Ramazan Bülbül, Stefan Reder and Jan-Peter Mund compared three raster-based and two point-cloud-based algorithms, that were developed for the segmentation of LiDAR point clouds with the aim of individual trees detection and segmentation in data products captured with a WLS using SLAM technology.

The performance test showed  the limits of raster-based segmentation approaches for WLS point clouds, especially in stands with high density. The same shortcomings were observed for the point-cloud-based region growth algorithm which is mainly depended on visible treetops. The ASM3D, developed for LiDAR point clouds, showed  the best performance and was able to detect most trees (>85%) up to 850 trees/ha and is only falling off slightly afterwards (1250 trees/ha; 75% detection rate). The proof-of-concept study was conducted as part of the WINMOL-Project (https://winmol.thuenen.de/) funded by the Fachagentur Nachwachsende Rohstoffe (FNR) and the Federal Ministry of Food and Agriculture, Germany (BMEL) (FNR funding number: 2220NR024A). The research was presented at the SilviLaser 2021 conference in Vienna and the full paper can be downloaded.