摘要
针对传统建筑物轮廓提取流程中的算法复杂和实用性受限等问题,该文提出一种基于投影后点云密度差异的建筑物轮廓提取方法。利用无人机搭载的激光雷达能够获取相对完整的建筑物侧面点云,使得建筑物轮廓和植被点云在水平投影上密度较高。采用统计滤波算法(SOR)剔除低密度点云区域。对于剩余的植被点云,利用相邻向量角度特征剔除植被内部点云。通过基于随机采样一致(RANSAC)的算法提取初始建筑物轮廓,最后用半径滤波算法剔除非建筑物点云线段。在3个测试区域中进行实验,召回率、F1分数、质量因子的最小值分别为84.07%、0.91、0.83。将区域3与实际地形图数据对比后,面积和周长平均相对百分比误差分别为0.23%、0.14%,坐标平均误差0.062m。验证了该方法能够有效提取建筑物轮廓。
Aiming at the problems of complex algorithms and limited practicality in the traditional building contour point cloud extraction process,a building contour extraction method based on the density difference of the point cloud after projection was proposed.The use of UAV-mounted LiDAR enabled the acquisition of relatively complete point clouds of the sides of buildings,resulting in a high density of building contours and vegetation point clouds in horizontal projection.A statistical outlier removal(SOR)based algorithm was used to reject low density point cloud regions.For the remaining vegetation point clouds,neighboring vector angle features were used to remove the vegetation interior point clouds.For the remaining vegetation point clouds,neighboring vector angle features were used to remove the vegetation interior point clouds.The initial building contours were extracted by an algorithm based on random sample consensus(RANSAC),and finally the non-building point cloud line segments were eliminated by a radius filtering algorithm.Experiments were conducted in three test regions and the minimum values of recall,F1-Score,and quality factor were 84.07%,0.91,and 0.83,respectively.After comparing area 3with the actual topographic map data,the average relative percentage errors in area and perimeter were 0.23%and 0.14%,coordinate average error of 0.062m,respectively.It was verified that this method could effectively extract the building contour.
作者
陈杰
刘茂华
赵占杰
李政霖
焦力
CHEN Jie;LIU Maohua;ZHAO Zhanjie;LI Zhenglin;JIAO Li(School of Transportation and Geomatics Engineering,Shenyang Jianzhu University,Shenyang 110168,China;Chinese Academy of Surveying and Mapping,Beijing 100036,China;College of Geodesy and Geomatics,Shandong University of Science and Technology,Qingdao,Shandong 266590,China)
出处
《测绘科学》
CSCD
北大核心
2023年第12期188-200,共13页
Science of Surveying and Mapping
基金
国家自然科学基金项目(4210010785)
住房和城乡建设部科学计划项目(2022-R-063)
辽宁省社会科学规划基金项目(L21BGL046)
