一种基于地面激光雷达点云的建筑物外轮廓自动提取方法

A method for automatic extraction of building outlines based on ground laser radar point clouds

本文针对基于地面点云数据的建筑物轮廓自动提取,设计了一种由粗到精的提取方法。首先,将点云降维,在图像中进行轮廓优化和粗提取;然后,将欧式距离聚类和全局最小二乘方法与RANSAC方法结合,对提取出的粗轮廓进行更精细地拟合,得到原始的线要素;最后,采用平面几何拓扑约束和经验模型的方法,对得到的线要素进行简化和连接,完成轮廓的提取。试验结果表明,本文方法能够有效提取出建筑物轮廓,与传统方法实测相比,轮廓顶点处的位置精度总体小于5 cm,能够满足大多数城市测量应用场景。

This paper designs a coarse-to-fine extraction method for automatic extraction of building outlines from ground point cloud.First,the point cloud is converted into a binary image,where outline optimization and coarse extraction are performed.Then,a combination of Euclidean distance clustering,global least squares method,and RANSAC is used to fit the coarsely extracted outline more precisely to obtain the original linear features.Finally,plane geometric topological constraints and empirical models are used to simplify and connect the obtained linear features to complete the outline extraction.Experimental results show that the algorithm can effectively extract building outlines.Compared with traditional methods,the position accuracy at the outline vertices is generally less than 5 cm,which can meet the requirements of most urban surveying application scenarios.