距离加权与冠层控制的LiDAR CHM凹坑去除方法  

A pits removal method for LiDAR CHM based on distance weighting and canopy control

作　　者：刘一成 王成[1,2,3,4] 习晓环 王金亮 张海清[5] LIU Yicheng;WANG Cheng;XI Xiaohuan;WANG Jinliang;ZHANG Haiqing(Faculty of Geography,Yunnan Normal University,Kunming 650500,China;Key Laboratory of Resources and Environmental Remote Sensing for Universitiesin Yunnan,Kunming 650500,China;Center for Geospatial Information Engineering and Technology of Yunnan Province,Kunming 650500,China;Key Laboratory of Digital Earth,Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China;Faculty of Land Resource Engineering of KUST,Kunming 650093,China)

机构地区：[1]云南师范大学地理学部,昆明650500 [2]云南省高校资源与环境遥感重点实验室,昆明650500 [3]云南省地理空间信息信息工程技术研究中心,昆明650500 [4]中国科学院空天信息创新研究院数字地球重点实验室,北京100094 [5]昆明理工大学国土资源工程学院,昆明650093

出　　处：《测绘科学》2021年第4期108-113,共6页Science of Surveying and Mapping

基　　金：国家自然科学基金面上项目(41871264);云南师范大学研究生科研创新基金项目(ysdyjs2019140)。

摘　　要：针对机载激光雷达(LiDAR)获取的高密度点云数据为森林冠层高度模型(CHM)构建提供了可能,但在应用中通常会出现局部无效值(凹坑)现象,影响森林高度等参数的提取精度的问题,该文提出基于距离加权和冠层控制的CHM凹坑去除方法。首先加权计算每个像素相似阈值判断可能凹坑,然后利用形态学运算控制冠层范围并确定最终凹坑,继而填入对应的CHM中值滤波后的值来去除CHM凹坑。通过与高斯滤波、均值滤波、中值滤波等方法比较,该文方法在凹坑的去除效果以及保留冠层形态方面均有优势,且对不同分辨率CHM具有普适性,直接提取单木数量精度可达96.7%,相比原始CHM以及另外4种方法处理的CHM所提取单木数量,精度分别提高2.38%、13.92%、17.81%、6.54%、2.38%;直接提取平均树高精度为99.41%,相比原始CHM及另外4种方法CHM精度分别提高0.16%、0.8%、0.56%、0.75%、0.16%。Aiming at the problem that the high-density point cloud data obtained by airborne light detection and ranging(LiDAR) provides the possibility of building a forest canopy height model(CHM),but local invalid values(pits) often occur in applications, which affects the extraction accuracy of parameters such as forest height. Therefore, a CHM pit removal method based on distance weighting and canopy control was proposed. Firstly, weighted calculation of each pixel similarity threshold to determine possible pits. Then use morphological operations to control the canopy range and determine the final pits.Finally,fill in the corresponding CHM median filtered value to remove CHM pits.Compared with Gaussian filtering,mean filtering,median filtering and other methods,this method had advantages in removing pits and retaining canopy morphology,and was universal for different resolution CHM,and directly extracted the number of single trees up to 96.7%.Compared with the number of single trees extracted by the original CHM and four filtering methods,the accuracy was increased by 2.38%,13.92%,17.81%,6.54%and2.38%,respectively;the accuracy of direct extraction of average tree height was 99.41%,compared with the original CHM and the other four methods,the accuracy of CHM was improved by 0.16%,0.8%,0.56%,0.75%and 0.16%,respectively.

关 键 词：激光雷达 冠层高度模型 凹坑 距离加权 冠层控制 

分 类 号：P237[天文地球—摄影测量与遥感]