Individual Tree Segmentation From Airborne LiDAR Data Based on Automatic Treetop Detection and Simulated Stem-Branch Points Compensation

Accurate retrieval of tree structural attributes is essential for effective forest management and resource assessment. Airborne laser scanning (ALS) provides detailed three-dimensional data from a top-down perspective, enabling the analysis of forest canopy structures at a large scale. However, due...

Full description

Saved in:
Bibliographic Details
Main Authors: Qingjun Zhang, Jiale Chen, Hanwen Qi, Xu Wang, Xinlian Liang
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11048925/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Accurate retrieval of tree structural attributes is essential for effective forest management and resource assessment. Airborne laser scanning (ALS) provides detailed three-dimensional data from a top-down perspective, enabling the analysis of forest canopy structures at a large scale. However, due to its limited penetration through dense canopies, ALS point clouds under the crown is often sparse or even missing, making individual tree segmentation particularly challenging. Therefore, we propose a novel individual tree segmentation method based on treetop detection and point compensation. First, a center shift algorithm is proposed to detect treetop candidates, which are then refined to identify reliable treetops through geometric-feature analysis of these candidates&#x2019; neighborhoods. Next, simulated stem-branch points are compensated with crown shape constraints to reconstruct sparse under-canopy structures. Finally, individual trees are segmented via a shortest-path algorithm applied to a graph that represents connectivity between adjacent points. The proposed method is evaluated using a public available dataset. The average extraction rate, completeness of treetop detection, and <inline-formula><tex-math notation="LaTeX">${{R}^2}$</tex-math></inline-formula> for tree height estimation were 108%, 51%, and 0.93, respectively. In comparison to existing methods, the proposed method gave a similar F-Score as the best method, i.e., only 1% lower, but achieves a significantly better extraction rate, demonstrating superior overall performance. The proposed method enables accurate individual tree parameter extraction, advancing forest inventory efficiency and providing a foundation for ecological studies.
ISSN:1939-1404
2151-1535