The goal of this study was to develop a decision rule heuristic that would incorporate theories of forest stand dynamics and crown competition into an automatic crown detection and crown size search algorithm. Specifically, we sought to develop new multi-dimensional template matching methods fused with knowledge-based decision rules that model spatial considerations for crown competition and develop multi-scale assessment criteria for appraisal of crown detection and crown size at the stand, local neighbourhood and individual crown levels. The decision rule approach to crown detection and crown size was tested on a mature mixed coniferous and deciduous forest typical of southern New England, USA. Multi-dimensional template matching was applied to a high resolution (30 cm per pixel side) colour infrared image of the study site. The decision rule heuristic developed for this study effectively reduced 2626 potential crown detections to 568 crowns, producing a 91% rate of crown detection when compared with the 516 field-measured crowns. The automatically derived crown size class distribution was shown to be statistically similar to the distribution of field crown size classes using the Kolomogorov-Smirnov statistic. Finally, a fuzzification of classification assignments to crown classes either one above or below actual crown size class resulted in an 80% match between individual field crowns and remotely sensed crowns within a 6 m spatial lag.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
No Article Media
Document Type: Research Article
US Coast Guard Research and Development Center, New London, CT, USA
School of Forestry and Environmental Studies, Yale University, New Haven, CT, USA
Metropolitan District Commission, Department of Forestry, Hartford, CT, USA
Publication date: April 1, 2010
More about this publication?