一种基于激光点云数据的房屋轮廓线提取方法  

Building Roof Boundary Extraction from LiDAR Point Cloud

作　　者：蔡湛[1,2] 李如仁[1,2] 李新科[3] 吴华意[3] 

机构地区：[1]辽宁工程技术大学测绘与地理科学学院,辽宁阜新123000 [2]辽宁工程技术大学时空分析与建模研究中心,辽宁阜新123000 [3]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079

出　　处：《地理与地理信息科学》2013年第5期17-21,25,共6页Geography and Geo-Information Science

基　　金：国家科技支撑计划项目(2013BAK06B00)

摘　　要：提出了一种基于激光点云数据提取房屋轮廓线的方法。首先在屋顶激光点云数据中搜索平面距离最远的两个激光点,选择其一为起始点,根据相邻边缘点连线的一侧不存在激光点的原则,检测屋顶所有的边缘点;然后对边缘点分组,利用最小二乘直线拟合方法分别拟合各条轮廓线并进行规则化;最后,选择轮廓线最外侧的激光点,对各条轮廓线进行平移外扩,并通过相邻轮廓线相交确定屋顶角点的坐标。实验结果表明,与通常采用的对屋顶激光点云构建不规则三角网(TIN)获取边缘点的方法相比,该算法运行效率明显提高,检测到的屋顶边缘点数量更多,获取的屋顶轮廓线精度更高。A new algorithm to extract building outline from laser cloud points is presented in this paper. Firstly, search for two points which have the max distance between each other and pick one as a starting point. According to the principle that there is no point out of the line segment between two neighbor-edge-points, detect all edge points successively. Divide those edge points in different groups and then fit outlines by least square methods respectively. The angles of building roof are generally right an- gles, so we need to do regularization for the lines fitted before. Finally, extend the outlines through moving each line across cor- responding outermost point to acquire the final roof outlines. Coordinates of real corner points are obtained by intersecting two neighbor roof outlines. Compared with general methods of obtaining edge points through constructing the TIN or Delaunay tri- angles, the algorithm presented in this paper gives faster running efficiency, more edge points detected and higher precision of roof outlines.

关 键 词：LIDAR 房屋轮廓线 边缘点检测 轮廓线扩展 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]