Detection of Tree Species in Mixed Broad-Leaved Stands of Caspian Forests Using UAV Images (Case study: Darabkola Forest)

Unmanned aerial vehicles (UAVs) images have high spatial resolution. They are a valuable source of information for mapping land cover and thematic information, particularly in the identification of tree species. The aim of this study was to investigate the capability of drone images and the base object method for detecting tree species in the Hyrcanian forests. For this purpose, part of an area in parcel 24 of district one in Mazandaran Darabkola forest was selected. The ground truth map was prepared through accurate recording with geographic coordinate’s algorithm using distance and azimuth in MATLAB software. Proper processing was done on the images and classification performed on images at three flight height; 55, 75 and 100 meters in two categories of one-step and hierarchical classifications. In object-based classification, the nearest neighbor method was used to classify three species. The accuracy of the maps derived from classifications was evaluated using 50% of the ground truth map. The results showed that the map of the hierarchical classification by the object based method at a flight height of 55 meters has the best ability to detect tree species in the three heights. These comparisons showed Kappa's coefficient of 0.81 accuracy of tree species classification in 55-meter height by UAV.