A Robust Skeleton Lines Extraction Method for Individual Tree Modeling Using Terrestrial LiDAR Point Clouds

Skeleton line extraction is a crucial step in individual tree modeling. However, extracting the skeleton from LiDAR point clouds still faces challenges with low accuracy and poor branch connection topology due to self-occlusion and noisy points. To tackle these issues, this paper introduces an innovative method for skeleton lines extraction. Initially, a revised Laplacian-based contraction method is presented, incorporating a weight adjustment factor. By considering the linearity of each point, the contracted point cloud shape closely resembles the one-dimensional skeleton shape of the tree. Additionally, the adaptive farthest distance spherical sampling method is used to extract skeleton points based on geometric features. To extract the skeleton line, a hybrid neighborhood determination model is devised to accurately connect branch nodes while maintaining a simple graph structure. Four individual trees with different geometric forms are employed for testing purposes. Experimental results indicate that the proposed method yields improved skeleton extraction outcomes.