학술논문
基于力场的点云树木骨架提取方法 / Skeleton extraction for point cloud trees based on force field
Document Type
Academic Journal
Author
Source
南京林业大学学报（自然科学版）. 40(2):160-166
Subject
Language
Chinese
ISSN
1000-2006
Abstract
树木骨架是树木仿真及建模的基础，笔者根据树木的拓扑原理，直接利用地面激光雷达扫描获得的单木点云数据，提出了一种基于物理学中力场概念的点云数据树木骨架提取方法：首先对点云树木模型运用空间层次剖分的方法进行分层，根据点云的邻域关系建立基于树木特征点的简化表示，然后根据计算点的测地距离对树木特征点进行连接，再运用力场将位于树木表面的骨架连线压缩至树木内部，最后根据骨架夹角阈值对骨架进行顺滑得到最终的树木骨架。研究显示，将该方法分别应用在含笑树和樱花树的骨架提取中效果较好，相比同类算法效率较高。
Skeleton extraction is a fundamental part in simulation and modeling of point cloud trees. According to topolo?gy, a novel skeleton extraction method, which based on physical force field, is first proposed in this paper. First, the point cloud tree is layered by space subdivision method, and a simplified representation of the feature points is estab?lished under the neighbor relationships. Next, the feature points are connected by calculating the geodesic distance. Then, the surface skeleton is compressed into trees by applying force field. Finally, the final skeleton will be acquired by polishing it according to a threshold setting. The experimental results indicate that the algorithm can provide an satisfac?tory result, and the algorithm performance is verified using datasets ofmichelia and cherry trees.
Skeleton extraction is a fundamental part in simulation and modeling of point cloud trees. According to topolo?gy, a novel skeleton extraction method, which based on physical force field, is first proposed in this paper. First, the point cloud tree is layered by space subdivision method, and a simplified representation of the feature points is estab?lished under the neighbor relationships. Next, the feature points are connected by calculating the geodesic distance. Then, the surface skeleton is compressed into trees by applying force field. Finally, the final skeleton will be acquired by polishing it according to a threshold setting. The experimental results indicate that the algorithm can provide an satisfac?tory result, and the algorithm performance is verified using datasets ofmichelia and cherry trees.