An Effective Approach to 3D Stem Modeling and Branch-Knot Localization in Multiscan TLS Data

Accurate three dimensional (3D) stem modeling is critical to individual tree analysis. Multi-scan Terrestrial Laser Scanning (TLS) systems accurately map fine 3D structural details of tree components including the stem, branches and leaves. State-of-the-art (SoA) methods to model stem are affected by problems such as occlusion and point density variation around the stem. Thus, we propose a voxel based approach to derive accurate 3D model of stem by exploiting the opacity-driven (to lasers) void volume formed within the stem. For each height slice, peaks detected in the external boundary of the point-density-map object generated jointly from points derived from the stem model and the proximal TLS data points correspond to branch-knots. All experiments were conducted on a set of 10 manually delineated trees belonging to pine and spruce. The preliminary results prove the effectiveness of the method to accurately model the 3D stem and localize branch-knots.