基于无人机载激光雷达点云数据的人工侧柏林单木分割研究  

Artificial Platycladus orientalis(L.)Franco Single Wood Segmentation Based on UAV-based Lidar Point Cloud Data

作　　者：李远航 笪志祥[3] 闫烨琛 LI Yuan-hang;DA Zhi-xiang;YAN Ye-chen(Yellow River Engineering Consulting Co.Ltd.,Zhengzhou 450003,Henan,China;Key Laboratory of Water Management and Water Security for Yellow River Basin,Ministry of Water Resources(Under Construction),Zhengzhou 450003,Henan,China;Tianjin Water Affairs Bureau.,Tianjin 300074,China;Tianjin Geo-Engineering Institute of Investigation&Surveying Design Co,Ltd.,Tianjin 300191,China)

机构地区：[1]黄河勘测规划设计研究院有限公司,河南郑州450003 [2]水利部黄河流域水治理与水安全重点实验室(筹),河南郑州450003 [3]天津市水务局,天津300074 [4]天津市地质工程勘测设计院有限公司,天津300191

出　　处：《西北林学院学报》2023年第6期171-179,共9页Journal of Northwest Forestry University

基　　金：天津市水务重点科技项目(KY2023-02)。

摘　　要：可持续的森林经营管理模式有助于实现生态文明建设的高质量发展,人工林单木分割结果的获取是森林经营管理的关键。目前无人机载激光雷达技术为单木位置的精准定位和树冠的精确划分提供了应用空间。利用冠层高度模型(CHM)检测和提取单木数量和树冠、树高等信息,评估和分析空间分辨率和点云密度对单木分割和树冠提取结果的影响和精度,该方法可准确分割人工侧柏林和提取树冠信息,总体上单木分割精确率均>75%,树冠轮廓提取精确率均>65%,实测数据与提取数据的回归决定系数均>0.6。适当的空间分辨率有助于提高单木分割精度,当分辨率为0.3 m时单木分割和树冠提取结果均为最优。同时研究发现随着点云密度的降低,单木数量的识别精度值随之下降。当点云密度为100%时,F为89%,当点云密度为10%时,其F降低至73%。基于CHM模型可以较好实现人工林的单木精确分割,对林木出现的树冠重叠、覆盖、偏移等现象均有一定的辨别能力;同时分析了空间分辨率和点云密度对单木分割和树冠提取结果的影响并评估了精确性,识别了单木分割时关键参数的最优选择。Sustainable forest management mode is helpful to realize the high quality development of ecological civilization construction.The key of forest management is to obtain the result of single tree division of plantation.At present,UAV-based LiDAR technology provides application space for the precise location of individual trees and the precise division of tree crowns.In this study,canopy height model(CHM)was used to detect and extract information about the number of individual trees,crown and tree height,and to evaluate and analyze the influence and accuracy of spatial resolution and point cloud density on the results of individual tree segmentation and crown extraction.The results showed that this method could accurately segmented P.orientalis plantation and extract the crown information.Overall,the accuracy rate of single tree segmentation was higher than 75%,the accuracy rate of tree crown contour extraction was higher than 65%,and the regression determination coefficients of measured data and extracted data were higher than 0.6.Appropriate spatial resolution was helpful to improve the accuracy of single tree segmentation.When the resolution was 0.3 m,the results of single tree segmentation and tree crown extraction were optimal.With the decrease of point cloud density,the recognition accuracy of the number of individual trees decreased.When the point cloud density was 100%,the F value was 89%,and when the point cloud density was 10%,the F value decreased to 73%.The CHM model could be used to accurately segment individual trees in the plantation,and could distinguish the overlapping,covering and deviation of tree crowns.Moreover,the effects of spatial resolution and point cloud density on the results of single tree segmentation and crown extraction were analyzed and the accuracy was evaluated,and the optimal selection of key parameters in single tree segmentation was identified.

关 键 词：无人机激光雷达 单木分割 结构参数信息 空间分辨率 点云密度 

分 类 号：S791.38[农业科学—林木遗传育种]