3D segmentation of forest structure using a mean-shift based algorithm

Consistent and accurate information on 3D forest canopy structure is required by many applications like forest inventory, management, logging, fuel mapping, habitat studies or biomass estimate. Compared to other remote sensing techniques (e.g., SAR or photogrammetry), airborne laser scanning is an adapted tool to provide such information by generating a three-dimensional georeferenced point cloud. Vertical structure analysis consists in detecting the number of layers within a forest stand and their limits. Until now, there is no approach that properly segments the different strata of a forest. In this study, we directly work on the 3D point cloud and we propose a mean shift (MS) based procedure for vertical forest segmentation. The approach that is carried out on complex forest plots improves the discrimination of vegetation strata.