关键词: 自适应, 骨架提取, 几何相似性, 聚类长度阈值, 曲线骨架

Abstract: To extract 3D art-designing tree skeleton, this paper proposes a self-adapting algorithm with three main steps: pre-processing, skeleton extraction and post-processing. The pre-processing sequentially implements the pre-
computing operations such as clustering all the sub-branches of the same geometry similarity, self-adaptively generating clustering-length threshold for each sub-branch, determining parent-child relationships between sub-branches. The skeleton extraction processing introduces the approaches to cluster cloud points, generate corresponding skeleton-nodes and form skeleton curve using these skeleton nodes for each sub-branch. The post-processing deletes all the separate skeleton nodes and lines and smoothes skeleton curve for the whole tree. This approach is running on computer platform fully automatically, without any intervention of users. The experimental results indicate that the algorithm in this paper can generate smoothly skeleton curve, accompanied with the complete shape and the correct topology structure of tree model.

Key words: self-adapting, skeleton extraction, geometry similarity, clustering-length threshold, curve skeleton