Abstract: To address the issue that traditional individual tree segmentation methods are susceptible to the accuracy of tree apex extraction,this paper proposes a bidirectional growth-based individual tree segmentation method using LiDAR point cloud data.The approach initially calculates the centers of multi-layer slices of trunk point cloud to perform linear fitting,thereby identifying the trunk axis and its intersection point with the ground to determine the tree position.Subsequently,based on the tree position,clustering is progressively conducted from bottom to top to locate the tree apex corresponding to each individual tree.Finally,a top-down progressive growth strategy is employed to segment individual tree point clouds.To validate the effectiveness of the proposed method,experiments were conducted in three distinct forest environments.The experimental results demonstrate F1 scores of 0.971,0.886,and 0.865 for the individual tree segmentation results in the three sample plots,respectively.Compared to conventional methods,our approach achieves the optimal individual tree extraction rate,matching rate,and the lowest omission error.These findings indicate that the proposed bidirectional growth individual tree segmentation method yields more accurate results and exhibits strong robustness.

Key words: LiDAR, single tree segmentation, tree vertex extraction, bidirectional growth