Use of forest structure to improve classification

This paper deals with forest classification in tropical and subtropical areas using multi-sources data fusion. Topological, environmental, structural and visual information are used to classify the samples. This study improves a previous classification by introducing airborne LiDAR information through the computation of the Digital Vegetation Elevation Model (DVEM). This kind of information is the first structural characteristic of the forest computed over the whole territory at the meter scale. The classification is based on decision trees and allows a significant improvement of the previous classification especially over the transition areas which are reduced and more precisely localized.