一种预拼装钢构件的点云自动分割算法  

An automatic point cloud data segmentation algorithm for pre-assembled steel structures

作　　者：李韦童 邓念武[1] LI Weitong;DENG Nianwu(School of Water Conservancy and Hydropower,Wuhan University,Wuhan 430072,China)

机构地区：[1]武汉大学水利水电学院,湖北武汉430072

出　　处：《武汉大学学报（工学版）》2022年第3期247-252,共6页Engineering Journal of Wuhan University

基　　金：国家自然科学基金项目(编号:51278387)。

摘　　要：针对由非完整圆柱和特征相似平面组成的预拼装钢构件,提出了一种点云自动分割算法。首先通过降维投影对预处理后的空间点云做基于边界识别的空间分区,再采用随机采样一致性算法估计各分区满足几何约束条件的初值模型,并结合最小二乘拟合完成模型点云的二次过滤,达成对平面及圆柱模型点云的完整提取。在处理平面点云数据时,引入基于密度的空间聚类以及抽样点距离约束来实现几何特征相似平面的有效分割。工程数据验证表明,该算法能够自动、准确地分割出构件的圆柱与平面点云数据,具有较强实用价值。An automatic point cloud segmentation algorithm is proposed for pre-assembled steel structures which are composed of incomplete cylinders and planes with similar characteristics.After dimensionality reduction projection,the pre-processed space point cloud is spatially partitioned by means of boundary recognition.The initial point sets of the planar and cylindrical elements in each sub-block are extracted separately by the random sampling consensus algorithm to satisfy geometrical constraints,and secondary filtering based on least squares fitting of the initial point set can make the complete extraction of plane and cylinder model point cloud.Densitybased spatial clustering and distance constraint of sampling point are introduced to extract planes with similar geometrical characteristics when processing plane point cloud data.The experimental result shows that the algorithm can accurately and automatically segment the cylindrical and planar point cloud,so it has a strong practical value.

关 键 词：三维激光扫描 点云分割 预拼装钢构件 随机采样一致性算法 

分 类 号：P232[天文地球—摄影测量与遥感]