文献类型：期刊文献

中文题名：结合无人机可见光和激光雷达数据的杉木树冠信息提取

英文题名：The Method of Extracting Information of Cunninghamia lanceolata Crown Combined with RGB and LiDAR Based on UAV

作者：王娟[1,2] 张超[1] 陈巧[2] 李华玉[1,2] 彭希[2,3] 白明雄[1] 徐志扬[2,4] 刘浩栋[2] 陈永富[2]

第一作者：王娟

机构：[1]西南林业大学林学院,云南昆明650233;[2]中国林业科学研究院资源信息研究所,国家林业和草原局林业遥感与信息技术重点实验室,北京100091;[3]四川农业大学林学院,四川成都611130;[4]国家林业和草原局华东调查规划设计院,浙江杭州310019

年份：2022

卷号：42

期号：1

起止页码：133-141

中文期刊名：西南林业大学学报：自然科学

外文期刊名：Journal of Southwest Forestry University:Natural Sciences

收录：CSTPCD;;北大核心:【北大核心2020】；

基金：中央级公益性科研院所基本科研业务费专项(CAFYBB2018SZ008)资助。

语种：中文

中文关键词：无人机;高分影像;激光雷达;树冠信息;面向对象;分水岭分割

外文关键词：UAV;high-resolution image;LiDAR;crown information;object-oriented;watershed segmentation

分类号：S771.5

摘要：以年珠实验林场为研究区,以无人机可见光正射影像和激光雷达数据为数据源,采用分水岭分割与面向对象结合的方法提取不同郁闭度下杉木单木树冠信息,并对提取精度进行验证首先采用面向对象法基于无人机可见光影像提取树冠区域,然后基于构建的CHM进行分水岭分割获取单木树冠初步分割结果,最后基于初步分割结果对树冠区域进行二次分割,提取单木树冠信息。结果表明:不同郁闭度林分条件下单木树冠信息提取效果较好,其中单木树冠提取F测度分别为88.07%~95.08%和78.57%~88.29%;提取的树冠面积与实测面积建立的线性回归模型,R2分别为0.859 1和0.736 7,RMSE分别为2.49 m^(2)和3.29 m^(2);提取的冠幅与实测冠幅建立的线性回归模型,R;分别为0.830 6和0.724 6,RMSE分别为0.46 m和0.57 m。基于无人机可见光影像采用面向对象多尺度分割法提取树冠区域很好的消除了样地内裸地及林下灌木等因素的影响;同时,无人机Li D-AR数据能够更加精确的区分单木信息,2种数据源结合发挥了二者的优势,提高了单木树冠的提取精度。本研究可为快速获取不同郁闭度林分下单木树冠信息提供参考。
With the Nianzhu forest field as the research area, the visible light image and lidar data of UAV were used as the data source, and the watershed segmentation algorithm and object-oriented method were used to extract single tree crown width information in different forest canopy densities and verify the extraction accuracy.First, tree crown was extracted from visible light image using object-oriented method which was to extract the canopy tree crown range. Then, the watershed segmentation method was used to obtain the preliminary segmentation results of single tree crown based on CHM. Finally, divided again to extract the information of single tree crown based on the preliminary segmentation boundary and crown area. The results of show that the extraction effect of individual tree crown and area is better in different density. The F-measure was between 88.07%–95.08%and 78.57%–88.29%;and a linear regression model was established between extracted crown area and the measured crown area, the R;were 0.859 1 and 0.736 7 and RMSE were 2.49 m^(2) and 3.29 m^(2). A linear regression model was established between extracted crown diameter and the measured crown diameter, the R;were 0.8306 and 0.724 6,RMSE of which were 0.46 m and 0.57 m. Based on the visible light image of the UAV, the tree crown area extracted by object-oriented multi-scale segmentation method has eliminated the influence of naked and shrub in the forest. At the same time, the UAV-LiDAR data can accurately distinguish individual tree. The combination of the2 data sources can take advantages of both and improves the extraction accuracy of single tree crown. This study can provide a reference for quickly obtaining individual tree crown information under different canopy density stands.