TLS Point Cloud Processing and Leaf–Wood Segmentation Using Open-Source Methods
Instructors: Jinyi Xia Description: Accurate segmentation of leaf and wood components is a critical step for tree structure analysis using Terrestrial Laser Scanning (TLS) data. This workshop introduces participants to open-source tools and workflows for preprocessing TLS point clouds and performing leaf-wood classification using both rule-based and machine learning approaches. Participants will gain hands-on experience with the following methods: i) TLSeparation - a heuristic algorithm based on geometric features; ii) DBSCAN - a density-based clustering algorithm; iii) Graph - identifies wood points by generating graph structure; iv) KPConv - a deep learning algorithm based on kernel points. The workflow includes point cloud normalization, filtering, segmentation, visualization, and evaluation using reference data. Comparative performance of the methods will be discussed in terms of accuracy and transferability across species. More details here.
