基于WorldView-3遥感影像与机载LiDAR的单木树高提取  

Extraction of individual tree height using WorldView-3 remote sensing images and airborne LiDAR

作　　者：胡天祺 王振锡[1,2] 郝康迪 曲延斌 马琪瑶 吕金城 HU Tianqi;WANG Zhenxi;HAO Kangdi;QU Yanbin;MA Qiyao;LV Jincheng(College of Forestry and Landscape Architecture,Xinjiang Agricultural University,Urumqi 830052;Key Laboratory of Forestry Ecology and Industrial Technology in Arid Areas,Xinjiang Department of Education,Urumqi 830052,China)

机构地区：[1]新疆农业大学林学与风景园林学院,乌鲁木齐830052 [2]新疆教育厅干旱区林业生态与产业技术重点实验室,乌鲁木齐830052

出　　处：《干旱区资源与环境》2022年第10期166-175,共10页Journal of Arid Land Resources and Environment

基　　金：2021年新疆林业改革发展基金项目(2020-942);2022年新疆维吾尔自治区研究生科研创新项目(XJ2022G144)资助

摘　　要：基于WorldView-3影像与机载LiDAR数据,结合地面每木定位调查,采用冠层高度模型法提取天山云杉单木树高,比较分析不同数据源提取单木树高的精度,为山地天然林单木高度提取提供方法依据。以新疆农业大学实习林场天山云杉为研究对象,实测单木树高,并用RTK每木定位。以WorldView-3遥感影像、LiDAR点云为数据源,分别通过立体像对与布料模拟滤波分类的方法提取LiDAR与WorldView-3两种数字表面模型(DSM),以LiDAR点云生成的数字高程模型(DEM)为基础,解算两种数据的冠层高度模型(CHM)获取单木高度。基于LiDAR与WorldView-3数据的天山云杉单木识别率较高,平均识别率为96.3%和93.7%,平均漏检率为3.7%和6.3%;利用LiDAR提取树高的精度较高,平均精度为90.50%,而采用WorldView-3提取树高的精度略低于前者,平均精度为83.28%。通过对机载激光雷达点云数据进行布料模拟滤波分类,能够获得高精度的DEM数据,通过二者叠加运算是一种提取单木树高可行的方法;基于WorldView-3影像进行立体像对可以获得精度较高的DSM,在拥有高精度DEM基础数据的情况下,可以作为调查范围较大、成本有限的森林资源调查树高提取方法。Based on WorldView-3 imagery and airborne LiDAR data,combined with the survey on the location of each tree on the ground,the canopy height model method is used to extract the single tree height of Picea schrenkiana var.tianschanica,and the accuracy of single tree height extraction from different data sources is compared and analyzed to provide a now method for extraction of single wood height in Tianshan Mountains.Taking Picea schrenkiana var.tianschanica growing in the internship forest farm of Xinjiang Agricultural University as the research object,the height of single tree was measured and the RTK was used to locate each tree.Taking WorldView-3 remote sensing image and LiDAR point cloud as data sources,two digital surface models(DSM)of LiDAR and WorldView-3 were extracted by stereo image pair and cloth simulation filter classification,respectively.Based on the digital elevation model(DEM)generated by LiDAR point cloud,above two DSMs were solved.By calculating the data of the two DSMs,canopy height models were obtained,and then the height of a single tree was calculated.The recognition rates of Picea schrenkiana var.tianschanica single tree based on LiDAR and WorldView-3 data were higher,the average recognition rates were 96.3%and 93.7%,respectively,and the average missed detection rates were 3.7%and 6.3%,respectively.The accuracy by using WorldView-3 to extract tree height was slightly lower than that by LiDAR.High-precision DEM data can be obtained by performing cloth simulation filtering and classification on the airborne lidar point cloud data.It is a feasible method to extract a single tree height through the superposition of the two.High precision DSM can be obtained by stereo image pairing based on WorldView-3 images,which can be applied as a tree height extraction method for forest resource survey with large survey scope and limited cost in the case of high-precision DEM data being available.

关 键 词：WorldView-3 激光雷达 RTK 立体像对 单木识别 树高提取 

分 类 号：S771.8[农业科学—森林工程] S758[农业科学—林学]