一种保留特征点的大数据量点云分类精简算法

A streamlined algorithm for large data point cloud classification with preserving feature points

针对桥梁等大型结构点云模型存在数据量过大、特征信息提取不准确等问题,该文提出了一种保留特征点分类精简的方法。对点云数据进行体素下采样预处理;空间栅格均匀划分点云数据,并以每个小栅格为单位计算曲率描述子并优化栅格,将点云数据快速划分为特征区和非特征区;利用曲率和K邻域法向量夹角的平均值把特征区点云划分为特征点和非特征点;对不同类点云采用不同的精简策略并整合在一起,较好地解决了传统精简算法在海量点云精简速度和精度上难以兼顾的不足。对某桥大型拼装构件的试验结果表明,在较高精简度下,所提算法较传统精简算法,能够有效避免孔洞现象发生,在精简速度和精度上都取得了较好的效果。

Aiming at the problems of excessively large data volume and inaccurate feature information extraction in point cloud models of large structures such as bridges,a method for classification and simplification of retaining feature points was proposed.First,voxel downsampling preprocessing on the point cloud data was performed;secondly,the spatial grid divided the point cloud data evenly,and calculated the curvature descriptor with each small grid as the unit,optimized the grid,and quickly divided the point cloud data into Characteristic area and non-characteristic area;Then,the point cloud in the characteristic area was divided into characteristic points and non-characteristic points using the average value of the curvature and the normal vector angle of the K-neighborhood;Finally,different simplification strategies were adopted for different types of point clouds and integrated Together,it better solved the shortcomings that traditional simplification algorithms could hardly take into account in the speed and accuracy of massive point cloud simplification.The test results of a large-scale assembly component of a bridge showed that,at a higher degree of simplicity,the proposed algorithm could effectively avoid the occurrence of holes compared with the traditional simplified algorithm,and had achieved better results in both the speed and accuracy of the streamlining.