基于无人机多光谱和HMLS的森林树种识别  

Forest tree species identification based on UAV multispectral and HMLS

作　　者：王俊杰 张思媛 滕鹏程 Wang Junjie;Zhang Siyuan;Teng Pengcheng(School of Mining Engineering,Heilongjiang University of Science&Technology,Harbin 150022,China;Yanzhou sinoma construction Co.Ltd.,Fuzhou 344706,China)

机构地区：[1]黑龙江科技大学矿业工程学院,哈尔滨150022 [2]兖州中材建设有限公司,江西抚州344706

出　　处：《黑龙江科技大学学报》2023年第5期774-778,共5页Journal of Heilongjiang University of Science And Technology

摘　　要：为探索多源遥感数据在森林识别中的潜力,以料甸林场无人机多光谱和手持移动激光雷达数据为数据源,识别森林树种。在对研究区实现单木分割的基础上提取单木特征,基于不同的特征组合方式,利用随机森林分类器进行树种识别,比较不同组合形式对树种识别精度的影响。结果表明:基于层堆叠种子点法的单木分割方法精度最高,F值达到0.84;点云结构特征特别是点云高度变量在树种识别中有良好的表现,手持点云数据和多光谱数据的结合较仅使用单一数据源的识别精度提高了10.42%。This paper intends to explore the potential of multi-source remote sensing data in forest identification by using UAV multispectral and handheld mobile lidar data of Liaodian forest farm as data source for researching tree species identification.The study involves extracting the single wood features based on the realization of single wood segmentation in study area,identifying tree species by random forest classifier based on different feature combination methods,and comparing the influence of different combination forms on tree species recognition accuracy.The results show that based on the layer stacking seed point method,the single-wood segmentation method has the highest accuracy with an F value 0.84;the structural characteristics of the point cloud,especially the point cloud height variable,have good performance in tree species identification,and compared with the single data source,the combination effort of handheld point cloud data and multispectral data improves the recognition accuracy by 10.42%.

关 键 词：树种识别 单木分割 激光雷达 随机森林 

分 类 号：P208[天文地球—地图制图学与地理信息工程]