TREETOP DETECTION IN RIVERINE BAMBOO FORESTS USING TWO TYPES OF DSMS

Abstract

 Treetop detection was conducted in a dense bamboo forest using Local Maximum Filtering (LMF) with digital surface models (DSMs), which were generated using airborne laser scanning data and UAV images processed with Structure-from-Motion/Multi-View-Stereo (UAV-SfM). Subsequently, LMF was applied to both DSMs and the optimal Window Size (WS) parameter for LMF was examined. The results showed that the optimal WS was smaller for the UAV-derived DSM (DSM-U) than for the ALS-derived DSM (DSM-A). This difference was attributed to smoother elevation roughness in the DSM-U than in the DSM-A. These results suggest that the types of DSMs generated by ALS or UAV-SfM data affect the application results of treetop detection. Therefore, it is important that the WS should be set considering the characteristics of the DSMs.