从三维激光点云中快速统计树木信息的方法

A Fast Statistical Method of Tree Information from 3D Laser Point Clouds

树木位置分布及胸径(DBH)是研究森林生态、管理林区的重要指标。激光雷达在获取树木相关数据方面有巨大潜力,因此,提出用手持移动激光雷达获取的三维点云快速统计树木信息的方法。手持移动激光雷达可近距离采集树木信息,获取更详细的单木立面信息。针对上述点云特点,提出分层聚类的点云处理方法,按不同高度对点云切片,形成一组切片截面图,再仅对切片截面图聚类;根据聚类结果使用随机抽样一致性算法拟合出圆,对比一组切片截面图的拟合结果,完成树木点云提取。这种先取样再计算的方法大大提高了运算速度。实验证明该方法树干提取准确率达94.4%,DBH计算平均误差3.4cm。本文方法可快速统计树木相关信息。

Location distribution and diameter at breast height（DBH）of trees are important indicators for studying forest ecology and managing forest areas.Lidar have great potential for obtaining tree-related data.Therefore,a fast statistics method of tree information from three-dimensional laser point clouds obtained by hand-held mobile laser is proposed.The hand-held mobile laser can collect the tree information from close distance and obtain the detailed information of a single tree facade.In view of the characteristics of the above point cloud,a method based on hierarchical clustering is proposed.A group of cross-section slices of point cloud are formed at different heights,and the segmentation process of each slice is performed by cluster analysis.We use random sample consensus algorithm to fit the circle according to the result of segmentation and complete the tree point cloud extraction by comparing the fitting results of a set of slice sections.This method of sampling and recalculating greatly improves the processing speed.The results show that the accuracy of trunk extraction is 94.4%,the average calculation error of DBH is 3.4 cm.The proposed method can quickly statistic tree information.