面向对象的车载LiDAR点云滤波方法  

Object-oriented filtering method for mobile LiDAR point clouds

作　　者：房华乐[1,2] 林祥国[1] 段敏燕[1] 张继贤[1] 

机构地区：[1]中国测绘科学研究院摄影测量与遥感研究所,北京100830 [2]山东农业大学信息科学与工程学院,山东泰安271018

出　　处：《测绘科学》2015年第4期92-96,共5页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41371405);中央级公益性科研院所基本科研业务费专项资金项目(7771413)

摘　　要：针对经典的机载LiDAR点云数据滤波方法不适用于车载LiDAR点云数据滤波的问题,该文提出了一种基于点云分割的不规则三角网渐进加密滤波方法。首先,剔除粗差;然后,进行点云分割;最后,以分割对象为基本判别单元,迭代地进行地面对象的识别。采用两个场景的车载LiDAR点云进行滤波实验。实验结果表明,本文滤波方法的总误差明显小于经典的不规则三角网渐进加密滤波方法。The existing filtering methods for the airborne LiDAR point cloud are often unsuitable for the mobile laser scanning（MLS）point cloud.A segmentation-based filtering method was proposed for MLS point cloud in this paper,where a segment instead of an individual point was the basic processing unit.First,the outliers were removed.Second,point cloud segmentation was made.At last,TIN progressive densification was performed.In the progress of densification,judgment was performed on the ratio of ground points to the whole points in the segment.Point cloud data of two scenes was employed to test the proposed method.Experimental results showed that the total error of our method was significantly less than the total error of the classical TIN progressive densification method,which proved that our method was effective and accurate.

关 键 词：车载激光扫描 滤波 不规则三角网 点云分割 

分 类 号：P234.1[天文地球—摄影测量与遥感]