to  active  over  Influence  i s able  t o do  the  the  t h i s may  take  for search  t h e i r eventual  as  much or of  are  use.  on  how  more  as  little  interpretation.  guarantee  Depending  current  Interpretation  progress  u s e r can  liking.  one  extreme,  In  the  It  current  the  u s e r may  is  not  that well  or  categorizer  classes  regions  to  automatic, execution  cycle  remain[2].  the the  less  will  images  device  was  the  t o examine t h e  as By  results automatic  effort  continue  flag  execution  cycle  network  was  system  h e / s h e draw a  he/she must h e l p  on  there  are  until  set  If that  not  possible  interrupt i t s  more  produced  i n d i c a t i n g an is  i s f i n i s h e d , the in greater  the  ways t o make t h i s  no  output  run  sketch  train  to assign  mode, MISSEE w i l l  available).  semantic  that  have a u t o m a t i c a l l y  (if a  l e t the  5.3.1.2.1)  image, but  terse  MISSEE w i l l  possibly  after  In but  required  (Section  i n the  too.  choose to  implementation,  intensity  then  Can  r o l e , the  automatically.  output  near  part.  At  and  user c o n t r o l  influence  i s performing,  his/her  map.  the  user  to h i s / h e r  regions  Furthermore, p o s s i b i l i t i e s  made e x p l i c i t , How  for  candidates a  protocol appropriate  insufficient, u s e r may  choose  detail.  [ 2 ] I n t h e c u r r e n t s y s t e m , t h a t would be a f t e r e v e r y image r e gion of suitable size and c a t e g o r i z a t i o n has been t r i e d w i t h e a c h of the f i v e b o t t o m l e v e l s c h e m a t a - - b r i d g e , m o u n t a i n , r i v e r , r o a d , and town.  Chapter  6.  Results  and  Evaluation  201 As  indicated in Section  processing  will  be  f o r some t a s k s .  adequate  significantly image. active  role  user  of  the  changes  provide  f a c e , more  lation  sketch  map  follow  three  types  4.4.1.  In  the  the  i f the  be  of  input  terms of  necessary  of  may  well  situation  scene over  user  takes  the  a more  have,  user  can  since  actions  the  nature  t o the  system  improving  only the  to  the  as  accu-  make  minor  possibilities  list.  i n c l u d e adding or d e l e t i n g messages.  i t is  been i n s t a n t i a t e d  possible  network d i r e c t l y  However,  about  they  help of  type  ensure there  by  c r e a t i n g and  that  the  i s no  require  (from  end  Lisp)  destroying  results  " n a t u r a l " user  the  user  of messages and  the  to  to  will inter-  have  schemata  more  manipu-  functions.  With system's  regard  to task  focus.  I f the  certain  type,  direct  MISSEE  p o n e n t s and  addition,  the  i t e m s of 6.  s u c h as to  priorities, user  r o a d s or  look  the  user  river  for only  interest.  Results  and  can  be  the  i s only  s p e c i a l i z a t i o n s ) and  In  Chapter  entities  extreme  knowledge  can  where s c e n e o b j e c t s have n o t  Thus, the  perfect.  user  will  i t may  semantic  instances.  a  m e a s u r e s would  non-existing  from t h i s  inaccurate, although  p a r a m e t e r s of m e s s a g e s on  extreme c a s e s  the  results  i n t e r p r e t a t i o n process.  in Section  i n the  modify  be  can  the  The  results  results,  More d r a s t i c  or  somewhat  drawing  i n the  described  racy  In  by  Even b e t t e r  The was  be  6.2,  to  very  can  narrow  interested in objects systems,  those  then  objects  ignore  (and  irrelevant  specific  Then, h e / s h e can Evaluation  user  about  the of  a  he/she  can  their  com-  messages.  where t o  examine s p e c i f i c  find parts  202 of  the  about  resulting  the The  user  also controls  fulfills  much i s i n c l u d e d to  be  interact  to  obtain  the  desired  information  objects.  interaction  is  schemata  i n each  tailoring  environment  h i s / h e r needs.  i n the  sent,  his/her  and  Global  p r o t o c o l s , i f and  whether  l o o p of  the  the  user  so  input  where  that  the  determines graphic  how  output  wants t o have a c h a n c e  execution  cycle.  would become even more c o n v e n i e n t  Such  to  environment  i f user  models were  utilized.  Appendix E c o n t a i n s with but  the  Ashcroft  i s paying  that  he  that  the  and  shown  slight  will  both  lateral,  to those  and  "user"  be  and  6.18.  E,  found The  has  no  modifications accurate.  are  the  in Figures  resulting  i t proceeds  map  images  ensure  are  generated: While only  5.15,  to  so  guidance  bottom-up p r o c e s s i n g  generated.  some of  as  priorities  i f necessary  Sketch  a l l sorts  user  session dealing  specific  interpretation  top-down and  i n Appendix  similar 6.16,  The  A l s o , messages of  bottom-up, is  make  results  trated.  image.  a t t e n t i o n to the  can  available  a p r o t o c o l f o r a sample  is  illus-  top-down,  the  produced  protocol would  be  5.16,  5.21,  6.10,  are  found  in  Figure  when,  where,  instances  6.11,  6.21 .  In and  summary, t h e  how  of  amount o f step  6.  can  interpretation  interaction  dialogue.  Chapter  user  The  R e s u l t s and  can  i n f l u e n c e the to  range  user  get  the  from a l m o s t  would  Evaluation  what,  desired results. none t o  g e n e r a l l y vary  a the  The  step  by  amount  of  203 interaction  d e p e n d i n g on t h e  priorities, automatic 6.6.  or  both.  This provides  of  processing,  a very  flexible  his/her adjunct.to  processing.  D e s c r i p t i v e Adequacy The  by  d e s c r i p t i v e adequacy of a v i s i o n  how w e l l  solution only  tem  interpretation  of p e r t i n e n t  employed  these  is  Section  scene  relating  3.3.  attributes cedures. after  intensity  an a s s e t  the  images  determined  t o advance t h e task but  i s not also  the  The schema-based  sys-  i n the p u r s u i t of both of  In  of  t o d e s c r i p t i v e a d e q u a c y were  response,  objects  Also,  that  the s l o t  interpretation  explicit  the  an a r e a  Facts  values  for  where e x t e n s i o n s  specific  an e x p l i c i t , to  the  schemata  6. R e s u l t s  the  the source  of  defining pro-  knowledge  information  p r o c e d u r e s do n o t takes  in  of the  i n v i a the attached  instantiation  place.  about make This  be made.  i n a uniform  schemata  t o them.  but an a t t e m p t occurring  manner.  manipulation  access  and E v a l u a t i o n  are  specific  The a t t a c h e d  uniform  names f o r commonly  Chapter  gathering  a r e g e n e r a l l y encoded by  These  provide  might  raised  VALUEd a n d LINK t y p e s  are f i l l e d  methods by w h i c h  maintained  enforces  the  facts explicit.  elements of t h e scene.  dard  In MISSEE,  information.  definitely  schemata make c r u c i a l  are  of  serve  is  goals. Questions  the  system  t h e d e s c r i p t i o n s o f t h e domain  of t h e p r o b l e m a t h a n d .  the  recovery  is  accuracy  The LINKS  system  VALUEd  which  types  are  has been made t o u s e s t a n -  attributes.  These  include  204  "sketchmapitem", corresponding tom  level  "regions",  and  "edges"  to  refer  f e a t u r e s from the i n f o r m a t i o n sources.  Since  bot-  schemata ( b r i d g e s , curbs, mountains, r i d g e s , r i v e r s ,  roads, and towns) a l l i n t e r f a c e to the i n t e n s i t y  image by  of regions and/or edges, the method of d e s c r i p t i o n and  to the  means  i s consistent  uniform. The  clear  knowledge  distinction  enables  becomes  available.  created  and  the  between  system  to  Instances  destroyed  based  general change  are on  s t e r e o t y p i c o b j e c t s are s t a t i c .  and  as  new i n f o r m a t i o n  hypothetical  the  and  evidence.  New f a c t s  particular  can  In c o n t r a s t ,  cause  slots  to  f i l l e d and an i m a g e - s p e c i f i c i n s t a n c e network to be b u i l t . c r u c i a l d i s t i n c t i o n between the g e n e r a l and the s p e c i f i c at a l l l e v e l s i n the h i e r a r c h i e s , not j u s t at the l e a f I n d i v i d u a l schemata are modular but dent.  Instantiation  independently contained,  from  having  of  bottom  other both  directly  from  the  because  procedural  However, higher l e v e l schemata  that  IS's must  be  This  i s made  nodes.  totally  indepen-  l e v e l schemata can take p l a c e  schemata a  not  be  they  are  self-  and a d e c l a r a t i v e p a r t .  can  not  be  instantiated  have knowledge about t h e i r com-  ponents or s p e c i a l i z a t i o n s so that c o n t r o l can be passed  to an  appropriate  During  schema  further  down  in  i n s t a n t i a t i o n , a d v i s o r y messages can purposes)  laterally  to  neighbouring  knowledge of t h e i r a t t r i b u t e s .  the h i e r a r c h i e s .  be  sent schemata  which  require  In b u i l d i n g the semantic  network  messages are sent up and down the h i e r a r c h i e s . Chapter  (for efficiency  6 . R e s u l t s and E v a l u a t i o n  205 Even  though  they  are  large  extent  The  sending  specifies  either  appropriate include such  as  type  1629",  and  on  rely  the  what t y p e "position receiver  IS's  s e c t i o n s have in  that  process.  uniformity, tinction  MISSEE  uniform  of p r o c e d u r e  (52  i t wants t o  the system  finds  context  . 70)",  to extract  the  to a  format.  the  parameters  i n f o r m a t i o n they  shown r e s u l t s  are  they are adequate  i n a way  tation  encode,  "sketchmapitem information i t  permits  adequate  modularity  facts  their'effective  of t h e  made between s t e r e o t y p i c  indicate  use  through  the  images.  from  disparate  i n the  interpre-  the  schemata and  and  that  for interpreting  for combining  This i s accomplished  and  that  explicitness, the c l e a r  dis-  h y p o t h e t i c a l knowledge.  P r o c e d u r a l Adequacy By  providing several  schema,  MISSEE e n s u r e s  interpretation. flexible,  easy  t y p e s of a t t a c h e d p r o c e d u r e s  that  S i n c e the p r o c e d u r e s  are  t o program  to a uniform  (compared  in conjunction with a d e c l a r a t i v e  effective  at  tightly  Chapter  6.  image  interpretation.  coupled with  for  t h e a p p r o p r i a t e method w i l l  operating  are  schemata  i s maintained a  the  of  the  utilize.  Moreover,  are  the  have  Furthermore,  "region  descriptions  for  modularity  top-down o r bottom-up, and  procedure.  Previous  _.]_•  to,  only  i n f o r m a t i o n about  *road-5", can  messages  "know" t h e names of  b e c a u s e t h e messages a l l  sender  address,  schemata must  R e s u l t s and E v a l u a t i o n  be  algorithms,  data  portion  of  used they  processor  structure),  A l s o , because the  the d e c l a r a t i v e  each  and  procedures the  schema  206 (as o p p o s e d modularity  t o a p r o g r a m and  of  them.  shifting  work,  I t has  the  focus  interpret  subsequent  ing  on  ate  sent into  uniform  a  to  specific from  of  to  the the  account  IS's)  modified  The  to  an  to give can  the  semantic  net-  level  provide  and  s c h e m a t a , and  complete  i n c l u d e much  intensity  of  Schemata m a i n t a i n the  context  only  be  the  Thus  a  for  whether  schema  Rather,  new  receive  influence  search  Results  of  and  need  map  to  the  the  guidance  message  current  procedure  is  context to  use.  (perhaps d e a l i n g  message w o u l d have t o  procedural at  processing.  a l l the  appropri-  with be  knowledge.  information  possibilities  rely-  instantiation  ( t h e c y c l e of p e r c e p t i o n  also maintains and  the  by  d o e s not  sketch  w h i c h bottom-up  r e c e i v e r of  the  for  t o use  schema  information  modularity  procedure  added t o a  the  the  message t o d e t e r m i n e  image a l o n e .  in determining  interpretation  6.  passed  a t t e n t i o n to b u i l d  IS--i.e.  c o n t r o l regime  user  Chapter  are  means  bottom-up  incorporate  cycle)  a  this  top-down p r o c e d u r e w h i c h t a k e s  and  tion  he/she  messages t h a t  effective  utilize.  More p r o c e d u r e s might new  of  r e c e i v e r s ' of  procedure  directly  the  i s an  t o decode them.  specify  structure), greater  processing.  the  instead  means of  been shown t h a t  bottom  Messages are required  data  i s maintained.  Schemata c o o p e r a t e by between  a separate  adequacy. timely  Evaluation  intervals  map  execu-  so  i t ranks  reasons,  sketch  the  It allows  Furthermore,  for e f f i c i e n c y regions,  and  even  the that the  though  instances,  and  207  the  like  6.7.J_.  would  be  sufficient.  Solutions  to Problems  Section models  3.2  described  i n model-based  system  is partly  i n Matching Data  four  vision  problems  systems.  d e t e r m i n e d by how  well  in  to  Models  matching  The  adequacy  it  solves  data  to  of a  vision  those  prob-  lems.  In  inconsistent  situations  model),  hypothetical  m o d e l s would  data  element,  one-to-one.  from a d i f f e r e n t model w e l l If  remain  specify  how  As an has  can  been  reduce  be  ranked  example, fixed  other  side  sides  of  a  edges  can  be  edge-1,  then the  there  a r e two  of the c o n f i d e n c e v a l u e s in a meaningful  Chapter  6.  R e s u l t s and  t o one,  those  values  which  side  model.  of a  so  way.  Evaluation  Since  the o r i e n t a t i o n s  a  part  the  the  of the  I f the  too d i s s i m i l a r  c a n be u s e d  that  bridge  c a n d i d a t e s f o r the  between a l t e r n a t i v e s .  i t was  f i t the network.  edge-3 a s d a t a e l e m e n t s ) .  of which  a  f o r each  not  the  instances  t h e c a s e where one  Otherwise, the s i m i l a r i t i e s  ordered  from  for  especially  from the IS's f i t the  edge-2 o r edge-3 was instance  does  confidence  s h o u l d be p a r a l l e l ,  used t o choose  of e i t h e r  tion  consider  datum  to account  instance  their  features  ( w i t h edge-2 and bridge  one  t h e number o f by  one  information,  i t t o be p r u n e d  ( t o edge-1) but  tation  troyed.  the  be c r e a t e d  show t h a t  not  closely  than  Additional  enough and c a u s e  p r u n i n g does  that  I S , may  (more  orien-  to that  can i n the  instances  be  of  des-  calculacan  be  208 The  ranking  in  consequence with the  form  of  attached  to the  likely In  the case  element)  that  indicates  data be  feature  the  the  resulting  handled  Other  instances  being  values.  For  used  6.  procedures  by  the  influence,  of  (no  that  is  existence  model.  of a r o a d ) can  the  pruned or w i l l  R e s u l t s and  model  a  lowered crucial  still  pruned. value  to  not  If  the  of  the  next  Evaluation  inconsistent  ones.  result  a  hypothetical i n some of  (or lower) t h e i r  of  be  p a r t of  neighbouring  situation.  Bridges are  roads.  than  the  confidence  the  intensity to  ele-  either  ambiguous c u r b m i g h t  average  data  of  raise  the  for a  creation  intensities  g e n e r a l l y found  one  manner s i m i l a r  information w i l l  a lower  be  of  be If  the c o n f i d e n c e  (more t h a n  to disambiguate  evidence  lowered a p p r o p r i a t e l y .  cause  The  the  corresponding  If there the  the  the curbs  then  confi-  first.  factor.  instantiate  a  example, an  water w h i c h has  Chapter  in  will  or a b r i d g e .  which a r e  to  situations  instance.  be  a  likelihood  i n s t a n c e would be  element  can  strong  be  the  ranked  is unsatisfied  definitive,  Each d a t a  road  be  another  advice (in  t h r e s h o l d s f o r a c c e p t i n g f e a t u r e s can  not  are  also  has Any  from  the h y p o t h e t i c a l i n s t a n c e w i l l  Ambiguous  a  can  ( f o r example,  then is  issued  a r e examined  where a model  i t i s "easier"  found,  is  instances w i l l  context  elements  ment)  two  values  execution cycle.  T h i s means,, s u b j e c t t o u s e r  data  so t h a t  confidence  that  possibilities  model, t h e n  of  r e s p e c t to the  messages)  dence v a l u e s . most  terms  urban  either regions next  to  regions  209 .The the  problem,  d a t a ) , has  cedures static  be  as  the  capable  incompleteness  of  (no m o d e l s t o a c c o u n t  solved.  M o d e l s and  by  user.  the  Generic  interpretation  of a b s t r a c t i o n  and  a convenient  their  process[3].  interpretation  result  provide  would  learning.  framework  attached  pro-  knowledge To  create  r e q u i r e the While  f o r such  for  this  is new  system system  research, i t  was  attempted.  6.8.  Robustness  The  results  gracefully (cf.  been  created  d u r i n g the  might not  not  must be  models to  final  described previously  when  it  Section 3.5).  corresponds  to  i s unable  Moreover, the  amount  to  the of  show t h a t  interpret  an  degradation  MISSEE image of  degrades completely  the  information available  results from  the  IS's.  If MISSEE tem  the  user maintains  can  produce  perfect  that  the  ensures  the  accuracy  from  the  the  s k e t c h map  decreases tion  of  as  further..  available  control  results.  user  results  i s not  over  chooses  the  The to  interpretation,  robustness  impart  However, even  i s the d i g i t i z e d  the v a l u e of  i f the  only  image, p a r t i a l  the  sys-  less information,  diminishes gradually.  available,  of  then  source  If  guidance  the  results  of  informa-  results  will  be  produced.  [ 3 ] A u s e r w i t h knowledge of t h e schemata manipulation t i o n s c a n c r e a t e o r m o d i f y schemata d u r i n g i n t e r p r e t a t i o n caping into Lisp.  Chapter  6.  R e s u l t s and E v a l u a t i o n  funcby e s -  210 This has  been  trates fect 36  range  the  an  error  image  using  ments. mation  sources.  Chapter  6.  The  that of  the  subject  of  basic conjectures  more i n f o r m a t i o n  is utilized  and  of  system illus-  r e s u l t s are i n 6 out  i n d i c a t e d how  results  the  map  good  to help  when  s e c t i o n 6.2  per-  of  the  results inter-  using  the  (where  the  24%).  by  the i n t e r p r e t a t i o n .  Results  the  of a s k e t c h The  of  Appendix E  information  6.1  of MISSEE i s one  It supports  results  guidance  were t h e  robustness  sections.  provided  Section  f o u n d t o be  robustness  information  only  the  image a l o n e  the  ensures that  ( e r r o r r a t e , 9%).  r a t e was  The  user  execution).  obtained  digitized  the  t h o u g h he  and  in previous  c a s e where t h e  c y c l e s of  pret  results  illustrated  (even  were  i n the  Evaluation  the  and  its  main  accomplish-  about m u l t i p l e the  infor-  more d i f f e r e n t  system, the  better will  the be  CHAPTER  Summary  7.j_.  A  Summary  1)Multiple this  work  several,  o f What  and  i s New  Information  has been  types  a i d in the interpretation  ers  have  raises  the  central  SEE  way  nature  is a  test  one  o f how  t o combine  usefulness  tures.  By  images,  i t has proved  using  and u s e f u l n e s s  of  input  of  images.  Other  information  source,  but  combine  of  combining  information  of v i s u a l  that  The d e s i g n  tools  which  mation  to solve  vision  system  effective  are useful  sources research-  this  work  I S ' s t o one o f  is  general  argreement  i s better),  to substantiate.  source  of adequacy development  different  Since  there  conjec-  and  robust-  of  several of  infor-  vision.  work.  the usefulness  to  interpret  types  this  ( t h e more  MIS-  implemented  i n model-based  the second  sym-  structure.  has been  motivating about  in their  i s available to  i n terms  of the problems  conjectures  i s harder  that  in reconciling  a r e two  (more  and  information  l e d to the  There  hypothesis  i n kind  that  of the m u l t i p l e  of MISSEE  many  vary  o b j e c t - o r i e n t e d data  the information  ness.  better)  unifying conjecture  to effectively  data  i s t o use an  schema-based  the  there  The  importance.  One bolic  than  issue  and I n t e r e s t i n g  the p o s s i b i l i t y  to  more  Conclusions  Sources:  possibly disparate,  used  7  While  of the  different, i s no  first the  satisfac-  21 1  212 tory  definition  determine less, than  how  two  images by  images. the  a  maintained the  a  user  i s even  in  map  and  the  sorts  an  it  more  that  difficult  are  are.  to  Nonethe-  more  different  viewpoints.  Information  different  than  that  found  in  to  from  images  alone  a l l  extract  MISSEE  i s able  r e s u l t s i n d i c a t e the of  is  sources  image  different  i t is possible  that  images,  information  from  information  disparate  two  sketch  recorded  While scene  information  disparate  intuitively  provided  of  of  information  to  derive,  advantage  available  of  from  i t  is  combining  the  various  sources.  .2)Schemata: along  with  objects  programs  of  objects.  for  MAIDS  schemata coming  cedures  are  a  cess.  Also,  the  and  tion  cycle)  Chapter  the  from  7.  a  means  of  via  the  Summary  and  of  a  relations  allows  reduces  driving  for  search,  a  to  be  Conclusions  understanding  for  generic are  examined.  repository  for  attached  pro-  interpretation  pro-  The  network  specializes of  at a l l  objects  been  uniform  the  particu-  static,  have  IS's.  method  In  between  that  These  schemata  hypothetical  from  semantic  allows  of  systems.  consistent, input  (MAIDS).  language  distinct  different  implemented  notions  relationships  construction  neighbours, that  vision  instances  and  provide  flexible  instances  to,  require  of  been  them  natural  i n t e r p r e t a t i o n tasks  information  of  and  needs  multiple  the  has  structuring  systems  hierarchies  schemata  manipulating  the  specific  Also,  necessary  for  of  systems  the  v a r i e t y of  expert  the  vision  levels  many  for to  new  system  adapt  developed  lar,  a  A  to,  control user  linking a l l decomposes (the  execu-  interaction,  213 and  promotes c o o p e r a t i o n 3) U s i n g  S k e t c h Maps:  a  vision  to  draw even  problem  they  contain  This  work  interpret  must be  much u s e f u l  i s the real  first  on  derived  from  representations ing of  the the  first  (regions).  map  different  quick an  as  and  easy  image,  yet  underlying  information  scene.  to  help  Image a n a l y s i s has  usu-  image  Because of  types  sources  systems t o  been u s e d  domain.  aspects  information  vision  the  d e r i v a t i o n of  different  provide  multiple  sketch  Region Data:  the  over  about  in a geographic  (edges) or h o m o g e n e i t i e s are  S k e t c h maps a r e  information  Edge and  concentrated  a n a l y s i s has  superimposed  t o use  imagery  schemata.  S k e t c h map  f o r some t i m e .  i f they  4) C o m b i n i n g ally  among t h e  in the  of  discontinuities practice image, t h e s e  information.  conjecture,  utilize  both  they  edge  two  Follow-  MISSEE i s and  one  region  data.  5) M i x e d C o n t r o l trol able does  ineffectively  utilize  a l l of  i n a model-based v i s i o n not  generally  down s y s t e m w i l l ing  Strategies: Strictly  from  not  take  system.  gained  uses a mixed c o n t r o l s t r a t e g y  (both  with  attention  i s focussed  where  among the  cooperating  schemata.  7.  structure  Summary and  A totally  t o use  it  of can The  Conclusions  con-  is avail-  bottom-up  knowledge.  the  from p r e v i o u s  combines  Chapter  the  able  models of  knowledge t h a t  a d v a n t a g e of model  u s u a l l y be  information  the  linear  context  processing.  system A  top-  resultMISSEE  top-down and  bottom-up)  that  the  network  that  be  semantic most  embodiment  usefully of  this  so  applied strategy  214 is  the  execution  to  give  it  i s not  tion  and  of  receive  necessary. various  sibilities  The  for  information  at  If d e s i r e d ,  the  types to the  l o c a t i o n s i n e i t h e r the reduces  discover  the  increased  the  appropriate user  can  modify  reflect  an  efficiency  sketch  map  or  search  entities  of  interest.  in contrast  to  the  the  s u f f i c i e n c y or  although  relate  informa-  the  of  list  own  pos-  priorities.  the  semantic  schemata c a n  image.  that  opportunity  points  his/her  "bottom l e v e l "  amount of  geographic  which a l l concern  user  cycle d i s t r i b u t e s control within  Messages d i r e c t e d at  locality  the  s y s t e m and  i n t e r p r e t a t i o n to  execution  network. tain  c y c l e which a l s o a l l o w s  con-  Advice  must  about  take p l a c e  This  results  advances mentioned  i n t e r p r e t i v e power  to in  above  of  the  system. Open I s s u e s There are sources has a  paradigm  dealt with data  that  other  prove use  mation  one  that way  s t r u c t u r e and  several also  several  aspects  of  have not  been a d d r e s s e d .  of  combining  that  effective.  For  sources,  erable  I f one  Chapter  aid.  7.  have been u s e d  Summary and  in  example, m e d i c a l  approaches to  a method of  could  This  research  schemata as  Al  There  that  diagnosis  t y p e s of  both are  might systems  test  infor-  domain.  When c o m p a r i n g d i f f e r e n t information  information  interpretation.  r u l e s combine d i f f e r e n t  in a non-perceptual  multiple  IS's--using  object-oriented  candidates  production  the  e v a l u a t i o n , w o u l d be  formalize  Conclusions  utilizing  the  notions  of  the  multiple of  consid-  amount  of  215 information determine the  available  the  that  tive  and  tion  Such a  information might  With  an  be  regard  formalism  redundant to the  or  in  add  Ongoing experience was  quite  will  i n d i c a t e whether  cult  to  will  l e a d t o an Future  there  the  straint  of so  IS's  lute  spatial  road  Chapter  only  or  useful  the  cumula-  increase  and  open  ques-  modularity  it  would  information  be  sources.  developing  continued  objects  (perhaps c l o u d s ? )  system  experimentation  IS's  that are  and  whether  in  complexity.  map  objects.  are  not  diffi-  increased  size  information runs over  7.  Summary and  a  it will  (e.g..  present  be  Conclusions  that  o r more m o d u l a r . In  the  approximately  current  registered  regarding  the  possible to  relax this  con-  use  "There  bridge").  system  information  system d i s c o v e r s  aligned,  that  t o be  spatial  It should  i f the  the  Constraint:  i s assumed  gives  to  more e f f i c i e n t ,  Registration  This  that  two  a  the  sketch  image.  location  C l a i m s of  elements to the  are  overwhelming  for  always  d e v e l o p e d , an  e a s y and  several extensions  l)Relaxing  to  was  objects  However,  w o u l d make i t more e f f e c t i v e ,  the  i s not  using  Directions  There are  system,  how  adding  positive.  incorporate  IS's  to  inconsistent.  e i t h e r new  with  possible  f o r e x t r a c t i n g and  i t s expandability.  i t i s unclear  to  i t w o u l d be  would have t o a c c o u n t  system t h a t  notwithstanding,  2-3_.  then  from d i f f e r e n t  e x i s t s concerning  general  IS,  e f f e c t i v e n e s s of a method  information.  fact  in  or  i s informed  relative  that  i n s t e a d of  i s a town t o t h e  the abso-  left  of  216 2) F e e d b a c k sketch  map  explained that  and  Mapsee2 i s run  s e n d s a l l of  i t s results  ( S e c t i o n 5.3.1), MISSEE can  i s u s e f u l to help disambiguate  schemata, used in  t o Mapsee2:  MISSEE,  i n Mapsee2, a r e  i t should  interpretation  of  simultaneously  to t h e i r  control  3) A  demonstration natural  sketch  system  the  map  MISSEE's  to point  used  methods of  different,  so  them.  the  as  of  a  a d d i t i o n of  for graphical  to p a r t s  that  proceed  usefulness  improved w i t h  Maya  so  can  However, the  a mechanism  user  Since  the  a  input, image  or  map.  the  model  declarative  knowledge  part  of  the  make  it  more  cause  the  instantiation  trend  in  of  I t i s p o s s i b l e t o move some  from the  schemata.  accessible,  other  descriptions tate  sketch  been  schemata,  to c o o r d i n a t e  4) A More D e c l a r a t i v e S c h e m a t a : of  t o MAIDS  has  the  information  map.  substantially  Interface: be  much  sketch  benefit.  i n t e r f a c e and  allowing  the  required  would  the  As  on  f o r them t o c o o p e r a t e  MISSEE a r e  User  language  as  mutual  work would be  Better  and  t o MISSEE.  provide  similar  possible  image  i n Mapsee2 and  considerable  such  the  be  to completion  attached Moving  modifiable,  method t o be  areas  of  objects.  Al  The  less  towards  procedures  the and  knowledge uniform,  flexible. more  into  would  but  would  There  explicit,  framework of MISSEE would  the  is  a  formal facili-  s u c h a move. 5) E f f i c i e n t l y  5.3.1.2, i t was  Chapter  7.  Finding  pointed  Summary and  out  Edges that  and  Regions:  edge d e t e c t i o n and  Conclusions  In  Section  region  merg-  217  i n g were  both  portions  of  control based confine  the  the  image  c a n be a p p l i e d  In a d d i t i o n , t o reduce  shifts  current  that  the focus context  and  of  erally  only  perform  dary  t o p o r t i o n s o f t h e image  uses the a l r e a d y effects  between  selected  search,  MISSEE's  in  6.4).  o f image a n a l y s i s  system:  image  Thus one c a n to  1)  i n a new way  c a l c u l a t e d region neighbouring  the  specific  them when n e e d e d .  r e q u i r e two c h a n g e s t o t h e c u r r e n t  have t o r e f e r  to  attention  ( c f . Section  the expensive o p e r a t i o n s of  would  operations  an image.  mechanism on  parts  local  one  This would  (MISSEE g e n -  n u m b e r s ) , and 2) boun-  s u b i m a g e s would have t o be  considered.  1_. 4.  Conclusions  Schema-based the as  systems a r e capable  a t t r i b u t e s of geographic well  as  Exploiting  directing information  results qualitatively  the  entities  process  in a cooperative  than  preting  the information  sources  Chapter  7. Summary and  Conclusions  those  describing  for their interpretation,  interpretation  sources  better  of s u f f i c i e n t l y  obtainable  separately.  itself.  fashion from  gives inter-  218  References  [1] A g i n , G.J. and B i n f o r d , T.O. ( 1 9 7 3 ) , Computer D e s c r i p t i o n o f Curved O b j e c t s , Proc. T h i r d I n t e r n a t i o n a l J o i n t Conference on A r t i f i c i a l I n t e l l i g e n c e , S t a n f o r d , 629-640. [2] A g i n , G.J. and Duda, R.O. Advanced Industrial Computer C o n f e r e n c e ,  ( 1 9 7 5 ) , SRI Vision Research for A u t o m a t i o n , P r o c . Second USA-Japan T o k y o , 113-117.  [3] A n s t i s , S.M. ( 1 9 7 0 ) , P h i Movement as a S u b t r a c t i o n V i s i o n R e s e a r c h 1 0 ( 1 2 ) , 1411-1430.  Process,  [4] B a j c s y , R. and L i e b e r m a n , L . I . ( 1 9 7 4 ) , Computer Description of Real Outdoor Scenes, Proc. Second I n t e r n a t i o n a l J o i n t C o n f e r e n c e on Pattern R e c o g n i t i o n , Copenhagen, 174-179. [5] B a j c s y , R. and J o s h i , A.K. (1978), A P a r t i a l l y Ordered World Model and N a t u r a l O u t d o o r S c e n e s , Computer V i s i o n S y s tems , A.R. Hanson & E.M. Riseman (eds.), Academic P r e s s , New Y o r k , 263-270. [6] B a j c s y , R. and T a v a k o l i , M. ( 1 9 7 3 ) , A Computer Recognition of Bridges, Islands, R i v e r s and L a k e s f r o m S a t e l l i t e P i c t u r e s , " P r o c . C o n f . on M a c h i n e P r o c e s s i n g o f R e m o t e l y Sensed Data, L a b . f o r A p p l i c a t i o n s of Remote S e n s i n g , P u r d u e , I n d i a n a , 2A-54 t o 2A-68. [7] B a j c s y , R. and T a v a k o l i , M. ( 1 9 7 6 ) , Computer R e c o g n i t i o n o f Roads From Satellite P i c t u r e s , P r o c . Second I n t e r n a t ional Joint Conference on Pattern Recognition, Copenhagen, 190-194. [8] B a k e r , H.H. and B i n f o r d , T.O. (1981), Depth From Edge and Intensity Based Stereo, Proc. Seventh I n t e r n a t i o n a l J o i n t C o n f e r e n c e on A r t i f i c i a l I n t e l l i g e n c e , V a n c o u v e r , 631-636. [9] B a l l a r d , D.H. ( 1 9 8 1 ) , P a r a m e t e r N e t w o r k s : Towards a Theory of Low-Level V i s i o n , P r o c . Seventh I n t e r n a t i o n a l J o i n t Conference on Artificial Intelligence, Vancouver, 1068-1078. [10] B a l l a r d , D.H. and Prentice Hall,  References  Brown, CM. (1982), Computer E n g l e w o o d C l i f f s , New J e r s e y .  Vision,  219  [11] B a l l a r d , D.H., Brown, C M . , a n d Feldman, J.A. (1978), An A p p r o a c h t o K n o w l e d g e - D i r e c t e d Image A n a l y s i s , Computer V i s i o n S y s t e m s, A.R. Hanson & E.M. Riseman (eds.), A c a d e m i c P r e s s , New York,. 271-282. [12] B a l l a r d , D.H. and Sabbah, D. (1981), On Shapes, Proc. Seventh International Joint C o n f e r e n c e on A r t i f i c i a l I n t e l l i g e n c e , V a n c o u v e r , 607-612. [13] B a r r , A. a n d Feigenbaum, E.A. ( e d s . ) ( 1 9 8 1 ) , The of A r t i f i c i a l I n t e l l i g e n c e , Volume J_, W i l l i a m Los A l t o s , C a l i f o r n i a .  Handbook Kaufman,  [14] Barrow, H.G. and Tenenbaum, J.M. ( 1 9 7 5 ) , R e p r e s e n t a t i o n a n d Use o f Knowledge i n V i s i o n , S i g a r t N e w s l e t t e r 52, 2-8. [15] Barrow, H.G. and Tenenbaum, J.M. ( 1 9 7 6 ) , M s y s : A System f o r R e a s o n i n g About S c e n e s , TN 121, AI C e n t e r , S R I . [16] Barrow, H.G. and Tenenbaum, J.M. ( 1 9 7 8 ) , R e c o v e r i n g I n t r i n sic Scene C h a r a c t e r i s t i c s From Images, Computer V i s i o n S y s t e m s , A.R. Hanson & E.M. Riseman (eds.), Academic P r e s s , New Y o r k , 3-26. [17] B a r t l e t t , F.C. ( 1 9 3 2 ) , Remembering, A S t u d y i n E x p e r i m e n t a l and Soc i a l P s y c h o l o g y , Cambridge U n i v e r s i t y P r e s s , Cambridge . [18] Bobrow, D.G. and W i n o g r a d , T. ( 1 9 7 7 ) , An O v e r v i e w o f KRL, A Knowledge R e p r e s e n t a t i o n Language, C o g n i t i v e S c i e n c e 1, 3-46. [19] B o l l e s , R.C. ( 1 9 7 5 ) , V e r i f i c a t i o n V i s i o n W i t h i n a Programmable Assembly S y s t e m : An I n t r o d u c t o r y D i s c u s s i o n , AIM 275, AI L a b , S t a n f o r d . [20] B o l l e s , R.C. ( 1 9 7 9 ) , Robust F e a t u r e M a t c h i n g T h r o u g h Maximal C l i q u e s , Proc. S o c i e t y of P h o t o - O p t i c a l Instrument a t i o n E n g i n e e r s 182, Imaging A p p l i c a t i o n s f o r A u t o m a t ed I n s p e c t i o n a n d A s s e m b l y , 140-149. [21] B o l l e s , R . C , Quam, L.H., F i s c h l e r , M.A., and Wolf, H.C. (1979), Automatic Determination of Image-to-Database Correspondences, Proc. Sixth , International Joint C o n f e r e n c e on A r t i f i c i a l I n t e l l i g e n c e , T o k y o , 73-78. [22] Brachman, R. ( 1 9 8 2 ) , What "ISA" I s and I s n ' t , P r o c . C o n f . o f t h e C a n a d i a n Soc i e t y f o r Computat i o n a l of I n t e l l i g e n c e , S a s k a t o o n , Canada, 212-221.  Fourth Studies  [23] B r i c e , C R . and Fennema, C.L. ( 1 9 7 0 ) , Scene A n a l y s i s R e g i o n s , A r t i f i c i a l I n t e l l i g e n c e 1 ( 3 ) , 205-226.  Using  References  220 [24] B r o o k s , R.A. (1981a), Model-Based Three Dimensional Interpretations o f Two D i m e n s i o n a l Images, P r o c . S e v e n t h I n t e r n a t i o n a l J o i n t C o n f e r e n c e on Artificial Intelligence , V a n c o u v e r , 619-624. [25] B r o o k s , R.A. and 2-D 348.  (-198lb). S y m b o l i c R e a s o n i n g Among 3-D Models Images, A r t i f i c i a l I n t e l l i g e n c e 1 7 ( 1 - 3 ) , 285-  [26] B r o o k s , R.A., G r e i n e r , R., and B i n f o r d , T.O. (1979), The ACRONYM M o d e l - B a s e d V i s i o n System, P r o c . S i x t h I n t e r n a tional Joint Conference on Art i f ic i a l Intelligence, T o k y o , 105-113. [27] Brown, J . S . and B u r t o n , R.R. (1975), Multiple tions of Knowledge f o r T u t o r i a l R e a s o n i n g , t i o n and Understanding, D.G. Bobrow & Teds.), A c a d e m i c P r e s s , New Y o r k , 311-349.  RepresentaRepresentaA. Collins  [28] Browse, R.A. (1982), Interpretation-Based Interaction Between L e v e l s of Detail, P r o c . F o u r t h C o n f . of t h e C a n a d i a n Soc i e t y f o r C o m p u t a t i o n a l S t u d i e s o f Intelliq e n c e , S a s k a t o o n , Canada, 27-32. [29] C l o w e s , M.B. ( 1 9 7 1 ) , On S e e i n g T h i n g s , ' A r t i f i c i a l gence 2 ( 1 ) , 79-112.  Intelli-  [30] C o l l i n s , A.M. and Loftus, E.F.A. (1975), A SpreadingA c t i v a t i o n T h e o r y of S e m a n t i c P r o c e s s i n g , P s y c h o l o g i c a l Review 8 2 ( 6 ) , 407-428. [31] D a v i s , L.S. ( 1 9 7 5 ) , A S u r v e y of Edge D e t e c t i o n Techniques, Computer G r a p h i c s and Image P r o c e s s i n g 4 ( 3 ) , 248-270. [32] D a v i s , L.S. and R o s e n f e l d , A. ( 1 9 8 1 ) , C o o p e r a t i n g P r o c e s s e s for Low-level V i s i o n : A Survey, A r t i f i c i a l I n t e l l i g e n c e 1 7 ( 1 - 3 ) , 245-263. [33] D a v i s , R. ( 1 9 8 0 a ) , M e t a - R u l e s : R e a s o n i n g t i f i c i a l I n t e l l i g e n c e 15, 179-222.  about  [34] D a v i s , R. ( 1 9 8 0 b ) , R e p o r t on t h e Workshop A l , SIGART N e w s l e t t e r 73, 42-52. [35] D a v i s , R. Do We  on  Control,  Ar-  Distributed  ( 1 9 8 2 ) , E x p e r t S y s t e m s : Where A r e We? And Go From H e r e ? , AI_ M a g a z i n e 3 ( 2 ) , 3-22.  Where  [36] Duda, R.O. and Hart, P.E. (1972), Use of the Hough Transformation t o D e t e c t L i n e s and C u r v e s i n P i c t u r e s , C o m m u n i c a t i o n s of t h e ACM 1 5 ( 1 ) , 11-15.  References  221  [37] D u d a n i , S.A. and Luk, A.L. ( 1 9 7 7 ) , Locating Straight-Line Edge Segments on O u t d o o r S c e n e s , P r o c . P a t t e r n R e c o g n i t i o n and Image P r o c e s s i n g C o n f e r e n c e , 367-377. [38] E i g e n , D.J., Fromm, F.R., and N o r t h o u s e , R.A. (1974), C l u s ter A n a l y s i s B a s e d on D i m e n s i o n a l I n f o r m a t i o n w i t h App l i c a t i o n s to Feature Selection and Classification, S y s t e m s , Man, and C y b e r n e t i c s SMC-4(3), 284-294. [39] Erman, L.D. and L e s s e r , V.R. (1975), A M u l t i - L e v e l Organization for Problem Solving Using Many, D i v e r s e , C o o p e r a t i n g S o u r c e s o f Knowledge, P r o c . F o u r t h I n t e r n a tional Joint Conference on A r t i f i c i a l Intelligence, T b i l i s s i , 483-490. [40] Erman, L.D., H a y e s - R o t h , F., L e s s e r , V.R., and Reddy, D.R. ( 1 9 8 0 ) , The HEARSAY-11 S p e e c h U n d e r s t a n d i n g S y s t e m : I n t e g r a t i n g Knowledge To R e s o l v e U n c e r t a i n t y , Computing S u r v e y s 1 2 ( 2 ) , 213-253. [41] F a l k , G. ( 1 9 7 2 ) , Dimensional 1 44.  I n t e r p r e t a t i o n of L i n e Data as Scene, A r t i f i c i a l Intelligence  a Three3 ( 2 ) , 101-  [42] F e i g e n b a u m , E. A. (1977), Themes and Case S t u d i e s of Knowledge Engineering, Proc. F i f t h International Joint Conference on A r t i fic i a l Intelligence, Cambridge, Mass., 1014-1029. [43] F i s c h l e r , M.A. and E l s c h l a g e r , R.A. tion and M a t c h i n g of P i c t o r i a l On Computers C - 2 2 ( 1 ) , 67-92.  ( 1 9 7 3 ) , The Structures,  RepresentaIEEE T r a n s .  [44] F i s c h l e r , M.A., Tenenbaum, J.M.,- and Wolf, H.C. (1981), Detection of Roads and Linear Structures in LowR e s o l u t i o n A e r i a l Imagery U s i n g a M u l t i s o u r c e Knowledge I n t e g r a t i o n T e c h n i q u e , Computer G r a p h i c s and Image P r o c e s s i n g 1 5 ( 3 ) , 201-223. [45] F l i n c h b a u g h , B.E. and C h a n d r a s e k a r a n , B. ( 1 9 8 1 ) , of Spatio-Temporal Aggregation for V i s i o n , I n t e l l i g e n c e 1 7 ( 1 - 3 ) , 387-407.  A Theory Artificial  [46] F o w l e r , R . J . and L i t t l e , J . J . ( 1 9 7 9 ) , A u t o m a t i c Extraction of I r r e g u l a r Network D i g i t a l T e r r a i n M o d e l s , P r o c . S I GRAPH, C h i c a g o , 199-207. [47] F r e u d e r , E.C. ( 1 9 7 6 ) , A Computer System f o r V i s u a l R e c o g n i t i o n U s i n g A c t i v e Knowledge, AI-TR-345, AI Lab, MIT.  References  222  [48] G a r v e y , T.D., L o w r a n c e , J.D., and F i s c h l e r , M.A. ( 1 9 8 1 ) , An Inference Technique for Integrating Knowledge From Disparate Sources, Proc. Seventh International Joint Conference on A r t i f i c i a l I n t e l l i g e n c e , V a n c o u v e r , 319325. [49] G e l e r t n e r , H. ( 1 9 6 3 ) , Realization of a Geometry-Theorem P r o v i n g M a c h i n e , Computers and T h o u g h t , E . F e i g e n b a u m & J . Feldman ( e d s . ) , M c g r a w - H i l l , New Y o r k , 134-152. [50] G l i c k s m a n , J . ( 1 9 8 2 ) , A S c h e m a t a - B a s e d System f o r U t i l i z i n g C o o p e r a t i n g Knowledge S o u r c e s i n Computer V i s i o n , P r o c . F o u r t h C o n f . o f t h e C a n a d i a n Soc i e t y f o r C o m p u t a t i o n a l S t u d i e s o f I n t e l l i g e n c e , S a s k a t o o n , Canada, 33-40. [51] G o e b e l , R. ( 1 9 7 7 ) , O r g a n i z i n g F a c t u a l Knowledge i n a Semantic Network, TR77-8, D e p t . o f Computer S c i e n c e , U. o f Alberta. [52] G r i m s o n , W.E.L. ( 1 9 8 1 ) , From Images t o S u r f a c e s : A Computational Study o f t h e Human E a r l y V i s u a l S y s t e m , MIT P r e s s , C a m b r i d g e , Mass. [53] H a l l , E . L . ( 1 9 7 9 ) , Computer Image P r o c e s s i n g t i o n , Academic P r e s s , New Y o r k .  and  Recogni-  [54] Hanson, A.R. a n d Riseman, E.M. ( 1 9 7 8 a ) , S e g m e n t a t i o n o f Natural Scenes, Computer V i s i o n S y s t e m s , A-.R. Hanson & E.M. Riseman ( e d s . ) , Academic P r e s s , New Y o r k , 129-164. [55] Hanson, A.R. and Riseman, E.M. ( 1 9 7 8 b ) , VISIONS: A Computer System for Interpreting S c e n e s , Computer V i s i o n S y s tems , A.R. Hanson & E.M. R i s e m a n (eds.), Academic P r e s s , New Y o r k , 303-334. [56] Havens, W.S. ( 1 9 7 8 ) , A P r o c e d u r a l M o d e l o f R e c o g n i t i o n f o r M a c h i n e P e r c e p t i o n , TR-78-3, D e p t . o f Computer S c i e n c e , U.B.C. [57] Havens, W.S. and Mackworth, A.K. (1980), Schemata-Based Understanding o f Hand-Drawn S k e t c h Maps, P r o c . T h i r d C o n f . o f t h e C a n a d i a n Soc i e t y f o r Computat i o n a l S t u d i e s of I n t e l l i g e n c e , V i c t o r i a , Canada, 172-178. [58] H a y e s , P., B a l l , Man-Machine 30.  E . , a n d Reddy, R. (1981), Breaking the C o m m u n i c a t i o n B a r r i e r , Computer 1 4 ( 3 ) , 19-  [59] H a y e s , P. a n d Reddy, R. (1979), Graceful Interaction in Man-Machine Communication, P r o c . Sixth International J o i n t Conference on A r i t f i c i a l Intelligence, Tokyo, 372-374.  References  223 [60] H i l d r e t h , E.C. ( 1 9 8 0 ) , I m p l e m e n t a t i o n o f a T h e o r y D e t e c t i o n , AI-TR-579, AI L a b , MIT.  of  Edge  [61] H e w i t t , C , B i s h o p , P., and S t e i g e r , R. ( 1 9 7 3 ) , A U n i v e r s a l Modular ACTOR Formalism for A r t i f i c i a l Intelligence, Proc. T h i r d I n t e r n a t i o n a l J o i n t Conference on Artific i a l I n t e l l i g e n c e , S t a n f o r d , 235-245. [62] H o r n , B.K.P. ( 1 9 7 5 ) , O b t a i n i n g Shape From S h a d i n g Information, The P s y c h o l o g y o f Computer V i s i o n , P.H. W i n s t o n ( e d . ) , M c G r a w - H i l l , New Y o r k , 115-155. [63] H o r n , B.K.P. and Bachman, B.L. ( 1 9 7 8 ) , U s i n g S y n t h e t i c ages t o R e g i s t e r R e a l Images w i t h S u r f a c e M o d e l s , m u n i c a t i o n s o f t h e ACM 2 1 ( 1 1 ) , 914-924.  ImCom-  [64] H u f f m a n , D.A. ( 1 9 7 1 ) , I m p o s s i b l e O b j e c t s a s Nonsense Sentences, M a c h i n e I n t e l l i g e n c e 6, B. M e l t z e r & D. M i c h i e • ( e d s . ) , A m e r i c a n E l s e v i e r , New Y o r k , 295-323. [65] J a i n , R. and Haynes, S. ( 1 9 8 2 ) , I m p r e c i s i o n i n Computer V i s i o n , Computer 1 5 ( 8 ) , 39-48. [66] Kanade, T. ( 1 9 8 0 ) , R e g i o n S e g m e n t a t i o n : S i g n a l v s . Semantics, Computer G r a p h i c s and Image P r o c e s s i n g 13, 279297. [67] Kanade, T. ( 1 9 8 1 ) , R e c o v e r y o f t h e T h r e e - D i m e n s i o n a l Shape of an O b j e c t from a S i n g l e View, A r t i f i c i a l Intelligence 1 7 ( 1 - 3 ) , 409-460. [68] K e l l y , M.D. ( 1 9 7 1 ) , Edge D e t e c t i o n i n P i c t u r e s by Computer U s i n g P l a n n i n g , M a c h i n e I n t e l l i g e n c e 6, B. M e l t z e r & D. M i c h i e ( e d s . ) , A m e r i c a n E l s e v i e r , New Y o r k , 397-409. [69] R e n d e r , J.R. ( 1 9 8 0 ) , Shape from Texture, D e p t . o f Computer S c i e n c e , CMU.  CMU-CS-81 -102,  [70] Koomen, J.A.G.M. ( 1 9 8 0 ) , The I n t e r l i s p V i r t u a l Machine: A Study of i t s D e s i g n and i t s I m p l e m e n t a t i o n a s M u l t i l i s p , M.Sc. d i s s e r t a t i o n , Dept. o f Computer Science, U.B.C. [71] Kuhn, T.S. ( 1 9 7 0 ) , The S t r u c t u r e o f S c i e n t i f i c R e v o l u t i o n s , e d i t i o n 2, U n i v e r s i t y o f C h i c a g o P r e s s , C h i c a g o . [72] K u i p e r s , B . J . ( 1 9 7 7 ) , R e p r e s e n t i n g Knowledge o f L a r g e - S c a l e Space, AI-TR-418, AI L a b , MIT . [73] L e b o u c h e r , G. and L o w i t z , G.E. ( 1 9 7 8 ) , What a H i s t o g r a m Can Really T e l l the C l a s s i f i e r , P a t t e r n R e c o g n i t i o n 10(4), 351-357.  References  224 [74] L e v e s q u e , H. and M y l o p o u l o s , J . ( 1 9 7 9 ) , A P r o c e d u r a l Semantics f o r S e m a n t i c N e t w o r k s , A s s o c i a t i v e N e t w o r k s : The R e p r e s e n t a t i o n a n d Use o f Knowledge by Computer, N.V. F i n d l e r ( e d . ) , A c a d e m i c P r e s s , New Y o r k , 93-120. [75] L e v i n e , M.D. a n d Shaheen, S . I . ( 1 9 8 1 ) , A M o d u l a r Computer Vision System f o r P i c t u r e S e g m e n t a t i o n a n d I n t e r p r e t a t i o n , P a t t e r n A n a l y s i s and M a c h i n e Intelligence PAMI3 ( 5 ) , 540-556. [76] L i l l e s a n d , T.M. a n d K i e f e r , R.W. ( 1 9 7 9 ) , Remote S e n s i n g and Image I n t e r p r e t a t i o n , John W i l e y a n d Sons, New Y o r k . [77] L i t t l e , J . J . ( 1 9 8 2 ) , A u t o m a t i c R e g i s t r a t i o n o f L a n d s a t MSS Images t o D i g i t a l E l e v a t i o n M o d e l s , P r o c . Workshop on Computer V i s i o n : R e p r e s e n t a t i o n a n d C o n t r o l , Rindge, New H a m p s h i r e , 178-184. [78] M a c k w o r t h , A.K. ( 1 9 7 6 ) , M o d e l - D r i v e n I n t e r p r e t a t i o n i n I n t e l l i g e n t V i s i o n S y s t e m s , P e r c e p t i o n 5, 349-370. [79] M a c k w o r t h , A.K. ( 1 9 7 7 a ) , C o n s i s t e n c y i n Networks t i o n s , A r t i f i c i a l I n t e l l i g e n c e 8 ( 1 ) , 99-118.  of  Rela-  [80] M a c k w o r t h , A.K. ( 1 9 7 7 b ) , How t o See a S i m p l e W o r l d : an Exegesis o f Some Computer Programs f o r Scene A n a l y s i s , M a c h i n e I n t e l l i g e n c e 8, E.W. E l c o c k & D. M i c h i e ( e d s . ) , John W i l e y , New Y o r k , 510-540. [81] M a c k w o r t h , A.K. ( 1 9 7 7 c ) , On Reading Sketch Maps, Proc. Fi fth I n t e r n a t i o n a l J o i n t C o n f e r e n c e on A r t i f i c i a l I n t e l l i g e n c e , C a m b r i d g e , Mass., 598-606. [82] M a c k w o r t h , A.K. ( 1 9 7 8 ) , Vision Research S t r a t e g y : Black M a g i c , M e t a p h o r s , Mechanisms, M i n i w o r l d s and Maps, Computer Vision Systems, A.R. Hanson & E.M. Riseman ( e d s . ) , Academic P r e s s , New Y o r k , 53-61. [83] M a c k w o r t h , A.K. a n d H a v e n s , W.S. ( 1 9 8 1 ) , S t r u c t u r i n g Domain Knowledge f o r V i s u a l P e r c e p t i o n , P r o c . S e v e n t h I n t e r n a tional Joint Conference on A r t i f i c i a l Intelligence, V a n c o u v e r , 625-627. [84] M a r r , D. ( 1 9 7 6 ) , E a r l y P r o c e s s i n g o f V i s u a l Information, Phil. Trans. Royal S o c i e t y o f London 2 7 5 B ( 9 4 2 ) , 483524. [85] M a r r , D. ( 1 9 8 2 ) , V i s i o n : A C o m p u t a t i o n a l I n v e s t i g a t i o n i n t o the Human R e p r e s e n t a t i o n and P r o c e s s i n g o f V i s u a l I n f o r m a t i o n , W.H. Freeman, San F r a n c i s c o . [86] M a r r , D. a n d H i l d r e t h , E.C. ( 1 9 8 0 ) , T h e o r y o f Edge Detect i o n , P r o c . R o y a l S o c i e t y o f London B ( 2 0 7 ) , 187-217.  References  225 [87] M a r r , D. and N i s h i h a r a , H.K. (1978), Representation and R e c o g n i t i o n of t h e S p a t i a l O r g a n i z a t i o n of T h r e e Dimens i o n a l S t r u c t u r e , P r o c . R o y a l Soc i e t y of London B ( 2 0 0 ) , 269-294. [88] M i n s k y , M. ( 1 9 7 5 ) , A Framework f o r R e p r e s e n t i n g Knowledge, The P s y c h o l o g y o f Computer V i s i o n , P.H. W i n s t o n (ed.), M c G r a w - H i l l , New Y o r k , 211-277. [ 8 9 ] M u l d e r , J.A. ( 1 9 7 9 ) , R e p r e s e n t a t i o n and C o n t r o l i n a gram that U n d e r s t a n d s L i n e S k e t c h e s o f a House, d i s s e r t a t i o n , D e p t . o f Computer S c i e n c e , U.B.C.  ProM.Sc.  [90] Nagao, M., Matsuyama, T., and I k e d a , Y. ( 1 9 7 8 ) , R e g i o n Extraction and Shape Analysis of A e r i a l Photographs, P r o c . F o u r t h I n t e r n a t i o n a l J o i n t C o n f e r e n c e on Pattern R e c o g n i t i o n , K y o t o , 620-628. [91] Nagao, M., Matsuyama, T., and M o r i , H. (1979), Structural Analysis of Complex Aerial Photographs, Proc. S i x t h I n t e r n a t i o n a l J o i n t C o n f e r e n c e on Ar i t f i c i a l Intelligence , Tokyo, 610-616. [92] N a z i f , A.M. and L e v i n e , M.D. (1982), S e t t i n g S t r a t e g y i n a Rule Based Vision System, P r o c . F o u r t h Conf. of the C a n a d i a n Soc i e t y f o r C o m p u t a t i o n a l S t u d i e s of Intelligence , S a s k a t o o n , Canada, 52-59. [93] N e i s s e r , U. York.  (1967),  Cognitive  Psychology,  Meredith,  [94] N e i s s e r , U. ( 1 9 7 6 ) , C o g n i t i o n and R e a l i t y : Principles I m p l i c a t i o n s of C o g n i t i v e P s y c h o l o g y , W.H. Freeman, Francisco.  New and San  [95] N e v a t i a , R. and P r i c e , K.E. ( 1 9 7 8 ) , L o c a t i n g S t r u c t u r e s in A e r i a l Images, P r o c . F o u r t h I n t e r n a t i o n a l J o i n t C o n f e r ence on P a t t e r n R e c o g n i t i o n , K y o t o , 686-690. [96] N e v a t i a , R. and P r i c e , K.E. ( 1 9 8 2 ) , L o c a t i n g S t r u c t u r e s in Aerial Images, Pattern A n a l y s i s and M a c h i n e I n t e l l i gence PAMI-4(5), 476-484. [97] N i i , H.P., Feigenbaum, E.A., Anton, J . J . , and Rockmore, A.J. ( 1 9 8 2 ) , S i g n a l - t o - S y m b o l T r a n s f o r m a t i o n : HASP/SIAP C a s e S t u d y , AI_ M a g a z i n e 3 ( 2 ) , 23-35. [98] N i l s s o n , cial  N. J . (1971), P r o b l e m - S o l v i n g Methods I n t e l l i g e n c e , M c G r a w - H i l l , New York.  _n  [99] N i l s s o n , N. J . (1980), P r i n c i p l e s of Artificial gence , T i o g a , P a l o A l t o , C a l i f o r n i a .  References  Artifi-  Intelli-  226  [100] O'Gorman, F. and Clowes, M.B. (1973), Finding Picture Edges Through Collinearity of Feature Points, Proc. T h i r d I n t e r n a t i o n a l J o i n t C o n f e r e n c e on A r t i f i c i a l Int e l l i g e n c e , S t a n f o r d , 543-555. [101] O h l a n d e r , R.B. ( 1 9 7 5 ) , A n a l y s i s o f N a t u r a l Scenes, d i s s e r t a t i o n , D e p t . of Computer S c i e n c e , CMU.  Ph.D.  [102] O h t a , Y., Kanade, T., and S a k a i , T. (1978), An Analysis System f o r Scenes Containing Objects with Substructures, Proc. Fourth I n t e r n a t i o n a l J o i n t Conference on P a t t e r n R e c o g n i t i o n , K y o t o , 752-754. [103]  Palmer, S. E. (1975), Visual Perception and World Knowledge: Notes on a M o d e l of S e n s o r y - C o g n i t i v e I n t e r a c t i o n , E x p l o r a t i o n s i n Cogn i t i o n , D.A. Norman & D.E. Rumelhart (eds.), W.H. Freeman, San F r a n c i s c o , 279-307.  [104]  P o s t a i r e , J.-G. and V a s s e u r , C P . A . ( 1 9 8 1 ) , An A p p r o x i m a t e S o l u t i o n t o Normal M i x t u r e I d e n t i f i c a t i o n w i t h A p p l i c a t i o n to Unsupervised Pattern Classification, Pattern A n a l y s i s and M a c h i n e I n t e l l i g e n c e PAMI-3(2), 163-179.  [105]  Ramer, U. ( 1 9 7 2 ) , An I t e r a t i v e P r o c e d u r e f o r t h e P o l y g o n a l Approximation of P l a n a r C u r v e s , Computer G r a p h i c s and Image P r o c e s s i n g 1 ( 3 ) , 244-256.  [106]  R i c h , E. (1979a), B u i l d i n g and Exploiting User Models, Proc• Sixth I n t e r n a t i o n a l J o i n t C o n f e r e n c e on A r t i f i c i a l I n t e l l i g e n c e , Tokyo, 720-722.  [107]  R i c h , E . (1979b), B u i l d i n g and Exploiting User CMU-CS-79-119, D e p t . o f Computer S c i e n c e , CMU.  [108]  R o b e r t s , L.G. ( 1 9 6 5 ) , M a c h i n e P e r c e p t i o n o f Three Dimensional S o l i d s , O p t i c a l and E l e c t r o - o p t i c a l I n f o r m a t i o n P r o c e s s i n g , J . T . T i p p e t t , D. B e r k o w i t z , L. Clapp, C. K o e s t e r , & A. V a n d e r b u r g h ( e d s . ) , MIT P r e s s , C a m b r i d g e , Mass., 159-197.  [109]  R o b e r t s , R.B. and G o l d s t e i n , AIM 409, AI L a b , MIT.  I.P.  (1977a),  The  FRL  Manual,  [110]  R o b e r t s , R.B. and G o l d s t e i n , AIM 408, AI Lab, MIT.  I.P.  (1977b),  The  FRL  Primer,  [111]  Robson, D. and G o l d b e r g , A. tem, B y t e 6 ( 8 ) , 36-47.  [112]  Rosenberg, 431, AI  References  S.T. Lab,  ( 1 9 8 1 ) , The  ( 1 9 7 7 ) , Frame-Based T e x t MIT.  Models,  Smalltalk-80  Processing,  Sys-  AIM  227 [113]  R o s e n t h a l , D. and B a j c s y , R. ( 1 9 7 8 ) , C o n c e p t u a l and V i s u a l Focussing in the Recognition Process a s I n d u c e d by Q u e r i e s , P r o c . F o u r t h I n t e r n a t i o n a l J o i n t C o n f e r e n c e on P a t t e r n R e c o g n i t i o n , K y o t o , 417-420.  [114] R u s s e l l , D.M. ( 1 9 7 9 ) , Where Do I Look Now?: M o d e l l i n g and Inferring Object L o c a t i o n s by C o n s t r a i n t s , P r o c . IEEE Comp. Soc i e t y C o n f . on P a t t e r n R e c o g n i t i o n and Image P r o c e s s i n g , C h i c a g o , 175-181. [115]  S a k a i , T. , Kanade, T., and O h t a , Y. Interpretation of O u t d o o r S c e n e s , t i o n a l J o i n t C o n f e r e n c e on P a t t e r n d a , C a l i f o r n i a , 581-585.  [116]  S c h a c t e r , B . J . , D a v i s , L.S., and Rosenfeld, A. (1976), Scene Segmentation by Cluster Detection in Color S p a c e s , SIGART N e w s l e t t e r 58, 16-17.  [117]  S c o t t , D.W. ( 1 9 7 9 ) , On O p t i m a l B i o m e t r i k a 6 6 ( 3 ) , 605-610.  [118]  S e l f r i d g e , P.G. and S l o a n , K.R. (1981), Reasoning About Images: Using Meta-Knowledge i n A e r i a l Image U n d e r s t a n d i n g , Proc. Seventh I n t e r n a t i o n a l J o i n t Conference on A r t i f i c i a l I n t e l l i g e n c e , V a n c o u v e r , 755-757.  [119]  Shirai, Y. (1975), Analyzing Intensity Arrays Using Knowledge About S c e n e s , The P s y c h o l o g y o f Computer V i s i o n , P.H. W i n s t o n ( e d . ) , M c G r a w - H i l l , New York, 93113.  [120]  S h o r t 1 i f f e , E.H. and Inexact Reasoning 23, 351-379.  [121]  S l o a n , K.R. ( 1 9 8 2 ) , A n a l y s i s o f "Dot P r o d u c t S p a c e " Shape Description, P a t t e r n A n a l y s i s and M a c h i n e I n t e l l i g e n c e P A M I - 4 ( 1 ) , 87-90.  [122]  S l o a n , K.R. and B a j c s y , R. (1977), World Model Driven Recognition of N a t u r a l S c e n e s , D e p t . o f Computer and Information S c i e n c e , Moore School of E.E., U. of Pennsylvania, Philadelphia.  [123]  S m i t h , R.G. and F r i e d l a n d , P. ( 1 9 8 0 ) , UNIT P a c k a g e G u i d e , T e c h n i c a l Memorandum 80/L, DREA.  [124]  S t e f i k , M.J. ( 1 9 7 9 ) , An E x a m i n a t i o n of a Frame-Structured Representation System, P r o c . S i x t h I n t e r n a t i o n a l J o i n t C o n f e r e n c e on A r t i f i c i a l I n t e l l i g e n c e , T o k y o , 845-852.  References  and  (1976), Model-Based Proc. T h i r d InternaR e c o g n i t i o n , Corona-  Data-based  Histograms,  Buchanan, B.G. (1975), A Model of in Medicine, Mathematical B i o s c i e n c e s  User's  228 [125]  S t e f i k , M.J. ( 1 9 8 0 ) , P l a n n i n g w i t h Constraints, STAN-CS80-784, D e p a r t m e n t of Computer S c i e n c e , S t a n f o r d .  [126]  S t o c h , L. ( 1 9 8 1 ) , I n t e r a c t i v e P o l y g o n F i l l i n g on a Raster Graphic D i s p l a y , M.Sc. d i s s e r t a t i o n , D e p t . o f Computer S c i e n c e , U.B.C.  [127]  Stockman, G. ( 1 9 7 8 ) , Toward A u t o m a t i c E x t r a c t i o n of C a r t o graphic F e a t u r e s , ETL-0153, L.N.K. C o r p o r a t i o n , S i l v e r Spring, Maryland.  [128]  T a v a k o l i , M. and R o s e n f e l d , A. ( 1 9 8 2 ) , B u i l d i n g and Road Extraction from A e r i a l P h o t o g r a p h s , S y s t e m s , Man,. and C y b e r n e t i c s SMC-120 ), 84-91.  [129]  Tenenbaum, J.M. ( 1 9 7 3 ) , On L o c a t i n g O b j e c t s by T h e i r Distinguishing F e a t u r e s i n M u l t i s e n s o r y Images, TN 84, AI C e n t e r , SRI.  [130]  Tenenbaum, J.M. and Barrow, H.G. (1977), Experiments in I n t e r p r e t a t i o n - G u i d e d Segmentation, A r t i f i c i a l I n t e l l i gence 8 ( 3 ) , 241-274.  [131]  Tenenbaum, J.M., F i s c h l e r , M.A., and W o l f , H.C. (1978), Scene-Analysis Approach t o Remote S e n s i n g , TN 173, C e n t e r , SRI.  [132]  T h o r n d y k e , P.W. ( 1 9 7 9 ) , H e u r i s t i c s f o r Knowledge Acquisit i o n from Maps, P r o c . S i x t h I n t e r n a t i o n a l J o i n t C o n f e r ence on A r t i f i c i a l I n t e l l i g e n c e , T o k y o , 880-883.  [133]  T o m i t a , F. ( 1 9 8 1 ) , H i e r a r c h i c a l D e s c r i p t i o n of Textures, P r o c . S e v e n t h I n t e r n a t i o n a l J o i n t C o n f e r e n c e on A r t i f i c i a l I n t e l l i g e n c e , V a n c o u v e r , 728-733.  [134]  U l l m a n , S. ( 1 9 7 8 ) , The I n t e r p r e t a t i o n MIT P r e s s , C a m b r i d g e , Mass.  of  Visual  A AI  Motion,  [135] W e s l e y , L.P. and Hanson, A.R. (1982), The Use of an Evidential-Based Model f o r R e p r e s e n t i n g Knowledge and R e a s o n i n g About Images i n the VISIONS System, Proc. Workshop on Computer V i s i o n : R e p r e s e n t a t i o n and Cont r o l , R i n d g e , New H a m p s h i r e , 14-25. ['136] Weymouth, T.E. ( 1 9 8 1 ) , E x p e r i m e n t s i n K n o w l e d g e - D r i v e n Interpretation of N a t u r a l Scenes, P r o c . Seventh I n t e r n a tional Joint Conference on A r t i f ic i a l Intelligence, V a n c o u v e r , 628-630. [137] W i n s t o n , P.H. ( 1 9 7 2 ) , The MIT Robot, M a c h i n e Intelligence 7, B. Meltzer & D. M i c h i e (eds.TI A m e r i c a n E l s e v i e r , New Y o r k , 431-463.  References  229 [ 138]  W i n s t o n , P.H. ( 1 9 7 7 ) , Artificial W e s l e y , R e a d i n g , Mass..  Intelligence,  Addison-  [139] W i t k i n , A.P. ( 1 9 8 1 ) , R e c o v e r i n g S u r f a c e Shape and O r i e n t a tion from Texture, Arti f i c i a l Intelligence 17(1-3), 17-45 . [140] Woodham, R . J . ( 1 9 8 0 a ) , P h o t o m e t r i c Method f o r D e t e r m i n i n g Surface Orientation from M u l t i p l e Images, O p t i c a l E n g i n e e r i n g 19, 139-144. [141] Woodham, R . J . ( 1 9 8 0 b ) , U s i n g D i g i t a l T e r r a i n Data t o Model Image Formation i n Remote S e n s i n g , P r o c . Soc i e t y o f P h o t o - O p t i c a l I n s t r u m e n t a t i o n E n g i n e e r s 238, 361-369. [142] Woodham, R . J . ( 1 9 8 1 ) , A n a l y z i n g Images o f C u r v e d S u r f a c e s , A r t i f i c i a l Intelligence 1 7 ( 1 - 3 ) , 117-140. [143]  Y a c h i d a , M., I k e d a , M., a n d T s u j i , S. ( 1 9 7 9 ) , A Knowledge Directed Line Finder f o r A n a l y s i s o f Complex S c e n e s , Proc. Sixth Internat i o n a l J o i n t Conference on Artific i a l I n t e l l i g e n c e , T o k y o , 984-991.  [144]  Y a k i m o v s k y , Y. a n d F e l d m a n , J.A. ( 1 9 7 3 ) , A S e m a n t i c s - B a s e d Decision Theory Region A n a l y z e r , Proc. T h i r d I n t e r n a tional Joint Conference on A r t i f i c i a l Intelligence, S t a n f o r d , 580-588.  [145]  Z u c k e r , S.W., R o s e n f e l d , A., and D a v i s , L . S . ( 1 9 7 5 ) , Gene r a l Purpose Models: E x p e c t a t i o n s about t h e Unexpected, P r o c . F o u r t h I n t e r n a t i o n a l J o i n t C o n f e r e n c e on Artific i a l I n t e l l i g e n c e , T b i l i s i , 716-721.  References  APPENDIX  A  The MAIDS System  Schemata  Schemata a r e s t o r e d is  the  name  on t h e p r o p e r t y l i s t  o f t h e atom whose pname  of t h e o b j e c t o r i n s t a n c e . T h e r e  a r e two t y p e s of  schemata: OBJECTS — s t e r e o t y p i c  or g e n e r a l knowledge:  They have % o b j e c t % as t h e f i r s t  p r o p e r t y on t h e i r  property  list. INSTANCES—particular  knowledge c r e a t e d as p a r t  of  interpretation: They a r e o f t h e form are  incremented  %instance%  The F o u r VALUEd  LINK  from  objectname-number  where  the  numbers  1 a s i n s t a n c e s a r e c r e a t e d . They have  as the f i r s t  p r o p e r t y on t h e p r o p e r t y  list.  P a r t s of a Schemata  type  type  CONFIDENCE  = attribute-name  = link-name  (%VALUE S - e x p r e s s e i o n ((%CONFIDENCE . number) (%DEFAULT . S-expression) (%IF-ADDED . form) (%IF-MODIFIED . form) (%IF-REMOVED . form) (%REQUIRED . predicate)))  {schema  | (schemal  schema2  ...)}  = CONFIDENCE number 230  231 CONF-ALG PROCEDURES  ->  form  p r o c e d u r e s t o be messages a r e  Global The  and  their =  inverses.  ((ako (apo (aio . .  %schemata: a l i s t  %instances:  records  i s of  the  a l l the  known  relation-  form  . specializes-to)(specializes-to. . decomposes-to) (decomposes-to . . instances) (instances . .)  of  all  the  i n the  a list  a l l the  of  loaded  generic  loaded  ako) apo) aio)  (stereotypic)  schemata  system  p a r t i c u l a r instances  available  in  system  Variables  the  procedures attached  VALUEd  types  REQUIRED), the variable  t o CONFIDENCE v a l u e s  (IF-ADDED,  IF-MODIFIED,  name of  current  the  (CONF-ALG)  IF-REMOVED,  schema c a n  be  and  IF-NEEDED,  found  in  the  %name.  Furthermore, has  It  available  the  In  appropriate  received  variable %link_types  %link_types  Local  when  Variables  global  ships  evaluated  just  i n VALUEd  type a t t a c h e d  been a d d e d , m o d i f i e d ,  etc.  procedures, is located  the  value  i n the  that  variable  %val.  Convent ions  In  the  following  Appendix  A  d e s c r i p t i o n of  functions,  all  arguments  are  232  evaluated enclosed  unless  explicitly  in "[]"'s,  and t h e  noted,  optional  arguments  following  suffixes  are  are  generally  used: A V L I 0  any A t t r i b u t e VALUEd t y p e a t t r i b u t e s LINK a t t r i b u t e s Instances Objects  Schemata M a n i p u l a t i o n  1.(add-link-type Returns: Side  Funct ions  link)  link  effect:  Adds a l i n k and i t s inverse to the g l o b a l variable % l i n k _ t y p e s : a l i s t o f a l l t h e known link types. The u s e r w i l l be q u e r i e d f o r t h e name of the i n v e r s e but i t i s not necessary f o r one t o be s u p p l i e d .  2.(add-link-inverse  link  inverse)  Returns: n i l Side  effect:  3.(ifneed  schema a t t r i b u t e )  Returns:  4.(saddl  Makes i n v e r s e t h e i n v e r s e of l i n k i n t h e glob a l v a r i a b l e % l i n k _ t y p e s . I f l i n k and i n v e r s e do n o t e x i s t i n % l i n k _ t y p e s , they are both added.  the r e s u l t of e v a l i n g the %if-needed a t t r i b u t e (a VALUEd t y p e ) i n schema.  schemal  Returns: Side  Appendix A  link  clause  of  schema2)  schema2  effects:  Adds a LINK from schemal t o schema2. In addition, the inverse of t h e LINK i s added from schema2 t o s c h e m a l . L i n k must be known t o %link_types.  233  5.(sanylink?  schemal  Returns:  6.(sattr  I f there i s a path between schemal and schema2 along any t y p e o f LINK i t r e t u r n s a l i s t o f a l l the l i n k types used i n the path (of t h e form (linkl link2 link3 ...)). I f t h e r e i s no p a t h , i t returns n i l .  schema  Where:  schema2)  type)  Type i s e i t h e r n i l ( r e t u r n a l l a t t r i b u t e s ) o r a l i s t of one o r more or t h e f o l l o w i n g types t o be returned: valued, l i n k , procedure, or other.  Returns: A l i s t type. 7.(sattrtype  of the a t t r i b u t e s of  schema  depending  on  schema a t t r i b u t e )  Returns:  t h e t y p e o f a t t r i b u t e i n schema. Type w i l l of v a l u e d , l i n k , p r o c e d u r e , o r o t h e r .  be one  8.(sconsist) Returns: n i l Side  effect:  9.(screate  Makes a d a t a b a s e consistent by a d d i n g a n d removing a p p r o p r i a t e LINKS a n d o b j e c t s . A f t e r i t i s f i n i s h e d , a l l LINK t y p e s w i l l point to existing o b j e c t s and t h e i r i n v e r s e s w i l l a l s o be i n p l a c e .  [name])  Where:  name i s t h e u n e v a l u a t e d name o f t h e schema t h a t w i l l be c r e a t e d . I f name i s n o t p r o v i d e d t h e u s e r w i l l be queried for i t . In a d d i t i o n , the user will be queried f o r t h e v a r i o u s p a r t s o f t h e schema and c a n i n p u t any a t t r i b u t e s t h a t a r e d e s i r e d .  R e t u r n s : The newly c r e a t e d Side  effect:  !0.(serasei  Appendix A  object.  The p r o p e r t y l i s t o f t h e named schemata will be m o d i f i e d t o i n c l u d e a l l t h e a t t r i b u t e s t h a t i t i s i n i t i a l l y endowed w i t h .  instance  [if-rem?  [splice?]])  234 Where: i f - r e m ? Returns: Side  12.(serrn  form  i s n o n - n i l or n i l .  t h e name o f t h e d e s t r o y e d  effect:  13.(sgeta  object.  Destroys the stereotype o b j e c t . Since t h i s i s not l i k e l y t o be done v e r y o f t e n , t h e u s e r i s prompted t o make s u r e he/she r e a l l y wants t o do i t . I f s p l i c e ? i s n o n - n i l , then the schemata adjacent to object have their links s p l i c e d around i t (see s p l i c e ) .  culprit)  Where: n i s a number Side  instance.  splice?)  Where: s p l i c e ?  Side  n o n - n i l or n i l .  Destroys instance as a schema (removes t h e a t t r i b u t e s from i t s p r o p e r t y l i s t ) and removes a l l t h e LINKs t o i t . I f i f - r e m ? i s non-nil, then a l l t h e e x i s t i n g i f - r e m o v e d c l a u s e s on VALUEd t y p e s a r e e v a l e d . If splice? i s nonn i l , t h e n t h e LINKs from a d j a c e n t schemata a r e s p l i c e d around i n s t a n c e (see s p l i c e ) .  object  Returns:-  are either  t h e name o f t h e d e s t r o y e d  effect:  11.(seraseo  and s p l i c e ?  effect:  from  1 t o 12.  E r r o r message n i s p r i n t e d u s i n g t h e f u n c t i o n form and v a l u e c u l p r i t . Debug i s t h e n c a l l e d on f o r m .  schema a t t r i b u t e  [inherit  [allorone  [links]]])  Where: i n h e r i t i s one o f "y" o r "n"; a l l o r o n e i s one o f " a " o r "o"; a n d l i n k s i s a l i s t of L I N K s . Returns:  Appendix A  The v a l u e o f a t t r i b u t e i n schema. I f the a t t r i bute i n schema h a s a v a l u e , t h e n i t i s r e t u r n e d . I f t h e r e i s no a t t r i b u t e i n schema, t h e n property inheritance will be u s e d ( i f inherit = y as o p p o s e d t o n ) . I n h e r i t a n c e w i l l be s o u g h t up t h e LINKs f o u n d i n t h e l i s t " l i n k s " . D e p e n d i n g on t h e v a l u e o f a l l o r o n e , e i t h e r the f i r s t v a l u e (from a b r e a d t h - f i r s t s e a r c h ) w i l l be r e t u r n e d ( a l l o r o n e = o) and t h e g l o b a l v a r i a b l e % w h e r e f r o m w i l l be s e t to t h e schema, where t h e a t t r i b u t e was f o u n d , o r a l l t h e v a l u e s and t h e i r schemata w i l l be r e t u r n e d in a list e,g. ( ( v a i l . s c h l ) ( v a l 2 . sch2)) ( i f  235  allorone Note:  14.(sgetc  The default links=(ako). schema  Returns:  I5.(sgetl  values  inherit=n,  allorone=o,  [attribute])  link)  t h e schema link.  schema a t t r i b u t e  Where:  are  t h e c o n f i d e n c e v a l u e of a schema ( i f a t t r i b u t e nil or " c o n f i d e n c e " ) o r of a VALUEd t y p e a t t r i bute .  schema  Returns:  16.(sgetv  = a).  ( o r schemata) " l i n k e d "  inherit  type  t o schema  allorone linksl  links2)  i n h e r i t i s one of "y" o r "n"; t y p e i s one of "v", "vd", or " d " ; a l l o r o n e i s one of " a " o r "o"; and l i n k s l and l i n k s 2 a r e l i s t s of LINKS.  Returns:  The v a l u e ( o r d e f a u l t ) of a t t r i b u t e (which must be a VALUEd type) i n schema. I f t h e r e i s no a t t r i b u t e i n schema, then property inheritance will be used ( i f i n h e r i t = y as opposed t o n ) . I n h e r i t a n c e w i l l be s o u g h t for either a value only (type = v, using linksl), a v a l u e or a d e f a u l t a t every step (type = vd, u s i n g linksl), or t h e whole s t r u c t u r e w i l l be s e a r c h e d f o r a v a l u e ( u s i n g l i n k s l ) and i f none i s found the whole structure will again be searched f o r a d e f a u l t ( u s i n g l i n k s 2 ) (type = d ) . Depending on the value of a l l o r o n e , e i t h e r the f i r s t value (from a b r e a d t h - f i r s t s e a r c h ) will be returned ( a l l o r o n e = o) and t h e g l o b a l v a r i a b l e % w h e r e f r o m w i l l be s e t t o t h e schema where t h e a t t r i b u t e was found, o r a l l t h e v a l u e s and t h e i r s c h e m a t a w i l l be r e t u r n e d i n a l i s t e . g . ( ( v a i l . s c h l ) ( v a l 2 sch2)) ( i f allorone = a). If a d e f a u l t as opposed t o a v a l u e i s r e t u r n e d , then t h e above description holds f o r d e f a u l t s plus the global variable %default i s set to t .  Note: d e f a u l t s a r e i n h e r i t = y, type = vd, a l l o r o n e l i n k s l = ( a k o ) , and l i n k s 2 = ( a k o ) .  17.(slink?  Appendix A  via  schemal  schema2  link)  =  o,  236  Returns:  I f t h e r e i s a p a t h f r o m schemal t o schema2 along link, then t h e minimum l e n g t h of t h e p a t h , e l s e nil. I f l i n k i s n i l , then sanylink? is called w h i c h l o o k s f o r a p a t h u s i n g any t y p e of l i n k .  I 8 . ( s m e r g e schemal  schema2  doconfalgorithm?)  Where: d o c o n f a l g o r i t h m ? Returns: Side  19.(snewi  effect:  20.(splice  p r i o r i t y queue. Combines t h e s l o t s of schema2 i n t o those of schemal, i . e . , f o r VALUEd types, if the a t t r i b u t e i n schema2 has a n o n - n i l v a l u e and t h e one i n schemal does n o t , put t h e v a l u e i n schemal's s l o t . For LINK types, take the u n i o n of t h e p o i n t e r s . I f d o c o n f a l g o r i t h m ? i s t r u e , e v a l u a t e the CONFIDENCE algorithm on the resulting schema. Also, a l l references t o schema2 i n p r e v i o u s messages i n t h e p r i o r i t y queue a r e c h a n g e d t o s c h e m a l .  object)  Returns: Side  the  i s n o n - n i l or n i l .  the  effect:  new  instance.  Creates a new instance of the stereotype object. It will have the name o b j e c t - n , where n is the next integer in sequence ( s t a r t i n g from 1).  schema l i n k  [inverse])  Returns: n i l Side  effect:  Removes t h e LINKs from schema to the nodes a t t a c h e d by l i n k and s p l i c e s t h e around i t . Thus: up  up  inverse schema link down Appendix  A  inverse goes-> to  link down  other LINKs  237  21.(sprint  schemaname)  Returns: n i l Side  effect: Pretty prints  22. ( s p r i n t n  schema  a schema.  link)  Returns: n i l Side  23. ( s p u t a  effect: Pretty prints using l i n k . schema a t t r i b u t e  Returns: Side  24.(sputc  of  schema  [value]) "value".  e f f e c t : P u t s a new d e f i n i t i o n of a t t r i b u t e i n schema and p l a c e s " v a l u e " as i t s i n i t i a l v a l u e . schema  Where:  [attribute  [cval  c v a l i s a number, n o n - n i l or n i l .  Returns: Side  the v a l u e  t h e t r e e s t r u c t u r e under  [spread?  and s p r e a d ?  the o l d c o n f i d e n c e  [skip?]]]]) and s k i p ? a r e  either  value.  e f f e c t : Changes t h e CONFIDENCE v a l u e o f a t t r i b u t e in schema to c v a l . I f a t t r i b u t e = n i l or "confidence" the value of CONFIDENCE at the schema l e v e l i s c h a n g e d ; o t h e r w i s e t h e VALUEd type a t t r i b u t e has i t s CONFIDENCE changed. I f s p r e a d ? i s non-NIL t h e n t h e e f f e c t s of t h e change ( i f t h e r e a r e any) a r e s p r e a d by r e evaling a l l t h e c o n f - a l g o r i t h m s up t h e apo and ako h i e r a r c h i e s . I f spread? i s non-NIL and s k i p ? i s , t h e n t h e CONF-ALG i s e v a l u a t e d for schema ( i . e . the value of c v a l is ignored).  2 5 . ( s p u t c l schema a t t r i b u t e spread?) Where:  spread?  Returns:  value  i s either  oldconfidence  newconfidence  n o n - n i l or n i l .  the o l d c o n f i d e n c e  value.  S i d e e f f e c t : R e p l a c e s o l d c o n f i d e n c e i n t h e l i s t of confidence values of attribute i n schema w i t h  Appendix A  238 newconfidence. I f spread? i s non-NIL then the e f f e c t s of t h e c h a n g e ( i f t h e r e a r e any) are spread by re-evaling a l l the confa l g o r i t h m s up t h e apo and ako h i e r a r c h i e s .  26.(sputv  schema a t t r i b u t e  Returns: Side  value)  value.  e f f e c t : P u t s a new v a l u e into the %value slot of a t t r i b u t e i n schema. D e p e n d i n g on t h e p r e v i ous and new values, certain demons may respond. previous value  new value ~nil nil ~ n i l ~ n i l nil  nil nil "nil ~nil 27.(sremovea Where:  schema a t t r i b u t e if-remclause?  Returns: Side  Returns:  Returns:  Appendix A  %if-added %if-modified %if-removed  i s evaled i s evaled i s evaled  [if-remclause?])  i s either  n o n - n i l or n i l .  of t h e removed  attribute.  schema  link)  the value of the "schema" by l i n k .  schema(ta)  pointed  e f f e c t : Removes a l l t h e LINKs attaching other schemata v i a l i n k .  29.(sremovel  Side  i s checked  e f f e c t : Removes the d e f i n i t i o n of attribute from schema. I f t h e i f - r e m c l a u s e ? i s n o n - n i l and t h e r e i s a n o n - n i l v a l u e , then the i f - r e m o v e d form ( i f i t e x i s t s ) i s e v a l e d .  28.(sremoveal  Side  the value  %required  schemal  link  to schema  from to  schema2)  schema2  e f f e c t : Removes t h e l i n k from schemal t o schema2 and the i n v e r s e l i n k from schema2 t o s c h e m a l . I f schema2 i s n i l , t h e n sremoveal i s called which removes a l l t h e nodes attached to  239 schemal 30. (sremovev  schema  by  link.  attribute)  Returns: n i l . S i d e e f f e c t : Removes t h e % v a l u e from a t t r i b u t e (a VALUEd t y p e ) i n schema (by s e t t i n g i t t o n i l ) . 31. ( s r e s t o r e Where:  [file]) file  i s ,not e v a l l e d .  Returns: n i l . Side  32. ( s s a v e  e f f e c t : R e s t o r e s t h e s t e r e o t y p e schemata from if file i s n i l , then the d e f a u l t "saved.d" i s used. [file])  Where:  file  Returns: Side  file. file  i s not e v a l l e d .  file.  e f f e c t : S a v e s a l l t h e known stereotype schemata on file. I f f i l e i s n i l , then the d e f a u l t file "saved.d" i s used.  33. ( s s p r i n t n  schema)  Returns: n i l Side  34. ( v c h e c k  e f f e c t : P r e t t y p r i n t s the t r e e following a l l links. schema a t t r i b u t e  Returns:  Appendix A  structure  under  schema  value)  the r e s u l t of e v a l l i n g the % r e q u i r e d clause on v a l u e ( i f i t e x i s t s ) of a t t r i b u t e (a VALUEd t y p e ) i n schema.  APPENDIX Edge  An  edge d e t e c t o r  Hildreth dreth,  theory  Laplacian This  Detection  h a s been  implemented  o f edge d e t e c t i o n  1980).  B  of  the  Gaussian  e d g e s and edge  function  specific  information  then  scaled  largest  integer  chosen  using  1980; H i l -  with  of  the  t h e image.  details  of  how  are calculated. on  - 2a ) * exp(-(x  so t h a t  mask  256 g r e y  out  curves  2  the value  < 15 s u c h t h a t  floating  by  This  2  theory,  to produce high  The  off  + y  2  the M a r r - H i l d r e t h  out  convolved  Marr-  the  origin,  is  gen-  from t h e f u n c t i o n  m(x,y) = ( x  to  the  crossings  implementation  A mask 31 p i x e l s s q u a r e , c e n t r e d  and  on  (Marr and H i l d r e t h ,  Edges a r e l o c a t e d a t t h e z e r o  appendix d e s c r i b e s  erated  based  point  + y )/2a )  2  2  o f m ( 0 , i ) = 0.5 where i i s t h e  m(0,i) i s non-zero.  only  one a v a l u e  r e s o l u t i o n edges.  In c o n t r a s t  i s u s e d , and i t  Arithmetic  is  is carried  arithmetic.  i s convolved  with the d i g i t i z e d  scale values).  Zero c r o s s i n g s  i n the convolved  produces  2  continuous  images  a r e f o u n d by t r a c i n g  image where t h e v a l u e s curves that  (128 by 128  change  signs.  are either closed  or run  t h e edge o f t h e image.  240  241 Since cally line  these  long  distinct  curves  objects,  line  1982).  i s drawn  the  curve  the  line  pixel  two new the  furthest  used) then  distance  threshold, value.  of  This  that  produces  a  1972;  process wherein a  I f the d i s t a n c e  i s subdivided in turn.  a l l the p o i n t s  guaranteeing  (Ramer,  than a t h r e s h o l d  segments a r e g e n e r a l i z e d  straight  of a c u r v e and t h e p o i n t  i s found.  i s greater  semanti-  into shorter  i s a recursive  the endpoints  the l i n e  several  generalization  from t h e l i n e  to the point  was  called  Generalization  between  include  they are broken  segments by a method  Little,  generally  from  (a v a l u e  at the p o i n t Recursion  the l i n e  the d e v i a t i o n  on from  of  1  and t h e  stops  when  i s l e s s than the  is  less  number o f c o n n e c t e d ,  than  that  " s t r a i g h t " seg-  ments.  From to  the  that  this  original  there  fying ture  contains  value  segment  segment  used  segment  Marr and H i l d r e t h ) . in  the  image.  Appendix  intensity  the  either  image  of.  (from  the region  This  t o the and  derivative  of the t h i r d  correspond  the d i f f e r e n c e  identi-  A separate data  0 t o 360 d e g r e e s  directional  that  0, meaning  p o i n t , , o r a number  t h e one on t h e l e f t ) ,  slope  corresponds  a s was  struc-  includes since  the  right  of  contrast.  across  the  suggested  I t i s c a l c u l a t e d by moving a l o n g  For each p o i n t ,  B  that  f o r e a c h edge segment.  than  i s the f i r s t  contains  at that  to i n d i c a t e that  (not  i s returned  Each p i x e l  orientation  i s brighter  Contrast  image  i t i s a part  information  length,  i s also  edge  image.  i s no edge segment  t h e edge  location,  the  process a l i n e  the  by  points  t o those of the  line  i s c a l c u l a t e d between t h e  242 pair  of  dicular  pixel to the  values  values  strength)  of  the  of  of)  the  level  functions.  B  all  the  away from t h e  the  segment.  p o i n t and  The  differences  average  i s the  perpenof  contrast  the (or  segment.  information  (one  Appendix  pixel  o r i e n t a t i o n of  absolute  This  one  forms a t y p e  intermediate  of  raw  primal  sketch.  representations accessed  by  It  is  higher  APPENDIX Region  Regions mentation general ity  of  pixels  i n the in  sity  images  This  that  applied  have been  regions  used can  only was  (a  region. values  on  a  significant  of a r e g i o n  regions  affin-  units  be  intensity  intensity  initial  25  S i n c e some  to the seed's  The  intensities.  4-connectedness)  n of the seed's  imple-  ( F r e u d e r , 1976).  initialization  (using  an  have t h e g r e a t e s t  i s not p a r t  w h i l e the  a  few  performed. is  chosen  whose value  intenof  10  This continues remain  similar.  (ranging  i n number  regions  that  from  t o 5959 f o r C r a n b r o o k ) .  n e x t phase towards  that  image).  different  a s e t of  3328 f o r A s h c r o f t  affinity  pixel  A l l neighbours  produces  The  350  a  outward  method  forming i n i t i a l  f o r n) a r e j o i n e d  iteratively  is  by  value i s within used  affinity  i n t h e image t h a t  a seed.  image a r e merged u s i n g  b a s e d m a i n l y on t h e i r  by  extent,  Each p i x e l  was  started  ( o u t o f a 300  features  Merging  i s t o merge r e g i o n s  each other,  Freuder  as  Freuder's  notion  towards  side  i n the i n t e n s i t y  C  continually each o t h e r .  t o i t and for  merges For each  each of i t s  their  affinity  each o t h e r .  tional  to the value returned  by  region  neighbours Affinity  (Ri), a  (Rj)  to  i s inversely  the f u n c t i o n .  The  have  an  function indicate propor-  function  used  i s  243  244  | I ( R i ) - KRj) | * ( A ( R i ) + A ( R j ) ) where I i s t h e a v e r a g e The  second  similarity  term  favours  of i n t e n s i t i e s  Each r e g i o n  t h e most a f f i n i t y  ity  values those  merging  This  the  was  regions  stage  values  merges  until  from  (usually  the  i t was f o u n d  and  tree  superimposed  tion.  Appendix C  By over  termination  affinto  Ra.  a r e canFreuder's  best  t h e ones 30  were  t h a t some o f t h e  merges were b e t t e r c a n d i -  would have been b y p a s s e d by  f o r t h e new r e g i o n s .  there to  neighbours  image).  it  affinity""pairs.  i s repeated  history  linked  satisfied"  that  point  departure  "high a f f i n i t y "  subsequent  but the  when Rb c a l c u l a t e s  i s only select  one r e g i o n regions  Freuder left  for  P r o c e s s i n g may s t o p on i t s own, however,  doubly  Houston  neighbour  r e g i o n s a r e merged, o n l y  done b e c a u s e  from  iterating  merge  another  linked  some o f t h e "lower  analysis. no  In  affinity  This process tinues  the  t h a t p o i n t t o each other  the lowest  for  ( R b ) . Now,  of neighbours  not a l l doubly  dates  with  pairs  merging.  area.  i s t h e major c o n s i d e r a t i o n .  i t may o r may n o t  method,  resulting  regions  f o r i t s neighbours,  for  merged).  t h e merging of s m a l l  with  didates  with  o f a r e g i o n and A i s i t s  (Ra) forms a l i n k  has  Only  intensity  t o merge  displaying  (this  the o r i g i n a l  image,  condition  was u s e d  an  and uses  subsequent i f there are  happened  the r e s u l t i n g  con-  with  the  r e g i o n s a t each "iterate  in this  until  implementa-  245 The o u t p u t f r o m detection. the  number  region,  C  image  identifying  a table  intensity,  Appendix  A region  r e g i o n merging  to that  i s p r o d u c e d where e a c h p i x e l  the r e g i o n  contains  i s similar  i t belongs t o .  i t s location,  and n e i g h b o u r i n g r e g i o n s .  area,  Also,  from  edge  contains f o r each  p e r i m e t e r , average  APPENDIX Using  The  sketch  D  Mapsee2  i s drawn o v e r an image  track  ball.  series  o f d r a w i n g commands ( p l o t s and  required Stoch,  then  graphic  by Mapsee2  (the algorithm  is  then  gotos),  t o do t h i s  (Havens and M a c k w o r t h ,  reencodes the curves  scene  using  a  translated into a the  input  form  was m o d i f i e d  1980; Mackworth a n d  as chains  from  hierarchy  The  are  Maya  pairs.  Maya, t h e a t t r i b u t e v a l u e s sgeta,  the generic  points,  links,  schemata  Since  decompo-  can  to lines,  towns, r o a d - s y s t e m s ,  to a f i l e Franz  be  accessed function.  so t h a t  Lisp.  chains,  which  are  made  up  MAIDS i s upwards c o m p a t i b l e  "get a t t r i b u t e "  mata a r e t h e n w r i t t e n Multilisp  and b u i l d s a  It  of the i n s t a n c e s .  instances  attribute/value  Havens,  and segments t h e image.  f i n d s i n t e r p r e t a t i o n s f o r the chains  sition  from  overlay  the  1981).  Mapsee2 1981)  The  of  MISSEE  using  The r e l e v a n t c a n be  with  sche-  transferred  The schemata o f i n t e r e s t a r e  bridges,  river-systems,  they  in  of  mountains,  roads,  rivers,  m t n - r a n g e s , and g e o s y s t e m s .  246  I  APPENDIX  E  A Sample P r o t o c o l  The  following  script  i n t e r m i x e s user  from  MISSEE t h a t  tion  sources a r e u s e d — t h e  the u s e r ' s in  constitutes  input.  input with  the p r o t o c o l .  intensity  The r e s u l t i n g  image,  the output  A l l three  informa-  t h e s k e t c h map, a n d  i n s t a n c e h i e r a r c h y c a n be  found  F i g u r e 6.21.  % missys —> c y c l e **  i  ***************************************  SUPERVISOR; I n p u t t h e s t a r t i n g v a l u e f o r t h e Queue o r n i l e.g. ( ( 1 0 0 t d % r o a d (smitem * r o a d - 3 ) ) ) nil  ; use the d e f a u l t  OBJECT: g e o s y s t e m - t o p - d o w n . SCHEDULER: a d d i n g  t o Queue  type: n i l v a l :  nil  (100 r i v e r - s y s t e m - t d n i l )  ** 2. *************************************** SCHEDULER: H e r e  i s the p r i o r i t y  queue:  %QUEUE = ( ( 1 0 0 r i v e r - s y s t e m - t d n i l ) ) Do you want  t o modify  it(y),  OBJECT: r i v e r - s y s t e m - t o p - d o w n , SCHEDULER: a d d i n g  t o Queue  be i n l i s p ( l ) , type: n i l v a l :  or q u i t ( q ) ? n nil  (100 b r i d g e - b u - s m ( * b r i d g e - l ) )  ** 3. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * SCHEDULER: H e r e  i s the p r i o r i t y  %QUEUE = ( ( 1 0 0 b r i d g e - b u - s m  queue:.  (*bridge-1))) 247  Do y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) ,  or q u i t ( q ) ? n  OBJECT: b r i d g e - b u - w i t h - s k e t c h - m a p f o r * b r i d g e - l b r i d g e - b u - s m : c r e a t e a new b r i d g e i n s t a n c e % b r i d g e - 1 b r i d g e - b u - s m : *** model c o n s i s t e n c y f o r b r i d g e from sm bridge-bu-sm: regions l i s t = (130 1186 1629 9 1750 2018) bridge-bu-sm: order l i s t = (0 S B B 0) % b r i d g e : v a l u e added t o smitem = *bridge-1 b r i d g e - b u - s m : *** model c o n s i s t e n c y f o r b r i d g e b r i d g e - b u - s m : edges l i s t = (88 11 205 206 241 242) b r i d g e - b u - s m : edges i n p r o p e r o r i e n t a t i o n (88 11 242) b r i d g e - b u - s m : new edge l i s t = ((88 89) (.10 11 12 13 14 15) (242 2 4 3 ) ) b r i d g e - b u - s m : toomany = 0 b r i d g e - b u - s m : c r e a t e new ( s ) c u r b i n s t a n c e % c u r b - 1 b r i d g e - b u - s m : edge segments a r e (88 89) b r i d g e - b u - s m : a n g l e s a r e (248 214) b r i d g e - b u - s m : l e n g t h = 20 m a x s t r e n g t h = 5 a v g s t r e n g t h = 3.5 b r i d g e - b u - s m : c r e a t e new ( s ) c u r b i n s t a n c e % c u r b - 2 b r i d g e - b u - s m : edge segments a r e (10 11 12 13 14 15) b r i d g e - b u - s m : a n g l e s a r e (90 57 79 45 90 27) b r i d g e - b u - s m : l e n g t h = 54 m a x s t r e n g t h = 134 a v g s t r e n g t h = 96 b r i d g e - b u - s m : c r e a t e new ( s ) c u r b i n s t a n c e % c u r b - 3 b r i d g e - b u - s m : edge segments a r e (242 243) b r i d g e - b u - s m : a n g l e s a r e (228 270) b r i d g e - b u - s m : l e n g t h = 20 m a x s t r e n g t h = 22 a v g s t r e n g t h = 18. SCHEDULER: a d d i n g t o Queue (68 r i v e r - t d (2018 r e g i o n ) ) SCHEDULER: a d d i n g t o Queue (68 r i v e r - t d (1750 r e g i o n ) ) SCHEDULER: a d d i n g t o Queue (68 r i v e r - t d (9 r e g i o n ) ) SCHEDULER: a d d i n g t o Queue (68 r i v e r - t d (130 r e g i o n ) ) bridge-bu-sm: neighbouring r i v e r regions are n i l SCHEDULER: a d d i n g t o Queue (68 r o a d - t d (1629 r e g i o n ) ) bridge-bu-sm: n e i g h b o u r i n g road r e g i o n s a r e n i l bridge-bu-sm: %bridge-1 i s model-consistent b r i d g e - b u - s m : % b r i d g e - 1 m u s t - b e - p a r t o f some r o a d - s y s t e m and r i v e r - s y s t e m SCHEDULER: a d d i n g t o Queue (73 r o a d - s y s t e m - b u ( % b r i d g e - l ) ) SCHEDULER: a d d i n g t o Queue (73 r i v e r - s y s t e m - b u (%bridge-l)) * *  A  * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *  SCHEDULER: H e r e %QUEUE = ( ( 7 3 (73 (68 (68 (68 (68 (68  i s the p r i o r i t y  road-system-bu (%bridge-1)) river-system-bu (%bridge-l)) r i v e r - t d (2018 r e g i o n ) ) r i v e r - t d (1750 r e g i o n ) ) r i v e r - t d (9 r e g i o n ) ) r i v e r - t d (130 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) )  Do y o u want t o m o d i f y  Appendix E  queue:  it(y),  be i n l i s p ( l ) ,  or q u i t ( q ) ? n  249 OBJECT: % r o a d - s y s t e m bottom-up f o r % b r i d g e - 1 %road-system-bu: e x i s t i n g systems n i l %road-system-bu: componentlist = (%curb-1 % c u r b - 2 % c u r b - 3 % b r i d g e - l ) % r o a d - s y s t e m - b u : new schema c r e a t e d % r o a d - s y s t e m - 1 % r o a d - s y s t e m - b u : component added % b r i d g e - 1 SCHEDULER: a d d i n g t o Queue (73 l a n d m a s s - b u ( % r o a d - s y s t e m - 1 ) )  ***************************************  **  SCHEDULER: Here %QUEUE =  ((73 (73 (68 (68 (68 (68 (68  i s the p r i o r i t y  queue:  river-system-bu (%bridge-1)) landmass-bu (%road-system-1)) r i v e r - t d (2018 r e g i o n ) ) r i v e r - t d (1750 r e g i o n ) ) r i v e r - t d (9 r e g i o n ) ) r i v e r - t d (130 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) )  Do y o u want t o m o d i f y i t ( y ) ,  be i n l i s p ( l ) ,  or q u i t ( q ) ? n  OBJECT: % r i v e r - s y s t e m bottom-up f o r % b r i d g e - 1 % r i v e r - s y s t e m - b u : e x i s t i n g systems n i l %river-system-bu: componentlist = (%curb-1 % c u r b - 2 % c u r b - 3 % b r i d g e - 1 ) % r i v e r - s y s t e m - b u : new schema c r e a t e d % r i v e r - s y s t e m - 1 % r i v e r - s y s t e m - b u : component added % b r i d g e - 1 SCHEDULER: a d d i n g t o Queue (73 w a t e r b o d y - b u ( % r i v e r - s y s t e m - 1 ))  * * 6. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * SCHEDULER: Here  i s the p r i o r i t y  queue:  %QUEUE = ( ( 7 3 l a n d m a s s - b u ( % r o a d - s y s t e m - 1 ) ) (73 w a t e r b o d y - b u (%river-system-1)) (68 r i v e r - t d (2018 r e g i o n ) ) (68 r i v e r - t d (1750 r e g i o n ) ) (68 r i v e r - t d (9 r e g i o n ) ) (68 r i v e r - t d (130 r e g i o n ) ) (68 r o a d - t d (1629 r e g i o n ) ) ) Do y o u want t o m o d i f y i t ( y ) ,  be i n l i s p ( l ) ,  or q u i t ( q ) ? n  OBJECT: % l a n d m a s s bottom-up f o r % r o a d - s y s t e m - 1 %landmass-bu: e x i s t i n g systems n i l %landmass-bu: componentlist = (%curb-1 % c u r b - 2 % c u r b - 3 % b r i d g e ~ 1 % r o a d - s y s t e m - 1 ) % l a n d m a s s - b u : new schema c r e a t e d %landmass-1 % l a n d m a s s - b u : component added % r o a d - s y s t e m - 1 SCHEDULER: a d d i n g t o Queue (73 g e o s y s t e m - b u (%landmass-1))  ** 7. *************************************** Appendix E  250  SCHEDULER: H e r e  i s the p r i o r i t y  queue:  %QUEUE = ( ( 7 3 w a t e r b o d y - b u (%river-system-1)) (73 g e o s y s t e m - b u ( % l a n d m a s s - 1 ) ) (68 r i v e r - t d (2018 r e g i o n ) ) (68 r i v e r - t d (1750 r e g i o n ) ) (68 r i v e r - t d (9 r e g i o n ) ) (68 r i v e r - t d (130 r e g i o n ) ) (68 r o a d - t d (1629 r e g i o n ) ) ) Do y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) ,  or q u i t ( q ) ? n  OBJECT: w a t e r b o d y - b o t t o m - u p f o r % r i v e r - s y s t e m - 1 w a t e r b o d y - b u : new schema c r e a t e d %waterbody-1 w a t e r b o d y - b u : a d d i n g component %waterbody-1 SCHEDULER: a d d i n g  t o Queue  (73 g e o s y s t e m - b u  (%waterbody-1))  ** 3^ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * SCHEDULER: Here %QUEUE = ( ( 7 3 (73 (68 (68 (68 (68 (68  i s the p r i o r i t y  queue:  geosystem-bu (%landmass-1)) geosystem-bu (%waterbody-1)) r i v e r - t d (2018 r e g i o n ) ) r i v e r - t d (1750 r e g i o n ) ) r i v e r - t d (9 r e g i o n ) ) r i v e r - t d (130 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) )  Do y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) ,  or q u i t ( q ) ? n  OBJECT: g e o s y s t e m - b o t t o m - u p f o r %landmass-1 geosystem-bu: e x i s t i n g systems n i l g e o s y s t e m - b u : new schema c r e a t e d %geosystem-1 geosystem-bu: c o m p o n e n t l i s t = (%curb-1 % c u r b - 2 % c u r b - 3 % b r i d g e - 1 % r o a d - s y s t e m - 1 %landmass-1 % g e o s y s t e m - 1 )  ** g  <  ***************************************  SCHEDULER: Here  i s the p r i o r i t y  queue:  %QUEUE = ( ( 7 3 g e o s y s t e m - b u ( % w a t e r b o d y - 1 ) ) (68 r i v e r - t d (2018 r e g i o n ) ) (68 r i v e r - t d (1750 r e g i o n ) ) (68 r i v e r - t d (9 r e g i o n ) ) (68 r i v e r - t d (130 r e g i o n ) ) (68 r o a d - t d (1629 r e g i o n ) ) ) Do y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) ,  OBJECT: g e o s y s t e m - b o t t o m - u p f o r %waterbody-1 geosystem-bu: e x i s t i n g systems %geosystem-1  Appendix E  or q u i t ( q ) ? n  g e o s y s t e m - b u : new schema c r e a t e d %geosystem-2 geosystem-bu: c o m p o n e n t l i s t = (%curb-1 % c u r b - 2 % c u r b - 3 % b r i d g e - 1 % r i v e r - s y s t e m - 1 %waterbody-1 %geosystem-2) g e o s y s t e m - b u : new s u p e r i o r s y s t e m c r e a t e d : % g e o s y s t e m - 3 c o n t a i n i n g %geosystem-2 and %geosystem-1 ; a t t h i s p o i n t , t h e e n t i r e s t r u c t u r e above ; % b r i d g e - 1 has been c o n s t r u c t e d **  10.  **************************************  SCHEDULER: Here %QUEUE =  ((68 (68 (68 (68 (68  i s the p r i o r i t y  r i v e r - t d (2018 r e g i o n ) ) r i v e r - t d (1750 r e g i o n ) ) r i v e r - t d (9 r e g i o n ) ) r i v e r - t d (130 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) )  Do y o u want t o m o d i f y Commands: d num a s-exp m n1 n2 q e 1. 2. 3. 4.  -> -> -> ->  5. ->  (68 (68 (68 (68  (68 r o a d - t d  Command:m 2 0 SCHEDULER: What  it(y),  (1629  or q u i t ( q ) ? y  region))  ; t r y region is"the  new  1750  first  priority  f o r 2?72  r i v e r - t d (1750 r e g i o n ) ) r i v e r - t d (2018 r e g i o n ) ) r i v e r - t d (9 r e g i o n ) ) r i v e r - t d (130 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) )  SCHEDULER: a d d i n g SCHEDULER: a d d i n g  Appendix  in l i s p ( l ) ,  (2018 r e g i o n ) ) (1750 r e g i o n ) ) (9 r e g i o n ) ) (130 r e g i o n ) )  OBJECT: Top Down on % r i v e r .  **  be  - - D e l e t e e l e m e n t num --Add s-exp t o t h e Queue u s i n g i t s p r i o r i — M o v e e l e m e n t n1 t o a f t e r n2 --Quit processing --End m o d i f y i n g Queue, c o n t i n u e  river-td river-td river-td river-td  1. -> (72 2. -> (68 3. -> (68 4. -> (68 5. -> (68 Command:e  queue:  11.  E  t o Queue t o Queue  type:  region  v a l : 1750  (72 r i v e r - b u - s m (72 r i v e r - b u - s m  (*river-5)) (*river-2))  **************************************  SCHEDULER: H e r e %QUEUE = ( ( 7 2 (72 (68 (68 (68 (68  i s the p r i o r i t y  river-bu-sm (*river-5)) river-bu-sm (*river-2)) r i v e r - t d (2018 r e g i o n ) ) r i v e r - t d (9 r e g i o n ) ) r i v e r - t d (130 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) )  Do y o u want t o m o d i f y i t ( y ) , SUPERVISOR: Do y o u r l i s p ?%printlevel 50 ?(setq % p r i n t l e v e l 1 1 ?d **  queue:  be i n l i s p ( l ) ,  thing—type  or q u i t ( q )  d or q t o stop  11)  12. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *  SCHEDULER: H e r e %QUEUE = ( ( 7 2 (72 (68 (68 (68 (68  i s the p r i o r i t y  queue:  river-bu-sm (*river-5)) river-bu-sm (*river-2)) r i v e r - t d (2018 r e g i o n ) ) r i v e r - t d (9 r e g i o n ) ) r i v e r - t d (130 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) )  Do y o u want t o m o d i f y i t ( y ) ,  be i n l i s p ( l ) ,  or q u i t ( q )  Commands: d num, a s-exp, m n1 n2, q, o r e 1. 2. 3. 4. 5. 6.  —> -> -> -> -> ->  (72 (72 (68 (68 (68 (68  Command:d 1. 2. 3. 4. 5.  river-bu-sm (*river-5)) river-bu-sm (*river-2)) r i v e r - t d (2018 r e g i o n ) ) r i v e r - t d (9 r e g i o n ) ) r i v e r - t d (130 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) 1  ; *river-5  i s an u n i n t e n d e d  —> n i l -> (72 r i v e r - b u - s m (*river-2)) -> (68 r i v e r - t d (2018 r e g i o n ) ) -> (68 r i v e r - t d (9 r e g i o n ) ) -> (68 r i v e r - t d (130 r e g i o n ) )  6. ->  (68 r o a d - t d  Command:d 4 1. —> n i l Appendix E  (1629 r e g i o n ) )  interpretation  253 2. 3. 4. 5. 6.  —> (72 r i v e r - b u - s m ( * r i v e r - 2 ) ) -> (68 r i v e r - t d (2018 r e g i o n ) ) -> n i l —> (68 r i v e r - t d (130 r e g i o n ) ) -> (68 r o a d - t d (1629 r e g i o n ) )  Command:e OBJECT: r i v e r - b u - w i t h - s k e t c h - m a p f o r * r i v e r - 2 r i v e r - b u - s m : c r e a t e a new r i v e r i n s t a n c e % r i v e r - 1 r i v e r - b u - s m : *** model c o n s i s t e n c y f o r r i v e r f r o m sm river-bu-sm: regions l i s t = (1750) r i v e r - b u - s m : i n t e r p r e t a t i o n s a r e (WATER) r i v e r - b u - s m : one i n t e r p r e t a t i o n must be c o n s i s t e n t w i t h w a t e r or f a i l r i v e r - b u - s m : added n e i g h b o u r i n g r e g i o n s a r e (2295 2095 2018) % r i v e r : v a l u e added t o r e g i o n s = (1750 2295 2095 2018) river-bu-sm: %river-1 i s model-consistent % r i v e r : . v a l u e added t o smitem = * r i v e r - 2 river-bu-sm: neighbouring bridge regions are (%bridge-1) r i v e r - b u - s m : % r i v e r - 1 m u s t - b e - p a r t of some r i v e r - s y s t e m SCHEDULER: a d d i n g t o Queue (73 r i v e r - s y s t e m - b u ( % r i v e r - 1 ) ) **  13.  **************************************  SCHEDULER: Here %QUEUE = ( ( 7 3 (68 (68 (68 Do  i s the p r i o r i t y  river-system-bu (%river-l)) r i v e r - t d (2018 r e g i o n ) ) r i v e r - t d (130 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) )  you want t o m o d i f y i t ( y ) ,  OBJECT: % r i v e r - s y s t e m bottom-up %river-system-bu: %river-system-bu:  existing %river-1  ; %river-1 ; contains **  14.  in lisp(l),  or q u i t ( q ) ? n  for %river-1  systems added t o  %river-system-1 %river-system-1  j o i n s t h e same r i v e r - s y s t e m t h a t %bridge-1  i s the p r i o r i t y  queue:  ((68 r i v e r - t d (2018 r e g i o n ) ) (68 r i v e r - t d (130 r e g i o n ) ) (68 r o a d - t d (1629 r e g i o n ) ) )  Do you want t o m o d i f y i t ( y ) , OBJECT: Top Down on % r i v e r .  Appendix  be  **************************************  SCHEDULER: Here %QUEUE =  queue:  E  be  type:  in l i s p ( l ) ,  or q u i t ( q ) ? n  r e g i o n v a l : 2018  SCHEDULER: a d d i n g **  t o Queue (68 r i v e r - b u ( 2 0 1 8 ) )  15. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *  SCHEDULER: Here  i s the p r i o r i t y  queue:  %QUEUE = ( ( 6 8 r i v e r - t d (130 r e g i o n ) ) (68 r o a d - t d (1629 r e g i o n ) ) (68 r i v e r - b u ( 2 0 1 8 ) ) ) ) Do y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) ,  or q u i t ( q ) ? y  Commands: d num, a s-exp, m n1 n2, q, o r e 1. —> 2. ->  (68 r i v e r - t d (130 r e g i o n ) ) (68 r o a d - t d (1629 r e g i o n ) )  3. ->  (68 r i v e r - b u ( 2 0 1 8 ) )  Command:m  3 0  SCHEDULER: What  i s t h e new p r i o r i t y  f o r 3?72  1. -> (72 r i v e r - b u ( 2 0 1 8 ) ) 2. -> (68 r i v e r - t d (130 r e g i o n ) ) 3. -> (68 r o a d - t d (1629 r e g i o n ) ) Command:e OBJECT: river-bu-image-alone r i v e r - b u - i m : r e g i o n 2018 h a s a l r e a d y with r e s u l t % r i v e r - 1 **  been  tried  16. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *  SCHEDULER: Here  i s the p r i o r i t y  queue:  %QUEUE = ( ( 6 8 r i v e r - t d (130 r e g i o n ) ) (68 r o a d - t d (1629 r e g i o n ) ) ) Do y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) ,  or q u i t ( q ) ? n  OBJECT: Top Down on % r i v e r . t y p e : r e g i o n v a l : 130 SCHEDULER: a d d i n g t o Queue (68 r i v e r - b u - s m (*river-5)) SCHEDULER: a d d i n g t o Queue (68 r i v e r - b u - s m (*river-1)) **  17. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *  SCHEDULER: Here  i s the p r i o r i t y  queue:  %QUEUE = ( ( 6 8 r o a d - t d (1629 r e g i o n ) ) (68 r i v e r - b u - s m (*river-5)) (68 r i v e r - b u - s m ( * r i v e r - 1 ) ) )  Appendix E  255 Do y o u want Commands:  t o modify i t ( y ) ,  be i n l i s p ( l ) ,  or quit(q)?y  d num, a s-exp, m n l n2, q , o r e  1. -> (68 r o a d - t d (1629 r e g i o n ) ) 2. —> (68 r i v e r - b u - s m ( * r i v e r - 5 ) ) 3. —> (68 r i v e r - b u - s m ( * r i v e r - l ) ) Command:d 2 1. -> (68 r o a d - t d (1629 r e g i o n ) ) 2. -> n i l 3. -> (68 r i v e r - b u - s m ( * r i v e r - ! ) ) Command:m 3 0 SCHEDULER: What  i s t h e new p r i o r i t y  f o r 3?75  1. -> (75 r i v e r - b u - s m ( * r i v e r - l ) ) 2. -> (68 r o a d - t d (1629 r e g i o n ) ) Command:e OBJECT: r i v e r - b u - w i t h - s k e t c h - m a p f o r * r i v e r - 1 r i v e r - b u - s m : c r e a t e a new r i v e r i n s t a n c e % r i v e r - 2 r i v e r - b u - s m : * * * model c o n s i s t e n c y f o r r i v e r f r o m sm river-bu-sm: regions l i s t = (130 943) r i v e r - b u - s m : i n t e r p r e t a t i o n s a r e (WATER WATER) r i v e r - b u - s m : one i n t e r p r e t a t i o n must be c o n s i s t e n t w i t h w a t e r or fail r i v e r - b u - s m : added n e i g h b o u r i n g r e g i o n s a r e (2065 1872 1779 1058 874 725 524 208 873) % r i v e r : v a l u e added t o r e g i o n s = (943 130 2065 1872 1779 1058 874 725 524 208 873) river-bu-sm: % r i v e r - 2 i s model-consistent % r i v e r : v a l u e a d d e d t o smitem = * r i v e r - 1 river-bu-sm: neighbouring bridge r e g i o n s a r e (%bridge-1) SCHEDULER: a d d i n g t o Queue (70 b r i d g e - t d ( 8 ) ) r i v e r - b u - s m : % r i v e r - 2 m u s t - b e - p a r t o f some r i v e r - s y s t e m SCHEDULER: a d d i n g t o Queue (73 r i v e r - s y s t e m - b u ( % r i v e r - 2 ) ) **  18. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *  SCHEDULER: H e r e  i s the p r i o r i t y  queue:  %QUEUE = ( ( 7 3 r i v e r - s y s t e m - b u ( % r i v e r - 2 ) ) (70 b r i d g e - t d ( 8 ) ) (68 r o a d - t d (1629 r e g i o n ) ) ) Do y o u want  t o modify i t ( y ) ,  be i n l i s p ( l ) ,  or q u i t ( q ) ? n  OBJECT: % r i v e r - s y s t e m bottom-up f o r % r i v e r - 2 % r i v e r - s y s t e m - b u : e x i s t i n g systems %river-system-1  Appendix E  256 %river-system-bu:  %river-2  added t o  %river-system-1  ** 19. ************************************** SCHEDULER: H e r e  i s the p r i o r i t y  queue:  %QUEUE = ( ( 7 0 b r i d g e - t d ( 8 ) ) (68  road-td  (1629 r e g i o n ) ) )  Do y o u want t o m o d i f y i t ( y ) ,  be i n l i s p ( l ) ,  or q u i t ( q ) ? y  Commands: d num, a s-exp, m n1 n2, q, o r e 1. ->  (70 b r i d g e - t d ( 8 ) )  2. ->  (68 r o a d - t d  (1629 r e g i o n ) )  Command:d 1 1. —> n i l 2. ->  (68 r o a d - t d  (1629 r e g i o n ) )  Command:e OBJECT: T o p Down on % r o a d . t y p e : r e g i o n v a l : 1629 SCHEDULER: a d d i n g t o Queue (68 road-bu-sm ( * r o a d - 9 ) ) SCHEDULER: a d d i n g t o Queue (68 road-bu-sm ( * r o a d - 5 ) ) ** 20. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * SCHEDULER: H e r e i s t h e p r i o r i t y q u e u e : %QUEUE = ( ( 6 8 road-bu-sm ( * r o a d - 9 ) ) (68 road-bu-sm ( * r o a d - 5 ) ) ) Do y o u want t o m o d i f y i t ( y ) ,  be i n l i s p ( l ) ,  or q u i t ( q ) ? y  Commands: d num, a s-exp, m n1 n2, q, o r e 1. —> (68 road-bu-sm 2. —> (68 road-bu-sm Command:d 1 1. —> n i l 2. —> (68 road-bu-sm Command:e  (*road-9)) (*road-5))  (*road-5))  ; *road-5 i s the road  that  runs over  OBJECT: r o a d - b u - w i t h - s k e t c h - m a p f o r * r o a d - 5 road-bu-sm: c r e a t e a new r o a d i n s t a n c e %road-1 road-bu-sm: *** model c o n s i s t e n c y f o r new-road road-bu-sm: d e v i a n c e = 53.4132 road-bu-sm: c r e a t e new c u r b i n s t a n c e % c u r b - 4  Appendix E  *bridge-1  from sm  257  road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm:  edge segments a r e (253 255) a n g l e s (153 l e n g t h = 16 m a x s t r e n g t h = 57 a v g s t r e n g t h c r e a t e new c u r b i n s t a n c e % c u r b - 5 edge segments a r e (263) a n g l e s (136) l e n g t h = 3 m a x s t r e n g t h = 34 a v g s t r e n g t h c r e a t e new c u r b i n s t a n c e % c u r b - 6 edge segments a r e (260 261) a n g l e s (136 l e n g t h = 15 m a x s t r e n g t h = 3 a v g s t r e n g t h  road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm:  140) = 32.5 = 34.0 172) = 2.5  c r e a t e new c u r b i n s t a n c e % c u r b - 3 6 edge segments a r e (351) a n g l e s (120) l e n g t h = 15 m a x s t r e n g t h = 42 a v g s t r e n g t h = 42.0 c r e a t e new c u r b i n s t a n c e % c u r b - 3 7 edge segments a r e (50) a n g l e s (142) l e n g t h = 10 m a x s t r e n g t h = 47 a v g s t r e n g t h = 47.0 regions l i s t = (2575 2432 1629 9 738 19 18) interpretations are ((URBAN H I L L S ) (URBAN H I L L S ) (ROAD MOUNTAIN) (URBAN H I L L S ) (URBAN H I L L S ) (URBAN H I L L S ) (ROAD MOUNTAIN)) road-bu-sm: one i n t e r p r e t a t i o n must be c o n s i s t e n t w i t h r o a d or f a i l % r o a d : v a l u e added t o smitem = * r o a d - 5 SCHEDULER: a d d i n g t o Queue (50 t o w n - t d (2575 r e g i o n ) ) SCHEDULER: a d d i n g t o Queue (50 t o w n - t d (2432 r e g i o n ) ) SCHEDULER: a d d i n g t o Queue (50 t o w n - t d (9 r e g i o n ) ) SCHEDULER: a d d i n g t o Queue (50 t o w n - t d (738 r e g i o n ) ) SCHEDULER: a d d i n g t o Queue (50 t o w n - t d (19 r e g i o n ) ) % r o a d : v a l u e added t o r e g i o n s = (18 1629) road-bu-sm: % r o a d - 1 i s m o d e l - c o n s i s t e n t road-bu-sm: n e i g h b o u r i n g b r i d g e r e g i o n s a r e ( % b r i d g e - l ) road-bu-sm: n e i g h b o u r i n g r o a d r e g i o n s a r e n i l SCHEDULER: a d d i n g t o Queue (50 r o a d - t d (18 r e g i o n ) ) SCHEDULER: a d d i n g t o Queue (50 r o a d - t d (1629 r e g i o n ) ) road-bu-sm: n e i g h b o u r i n g town r e g i o n s a r e n i l road-bu-sm: %road-1 m u s t - b e - p a r t o f some r o a d - s y s t e m SCHEDULER: a d d i n g t o Queue (53 r o a d - s y s t e m - b u ( % r o a d - l ) ) **  21  **************************************  SCHEDULER: H e r e %QUEUE = ( ( 5 3 (50 (50 (50 (50 (50 (50 (50  Appendix  E  i s the p r i o r i t y  queue:  road-system-bu (%road-1)) t o w n - t d (2575 r e g i o n ) ) t o w n - t d (2432 r e g i o n ) ) t o w n - t d (9 r e g i o n ) ) t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) )  Do y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) ,  or quit(q)?n  OBJECT: % r o a d - s y s t e m bottom-up f o r %road-1 %road-system-bu: e x i s t i n g systems %road-system-1 % r o a d - s y s t e m - b u : %road-1 added t o % r o a d - s y s t e m - 1 **  22.  *************************************  SCHEDULER: H e r e i s t h e p r i o r i t y %QUEUE = ( ( 5 0 (50 (50 (50 (50 (50 (50  town-td town-td town-td town-td town-td road-td road-td  queue:  (2575 r e g i o n ) ) (2432 r e g i o n ) ) (9 r e g i o n ) ) (738 r e g i o n ) ) (19 r e g i o n ) ) (18 r e g i o n ) ) (1629 r e g i o n ) ) )  Do y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) ,  Commands: d numi , a s-exp, m n1 n2 1 . 2. 3. 4. 5. 6. 7.  —> (50 (50 —> (50 —> (50 —> (50 —> (50 —> (50 —>  town- t d town- t d town- t d town- t d town- t d road- td road- td  il —>n(50 —> (50 —> (50 —> (50 —> (50 —> (50 —>  town- t d town- t d town- t d town- t d road- td road- td  Command:d 1 . 2. 3. 4. 5. 6. 7.  Command:d 1 . 2. 3. 4. 5. 6. 7.  —>nn ii ll —> (50 —> (50 —> (50 —> (50 —> (50 —>  or e  (2575 r e g i o n ) ) (2432 r e g i o n ) ) (9 r e g i o n ) ) (738 r e g i o n ) ) (19 r e g i o n ) ) (18 r e g i o n ) ) (1629 region))  1 (2432 r e g i o n ) ) (9 r e g i o n ) ) (738 r e g i o n ) ) (19 r e g i o n ) ) (18 r e g i o n ) ) (1629 region))  2  town- t d town- t d town- t d road- td road- td  (9 r e g i o n ) ) (738 r e g i o n ) ) (19 r e g i o n ) ) (18 r e g i o n ) ) (1629 region))  Command:e  OBJECT: Top Down on %town. t y p e :  Appendix E  region v a l : 9  or q u i t ( q ) ? y  259  SCHEDULER: a d d i n g  t o Queue  (50 town-bu-sm  (*town-l))  ** 23. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * SCHEDULER: H e r e  i s the p r i o r i t y  queue:  %QUEUE = ( ( 5 0 t o w n - t d (738 r e g i o n ) ) (50 t o w n - t d (19 r e g i o n ) ) (50 r o a d - t d (18 r e g i o n ) ) (50 r o a d - t d (1629 r e g i o n ) ) (50 town-bu-sm (*town-1))) Do y o u want t o m o d i f y i t ( y ) ,  be i n l i s p ( l ) ,  or q u i t ( q ) ? y  Commands: d num, a s-exp, m n1 n2, q, o r e 1. 2. 3. 4. 5.  -> -> -> -> —>  (50 (50 (50 (50 (50  Command:m  t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) town-bu-sm ( * t o w n - l ) ) 5 0  SCHEDULER: What 1. 2. 3. 4. 5.  i s t h e new p r i o r i t y  f o r 5?70  —> (70 town-bu-sm (*town-1)) -> (50 t o w n - t d (738 r e g i o n ) ) —> (50 t o w n - t d (19 r e g i o n ) ) -> (50 r o a d - t d (18 r e g i o n ) ) -> (50 r o a d - t d (1629 r e g i o n ) )  Command:e OBJECT: t o w n - b u - w i t h - s k e t c h - m a p f o r *town-1 town-bu-sm: c r e a t e a new town i n s t a n c e %town-1 town-bu-sm: *** model c o n s i s t e n c y f o r town f r o m sm town-bu-sm: r e g i o n s l i s t = (9) town-bu-sm: i n t e r p r e t a t i o n s a r e ((URBAN H I L L S ) ) town-bu-sm: one i n t e r p r e t a t i o n must be c o n s i s t e n t w i t h or f a i l town-bu-sm: %town-1 i s m o d e l - c o n s i s t e n t %town: v a l u e added t o r e g i o n s = (9) town-bu-sm: c e n t r e o f town a t (51 . 81) %town: v a l u e added t o smitem = *town-1 town-bu-sm: n e i g h b o u r i n g r o a d r e g i o n s a r e n i l SCHEDULER: a d d i n g t o Queue (50 r o a d - t d (8)) SCHEDULER: a d d i n g t o Queue (50 r o a d - t d ( 1 0 0 6 ) ) town-bu-sm: %town-1 m u s t - b e - p a r t o f some landmass SCHEDULER: a d d i n g t o Queue (53 l a n d m a s s - b u (%town-1)) DEMON: a d d i n g n e i g h b o u r s l i n k f r o m %road-1 t o %town-1 ** 24  Appendix E  **************************************  urban  260  SCHEDULER: Here %QUEUE = ( ( 5 3 (50 (50 (50 (50 (50 (50  i s the p r i o r i t y  landmass-bu (%town-l)) t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) road-td (8)) road-td (1006)))  Do y o u want t o m o d i f y i t ( y ) , SUPERVISOR: Do y o u r l i s p ?(sprint  queue:  be i n l i s p ( l ) , o r  thing--type  quit(q)?l  d or q t o stop  '%town-1)  ****instance:  %town-1  centre:  v a l u e : (51  regions:  v a l u e : (9)  *** f G  81)  %if-added:  smitem:  value:  %town  %confidence:  100  %confidence:  100  (prog n i l ( p r i n t l b 8 "%town: v a l u e a d d e d t o " " r e g i o n s =" % v a l ) )  *town-1  %if-added:  %confidence: (prog n i l ( p r i n t l b 8 "%town: a d d e d t o " "smitem  nil apo-> decomposes-to—> n i l neighbours—> %road-1 conf-alg: (prog n i l ( r e t u r n (cond confidence:  stereotype:  ( ( s g e t v %name (t 2 5 ) ) ) )  100 value =" %val))  'smitem  50  nil ?d  ** 25.  **************************************  SCHEDULER: Here %QUEUE = ( ( 5 3 (50 (50 (50 (50  Appendix E  i s the p r i o r i t y  queue:  landmass-bu (%town-l)) t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) )  'n) 50)  (50 r o a d - t d ( 8 ) ) (50 r o a d - t d ( 1 0 0 6 ) ) ) Do y o u want t o m o d i f y  it(y),  Commands: d num, a s-exp, 1. 2. 3. 4. 5. 6. 7.  —> -> -> -> -> -> ->  (53 (50 (50 (50 (50 (50 (50  be i n l i s p ( l ) ,  m n1 n2, q, o r e  landmass-bu (%town-l)) t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) road-td (8)) r o a d - t d (1006))  Command:d 7 1. 2. 3. 4. 5. 6. 7.  —> (53 -> (50 —> (50 -> (50 -> (50 -> (50 —> n i l  landmass-bu (%town-l)) t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) ) road-td (8))  Command:d 6 1. 2. 3. 4. 5. 6. 7.  —> (53 -> (50 -> (50 -> (50 -> (5.0 —> n i l —> n i l  landmass-bu (%town-l)) t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) r o a d - t d (1629 r e g i o n ) )  Command:d 5 1. 2. 3. 4. 5. 6. 7.  -> (53 -> (50 -> (50 -> (50 —> n i l —> n i l —> n i l  l a n d m a s s - b u (%town-1)) t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) ) r o a d - t d (18 r e g i o n ) )  Command:m 4 0 SCHEDULER: What i s t h e new p r i o r i t y 1. 2. 3. 4.  -> —> -> ->  (60 (53 (50 (50  Appendix  E  r o a d - t d (18 r e g i o n ) ) landmass-bu (%town-l)) t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) )  f o r 4?60  or q u i t ( q )  Commandite  ; region  18 c o n t a i n s s e v e r a l  OBJECT: T o p Down on % r o a d . SCHEDULER: a d d i n g t o Queue SCHEDULER: a d d i n g t o Queue SCHEDULER: a d d i n g t o Queue SCHEDULER: a d d i n g t o Queue SCHEDULER: a d d i n g t o Queue SCHEDULER: a d d i n g t o Queue  roads  t y p e : r e g i o n v a l : 18 (60 road-bu-sm ( * r o a d - 3 ) ) (60 road-bu-sm ( * r o a d - 4 ) ) (60 road-bu-sm ( * r o a d - 8 ) ) (60 road-bu-sm ( * r o a d - 7 ) ) (60 road-bu-sm ( * r o a d - 6 ) ) (60 road-bu-sm ( * r o a d - 5 ) )  ** 26. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * SCHEDULER: Here ^QUEUE =  i s the p r i o r i t y  queue:  ( (60 r o a d - b u - sm ( * r o a d - 3 ) ) (60 r o a d - b u - sm ( * r o a d - 4 ) ) (60 r o a d - b u - sm ( * r o a d - 8 ) ) (60 r o a d - b u - sm ( * r o a d - 7 ) ) (60 r o a d - b u - sm ( * r o a d - 6 ) ) (60 r o a d - b u - sm ( * r o a d - 5 ) ) (53 landmass -bu ( % t o w n - l ) ) (50 t o w n - t d (738 r e g i o n ) ) (50 t o w n - t d (1 9 r e g i o n ) ) )  Do y o u want t o m o d i f y i t ( y ) ,  be i n l i s p ( l ) ,  Commands: d num, a s-exp, m n1 n2, q , o r e 1 . 2. 3. 4. 5. 6. 7. 8. 9.  —> (60 —> (60 —> (60 —> (60 —> (60 —> (60 —> (53 —> (50 —> (50  Command:m  road-bu-sm *road-3)) road-bu-sm *road-4)) *road-8)) road-bu-sm *road-7)) road-bu-sm *road-6)) road-bu-sm *road-5)) road-bu-sm l a n d m a s s - b u (%town-1)) t o w n - t d (738 r e g i o n ) ) t o w n - t d (19 r e g i o n ) ) 8 0  SCHEDULER: What 1. 2. 3. 4. 5. 6. 7. 8. 9.  —> (70 (60 — > —> (60 —> (60 —> (60 —> (60 —> (60 —> (53 —> (50  Command:e  Appendix E  i s t h e new p r i o r i t y  town- t d (738 r e g i o n ) ) r o a d - bu- sm ( * r o a d - 3 ) ) r o a d - bu- sm ( * r o a d - 4 ) ) r o a d - bu- sm ( * r o a d - 8 ) ) r o a d - bu- sm ( *.road-7) ) r o a d - bu- sm ( * r o a d - 6 ) ) r o a d - bu- sm ( * r o a d - 5 ) ) landmass -bu (%town-1)) town- t d (19 r e g i o n ) )  f o r 8?70  or q u i t ( q )  OBJECT: Top Down on %town. t y p e : r e g i o n v a l : 738 SCHEDULER: a d d i n g t o Queue (70 town-bu-im ( 7 3 8 ) )  **  27  **************************** *.* ********  SCHEDULER: H e r e ((70 (60 (60 (60 (60 (60 (60 (53 (50  %QUEUE =  Do y o u want  i s the p r i o r i t y  town-bu-im road-bu-sm road-bu-sm road-bu-sm road-bu-sm road-bu-sm road-bu-sm landmass-bu t o w n - t d (19 to modify  queue:  738)) *road-3)) *road-4)) *road-8)) *road-7)) *road-6) ) *road-5) ) (%town-1 )) region)))  it(y),  be  in l i s p ( l ) ,  or q u i t ( q ) ? n  OBJECT: t o w n - b u - i m a g e - a l o n e town-bu-im: r e g i o n = 738 a r e a 2500 town-bu-im: c r e a t e a new town i n s t a n c e %town-2 town-bu-im: *** model c o n s i s t e n c y f o r town from sm %town: v a l u e added t o r e g i o n s = 738 town-bu-im: c e n t r e of town a t (33 . 66) town-bu-im: %town-2 i s m o d e l - c o n s i s t e n t town-bu-im: %town-2 m u s t - b e - p a r t o f some l a n d m a s s SCHEDULER: a d d i n g t o Queue (68 l a n d m a s s - b u (%town-2)) town-bu-im: n e i g h b o u r i n g r o a d r e g i o n s a r e n i l DEMON: a d d i n g n e i g h b o u r s l i n k f r o m % r o a d ~ 1 t o %town-2 **  28. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *  SCHEDULER: H e r e %QUEUE =  ((68 (60 (60 (60 (60 (60 (60 (53 (50  Do y o u want  —> -> —> ->  (68 (60 (60 (60  Appendix  queue:  l a n d m a s s - b u (%town-2)) road-bu-sm ( * r o a d - 3 ) ) road-bu-sm ( * r o a d - 4 ) ) road-bu-sm ( * r o a d - 8 ) ) road-bu-sm ( * r o a d - 7 ) ) road-bu-sm ( * r o a d - 6 ) ) road-bu-sm ( * r o a d - 5 ) ) l a n d m a s s - b u (%town-1)) t o w n - t d (19 r e g i o n ) ) ) to modify  Commands: d num, 1. 2. 3. 4.  i s the p r i o r i t y  it(y),  be  a s-exp, m n1 n2, q, o r e  l a n d m a s s - b u (%town-2)) road-bu-sm ( * r o a d - 3 ) ) road-bu-sm ( * r o a d ~ 4 ) ) road-bu-sm ( * r o a d - 8 ) ) E  in l i s p ( l ) ,  or q u i t ( q ) ? y  264  5. 6. 7. 8. 9.  -> —> —> -> ->  (60 (60 (60 (53 (50  road-bu-sm road-bu-sm road-bu-sm landmass-bu t o w n - t d (19  (*road-7)) (*road-6)) (*road-5)) (%town-l)) region))  Command:m 2 0 SCHEDULER: What 1. 2. 3. 4. 5. 6. 7. 8. 9.  —> -> —> —> —> —> —> —> ->  (75 (68 (60 (60 (60 (60 (60 (53 (50  i s t h e new  road-bu-sm landmass-bu road-bu-sm road-bu-sm road-bu-sm road-bu-sm road-bu-sm landmass-bu t o w n - t d (19  Command:e  priority  f o r 2?75  (*road-3)) (%town-2)) (*road-4)) (*road-8)) (*road-7)) (*road~6)) (*road-5)) (%town-l)) region))  ; *road-3 runs " n o r t h "  from *town-1  OBJECT: r o a d - b u - w i t h - s k e t c h - m a p f o r * r o a d - 3 r e a t e a new r o a d i n s t a n c e % r o a d - 2 r o a d - bu- sm: r o a d - bu- sm: r o a d - bu- sm: r o a d - bu- sm: c r e a t e new c u r b i n s t a n c e % c u r b ~ 3 8 r o a d - bu- sm: edge segments a r e (412) a n g l e s (104) r o a d - bu- sm: l e n g t h = 6 m a x s t r e n g t h = 73 a v g s t r e n g t h = 73.0 r o a d - bu- sm: one o r i e n t a t i o n must be c o n s i s t e n t w i t h o t h e r c u r b s or f a i l r o a d - bu- sm: c r e a t e new c u r b i n s t a n c e % c u r b ~ 3 9 r o a d - bu- sm: edge segments a r e (408) a n g l e s (90) r o a d - bu- sm: l e n g t h = 3 m a x s t r e n g t h = 43 a v g s t r e n g t h = 43.0 r o a d - bu- sm: max c o n f i d e n c e i s 50.0 (18) r o a d - bu- sm: r e g i o n s l i s t = r o a d - bu- sm: i n t e r p r e t a t i o n s a r e ((ROAD MOUNTAIN)) r o a d - bu- sm: one i n t e r p r e t a t i o n must be c o n s i s t e n t w i t h r o a d or fail % r o a d : v a l u e added t o smitem = * r o a d - 3 SYSTEM: W a r n i n g , more t h a n one i n t e r p r e t a t i o n f o r r e g i o n 18 SYSTEM: A d d i n g % r o a d % r o a d ~ 2 t o (%road % r o a d - 1 ) % r o a d : v a l u e added t o r e g i o n s = (18) road-bu-sm: % r o a d - 2 i s m o d e l - c o n s i s t e n t road-bu-sm: n e i g h b o u r i n g b r i d g e r e g i o n s a r e n i l road-bu-sm: n e i g h b o u r i n g r o a d r e g i o n s a r e n i l SCHEDULER: a d d i n g t o Queue (67 r o a d - t d (18 r e g i o n ) ) road-bu-sm: n e i g h b o u r i n g town, r e g i o n s a r e (%town-1) road-bu-sm: % r o a d - 2 m u s t - b e - p a r t o f some r o a d - s y s t e m SCHEDULER: a d d i n g t o Queue (70 r o a d - s y s t e m - b u ( % r o a d - 2 ) ) ** 29.  Appendix  E  **************************************  265  SCHEDULER: H e r e i s t h e p r i o r i t y %QUEUE =  ((70 (68 (67 (60 (60 (60 (60 (60 (53 (50  queue:  r o a d - s y s t e m -bu ( % r o a d - 2 ) ) l a n d m a s s - b u (%town-2)) r o a d - t d (18 r e g i o n ) ) road-bu-sm ( * r o a d - 4 ) ) road-bu-sm (*road-8) ) road-bu-sm ( * r o a d - 7 ) ) road-bu-sm ( * r o a d - 6 ) ) road-bu-sm ( * r o a d - 5 ) ) l a n d m a s s - b u (%town-1 ) ) t o w n - t d (19 r e g i o n ) ) )  Do y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) ,  or q u i t ( q ) ? y  Commands: d num, a s-exp, m n1 n2, q, o r e 1. —> 2. —> 3. -> 4. —> 5. —> 6. —> 7. —> 8. —> 9. —> 10. ->  (70 r o a d - s y s t e m -bu ( % r o a d - 2 ) ) (68 l a n d m a s s - b u (%town-2)) (67 r o a d - t d (18 r e g i o n ) ) (60 road-bu-sm ( * r o a d - 4 ) ) (60 road-bu-sm ( * r o a d - 8 ) ) (60 road-bu-sm ( * r o a d - 7 ) ) (60 road-bu-sm ( * r o a d - 6 ) ) (60 road-bu-sm (*road-5) ) (53 l a n d m a s s - b u (%town-1)) (50 t o w n - t d (1 9 r e g i o n ) )  Command:a 1. —> 2. —> 3. —> 4. -> 5. —> 6. —> 7. —> 8. —> 9. —> 10. —> 11. ->  (77 road-bu-sm  (*road-4))  (77 road-bu-sm ( * r o a d - 4 ) ) (70 r o a d - s y s t e m - bu ( % r o a d - 2 ) ) (68 l a n d m a s s - b u (%town-2) ) (67 r o a d - t d (18 r e g i o n ) ) (60 road-bu-sm ( * r o a d - 4 ) ) (60 road-bu-sm ( * r o a d - 8 ) ) (60 road-bu-sm ( * r o a d - 7 ) ) (60 road-bu-sm ( * r o a d - 6 ) ) (60 road-bu-sm ( * r o a d - 5 ) ) (53 l a n d m a s s - b u (%town-1 )) (50 t o w n - t d (19 r e g i o n ) )  Command:e  ; * r o a d - 4 s t a r t s a t * r o a d - 5 a n d r u n s by ; t h e end o f * r o a d - 2  OBJECT: r o a d - b u - w i t h - s k e t c h - m a p f o r * r o a d - 4 road-bu-sm: c r e a t e a new r o a d i n s t a n c e % r o a d - 3 road-bu-sm: *** model c o n s i s t e n c y f o r r o a d from sm road-bu-sm: d e v i a n c e = 0.921442 road-bu-sm: c r e a t e new c u r b i n s t a n c e % c u r b - 4 0 road-bu-sm: edge segments a r e (422) a n g l e (24) road-bu-sm: l e n g t h = 15 m a x s t r e n g t h = 74 a v g s t r e n g t h = 74.0 road-bu-sm: one o r i e n t a t i o n must be c o n s i s t e n t w i t h o t h e r  Appendix E  266 curbs or f a i l road-bu-sm: c r e a t e new c u r b i n s t a n c e % c u r b - 4 1 road-bu-sm: edge segments a r e (411) a n g l e s (162) road-bu-sm: l e n g t h = 5 m a x s t r e n g t h = 41 a v g s t r e n g t h  =  41.0  road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm: road-bu-sm:  c r e a t e new c u r b i n s t a n c e % c u r b - 5 0 edge segments a r e (146) a n g l e s (39) l e n g t h = 18 m a x s t r e n g t h = 14 a v g s t r e n g t h = 14.0 c r e a t e new c u r b i n s t a n c e % c u r b - 5 l edge segments a r e (160 159 161) c u r b a n g l e s a r e (45 180 180) l e n g t h = 12 m a x s t r e n g t h = 86 a v g s t r e n g t h = 74.667 max c o n f i d e n c e i s 53.986 regions l i s t = (438 738 18) i n t e r p r e t a t i o n s a r e ((URBAN H I L L S ) (URBAN H I L L S ) (ROAD MOUNTAIN)) road-bu-sm: one i n t e r p r e t a t i o n must be c o n s i s t e n t w i t h r o a d or fail, % r o a d : v a l u e added t o smitem = * r o a d - 4 SCHEDULER: a d d i n g t o Queue (64 t o w n - t d (438 r e g i o n ) ) SYSTEM: W a r n i n g , more t h a n one i n t e r p r e t a t i o n f o r r e g i o n 18 SYSTEM: A d d i n g % r o a d % r o a d - 3 t o (%road % r o a d - 2 % r o a d % r o a d - 1 ) % r o a d : v a l u e added t o r e g i o n s = (18) road-bu-sm: % r o a d ~ 3 i s m o d e l - c o n s i s t e n t road-bu-sm: n e i g h b o u r i n g b r i d g e r e g i o n s a r e n i l road-bu-sm: n e i g h b o u r i n g r o a d r e g i o n s a r e (%road-1 % r o a d - 2 ) SCHEDULER: a d d i n g t o Queue (64 r o a d - t d (18 r e g i o n ) ) road-bu-sm: n e i g h b o u r i n g town r e g i o n s a r e n i l road-bu-sm: % r o a d - 3 m u s t - b e - p a r t of some r o a d - s y s t e m SCHEDULER: a d d i n g t o Queue (67 r o a d - s y s t e m - b u ( % r o a d - 3 ) ) **  30. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *  SCHEDULER: Here %QUEUE = ((70 (68 (67 (67 (64 (64 (60 (60 (60 (60 (60 (53 (50  i s the p r i o r i t y  road-system-bu (%road-2)) l a n d m a s s - b u (%town-2)) road-system-bu (%road-3)) r o a d - t d (18 r e g i o n ) ) t o w n - t d (438 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) road-bu-sm ( * r o a d - 4 ) ) road-bu-sm ( * r o a d - 8 ) ) road-bu-sm ( * r o a d - 7 ) ) road-bu-sm ( * r o a d - 6 ) ) road-bu-sm ( * r o a d - 5 ) ) landmass-bu (%town-l)) t o w n - t d (19 r e g i o n ) ) )  Do y o u want t o m o d i f y  Appendix  E  queue:  it(y),  be  in l i s p ( l ) ,  or  quit(q)?l  267  SUPERVISOR: Do y o u r ?(sprint  lisp  thing—type  d or q to stop  '%road-3)  ****instance:  %road-3  *** f 0  stereotype:  %road  %confidence: ( p r o g ...  100  smitem:  v a l u e : *road-4 % i f-added:  ends:  v a l u e : ((41 . 62) (29 . 126)) %conf idence: 100  separation:  value: n i l  %confidence: n i l  deviance:  value:  %confidence: n i l  regions:  v a l u e : (18) %if-added:  0.921442  %confidence: ( p r o g ...  100  nil (%curb-40 % c u r b - 4 1 % c u r b - 4 2 % c u r b - 4 3 % c u r b - 4 4 %curb-45 %curb-46 %curb-47 %curb-48 %curb-48 %curb-49 %curb-50 %curb-51) (%road-1 % r o a d - 2 ) (prog ( v a l 1 s t ) 64  apo—> decomposes-to—> neighbours—> conf-alg: conf i d e n c e : nil ?d ** 3 1  **************************************  SCHEDULER: Here %QUEUE = ( ( 7 0 (68 (67 (67 (64 (64 (60 (60 (60 (60 (60 (53 (50  i s the p r i o r i t y  queue:  r o a d - s y s t e m - bu ( % r o a d - 2 ) ) l a n d m a s s - b u (%town-2) ) r o a d - s y s t e m - bu ( % r o a d - 3 ) ) r o a d - t d (18 r e g i o n ) ) t o w n - t d (438 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) road-bu-sm ( * r o a d - 4 ) ) road-bu-sm ( * r o a d - 8 ) ) road-bu-sm ( * r o a d - 7 ) ) road-bu-sm ( * r o a d - 6 ) ) road-bu-sm ( * r o a d - 5 ) ) l a n d m a s s - b u (%town-1 )) t o w n - t d (19 r e g i o n ) ) )  Do. y o u want t o m o d i f y  it(y),  be i n l i s p ( l ) , o r q u i t ( q ) ? n  OBJECT: % r o a d - s y s t e m bottom-up f o r % r o a d - 2 %road-system-1 %road-system-bu: e x i s t i n g systems  Appendix  E  % r o a d - s y s t e m - b u : new schema c r e a t e d % r o a d - s y s t e m - 2 % r o a d - s y s t e m - b u : component a d d e d % r o a d - 2 SCHEDULER: a d d i n g t o Queue (68 l a n d m a s s - b u ( % r o a d - s y s t e m - 2 ) ) ** 32.  **************************************  SCHEDULER: H e r e %QUEUE = ( ( 6 8 (68 (67 (67 (64 (64 (60 (60 (60 (60 (60 (53 (50 Do  you want  i s the p r i o r i t y  queue:  landmass-bu (%road-system-2)) l a n d m a s s - b u (%town-2)) road-system-bu (%road-3)) r o a d - t d (18 r e g i o n ) ) t o w n - t d (438 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) road-bu-sm ( * r o a d - 4 ) ) road-bu-sm ( * r o a d - 8 ) ) road-bu-sm ( * r o a d - 7 ) ) road-bu-sm ( * r o a d - 6 ) ) road-bu-sm ( * r o a d - 5 ) ) landmass-bu (%town-l)) t o w n - t d (19 r e g i o n ) ) ) to modify i t ( y ) ,  Commands: d num,  a s-exp, m n1  be  in l i s p ( l ) ,  n2, q, o r e  landmass-bu (%road-system —> (68 (68 l a n d m a s s - b u (%town-2)) —> (67 r o a d - s y s t e m - bu ( % r o a d - 3 ) ) —> (67 r o a d - t d (18 r e g i o n ) ) —> (64 t o w n - t d (438 r e g i o n ) ) —> r o a d - t d (18 r e g i o n ) ) —> (64 (60 road-bu-sm ( * r o a d - 8 ) ) —> (60 road-bu-sm ( * r o a d - 7 ) ) —> (60 road-bu-sm ( * r o a d - 6 ) ) >> (60 road-bu-sm ( * r o a d - 5 ) ) .—— . -> (53 l a n d m a s s - b u (%town-1)) . -> (50 t o w n - t d (19 r e g i o n ) ) ) )  1. 2. 3. 4. 5. 6. 7. 8. 9. 10 1 1 12  1  Command:d  il —> n(68 (%town-2)) —> (67 lr ao na dd m- as syss-tbeum - bu ( % r o a d - 3 ) ) —> (67 r o a d - t d (18 r e g i o n ) ) —> (64 t o w n - t d (438 r e g i o n ) ) —> (64 r o a d - t d (18 r e g i o n ) ) —> (60 road-bu-sm ( * r o a d - 8 ) ) —> (60 road-bu-sm ( * r o a d - 7 ) ) —> (60 road-bu-sm ( * r o a d - 6 ) ) —> (60 road-bu-sm ( * r o a d - 5 ) )  1 . 2. 3. 4. 5. 6. 7. 8. 9. 10. —> 11. —> 12. ->  (53 l a n d m a s s - b u ( % t o w n - l ) ) (50 t o w n - t d (19 r e g i o n ) ) ) )  Appendix  E  or  quit(q)?y  269  Command:e OBJECT: % l a n d m a s s bottom-up f o r %town-2 %landmass-bu: e x i s t i n g systems (%landmass-1) % l a n d m a s s - b u : %town-2 added t o %landmass-1 ** 33.  **************************************  SCHEDULER: H e r e %QUEUE =  Do  ((67 (67 (64 (64 (60 (60 (60 (60 (60 (53 (50  i s the p r i o r i t y  queue:  r o a d - s y s t e m - bu ^ r o a d - 3)) r o a d - t d (18 r e g i o n ) ) t o w n - t d (438 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) road-bu-sm ( * r o a d - 4 ) ) road-bu-sm ( * r o a d - 8 ) ) road-bu-sm ( * r o a d - 7 ) ) road-bu-sm ( * r o a d - 6 ) ) road-bu-sm ( * r o a d - 5 ) ) l a n d m a s s - b u (%town-1)) t o w n - t d (19 r e g i o n ) ) )  you want t o m o d i f y  Commands: d num,  it(y),  a s-exp, m n1  be  in l i s p ( l ) ,  n2, q, o r e  —> (67 —> (67 — > (64 -> (64  road-system-bu (%road-3)) r o a d - t d (18 r e g i o n ) ) t o w n - t d (438 r e g i o n ) ) r o a d - t d (18 r e g i o n ) ) (60 road-bu-sm (*road-4)) -> (60 road-bu-sm (*road-8)) (60 road-bu-sm (*road-7)) (60 road-bu-sm (*road-6)) 9. -> (60 road-bu-sm (*road-5)) —> (53 l a n d m a s s - b u (%town-1)) 10 1 1 —> (50 t o w n - t d (19 r e g i o n ) )  —> —> —>  Command:m  10 1  SCHEDULER: What  i s t h e new  priority  r o a d - s y s t e m - bu ( % r o a d - 3 ) ) —> (67 (67 l a n d m a s s - b u (%town-1 ) ) —> (67 r o a d - t d (18 r e g i o n ) ) —> (64 t o w n - t d (438 r e g i o n ) ) —> (64 r o a d - t d (18 r e g i o n ) ) —> (60 road-bu-sm ( * r o a d - 4 ) ) —> (60 road-bu-sm ( * r o a d - 8 ) ) —> (60 road-bu-sm ( * r o a d - 7 ) ) —> (60 road-bu-sm ( * r o a d - 6 ) ) • .—>-> (60 road-bu-sm ( * r o a d - 5 ) ) . -> (50 t o w n - t d (19 r e g i o n ) )  1. 2. 3. 4. 5. 6. 7. 8. 9. 10 1 1  Appendix  E  f o r 10767  or  quit(q)?y  Command:e OBJECT: % r o a d - s y s t e m b o t t o m - u p f o r % r o a d - 3 (%road-system-2 %road-system-bu: e x i s t i n g systems road-system-1 ) % r o a d - s y s t e m - b u : % r o a d - 3 added t o % r o a d - s y s t e m - 2 %road-system-bu: merging %road-system-1 i n t o %road-system-2  ** 34  **************************************  SCHEDULER: Here %QUEUE =  i s t h e p r i o r i t y queue:  ( (67 landmass -bu ( % t o w n - l ) ) (67 r o a d - t d (18 r e g i o n ) ) (64 t o w n - t d (438 r e g i o n ) ) (64 r o a d - t d (18 r e g i o n ) ) (60 r o a d - b u - sm ( * r o a d - 4 ) ) (60 r o a d - b u - sm ( * r o a d - 8 ) ) (60 r o a d - b u - sm ( * r o a d - 7 ) ) (60 r o a d - b u - sm ( * r o a d - 6 ) ) (60 r o a d - b u - sm ( * r o a d - 5 ) ) (50 t o w n - t d (19 r e g i o n ) ) )  Do y o u want t o m o d i f y i t ( y ) ,  be i n l i s p ( l ) ,  or q u i t ( q ) ? n  OBJECT: % l a n d m a s s b o t t o m - u p f o r %town-1 %landmass-bu: e x i s t i n g systems (%landmass-1) % l a n d m a s s - b u : %town-1 a d d e d t o %landmass-1 ** 35,  **************************************  SCHEDULER: H e r e %QUEUE =  ((67 (64 (64 (60 (60 (60 (60 (60 (50  i s the p r i o r i t y queue:  road-td town-td road-td road-buroad-buroad-buroad-buroad-butown-td  (18 r e g i o n ) ) (438 r e g i o n ) ) (18 r e g i o n ) ) sm ( * r o a d - 4 ) ) sm ( * r o a d - 8 ) ) sm ( * r o a d - 7 ) ) sm ( * r o a d - 6 ) ) sm ( * r o a d - 5 ) ) (19 r e g i o n ) ) )  Do y o u want t o m o d i f y i t ( y ) , ?(sprint '%bridge-l) * * * * i n s t a n c e : %bridge-1  ***of  be i n l i s p ( l ) , stereotype:  smitem:  value: *bridge-1 % i f-added:  orderlist  value:  Appendix E  (0 S B B 0)  or q u i t ( q ) ? l %bridge  %confidence: (prog . . .  100  %confidence:  100  neighbour-regions:  v a l u e : (2018 %confidence:  shadowregions:  value:  (1186)  %confidence:  100  roadregions:  value:  (1629)  %confidence:  100  decomposes-to—> neighbours-> apo—> conf-alg: confidence:  1750  9  130) 100  (%curb-3 % c u r b - 2 % c u r b - l ) (%road-1 % r i v e r - 2 % r i v e r - 1 ) (%road-system-2 %river-system-1) ( p r o g ... 68  nil ?(sprint '%curb-l) * * * * i n s t a n c e : %curb-1  ***of  stereotype:  %curb  avgstrength:  value:  3.5  % c o n f i d e n c e : 75  maxstrength:  value:  5  % c o n f i d e n c e : 75  length:  value:  20  % c o n f i d e n c e : 100  angles:  value:  (248  edgesegs:  value:  (88  type:  value:  shadow  apo—> conf-alg: conf idence:  214) 89)  %conf idence:  (80)  % c o n f i d e n c e : 100 % c o n f i d e n c e : 100  %bridge-1 (prog . 85  nil ? ( s p r i n t '%curb-2) * * * * i n s t a n c e : %curb-2  *** o f :  stereotype:  %curb  avgstrength:  value:  96.33  % c o n f i d e n c e : 100  maxstrength:  value:  1 34  % c o n f i d e n c e : 100  length:  value:  54  % c o n f i d e n c e : 100  angles:  v a l u e : (90 57 %conf idence:  79 45 90 27) (80)  edgesegs:  v a l u e : (10 11 %conf idence:  12 13  type:  value:  Appendix  E  road  14 15) 100  % c o n f i d e n c e : 100  272  apo-> conf-alg: conf idence:  %bridge-1 (prog 93  nil ? ( s p r i n t '%curb-3) * * * * i n s t a n c e : %curb-3  ***of  stereotype:  %curb  avgstrength:  value:  18.0  %conf idence:  100  maxstrength:  value:  22  %conf idence:  100  length:  value:  20  % c o n f i d e n c e : 100  angles:  value:  (228  270)  %conf idence:  (80)  edgesegs:  value:  (242  243)  %conf idence:  100  type:  value:  road  apo—> conf-alg: conf idence:  % c o n f i d e n c e : 100  %bridge-1 (prog 93  nil ?%instances (%landmass-1 % r o a d - s y s t e m - 2 % c u r b - 9 % c u r b - 8 % r o a d - 3 % c u r b - 7 % c u r b - 6 % r o a d - 2 %town-2 %town-1 % c u r b - 5 % c u r b - 4 % r o a d - 1 % r i v e r - 2 % r i v e r - 1 %geosystem-3 % g e o s y s t e m - 2 %geosystem-1 %waterbody-1 % r i v e r - s y s t e m - 1 %curb-3 %curb-2 %curb-1 %bridge-l) ?(sprint '%road-system-2) * * * * i n s t a n c e : %road-system-2 * * * f stereotype: %road-system Q  neighbours—> decomposes-to-> apo—> conf-alg: confidence:  n i l (%bridge-1 %road-1 %landmass-1 ( p r o g ... 63  ?(sprint '%river-l) ****instance: %river-1  *** f 0  %road-3  stereotype:  %road-2)  %river  v a l u e : (1750 2295 2095 2018) %confidence: 100 %if-added: ( p r o g ...  regions:  Appendix  ^confidence: n i l ( p r o g ...  value: n i l % i f-added:  smitem:  E  value: *river-2 %if-added:  smitem:  apo—> decomposes-to—> neighbours—> conf-alg: conf idence:  E  stereotype:  %landmass  %confidence: n i l ( p r o g ...  nil (%town-1 % r o a d - s y s t e m - 2 )%geosystem-1 ( p r o g ... 63  ? ( s s p r i n t n '%geosystem-3) %geosystem-3 aio %geosystem instances %geosystem-3 %geosystem-2 apo %geosystem-3 spec i a l i z e s - t o %waterbody-1 ako %geosystem-2 aio %waterbody instances %waterbody-1 specializes-to %river-system-1 aio %river-system instances % r iver-system-1 ako %waterbody-1 spec i a l i z e s - t o %river-2 apo % r iver-system-1 aio %river  Appendix  ***of  value: n i l %if-added:  neighbours—> decomposes-to—> ako—> conf-alg: confidence:  100  % r iver-system-1 nil %bridge-1 ( p r o g ... 70  ? ( s p r i n t '%landmass-1) * * * * i n s t a n c e : %landmass-1 smitem:  %confidence: ( p r o g ...  instances %river-2 %river-1 apo %river-system-1 aio %river neighbours %bridge-1 aio %bridge instances %bridge-1 decomposes-to %curb-3 apo %bridge-1 aio %curb instances %curb-51 apo %road-3 apo %road-system-2 apo %landmass-1 ako %geosystem-1 apo %geosystem-3 specializes %landmass-1 aio %geosystem aio %landmass instances-%landmass-1 decomposes-to %road-system-2 %town-2 apo %landmass-1 aio %town instances %town-2 %town-1 apo %landmass aio %town  Appendix E  neighbours %road-2 apo %road-system-2 aio %road instances %road-3 %road-2 %road-1 aio %road decomposes-to%curb-37 apo %road-1 aio %curb  •  %curb-4 apo %road-1 aio %curb neighbours %road-3 %town-2 %town-1 %bridge-1 apo %road-system-2 decomposes-to %curb-39 apo %road-2 aio %curb %curb-38 apo %road-2 aio %curb neighbours %road-3 %town-1 %road-1 neighbours %road-1 %town-1 aio %road-system  Appendix  E  instances— %road-system-2 decomposes-to %bridge-1 %road-1 %road-3 %road-2 aio %road decomposes-to~%curb-5l %curb-50 apo %road-3 aio %curb  •  %curb-40 apo %road-3 aio %curb neighbours %road-2 %road-1 aio %curb %curb-50  %curb-2 apo %bridge-1 aio %curb %curb-1 apo—%bridge-1 aio %curb %curb-2 %curb-1 neighbours %road-1 %river-2 %river-1 apo %road-system-2 %r i v e r - s y s t e m - 1 neighbours Appendix  E  %bridge-1 %river-1 %bridge-1 aio %geosystem %geosystem-1 decomposes-to %geosystem-1 %geosystem-2 nil ? ( s p r i n t n '%river-system-1 'decomposes-to) %river-system-1 %r iver-1 %river-2 % b r idge-1 %curb-1 %curb-2 %curb-3 nil ? ( s p r i n t n '%landmass-1 'decomposes-to) %landmass-1 %town-1 %road-system-2 %bridge-1 %curb-1 %curb-2 %curb-3 %road-l %curb-4  %curb-37 %road-2 %curb-38 %curb-39 %road-3 %curb-40  %curb-51 nil ? ( s p r i n t n '%geosystem-3 %geosystem-3 %geosystem^1 %geosystem-2 nil ?d ** 3g  **************************************  SCHEDULER: Here Appendix  E  'decomposes-to)  i s the p r i o r i t y  queue:  %QUEUE = ( ( 6 7 r o a d - t d (18 r e g i o n ) ) (64 t o w n - t d (438 r e g i o n ) ) (64 r o a d - t d (18 r e g i o n ) ) (60 road-bu-sm (*road-4)) (60 road-bu-sm (*road-8)) (60 road-bu-sm (*road-7)) (60 road-bu-sm (*road-6)) (60 road-bu-sm (*road-5)) (50 t o w n - t d (19 r e g i o n ) ) ) Do y o u want t o m o d i f y i t ( y ) , be i n l i s p ( l ) , SUPERVISOR: C y c l e b r o k e n by u s e r - - b y e nil -> %  Appendix  E  or q u i t ( q )  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0051869/manifest

Comment

Related Items