直线段检测算法在机载LiDAR建筑物轮廓线提取中的应用  

Application of line segment detector algorithm in extracting building contour lines from airborne LiDAR

作　　者：刘建兴 胡全一 乐美燕 舒斯红 LIU Jianxing;HU Quanyi;LE Meiyan;SHU Sihong(Jiangxi Nuclear Industry Surveying and Mapping Institute Group Co.,Ltd.,Nanchang 330038,China;Jiangxi Nuclear Industry Geological Survey,Nanchang 330038,China)

机构地区：[1]江西核工业测绘院集团有限公司,江西南昌330038 [2]江西省核工业地质调查院,江西南昌330038

出　　处：《测绘通报》2023年第10期145-149,共5页Bulletin of Surveying and Mapping

基　　金：南昌市高层次科技创新人才“双百计划”创新团队(培养类)(洪科字〔2021〕156号)。

摘　　要：针对机载LiDAR点云数据提取建筑物轮廓线耗时多且精度不高的问题,本文提出了一种基于直线段检测(LSD)的机载LiDAR建筑物轮廓线提取方法。该方法首先对已分类的建筑物点云进行栅格化得到二值图;然后对二值图进行膨胀、腐蚀操作,消除二值图中因栅格化产生的空洞;最后利用LSD算法进行直线检测获取规则的建筑物轮廓线。经过实测数据的验证,本文方法可以检测到亚像素级的建筑物轮廓线,与传统的Canny算法相比能够提高约50倍的效率。To address the problem of time-consuming and low accuracy of extracting building contour lines from airborne LiDAR point cloud data,this paper proposes an(line segment detector)LSD-based method for extracting building contour lines from airborne LiDAR.The method first rasterizes the classified building point clouds to obtain the binary image.Then,the binary image is expanded and eroded to eliminate the voids generated by the rasterization.Finally,the LSD algorithm is used for straight line detection to obtain the regular building contour lines.Using the measured data for verification,the proposed method can detect the building contour lines with sub-pixel accuracy,which is about 50 times more efficient than the traditional Canny algorithm.

关 键 词：二值图 LSD算法 建筑物轮廓线 机载激光雷达 点云 

分 类 号：P237[天文地球—摄影测量与遥感]