An Individual Tree Segmentation Algorithm Without Prior-Knowledge Based on Airborne Lidar Data

This paper aims at improving the individual tree segmentation method proposed by Liu et al. (2019), which is based on the 3-D distribution of point cloud without any prior knowledge or manual threshold setting. The improved algorithm has been applied to 22 plots of single-layered coniferous forest, single-layered broad-leaved forest, single-layered coniferous broad-leaved mixed forest, multi-layered coniferous forest and multi-layered coniferous broad-leaved mixed forest. The results showed that the algorithm can segment trees effectively, with an overall root mean square extraction rate of 0.95, and a root mean square F_Score of 0.73. It was found that the algorithm performed better in coniferous forest and single forest type of single layer with a stand density less than 300 plants/ha.