Object-based Classification using UltraCam-D Images for Forest Tree Identification (Case study: Hyrcanian Forest of Iran)

摘要

This research has been conducted to evaluate the object-based method with high spatial resolution airborne remote sensing data in tree species identification and mapping. It has been done in 2 areas; include the damaged natural broadleaved forest and the broadleaved and coniferous mixed forestation in the Northern forests of Iran. After pre-processing the imagery a trial and error method was employed to reach the ideal segmentation results. Subsequent to class definition, sample objects were selected as representative of defined classes and NN classifier was accomplished using integration of a broad spectrum of different object features. Accuracy assessment of the produced maps, comparing with field reference data shows the overall accuracies and Kappa statistics of 0.79, 0.61 (Areal) and 0.76, 0.69 (Area2) respectively. Achieved relatively low accuracies demonstrated that the standalone optical remote sensing methods are insufficient for tree species discrimination of such complex forest structures.