Tree segmentation of LiDAR point clouds using a graph-based algorithm

LiDAR is one of the most promising scanning technologies with a wide range of applications. It is commonly used for geodetic and cartographic tasks, allowing for the scanning and reconstruction of various terrains and volumetric models, but also finds application as sensors in other fields. LiDAR technology has gained widespread adoption in the forestry industry, where it is employed to monitor changes in the structure of forested areas and the conditions of individual trees. In this study, an algorithm specialized in tree clustering was implemented for point cloud segmentation. The algorithm was tested on data obtained from scanning a carbon polygon located in the Fryanovsky forest area. To enhance data processing quality, the point cloud was cleaned with height distribution alignment. Results from dataset segmentation were compared with the outcomes of other algorithms. It is anticipated that the algorithm will demonstrate superior performance compared to DBSCAN and OPTICS.