摘要
针对建筑物点云提取不完整问题,笔者采用一种组合方法从影像密集匹配获得的多视图像(MVS)点云中提取建筑物点云。首先运用布料模拟滤波(CSF)算法进行地面点滤波,去除MVS点云中的地面点;然后根据MVS点云的颜色信息,利用过绿指数(EXG)和植被密集成块特性将植被点云剔除;最后使用密度聚类从剩余的点云中分割出建筑物点云。结果表明采用该方法提取建筑物点云的正确性为98.06%,完整性为98.20%,质量为96.34%。相较于使用单波段阈值分割剔除植被点的组合方法,该方法在建筑物完整性和质量上提升超过26%。
To solve the problem of incomplete extraction of building point clouds,a combined method was used to extract the building point clouds from the multiple view stereo(MVS)point clouds that is obtained by image dense matching.The ground points in MVS point clouds were removed by the cloth simulation filter(CSF)algorithm.Then according to the color information of MVS point cloud,the vegetation point clouds were removed by the excess green index(EXG)and the characteristic of vegetation gathering together.After that,the building point cloud was separated from the remaining point cloud by density clustering.The results showed that the accuracy,completeness and quality of this method was 98.06%,98.20%and 96.34%,respectively.Compared with the combined method of single-band threshold segmentation to remove vegetation points,this method improves completeness and quality of the building by more than 26%.
作者
余和顺
刘荣
YU He-shun;LIU Rong(School of Geomatics,East China University of Technology,Nanchang 330013,China)
出处
《东华理工大学学报（自然科学版）》
CAS
2021年第2期168-173,共6页
Journal of East China University of Technology(Natural Science)
基金
国家自然科学基金项目(41206078)。
关键词
倾斜摄影
密集匹配
滤波
过绿指数
建筑物点云
oblique photography
dense matching
filtering
EXG
building point cloud
