泥石流沟谷地形的机载激光雷达点云滤波算法对比  

A comparative study of airborne light detection and ranging point cloud filtering algorithms for mudslide gully terrains

作　　者：王泽华 甘淑[1] 吕杰 杨在松 WANG Ze-hua;GAN Shu;LüJie;YANG Zai-song(Faculty of Land Resources Engineering,Kunming University of Science and Technology,Kunming 650093,China;City College,Kunming University of Science and Technology,Kunming 650051,China)

机构地区：[1]昆明理工大学国土资源工程学院,昆明650093 [2]昆明理工大学城市学院,昆明650051

出　　处：《兰州大学学报（自然科学版）》2023年第6期770-777,共8页Journal of Lanzhou University(Natural Sciences)

基　　金：国家自然科学基金项目(62266026);教育部产学合作协同育人项目(202101096033);云南省教育厅基金项目(2021J0047)。

摘　　要：为探究不同滤波算法对泥石流沟谷复杂地形机载激光雷达点云数据处理的普适性,选取昆明市东川区大白泥河沟作为研究区,采用改进的渐进加密三角网滤波(IPTDF)、布料模拟滤波(CSF)、基于坡度的滤波(SBF)3种常用滤波算法对研究区内的15个样本机载激光雷达点云数据进行滤波试验,结合国际摄影测量和遥感学会提出的交叉表评价体系对滤波的结果进行量化分析,探究坡度、植被覆盖度两类因素与滤波精度之间的关系.结果表明,当地形平坦、坡度较小时(样本1、2、6、7、10),IPTDF、SBF和CSF算法的平均总误差分别为8.73%、8.81%、8.90%,数值上十分接近,3种算法都能达到分类地面点的目的,当地形复杂,坡度较大时(样本5、8、9、12、14),IPTDF、SBF和CSF算法的平均总误差分别为13.4%、36.4%、30.2%,CSF与SBF算法在地形复杂的陡壁区域精度较低,IPTDF算法平均总误差远低于其他两种算法,能够很好地适应泥石流沟谷的复杂地形地貌特征;研究区内坡度和植被覆盖度呈显著正相关,坡度和植被覆盖度对CSF和SBF的误差影响较为明显,各算法Ⅰ类误差和总误差整体上随着坡度的增大而增加,且在植被覆盖度近似的样本内,坡度越大的区域各算法总误差也越大.In order to investigate the general applicability of different filtering algorithms to the processing of airborne light detection and ranging point cloud data in the complex terrain of mudflow gully,Dabaini River in Dongchuan District,Kunming City was selected as the study area,and three common filtering algorithms,i.e.improved progressive triangular irregular network densification filtering(IPTDF),cloth simulation filtering(CSF)and slope-based filtering(SBF),were used to filter 15 samples of airborne light detection and ranging point cloud data in the study area,and the cross-tabulation evaluation system proposed by International Society for Photogrammetry and Remote Sensing was combined to quantify the filtering.The results were quantified and analyzed to investigate into the relationship between slope and vegetation coverage and filtering accuracy,showing that when the terrain was flat and the slope small(samples 1,2,6,7,and 10),the average total errors of IPTDF,SBF,and CSF algorithms were 8.73%,8.81%,and 8.90%,respectively,which were numerically very close,and all three algorithms could achieve the purpose of classifying ground points,but when the terrain was complex and the slope large(samples 5,8,9,12,and 14),the average total errors of IPTDF,SBF,and CSF algorithms were SBF and IPTDF algorithms had average total errors of 13.4%,36.4%,and 30.2%,respectively,CSF and SBF algorithms had a lower accuracy in steep-walled areas with a complex terrain;IPTDF algorithm had a much lower average total error than the other two algorithms,and could adapt well to the complex topographic features of mudslide gullies;slope and vegetation coverage in the study area were significantly positively correlated,and slope and vegetation coverage had rather evident influence on the errors of CSF and SBF.The type I error and total error of each algorithm increased with the increase in slope,and the total error of each algorithm was larger in the area with a large slope in the sample of a similar vegetation coverage.The experimental r

关 键 词：滤波算法 泥石流沟谷 机载激光雷达点云 坡度 植被覆盖度 误差分析 

分 类 号：P237[天文地球—摄影测量与遥感]