A Unified Approach for Dynamic Tree Segmentation using LiDAR Point Cloud Data

Accurate segmentation of individual trees from LiDAR point clouds is essential for forest resource management and ecological studies. This research introduces a novel methodology that combines a dynamic Euclidean clustering approach for extracting tree coordinates and Gaussian kernel-based dynamic thresholding to enhance segmentation precision. The method was demonstrated using LiDAR data from the Shivamogga forest region (Shimoga), Karnataka, India, and validated using extensive field inventory data. The performance of the tree height estimation was analyzed by box plot analysis, with a comparison of observed and estimated heights showing a strong alignment in most cases. However, overestimation was observed in denser vegetation areas. The approach demonstrated robustness in handling variable forest conditions, including dense canopy structures and diverse tree sizes. This work contributes to LiDAR-based forest analysis, offering scalable solutions for large-scale ecological assessments and resource management.