International Construction Specialty Conference of the Canadian Society for Civil Engineering (ICSC) (5th : 2015)

Development of an automated 3D/4D as-built model generation system for construction progress monitoring… Maalek, Reza; Lichti, Derek; Ruwanpura, Janaka Jun 30, 2015

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

Item Metadata

Download

Media
52660-Maalek_R_et_al_ICSC15_312_Development_Of_An.pdf [ 2.71MB ]
52660-Maalek_R_et_al_ICSC15_312_Development_Of_An_slides.pdf [ 5MB ]
Metadata
JSON: 52660-1.0076418.json
JSON-LD: 52660-1.0076418-ld.json
RDF/XML (Pretty): 52660-1.0076418-rdf.xml
RDF/JSON: 52660-1.0076418-rdf.json
Turtle: 52660-1.0076418-turtle.txt
N-Triples: 52660-1.0076418-rdf-ntriples.txt
Original Record: 52660-1.0076418-source.json
Full Text
52660-1.0076418-fulltext.txt
Citation
52660-1.0076418.ris

Full Text

5th International/11th Construction Specialty Conference 5e International/11e Conférence spécialisée sur la construction    Vancouver, British Columbia June 8 to June 10, 2015 / 8 juin au 10 juin 2015   DEVELOPMENT OF AN AUTOMATED 3D/4D AS-BUILT MODEL GENERATION SYSTEM FOR CONSTRUCTION PROGRESS MONITORING AND QUALITY CONTROL Reza Maalek1,3 Derek Lichti2 and Janaka Ruwanpura1 1 Department of Civil Engineering, University of Calgary, Canada 2 Department of Geomatics Engineering, University of Calgary, Canada 3 rmaalek@ucalgary.ca Abstract: Automating the progress monitoring and control process is of great interest to industry practitioners to help improve the limitations associated with the current manual data collection and analysis practices. Two remote sensing technologies, namely, Light Detection and Ranging (LiDAR) and digital camera, are widely used to acquire 3D point clouds as a means of measuring the “scope of the work performed” of structural elements. However, to assign the collected 3D point clouds to their corresponding structural element, current object-based recognition models use the as-planned 4D model, which may not be reliable in cases where the locations of the as-built structure differ from those of the planned, and/or the planned 4D model is not available with sufficient detail. Here, a novel method is proposed to eliminate the dependency on the as-planned data by automatically generating the 3D/4D as-built model through a robust Principal Component Analysis-based (PCA) segmentation algorithm. The proposed system is also independent of the technology used to capture the 3D point clouds. To evaluate the reliability of the proposed automated as-built model generation procedure, two sets of LiDAR data from the "Mechanics of Materials" laboratory and the "Graduate Student Hall of Residence" construction site at the University of Calgary were collected. A novel method of automated registration of the as-built model to the planned model coordinate system is also proposed through which the compliance of the planned vs. actual dimensions of corresponding structural elements are examined. The results of the two experiments demonstrate the applicability of the proposed methods for the automatic generation of the 3D/4D as-built model and the dimension compliance control of structural elements. 1 INTRODUCTION Project monitoring and control are vital to facilitate decision makers identify deviations between the planned vs. as-built states of the project and take timely measures where required (Maalek and Sadeghpour, 2012). Monitoring is the process of collecting onsite data as a means of measuring the Project Performance Indicators (PPI). Traditionally, onsite data are collected manually, a time consuming, error-prone and labor intensive task particularly on large scale projects (Golparvar-Fard et al. 2009a). In practice, to justify the time and cost associated with such manual approaches, a limited amount (and/or frequency) of onsite data are collected, which diminishes the ability of the project manager to identify the causes of delays and cost overruns on time. In addition, most onsite data collection processes are anecdotal without a proper monitoring plan/strategy (Golparvar-Fard et al. 2011), which influences the time, cost and reliability of the collected PPIs. 312-1 Project control involves the processing of the accumulated data for the determination of the performance of the current state of the project. Therefore, the reliability of the determination of the performance of the project is highly dependent on the strategy as well as the accuracy of the collected data during the monitoring process (Saadat and Cretin, 2002). Currently, site supervisory personnel spend 30-50% of their time on manually monitoring and controlling onsite data (McCullouch, 1997, Golparvar-Fard et al. 2009). If this time is reduced by means of a novel approach to onsite data collection and analysis, more time can be allocated to improving vital construction related concerns such as safety (Maalek & Sadeghpour, 2011), and workforce productivity and communications (Choy and Ruwanpura, 2007). Therefore, automating the monitoring and control process is proposed in recent years to help overcome the aforementioned limitations of current manual practices. 2 LITERATURE REVIEW The percentage of completion of an activity is suggested as the Key Performance Indicator (KPI) capable of providing progress information in activity-level (Maalek et al. 2014). In order to automatically extract this metric, the scope of the work performed for each activity is required to be identified by means of a remote sensing technology. Currently, two remote sensing technologies, namely, digital camera and LiDAR, are widely used to generate 3D coordinates of the surrounding surfaces. The overview of the previous research related to the application of these two technologies for progress monitoring of construction activities are presented in the following. 2.1 Digital Camera In practice, photographs are commonly taken to record the progress of specific activities and/or to help minimize disputes/claims (Golparvar-Fard et al. 2009a). These images are stored without proper documentation and indexing (Brilakis et al. 2006). In addition, the KPIs are manually extracted from the large number of unordered/randomized images, constituting a costly, lengthy and challenging procedure. Current research aims at improving these aspects by means of automating the extraction of meaningful information from the accumulated images. These research studies can be subcategorized into two groups, namely, those using a single pre-calibrated fixed-location camera, and those using multiple cameras to determine the 3D coordinates of the surrounding surfaces. 2.1.1 Single fixed-location camera Using a single camera at fixed locations, it is not possible to determine the 3D coordinates of the structural elements (Rougier and Meunier 2010); however, some researchers have innovatively used time-lapsed images to determine the completion and production rate of certain construction activities. Lukins and Trucco (2007), Zhang et al. (2009) Ibrahim et al. (2009) aimed at determining the completion of certain activities on site by detecting the changes between consecutive images. Golparvar-Fard et al. (2009a) aimed at visually identifying the deviations between the plan and the actual states of the construction work by means of color-coding the identified differentiations. Ranaweera et al. (2013) developed a system to automatically determine the productivity of tunnel construction by identifying the number of liners lowered into a tunnel during a shift. 2.1.2 Multiple cameras As mentioned in the previous section, it is not possible to quantify the “scope of the work performed” (i.e. the progress) of construction activities using a single camera. In addition, since construction sites are dynamic environments with many moving objects, the presence of newly added obstacles may block the Line-Of-Sight (LOS) of the camera. Therefore, additional cameras are required to correctly determine the progress of construction activities. At least two camera exposures taken from different locations are required to estimate the 3D coordinates of a point relative to the image coordinate system using collinearity equations and triangulation (Kraus, 1993). In order to determine the position and orientation of each exposure station, at least three (3) tie-points (point correspondences) are required per image. Recent research studies aim at automatically matching similar features (tie-points) from unordered construction site photographs to generate the 3D coordinates of the surrounding surfaces (point clods). 312-2 Golparvar-Fard et al. (2009b, 2011a, 2011b, and 2015) developed an automated progress monitoring system, called, D4AR, which uses the accumulated unorganized photographs to determine the 3D coordinates of structural elements on site. The system uses a dense pixel to pixel matching algorithm to link the similar features between every two image to create a 3D point cloud of the site at the time the monitoring is performed. The system then estimates the translation and rotation of each of the camera exposures with an arbitrarily defined scale via a bundle adjustment. The scale of the measurement is then identified by manually registering some corresponding key points from the planned to the as-built space. At least three point correspondences are required to solve for the seven parameters (including scale; Horn, 1987). It is to emphasize that the manual registration is mandatory since the true 3D coordinates of the point clouds can only be estimated if the scale is defined with respect to the true space. However, the method assumes that the 3D coordinates of the selected key points remain unchanged between the as-built and the as-planned models. In other words, the potential errors in construction are neglected. Therefore, it is likely that the expected accuracy of the generated 3D coordinates drops since construction errors may cause differences between the as-planned and the as-built structure, which consequently affects the scale of the measured point clouds. In order to overcome this challenge, the correctly-scaled coordinates of the key points should ideally be measured by means of an external measurement system (such as a reflector-less total station or a scale bar), which increases the time and cost of data collection as well as possibility of interruption with construction activities. In (Golparvar-fard et al, 2011a, 2011b and 2015), the constructed point clouds from the D4AR system, proposed above, are assigned to their corresponding structural elements by superimposing the as-planned 4D BIM model to the constructed point clouds. As indicated above, at least two camera exposures are required to estimate the 3D coordinates of a point. Considering the limited Field of View (FOV) of a camera, a large number of manual photographs must be captured in order to cover every structural element on the site at least twice, which increases the time and cost of onsite data collection and analysis (Maalek et al. 2014). In addition, finding point-to-point correspondences between every image pair requires additional processing. In Golparvar-fard et al. (2011a), the processing of 288 images, constructing only 62,000 points, is shown to take approximately 7 hours. Furthermore, the quality of the images taken by a camera is a function of the lighting conditions and thus the accuracy of the generated point cloud can be highly affected by the lighting conditions (Golparvar-fard et al. 2011a). To summarize, the large number of images required due to limited FOV, the additional processing time due to the correspondence problem between the accumulated images and the need for adequate illumination (Dario et al. 2013) has led researchers to use LiDAR to help overcome the aforementioned limitations of implementing cameras on construction sites. 2.2 LiDAR LiDAR is a remote sensing technology used to collect 3D coordinates of the surrounding surfaces in LOS using only a single scan-station without additional processing (i.e. directly). In addition, LiDAR is more likely to achieve more accurate data compared to those provided by photogrammetry (Golparvar-Fard et al. 2011b and Bhatla et al. 2012). Therefore, the feasibility of preparation of as-built 3D/4D models using LiDAR technology has been of great interest to researchers in recent years. In the work of (Bosché, 2010, Bosché et al. 2009, 2015, and Turkan et al. 2013), the as-planned 4D CAD model is used to assign the accumulated point clouds to their corresponding structural elements. In their approach, first, the collected point clouds are manually registered to the planned model coordinate system. For each registered scan-station, the point clouds corresponding to the planned model are then generated by considering the potential random errors associated with the LiDAR equipment. If the distance between the two points is smaller than a threshold, the two points are equivalent. For the corresponding points, the Iterative Closest Point (ICP) registration is used to improve the results of the manual registration. In (Turkan et al. 2013), an earned value analysis on the data captured and processed in (Bosché, 2010) was performed. The point clouds were assigned to their corresponding structural elements; however, the “scope of the work performed” for each object was determined manually, which is a time consuming process, especially for in large scale construction projects. In (Bosché et al. 2015), the same procedure presented above was used to first generate the 3D as-planned point clouds on a piping project. A 3D Hough transform was then performed on both the as-planned and 312-3 as-built point clouds to determine the circular cross sections. These cross sections were then matched to identify the dimension compliance, location and the completion of the pipes. Zhang and Arditi (2013) also used the as-planned model in order to determine the completion of an object by counting the number of point clouds inside two predetermined boundaries, representing the tolerance region, of the object. Kim et al. (2013) also used the as-planned 4D Building Information Model (BIM) in order to report the progress of construction activities. In their approach the use of connectivity between components as well as sequence of activities were suggested in order to improve the classification results and to deal with misclassifications caused by missing data. 2.3 Some Limitations of Current State of Research Current object-based recognition models use the planned 4D model as a-priori knowledge to assign the collected 3D point clouds to a structural element (Golparvar-Fard et al. 2009a, 2011a, 2012, 2015 and Bosché et al. 2009, 2015; Bosché, 2010; and Zhang and Arditi 2013), which may not be reliable in cases where the locations of the as-built structure differ from those of the planned (Shahi et al. 2013) or the Issue for Construction (IFC) plan with sufficient detail is not readily available. In other words, the assumption that there are no significant deviations between the as-built and the as-planned states of a project is contradictory to the nature of monitoring and control. In order to reduce this dependency on the details of the planned model, here, it is proposed to summarize the information carried by the accumulated point clouds (regardless of the method the point clouds are generated) into meaningful information that is comparable to the details presented in the planned model (not vice versa). The procedure is explained in more detail in the following sections. 3 OBJECTIVE AND METHODOLOGY As mentioned previously, the goal of this research is to automatically summarize the accumulated point clouds into vertices that represent the boundaries of the structural elements, which can be used to determine the scope of the work performed. In other words, a novel method is proposed to automatically generate the 3D as-built model of structural elements. For this matter, the geometric primitives (only the 3D coordinates) are used to determine structural vertices from the collected point clouds. The procedure is as depicted in Figure 1. Each element of Figure 1 is explained in more detail in the following sections.   Figure 1: Point clouds summarization into meaningful vertices 3.1 Point Cloud Classification Point cloud classification is the process of labeling points with similar physical attributes into predefined classes. Since the most generic building elements as well as most man-made objects (Nunnally, 2010; Vosselman et al. 2004) are constructed from the intersection of planar surfaces, the classification of point clouds into planar surfaces is the major focus of this study. There are two methods commonly used to classify point clouds into planar surfaces, namely, the Hough transform and Principal Component Analysis (PCA). However, the use of Hough transformation for planar classification is computationally expensive and the results of the classification are highly affected by the presence of outlying data (Lari, 2014). Therefore, PCA-based classification is used in this study. PCA is used to summarize the variation of a multivariate data set into independent (orthogonal) axes. These axes are regarded as the principal components. The magnitudes of these axes represent the variation of the data set in the direction of the axis. This is accomplished by decomposing the covariance matrix of the data set into its eigenvalues and eigenvectors. In case of a 3-dimensional data set such as a point cloud, three orthogonal axes can be determined, which represent the maximum variations of the data set. For noise-free, coplanar points, the data has no variation in the direction of the surface normal. In other words, the eigenvalue corresponding to the direction of the surface normal is equal to zero. For Robust PCA Accumulated Point Clouds Classification Segmentation Surface Intersection Complete Linkage Convex Hull  312-4 system is a stand-alone method that can be used to assess the progress of construction activities even at times when the planned model is not available. The planned model is however, required to assess the deviations between the planned and the actual states of the project. This approach uses a novel robust PCA to first classify coplanar point clouds. The classified points are then segmented using the complete linkage clustering method. The boundary points of each cluster are determined to break discontinuous surfaces into smaller segments. The planar surfaces are then intersected to determine the vertices of the objects of desire. A novel method is also proposed to robustly extract the floor and flat slab ceilings without the need for the aforementioned approach, which can help reduce the calculation time significantly. A new method of automated registration and point to point correspondence search is also proposed. The two experiments show the effectiveness of the proposed system in automatically generating the 3D as-built model in both a highly occluded laboratory and an actual construction site environment. Acknowledgements The authors would like to acknowledge the National Science and Engineering Research Council (NSERC) for their funding of this research and the CANA construction Ltd for their collaborations.  References Bhatla, A., Choe, S. Y., Fierro, O., and Leite, F. (2012), “Evaluation of accuracy of as-built 3D modeling from photos taken by handheld digital cameras”, Automation in Construction 28, 116-127. Bosché, F. (2010), “Automated recognition of 3D CAD model objects in laser scans and calculation of as-built dimensions for dimensional compliance control in construction”, Adv. Eng. Informatics 24 (1), 107-118.  Bosché, F., Ahmed, M.,Turkan, Y., Haas, C. T., Haas, R. (2015), “The value of integrating Scan-to-BIM and Scan-vs-BIM techniques for construction monitoring using laser scanning and BIM: The case of cylindrical MEP components”, Automation in Construction 44,  212–226. Bosché, F., Haas, C. T., and Akinci, B. (2009), “Automated recognition of 3D CAD objects in site Laser scans for project 3D status visualization and performance control”, J. Comput. Civ. Eng. 23, 311-318. Brilakis, I. K., Soibelman, L., Shinagawa, Y. (2006), “Construction site image retrieval based on material cluster recognition”, Advanced Eng. Informatics 20(4), 443-452 Golparvar-Fard, M., Bohn, J., Teizer, J., Savarese, S., Peña-Mora, F. (2011a), “Evaluation of image-based modeling and laser scanning accuracy for emerging automated performance monitoring techniques”, Automation in Construction 20(8), 1143-1155.  Golparvar-Fard, M., Feniosky, P. M., and Savarese, S. (2009b), D4AR - A 4-dimentional augmented reality model for automating construction progress monitoring data collection, processing and communication, J. of Inf. Tech. in Constr. 14, 129-154. Golparvar-Fard, M., Feniosky, P. M., and Savarese, S. (2012), “Automated progress monitoring using unordered daily construction photographs and IFC-based building information models”, J. Comput. Civ. Eng.  Golparvar-Fard, M., Peña-Mora, F., and Savarese, S. (2010). “D4AR—4 dimensional augmented reality—Tools for automated remote progress tracking and support of decision-enabling tasks in the AEC/FM industry.”, Proc., 6th Int. Conf. on Innovations in AEC Special, Emerald, College Park, PA.  Golparvar-Fard, M., Peña-Mora, F., and Savarese, S. (2011b). ”Integrated Sequential As-Built and As-Planned Representation with D4AR Tools in Support of Decision-Making Tasks in the AEC/FM Industry.” J. Constr. Eng. Manage., 137(12), 1099–1116.  Golparvar-Fard, M., Peña-Mora, F., and Savarese, S. (2015). ”Automated Progress Monitoring Using Unordered Daily Construction Photographs and IFC-Based Building Information Models.” J. Comput. Civ. Eng., 29(1), 04014025. Golparvar-Fard, M., Peña-Mora, F., Arboleda, C. A., and Lee, S. H. (2009a), “Visualization of construction progress monitoring with 4D simulation model overlaid on time-lapsed photographs, J. Comput. Civ. Eng. 23, 391-404. Horn, B. K. (1987), “Closed-form solution of absolute orientation using unit quaternions”, Optical Society of America 4(4), 629-642 312-9 Hubert, M., Rousseeuw, P. J., and Verdonck, T. (2012), “A Deterministic algorithm for robust location and scatter”, Journal of Computational and Graphical Statistics 21(3), 618-637 Ibrahim, Y. M., Lukins, T. C., Zhang, X., Trucco, E., Kaka, A. P. (2009), “Towards automated progress assessment of workpackage components in construction projects using computer vision”, Adv. Eng. Informatics 23(1), 93-103 Jain, A. K., and Dubes, R. C. (1988), “Algorithms for Clustering Data”, Prentice-Hall advanced reference series. Prentice-Hall, Inc., Upper Saddle River, NJ. Johnson, R.A., and Wichern D. W. (2007), “Applied Multivariate Statistical Analysis”, 6th edition. Upper Saddle River, N.J., Pearson Prentice Hall. Kim, C., Son, H., and Kim, C. (2013). “Automated construction progress measurement using a 4D building information model and 3D data.” Autom. Constr., 31, 75–82. Kraus, K. (1993), “Photogrammetry, Volume I: Fundamentals and standard processes”, Dümmler, Bonn.  Lari, Z. (2014), “Adaptive Processing of Laser Scanning Data and Texturing of the Segmentation Outcome”, Ph.D. Thesis University of Calgary. Leica HDS 6100 (2009), “Product specification sheet”, http://archive.cyark.org/temp/LeicaHDS6100Datasheetus.pdf Lukins, T. C., and Trucco, E. (2007), “Towards automated visual assessment of progress in construction projects”, Proc. of the British Machine Conf., 18.1-18.10 Maalek, R., and Sadeghpour, F. (2011), “A Comparative study for automated tracking tools on construction sites”, CSCE, Ottawa.  Maalek, R., and Sadeghpour, F. (2012), “Reliability assessment of Ultra-Wide Band for indoor tracking of static resources on construction sites”, CSCE, Edmonton. Maalek, R., Ruwanpura, J., and Ranaweera, K. (2014), “Evaluation of the State-of-the-Art Automated Construction Progress Monitoring and Control Systems”, Construction Research Congress (CRC) Conference, Atlanta, Georgia, USA. McCullouch, B. (1997), “Automating field data collection on construction organizations”, Construction Congress V: Construction Congress V: Managing Engineered Construction in Expanding Global Markets, ASCE, 957-963. Milligan, G. W. and Isaac, P. D., (1980), “The validation of four ultrametric clustering algorithms”, Pattern Recognition, 12, 41–50. Nunnally, S. W., “Construction Methods and Management”, eighth ed., Pearson Education, USA, 2010, pp. 517-518. Ranaweera, K., Ruwanpura, J., and Fernando, S. (2013). ”Automated Real-Time Monitoring System to Measure Shift Production of Tunnel Construction Projects.” J. Comput. Civ. Eng., 27(1), 68–77. Rougier, C., and Meunier, J. (2010), “3D Head Trajectory using a single camera”, Proc. of the 4th ICISP, Berlin, 505-512 Saadat, M., and Cretin, L. (2002), “Measurement systems for large aerospace components”, Sensor Review, 22(3), 199-206. Sampath, A. and Shan, J., (2007), “Building boundary tracing and regularization from airborne LiDAR point clouds”, Photogrammetric Engineering and Remote Sensing, 73(7), 805–811. Shahi, A., West, J. S., Haas, C. T. (2013), “On site 3D marking for construction activity tracking”, Automation in Construction 30, 136-143. Shapira, L., Avidan, S., Shamir, A. (2009), “Mode-detection via median-shift”, In: IEEE Computer Vision and Pattern Recognition (CVPR), 1909–1916. Turkan, Y., Bosche, F., Haas, C. T., Haas, R. (2013), “Toward automated earned value tracking using 3d imaging tools”, J. Constr. Eng. Manage. 139, 423–433. Wang, J., Shan, J., (2009), “Segmentation of LiDAR point clouds for building extraction”, In: Proceedings of ASPRS Annual Conference. Presented at the ASPRS Annual Conference, Baltimore, Maryland, USA Zhang, C., Arditi, D. (2013), “Automated progress control using laser scanning technology”, Automation in Construction 36, 108-116. Zhang, X., Bakis, N., Lukins, T. C., Ibrahim, Y. M., Wu, S., Kagioglou, M., Aouad, G., Kaka, A. P., and Trucco, E. (2009), “Automating progress measurement of construction projects”, Automation in Construction 18(3), 294-301.  312-10  5th International/11th Construction Specialty Conference 5e International/11e Conférence spécialisée sur la construction    Vancouver, British Columbia June 8 to June 10, 2015 / 8 juin au 10 juin 2015   DEVELOPMENT OF AN AUTOMATED 3D/4D AS-BUILT MODEL GENERATION SYSTEM FOR CONSTRUCTION PROGRESS MONITORING AND QUALITY CONTROL Reza Maalek1,3 Derek Lichti2 and Janaka Ruwanpura1 1 Department of Civil Engineering, University of Calgary, Canada 2 Department of Geomatics Engineering, University of Calgary, Canada 3 rmaalek@ucalgary.ca Abstract: Automating the progress monitoring and control process is of great interest to industry practitioners to help improve the limitations associated with the current manual data collection and analysis practices. Two remote sensing technologies, namely, Light Detection and Ranging (LiDAR) and digital camera, are widely used to acquire 3D point clouds as a means of measuring the “scope of the work performed” of structural elements. However, to assign the collected 3D point clouds to their corresponding structural element, current object-based recognition models use the as-planned 4D model, which may not be reliable in cases where the locations of the as-built structure differ from those of the planned, and/or the planned 4D model is not available with sufficient detail. Here, a novel method is proposed to eliminate the dependency on the as-planned data by automatically generating the 3D/4D as-built model through a robust Principal Component Analysis-based (PCA) segmentation algorithm. The proposed system is also independent of the technology used to capture the 3D point clouds. To evaluate the reliability of the proposed automated as-built model generation procedure, two sets of LiDAR data from the "Mechanics of Materials" laboratory and the "Graduate Student Hall of Residence" construction site at the University of Calgary were collected. A novel method of automated registration of the as-built model to the planned model coordinate system is also proposed through which the compliance of the planned vs. actual dimensions of corresponding structural elements are examined. The results of the two experiments demonstrate the applicability of the proposed methods for the automatic generation of the 3D/4D as-built model and the dimension compliance control of structural elements. 1 INTRODUCTION Project monitoring and control are vital to facilitate decision makers identify deviations between the planned vs. as-built states of the project and take timely measures where required (Maalek and Sadeghpour, 2012). Monitoring is the process of collecting onsite data as a means of measuring the Project Performance Indicators (PPI). Traditionally, onsite data are collected manually, a time consuming, error-prone and labor intensive task particularly on large scale projects (Golparvar-Fard et al. 2009a). In practice, to justify the time and cost associated with such manual approaches, a limited amount (and/or frequency) of onsite data are collected, which diminishes the ability of the project manager to identify the causes of delays and cost overruns on time. In addition, most onsite data collection processes are anecdotal without a proper monitoring plan/strategy (Golparvar-Fard et al. 2011), which influences the time, cost and reliability of the collected PPIs. 312-1 Project control involves the processing of the accumulated data for the determination of the performance of the current state of the project. Therefore, the reliability of the determination of the performance of the project is highly dependent on the strategy as well as the accuracy of the collected data during the monitoring process (Saadat and Cretin, 2002). Currently, site supervisory personnel spend 30-50% of their time on manually monitoring and controlling onsite data (McCullouch, 1997, Golparvar-Fard et al. 2009). If this time is reduced by means of a novel approach to onsite data collection and analysis, more time can be allocated to improving vital construction related concerns such as safety (Maalek & Sadeghpour, 2011), and workforce productivity and communications (Choy and Ruwanpura, 2007). Therefore, automating the monitoring and control process is proposed in recent years to help overcome the aforementioned limitations of current manual practices. 2 LITERATURE REVIEW The percentage of completion of an activity is suggested as the Key Performance Indicator (KPI) capable of providing progress information in activity-level (Maalek et al. 2014). In order to automatically extract this metric, the scope of the work performed for each activity is required to be identified by means of a remote sensing technology. Currently, two remote sensing technologies, namely, digital camera and LiDAR, are widely used to generate 3D coordinates of the surrounding surfaces. The overview of the previous research related to the application of these two technologies for progress monitoring of construction activities are presented in the following. 2.1 Digital Camera In practice, photographs are commonly taken to record the progress of specific activities and/or to help minimize disputes/claims (Golparvar-Fard et al. 2009a). These images are stored without proper documentation and indexing (Brilakis et al. 2006). In addition, the KPIs are manually extracted from the large number of unordered/randomized images, constituting a costly, lengthy and challenging procedure. Current research aims at improving these aspects by means of automating the extraction of meaningful information from the accumulated images. These research studies can be subcategorized into two groups, namely, those using a single pre-calibrated fixed-location camera, and those using multiple cameras to determine the 3D coordinates of the surrounding surfaces. 2.1.1 Single fixed-location camera Using a single camera at fixed locations, it is not possible to determine the 3D coordinates of the structural elements (Rougier and Meunier 2010); however, some researchers have innovatively used time-lapsed images to determine the completion and production rate of certain construction activities. Lukins and Trucco (2007), Zhang et al. (2009) Ibrahim et al. (2009) aimed at determining the completion of certain activities on site by detecting the changes between consecutive images. Golparvar-Fard et al. (2009a) aimed at visually identifying the deviations between the plan and the actual states of the construction work by means of color-coding the identified differentiations. Ranaweera et al. (2013) developed a system to automatically determine the productivity of tunnel construction by identifying the number of liners lowered into a tunnel during a shift. 2.1.2 Multiple cameras As mentioned in the previous section, it is not possible to quantify the “scope of the work performed” (i.e. the progress) of construction activities using a single camera. In addition, since construction sites are dynamic environments with many moving objects, the presence of newly added obstacles may block the Line-Of-Sight (LOS) of the camera. Therefore, additional cameras are required to correctly determine the progress of construction activities. At least two camera exposures taken from different locations are required to estimate the 3D coordinates of a point relative to the image coordinate system using collinearity equations and triangulation (Kraus, 1993). In order to determine the position and orientation of each exposure station, at least three (3) tie-points (point correspondences) are required per image. Recent research studies aim at automatically matching similar features (tie-points) from unordered construction site photographs to generate the 3D coordinates of the surrounding surfaces (point clods). 312-2 Golparvar-Fard et al. (2009b, 2011a, 2011b, and 2015) developed an automated progress monitoring system, called, D4AR, which uses the accumulated unorganized photographs to determine the 3D coordinates of structural elements on site. The system uses a dense pixel to pixel matching algorithm to link the similar features between every two image to create a 3D point cloud of the site at the time the monitoring is performed. The system then estimates the translation and rotation of each of the camera exposures with an arbitrarily defined scale via a bundle adjustment. The scale of the measurement is then identified by manually registering some corresponding key points from the planned to the as-built space. At least three point correspondences are required to solve for the seven parameters (including scale; Horn, 1987). It is to emphasize that the manual registration is mandatory since the true 3D coordinates of the point clouds can only be estimated if the scale is defined with respect to the true space. However, the method assumes that the 3D coordinates of the selected key points remain unchanged between the as-built and the as-planned models. In other words, the potential errors in construction are neglected. Therefore, it is likely that the expected accuracy of the generated 3D coordinates drops since construction errors may cause differences between the as-planned and the as-built structure, which consequently affects the scale of the measured point clouds. In order to overcome this challenge, the correctly-scaled coordinates of the key points should ideally be measured by means of an external measurement system (such as a reflector-less total station or a scale bar), which increases the time and cost of data collection as well as possibility of interruption with construction activities. In (Golparvar-fard et al, 2011a, 2011b and 2015), the constructed point clouds from the D4AR system, proposed above, are assigned to their corresponding structural elements by superimposing the as-planned 4D BIM model to the constructed point clouds. As indicated above, at least two camera exposures are required to estimate the 3D coordinates of a point. Considering the limited Field of View (FOV) of a camera, a large number of manual photographs must be captured in order to cover every structural element on the site at least twice, which increases the time and cost of onsite data collection and analysis (Maalek et al. 2014). In addition, finding point-to-point correspondences between every image pair requires additional processing. In Golparvar-fard et al. (2011a), the processing of 288 images, constructing only 62,000 points, is shown to take approximately 7 hours. Furthermore, the quality of the images taken by a camera is a function of the lighting conditions and thus the accuracy of the generated point cloud can be highly affected by the lighting conditions (Golparvar-fard et al. 2011a). To summarize, the large number of images required due to limited FOV, the additional processing time due to the correspondence problem between the accumulated images and the need for adequate illumination (Dario et al. 2013) has led researchers to use LiDAR to help overcome the aforementioned limitations of implementing cameras on construction sites. 2.2 LiDAR LiDAR is a remote sensing technology used to collect 3D coordinates of the surrounding surfaces in LOS using only a single scan-station without additional processing (i.e. directly). In addition, LiDAR is more likely to achieve more accurate data compared to those provided by photogrammetry (Golparvar-Fard et al. 2011b and Bhatla et al. 2012). Therefore, the feasibility of preparation of as-built 3D/4D models using LiDAR technology has been of great interest to researchers in recent years. In the work of (Bosché, 2010, Bosché et al. 2009, 2015, and Turkan et al. 2013), the as-planned 4D CAD model is used to assign the accumulated point clouds to their corresponding structural elements. In their approach, first, the collected point clouds are manually registered to the planned model coordinate system. For each registered scan-station, the point clouds corresponding to the planned model are then generated by considering the potential random errors associated with the LiDAR equipment. If the distance between the two points is smaller than a threshold, the two points are equivalent. For the corresponding points, the Iterative Closest Point (ICP) registration is used to improve the results of the manual registration. In (Turkan et al. 2013), an earned value analysis on the data captured and processed in (Bosché, 2010) was performed. The point clouds were assigned to their corresponding structural elements; however, the “scope of the work performed” for each object was determined manually, which is a time consuming process, especially for in large scale construction projects. In (Bosché et al. 2015), the same procedure presented above was used to first generate the 3D as-planned point clouds on a piping project. A 3D Hough transform was then performed on both the as-planned and 312-3 as-built point clouds to determine the circular cross sections. These cross sections were then matched to identify the dimension compliance, location and the completion of the pipes. Zhang and Arditi (2013) also used the as-planned model in order to determine the completion of an object by counting the number of point clouds inside two predetermined boundaries, representing the tolerance region, of the object. Kim et al. (2013) also used the as-planned 4D Building Information Model (BIM) in order to report the progress of construction activities. In their approach the use of connectivity between components as well as sequence of activities were suggested in order to improve the classification results and to deal with misclassifications caused by missing data. 2.3 Some Limitations of Current State of Research Current object-based recognition models use the planned 4D model as a-priori knowledge to assign the collected 3D point clouds to a structural element (Golparvar-Fard et al. 2009a, 2011a, 2012, 2015 and Bosché et al. 2009, 2015; Bosché, 2010; and Zhang and Arditi 2013), which may not be reliable in cases where the locations of the as-built structure differ from those of the planned (Shahi et al. 2013) or the Issue for Construction (IFC) plan with sufficient detail is not readily available. In other words, the assumption that there are no significant deviations between the as-built and the as-planned states of a project is contradictory to the nature of monitoring and control. In order to reduce this dependency on the details of the planned model, here, it is proposed to summarize the information carried by the accumulated point clouds (regardless of the method the point clouds are generated) into meaningful information that is comparable to the details presented in the planned model (not vice versa). The procedure is explained in more detail in the following sections. 3 OBJECTIVE AND METHODOLOGY As mentioned previously, the goal of this research is to automatically summarize the accumulated point clouds into vertices that represent the boundaries of the structural elements, which can be used to determine the scope of the work performed. In other words, a novel method is proposed to automatically generate the 3D as-built model of structural elements. For this matter, the geometric primitives (only the 3D coordinates) are used to determine structural vertices from the collected point clouds. The procedure is as depicted in Figure 1. Each element of Figure 1 is explained in more detail in the following sections.   Figure 1: Point clouds summarization into meaningful vertices 3.1 Point Cloud Classification Point cloud classification is the process of labeling points with similar physical attributes into predefined classes. Since the most generic building elements as well as most man-made objects (Nunnally, 2010; Vosselman et al. 2004) are constructed from the intersection of planar surfaces, the classification of point clouds into planar surfaces is the major focus of this study. There are two methods commonly used to classify point clouds into planar surfaces, namely, the Hough transform and Principal Component Analysis (PCA). However, the use of Hough transformation for planar classification is computationally expensive and the results of the classification are highly affected by the presence of outlying data (Lari, 2014). Therefore, PCA-based classification is used in this study. PCA is used to summarize the variation of a multivariate data set into independent (orthogonal) axes. These axes are regarded as the principal components. The magnitudes of these axes represent the variation of the data set in the direction of the axis. This is accomplished by decomposing the covariance matrix of the data set into its eigenvalues and eigenvectors. In case of a 3-dimensional data set such as a point cloud, three orthogonal axes can be determined, which represent the maximum variations of the data set. For noise-free, coplanar points, the data has no variation in the direction of the surface normal. In other words, the eigenvalue corresponding to the direction of the surface normal is equal to zero. For Robust PCA Accumulated Point Clouds Classification Segmentation Surface Intersection Complete Linkage Convex Hull  312-4 system is a stand-alone method that can be used to assess the progress of construction activities even at times when the planned model is not available. The planned model is however, required to assess the deviations between the planned and the actual states of the project. This approach uses a novel robust PCA to first classify coplanar point clouds. The classified points are then segmented using the complete linkage clustering method. The boundary points of each cluster are determined to break discontinuous surfaces into smaller segments. The planar surfaces are then intersected to determine the vertices of the objects of desire. A novel method is also proposed to robustly extract the floor and flat slab ceilings without the need for the aforementioned approach, which can help reduce the calculation time significantly. A new method of automated registration and point to point correspondence search is also proposed. The two experiments show the effectiveness of the proposed system in automatically generating the 3D as-built model in both a highly occluded laboratory and an actual construction site environment. Acknowledgements The authors would like to acknowledge the National Science and Engineering Research Council (NSERC) for their funding of this research and the CANA construction Ltd for their collaborations.  References Bhatla, A., Choe, S. Y., Fierro, O., and Leite, F. (2012), “Evaluation of accuracy of as-built 3D modeling from photos taken by handheld digital cameras”, Automation in Construction 28, 116-127. Bosché, F. (2010), “Automated recognition of 3D CAD model objects in laser scans and calculation of as-built dimensions for dimensional compliance control in construction”, Adv. Eng. Informatics 24 (1), 107-118.  Bosché, F., Ahmed, M.,Turkan, Y., Haas, C. T., Haas, R. (2015), “The value of integrating Scan-to-BIM and Scan-vs-BIM techniques for construction monitoring using laser scanning and BIM: The case of cylindrical MEP components”, Automation in Construction 44,  212–226. Bosché, F., Haas, C. T., and Akinci, B. (2009), “Automated recognition of 3D CAD objects in site Laser scans for project 3D status visualization and performance control”, J. Comput. Civ. Eng. 23, 311-318. Brilakis, I. K., Soibelman, L., Shinagawa, Y. (2006), “Construction site image retrieval based on material cluster recognition”, Advanced Eng. Informatics 20(4), 443-452 Golparvar-Fard, M., Bohn, J., Teizer, J., Savarese, S., Peña-Mora, F. (2011a), “Evaluation of image-based modeling and laser scanning accuracy for emerging automated performance monitoring techniques”, Automation in Construction 20(8), 1143-1155.  Golparvar-Fard, M., Feniosky, P. M., and Savarese, S. (2009b), D4AR - A 4-dimentional augmented reality model for automating construction progress monitoring data collection, processing and communication, J. of Inf. Tech. in Constr. 14, 129-154. Golparvar-Fard, M., Feniosky, P. M., and Savarese, S. (2012), “Automated progress monitoring using unordered daily construction photographs and IFC-based building information models”, J. Comput. Civ. Eng.  Golparvar-Fard, M., Peña-Mora, F., and Savarese, S. (2010). “D4AR—4 dimensional augmented reality—Tools for automated remote progress tracking and support of decision-enabling tasks in the AEC/FM industry.”, Proc., 6th Int. Conf. on Innovations in AEC Special, Emerald, College Park, PA.  Golparvar-Fard, M., Peña-Mora, F., and Savarese, S. (2011b). ”Integrated Sequential As-Built and As-Planned Representation with D4AR Tools in Support of Decision-Making Tasks in the AEC/FM Industry.” J. Constr. Eng. Manage., 137(12), 1099–1116.  Golparvar-Fard, M., Peña-Mora, F., and Savarese, S. (2015). ”Automated Progress Monitoring Using Unordered Daily Construction Photographs and IFC-Based Building Information Models.” J. Comput. Civ. Eng., 29(1), 04014025. Golparvar-Fard, M., Peña-Mora, F., Arboleda, C. A., and Lee, S. H. (2009a), “Visualization of construction progress monitoring with 4D simulation model overlaid on time-lapsed photographs, J. Comput. Civ. Eng. 23, 391-404. Horn, B. K. (1987), “Closed-form solution of absolute orientation using unit quaternions”, Optical Society of America 4(4), 629-642 312-9 Hubert, M., Rousseeuw, P. J., and Verdonck, T. (2012), “A Deterministic algorithm for robust location and scatter”, Journal of Computational and Graphical Statistics 21(3), 618-637 Ibrahim, Y. M., Lukins, T. C., Zhang, X., Trucco, E., Kaka, A. P. (2009), “Towards automated progress assessment of workpackage components in construction projects using computer vision”, Adv. Eng. Informatics 23(1), 93-103 Jain, A. K., and Dubes, R. C. (1988), “Algorithms for Clustering Data”, Prentice-Hall advanced reference series. Prentice-Hall, Inc., Upper Saddle River, NJ. Johnson, R.A., and Wichern D. W. (2007), “Applied Multivariate Statistical Analysis”, 6th edition. Upper Saddle River, N.J., Pearson Prentice Hall. Kim, C., Son, H., and Kim, C. (2013). “Automated construction progress measurement using a 4D building information model and 3D data.” Autom. Constr., 31, 75–82. Kraus, K. (1993), “Photogrammetry, Volume I: Fundamentals and standard processes”, Dümmler, Bonn.  Lari, Z. (2014), “Adaptive Processing of Laser Scanning Data and Texturing of the Segmentation Outcome”, Ph.D. Thesis University of Calgary. Leica HDS 6100 (2009), “Product specification sheet”, http://archive.cyark.org/temp/LeicaHDS6100Datasheetus.pdf Lukins, T. C., and Trucco, E. (2007), “Towards automated visual assessment of progress in construction projects”, Proc. of the British Machine Conf., 18.1-18.10 Maalek, R., and Sadeghpour, F. (2011), “A Comparative study for automated tracking tools on construction sites”, CSCE, Ottawa.  Maalek, R., and Sadeghpour, F. (2012), “Reliability assessment of Ultra-Wide Band for indoor tracking of static resources on construction sites”, CSCE, Edmonton. Maalek, R., Ruwanpura, J., and Ranaweera, K. (2014), “Evaluation of the State-of-the-Art Automated Construction Progress Monitoring and Control Systems”, Construction Research Congress (CRC) Conference, Atlanta, Georgia, USA. McCullouch, B. (1997), “Automating field data collection on construction organizations”, Construction Congress V: Construction Congress V: Managing Engineered Construction in Expanding Global Markets, ASCE, 957-963. Milligan, G. W. and Isaac, P. D., (1980), “The validation of four ultrametric clustering algorithms”, Pattern Recognition, 12, 41–50. Nunnally, S. W., “Construction Methods and Management”, eighth ed., Pearson Education, USA, 2010, pp. 517-518. Ranaweera, K., Ruwanpura, J., and Fernando, S. (2013). ”Automated Real-Time Monitoring System to Measure Shift Production of Tunnel Construction Projects.” J. Comput. Civ. Eng., 27(1), 68–77. Rougier, C., and Meunier, J. (2010), “3D Head Trajectory using a single camera”, Proc. of the 4th ICISP, Berlin, 505-512 Saadat, M., and Cretin, L. (2002), “Measurement systems for large aerospace components”, Sensor Review, 22(3), 199-206. Sampath, A. and Shan, J., (2007), “Building boundary tracing and regularization from airborne LiDAR point clouds”, Photogrammetric Engineering and Remote Sensing, 73(7), 805–811. Shahi, A., West, J. S., Haas, C. T. (2013), “On site 3D marking for construction activity tracking”, Automation in Construction 30, 136-143. Shapira, L., Avidan, S., Shamir, A. (2009), “Mode-detection via median-shift”, In: IEEE Computer Vision and Pattern Recognition (CVPR), 1909–1916. Turkan, Y., Bosche, F., Haas, C. T., Haas, R. (2013), “Toward automated earned value tracking using 3d imaging tools”, J. Constr. Eng. Manage. 139, 423–433. Wang, J., Shan, J., (2009), “Segmentation of LiDAR point clouds for building extraction”, In: Proceedings of ASPRS Annual Conference. Presented at the ASPRS Annual Conference, Baltimore, Maryland, USA Zhang, C., Arditi, D. (2013), “Automated progress control using laser scanning technology”, Automation in Construction 36, 108-116. Zhang, X., Bakis, N., Lukins, T. C., Ibrahim, Y. M., Wu, S., Kagioglou, M., Aouad, G., Kaka, A. P., and Trucco, E. (2009), “Automating progress measurement of construction projects”, Automation in Construction 18(3), 294-301.  312-10  Development of an Automated Monitoring and Control System for Construction Sites Prepared by: Reza Maalek Supervisor(s):  Prof. Janaka Ruwanpura  & Prof. Derek Lichti 1 June 2015 Problem Statement •  Current Monitoring Practices are mostly Manual [1] :  –  Time-consuming, costly and error-prone –  Limited onsite data collection •  Untimely identification of causes of delays and cost overruns [2] •  Site supervisors spend 30-50% on analysing the data [3 & 4] Plan Monitor Control Assess the Performance Revised Plan 2 Identification of Deviations •  Safety à Risks of most common site accidents are not decreased [5] •  Productivity & Communicationà 40-60% of tool time is wasted [6] TQM Research Objective •  To Automate the Monitoring and Control Process –  Automated Monitoring to determine: 3 Overview of LiDAR 4 Research Method 5 As-built Model As-built vs. Planned As-built vs. Planned On Behind Ahead Rework 3 Main Questions 6 Planned Model As-built Model Deviations? As-built Model Automated As-built Model Generation 7 Robust PCA Robust Complete Linkage Convex Hull As-built Model 2. As-built Model Generation a.  Point Cloud Classification b.  Point Cloud Segmentation c.  Boundary Detection d.  Data Summarization and Intersection 8 2a. Point Cloud Classification: Planes   9 No variation in the direction of the surface normal Smallest Eigen value is zero Perform the PCA For a Planar Surface 2a. Point Cloud Classification: Lines   10 100% of the data in the maximum eigenvalue Perform the PCA Majority of the variation in the direction of the longitude Reality 11 Due to data artifacts caused by:  1.  Occlusions 2.  Moving objects 3.  Dust  Outliers are present in the data  Classical PCA is very sensitive to outliers [6, 7, 8, 9] à  Searching for a Covariance estimation Robust to outliers •  Minimum Covariance Determinant (MCD) [8, 9, 10] Mixed Pixel 12 Classical PCA 0-60% Robust PCA 47-93% 2b. Point Cloud Segmentation 13 Plane parameters Complete Linkage 14 Within the Attribute Space: No prior knowledge of the number of clusters 2c. Discontinuous Surfaces 15 Use the “Modified Convex Hull” algorithm Boundary Point extraction Modified Convex Hull algorithm: 16 2d. Vertices and Intersection 17 Intersect Segments with Closest Boundaries Experiment 1: Laboratory Testing the proposed method in a highly occluded area: •  A set of LiDAR data was collected using Leica HDS6100   18 13.5 m 9.4 m 30 Million Points Removing Floor and Flat Slab Ceiling •  To improve Calculation time:  19 Distribution Elevation Points on floor Points on ceiling Pf Pc rf rc Median-Shift rf rc Robust PCA using MCD Results 20 Elevation (m) Number of Points Floor Precision = 91.5% Ceiling Precision = 92.5% Recall=100% Accounting for more than half of the points Experiment 1: Robust Segmentation 21 95% of the points correctly segmented 30% improvement to current available method Experiment 1: As-built Model 22 As-built X Y Registration? 23 X Y As-built Model As-planned Model 2 Metrics for every point 24 Cluster Determine 3 points or 2 lines Rigid Body Transformation As-built As-planned Experiment 1: Results 25 Accuracy Assessment:  XRSE = 7cm YRSE = 6cm ZRSE = 1.2cm MRSE = 9.4cm Dimension Compliance Control of Walls:  Horizontal Direction = 7.5cm Vertical Direction = 2.4cm As-built vs. Total-station MRSE = 0.7 cm 9.4 cm is the errors during construction Experiment 2: Construction Site 26 150 million points from 4 scan-stations Graduate Student Hall of Residence Project: Experiment 2: Robust Planar Segmentation 27 Planar Segmentation:  95.2% Accuracy  Using Region Growing and Classical PCA Over-segmentation Planar Segmentation:  73% Accuracy  Using Our Robust Method Experiment 2: Robust Linear Segmentation 28 Re-bars of Elevator Shaft Linear Segmentation:  91.4% Accuracy  Inclined Inclined Experiment 2: As-built modelling of Elevator shaft 29 As-built Dimension Compliance: ±1.4 cm in X ±1.6 cm in Y  Contractor suggested tolerance of ±2 cm  X Y Future Work: Progress Monitoring 30 Graduate Student Hall of Residence Project: Future Work: Dimension Compliance 31 Removing the Bracing Settlement of the Cantilever Taylor Institute of Teaching and Learning Project: Cantilever Truss Contributions •  Fully Automated Monitoring and Control Process •  Automated As-built Model Generation Process •  Novel Robust Planar and Linear Classification Method •  Novel Robust Planar and Linear Segmentation Method Construction Industry benefits: •  Reducing monitoring time, cost and quality •  Reducing rework due to poor quality •  Reducing construction claims and disputes •  Quality assurance, and dimension compliance control 32 Thank You for Your Attention     Questions? 33 References 1.  Maalek, R. and Sadeghpour, F. (2011), “A Comparative Overview of Radio Frequency-Based Technologies Applicable to Locating Resources on Construction Sites”. 2.  Semple, C., Hartman, F. T., and Jergeas, G. (1994), “Construction claims and disputes: causes and cost/time overruns”. 3.  McCullouch, B., (1997), “Automating field data collection on construction organizations”. 4.  Golparvar-Fard, M., Feniosky, P. M., and Savarese, S. (2009), “D4AR - A 4-dimentional augmented reality model for automating construction progress monitoring data collection, processing and communication”. 5.  Maalek, R. and Sadeghpour, F. (2014), “Accuracy Assessment of Dynamic Resources on Construction Sites”, Automation in Construction (in-press) 6.  Serneels, S., and Verdonck, T. (2008), “Principal Component Analysis for data containing outliers and missing elements”. 7.  Stanimirova, I., Daszykowski, M., and Walczak, B. (2007), “Dealing with missing values and outliers in principal component analysis”. 8.  Hubert, M., Rousseeuw, P. J., and Verdonck, T. (2012), “Deterministic algorithm for MCD”. 9.  Nurunnabi, A., Belton, D., and West, G. (2012), “Robust segmentation for multiple planar surface extraction in Laser scanning 3D point cloud data”. 10.  Rousseeuw, P. J., and Driessen, K. N. (1999), “A fast algorithm for the Minimum Covariance Determinant Estimator” 34 

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.52660.1-0076418/manifest

Comment

Related Items