Skeleton extraction for tree models

Abstract

The curve skeleton extraction for a given laser-scanned tree model plays an important role in many virtual agricultural applications, such as modelling, animation and growth simulation. To extract the curve skeletons of various tree models, which can capture the essential topology structures, a simple and robust algorithm based on point cloud contraction using constrained Laplacian smoothing is proposed. The adaptive sampling and post-processing steps designed for complex tree-like models can effectively reduce the computation time and reconstruct the correct curve skeleton from a contracted point cloud. The experimental results show that the curve skeletons extracted by this algorithm are faithful and smooth, and they can be well utilized in many fields.