This thesis assessed object-based image analysis techniques using multi-source remote sensing data in order to automatically delineate canopy gaps and individual tree crowns (ITCs). Image segmentation is much more complex when conducted on data covering deciduous, un-even aged forests like those in Central Ontario. To delineate canopy gaps high spatial resolution multispectral ADS40 aerial imagery and a LiDAR CHM were assessed both separately and jointly. To delineate ITCs two commonly used segmentation approaches were assessed – region growing and watershed segmentation. Ground based measurements and manually delineated data were used as reference to evaluate results. Using multi-source data to delineate canopy gaps produced an average overall accuracy of 99.35%, whereas using the imagery and CHM individually resulted in average overall accuracies of 81.41% and 82.45%, respectively. For the delineation of ITCs, the watershed and region growing segmentations resulted in average overall accuracies of 67.5% and 65.5%, respectively.
| Identifer | oai:union.ndltd.org:TORONTO/oai:tspace.library.utoronto.ca:1807/65603 |
| Date | 04 July 2014 |
| Creators | Saliola, Assunta |
| Contributors | He, Yuhong, Caspersen, John |
| Source Sets | University of Toronto |
| Language | en_ca |
| Detected Language | English |
| Type | Thesis |
Page generated in 0.002 seconds