基于点云孔洞边界检测的暗涵排口信息提取

Information Extraction of Culvert Outlet Based on Point Cloud Hole Boundary Detection

针对暗涵三维激光扫描模型的点云数据量庞大且噪音干扰严重,手动提取其排口信息效率低且易误判的情况,提出一种基于点云孔洞边界检测的暗涵排口信息提取方法。首先,根据点云平均距离建立球形邻域,并投影至邻域的微切平面内;其次,利用最大夹角和引力的复合准则判别边界点,并运用边界点聚类及平面拟合方法去除噪点和非孔洞边界;最后,根据边界点的凸包提取暗涵排口坐标、高程与尺寸信息。结果表明,球形邻域比kd-tree邻域表现出更强的鲁棒性;复合准则在计算时间仅增加5%的情况下,边界检测准确率比单一准则提高了3%,且误判率降低了近一半。该方法为暗涵排查信息化提供了一定参考。

In view of the large amount of point cloud data and serious noise interference of the 3D laser scanning model of the culvert,the manual extraction of outlet information is inefficient and easy to misjudge.In this paper,a method of extracting the information of the culvert outlet based on the hole boundary detection of the point cloud is proposed.Firstly,the spherical neighborhood is established according to the average distance of the point cloud,and projected into the tangent plane.Then,the boundary points are detected by the combined criterion of the maximum angle and force,and the noise points,and non-hole boundaries are removed by clustering and plane fitting.Finally,the coordi⁃nates,elevation and size information of the culvert outlet are extracted according to the convex hull of the boundary points.The results show that the spherical neighborhood is more robust than the kd-tree.When the calculation time of the combined criterion is only increased by 5%,the accuracy of boundary detection is 3%higher than that of the single criterion,and the misjudgment rate is reduced by nearly half.This method provides a certain reference for the informatization of culvert investigation.