机载LiDAR测量复杂地形中滤波算法的比较研究

(1.西安科技大学 测绘科学与技术学院, 西安 710054; 2.西北农林科技大学 黄土高原土壤侵蚀与旱地农业国家重点实验室, 陕西 杨凌 712100; 3.中国科学院 水利部 水土保持研究所, 陕西 杨凌 712100)

机载LiDAR; 滤波算法; 地形表达; 微地形变化监测; 黄土高原

Accuracy of Airborne LiDAR Point Cloud Filtering for Areas with Complex Terrain
LI Dou1, LI Pengfei1, MU Xingmin2,3, YAO Wanqiang1, TANG Fuquan1, LI Ting1

(1.College of Geomatics, Xi'an University of Science and Technology, Xi'an 710054, China; 2.State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau, Northwest A&F University, Yangling, Shaanxi 712100, China; 3.Institute of Soil and Water Conservation, CAS&MWR, Yangling, Shaanxi 712100, China)

Airborne LiDAR; filtering algorithms; terrain expression; microtopographic change monitoring; the Loess Plateau

备注

为了探究适合地形复杂区域机载LiADR点云数据的滤波算法,为微地形变化、地表过程的监测提供有效手段,以黄土高原沟壑区董庄沟流域为研究区,使用MCC,ETEW,ATIN,PM,SBF这5种滤波算法对流域内12个样本区域无人机LiDAR点云数据进行滤波,以国际摄影测量与遥感学会推荐的方法评价滤波结果的精度,分析了坡度、植被、点云密度对滤波精度的影响。结果表明:MCC和PM的Ⅰ类(地面点云错分比例)误差通常最低,ETEW和SBF最高,ATIN居中,各算法的Ⅱ类误差(非地面点云错分比例)和总误差(总体点云错分比例)大小顺序与Ⅰ类误差大致相反,PM在陡峭区域的点云滤波效果最优。坡度和植被覆盖度显著影响各算法的Ⅰ类误差。MCC和PM的Ⅰ类误差随坡度上升和植被覆盖度下降的速率最小,ETEW最大,SBF和ATIN居中。随着点云密度增加,MCC,PM,ATIN的Ⅰ类误差无明显变化,Ⅱ类误差和总误差缓慢下降,SBF和ETEW的Ⅰ类误差增加,Ⅱ类误差和总误差下降。总之,PM滤波算法最适用于地形复杂区域的滤波算法。
In order to explore the filtering algorithm which is suitable for the airborne LIADR point cloud data in the complex terrain area and provide an effective means for the monitoring of microtopographic changes and surface processes. In this article, five commonly-used algorithms(MCC, ETEW, ATIN, PM, and SBF)were employed to filter the point cloud data of 12 sample areas in a small catchment(i.e. Dongzhuanggou)of the gullied Loess Plateau. The accuracy of the five algorithms was evaluated based on the method suggested by the International Society for Photogrammetry and Remote Sensing, while its relationship with impacting factors(slope gradient, vegetation coverage, point density)were also assessed. Results showed that the type Ⅰ errors(misclassification ratio of ground points)of MCC and PM were generally lowest, type Ⅰ error of the ETEW and SBF were highest, and those of ATIN were in the middle. The sequence of type Ⅱ(misclassification ratio of non-ground points)and total errors(misclassification ratio of total points)of the algorithms were roughly opposite to that of the type I error. The PM, compared to other algorithms, yielded the best filtering results for steep-sloping areas. Slope gradients and vegetation coverage significantly affected type Ⅰ errors of the employed algorithms rather than type Ⅱ and total errors. The increase(decrease)rate of type Ⅰ error with slope gradients(vegetation coverage)was lowest for MCC and PM, and highest for ETEW, with that of SBF and ATIN being in the middle. With the increase of the point density, the type Ⅰ errors of MCC, ATIN and PM did not change apparently, and the type Ⅱ and total errors slowly decreased, while the type Ⅰ errors of SBF and ETEW increased, and the type Ⅱ errors and total errors decreased.