基于体元逐层聚类的TLS点云数据单木分割算法

摘要

为了提高地基激光雷达(Terrestrial Laser Scanning,TLS)点云数据的单木分割精度及分割效率,以落叶期的蒙古栎人工林为研究对象,以TLS为基础数据,在三维点云数据体元化的基础上提出一种基于体元逐层聚类的单木分割算法,通过分析体元在竖直方向的z值序列准确获取单木的位置,然后利用模糊C均值算法以单木位置为初始聚类中心从最底层体元开始进行逐层聚类,最终实现样地水平蒙古栎单木分割.研究结果表明,通过分析体元在竖直方向的z值序列能准确获取单木的位置,本研究提出的单木分割算法能够实现样地水平单木的精确分割.%To improve the accuracy and efficiency of individual tree segmentation with terrestrial laser scanning (TLS) data,a new method based on clustering of voxels layer by layeris proposed in this paper.Additionally,the TLS data of Mongolian Oak plantation in deciduous stage is obtained and processed.Then the position of single tree is obtained by analyzing the z-sequence of voxels in the vertical direction.The position of each tree is used as the initial clustering center to start clustering from the bottom voxel layer by fuccy c-means clustering method.The experimental results demonstrate that the position of single tree can be obtained accurately by analyzing the z-value sequence in the vertical direction,the proposed algorithm can segment each individual tree accurately.