Open Collections

Determining stand ages in a hyperspectral image using artificial neural networks

University of British Columbia

An Airborne Visible / Infrared Imaging Spectrometer (AVIRIS) image was analysed to assess the feasibility of using feed-forward artificial neural networks to determine stand ages in the Sooke watershed located in southern Vancouver Island. Training data were systematically collected at the pixel level from photo interpreted forest cover data of the site. Artificial neural networks having a variety of architectures were trained using these data in conjunction with several different learning parameters. Once the neural networks were trained, all pixels in the area of interest were analysed and assigned an age. These pixel-level ages were used to extrapolate the ages of the forest cover polygons, and the ages were compared to the photointerpreted polygon ages. Results showed that a good correspondence existed, especially for larger polygons. However, accurate results could be obtained using a subset of the bands and data simulating Landsat TM bands 2 through 5 and 7. The use of topographic and view angle data in addition to the spectral data produced equivalent or slightly poorer results than the spectral data on its own. The use of a first difference image did not improve the prediction accuracy. The results of this study suggest that the technique used may be a reasonable means of determining the ages of forest stands, and may be a good choice when ground data on stand ages are available but collecting and interpreting aerial photographs is prohibitively expensive. This will especially be the case when data from satellite-mounted hyperspectral sensors become available in late 2000.

Thesis/Dissertation

application/pdf

2009-07-09

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/10567

Forestry

University of British Columbia

UBCV

DSpace