机载激光点云单木分割方法对比及精度分析  

Comparison and Accuracy Analysis of Airborne Laser Point Cloud Individual Tree Segmentation Methods

作　　者：张丽[1] 王健[1] 曲相屹 邵为真 ZHANG Li;WANG Jian;QU Xiangyi;SHAO Weizhen(College of Geodesy and Geomatics,Shandong University of Science and Technology,Qingdao 266590,China;Jiaojia Gold Mine of Shandong Gold Mining(Laizhou)Co.,Ltd.,Yantai 261441,China;Shandong Xinhui Construction Group Co.,Ltd.,Dongying 257091,China)

机构地区：[1]山东科技大学测绘与空间信息学院,山东青岛266590 [2]山东黄金矿业(莱州)有限公司焦家金矿,山东烟台261441 [3]山东新汇建设集团有限公司,山东东营257091

出　　处：《测绘与空间地理信息》2023年第5期34-37,42,共5页Geomatics & Spatial Information Technology

基　　金：多平台点云矿山实景三维重建及地表沉陷分析(LNYS-2021-Z07);高端外国专家引进计划——生态环境与自然资源遥感监测关键技术(G2021025006L)资助。

摘　　要：通过对机载激光点云基于冠层高度模型、基于点云以及基于层堆叠种子点分割方法,对针叶林、阔叶林、针阔混交林3种不同类型林分进行单木分割,并通过计算单木分割的检测率、正确率、F-score等精度指标,探究不同分割方法在不同类型林分的适用性。实验结果表明,对于针叶林,基于点云分割方法的分割精度最高,基于层堆叠种子点分割方法对树木分割的正确率最高;对于阔叶林和针阔混交林,基于层堆叠种子点分割方法的分割精度和检测率较高,优于其他两种分割方法。Based on canopy height model,point cloud and layer stacked seed point,this paper conducts individual tree segmentation on airborne laser point cloud for three different forest stands:coniferous forest,broad-leaved forest and coniferous-broadleaf mixed forest,and by calculating the recall,accuracy,F-score and other accuracy indexes of individual tree segmentation,the applicability of different segmentation methods in different types of forest stands is investigated.The experimental results show that for coniferous forest,the method based on point cloud segmentation has the highest segmentation accuracy,the method based on layer stacking seed point segmentation algorithm has the highest accuracy of tree segmentation;For broad-leaved forest and coniferous-broadleaf mixed forest,the layer stacking seed point segmentation algorithm has higher segmentation accuracy and recall than the other two segmentation methods.

关 键 词：机载激光点云 单木分割 冠层高度模型 点云分割 层堆叠算法 

分 类 号：P228.5[天文地球—大地测量学与测量工程]