基于VG-DBSCAN算法的大场景散乱点云去噪

Large-Scale Scattered Point-Cloud Denoising Based on VG-DBSCAN Algorithm

针对城市环境下三维激光雷达(LiDAR)点云数据密度不均匀、离群噪点多而不利于后期点云帧间匹配的问题,提出一种应用于城市环境下大规模散乱LiDAR点云的离群噪点滤除算法。该算法对传统的基于密度的噪声应用空间聚类(DBSCAN)算法进行改进,通过对三维点云进行体素栅格划分,创建了一个由栅格单元组成的集合,以此大幅减小每个对象在数据空间中邻域的搜索范围。改进后的算法能够快速发现各个聚类,使目标点云与离群点分离,从而剔除点云中的离群噪点。实验结果表明:所提算法能够实时处理点云数据,在保证点云三维几何特征的同时能有效识别并滤除点云中的离群噪点,降低点云规模,加快点云后续处理的效率,使帧间匹配的精确度提高了2倍,且匹配耗时仅为去噪处理前的1/3。

Non-uniform 3 Dlight detection and ranging（LiDAR）point-cloud data with outlier noises are not conducive to interframe point-cloud-matching in urban environments.Thus,an outlier noise filtering algorithm for large-scale scattered LiDAR point-cloud in urban environments is proposed.This algorithm improves the traditional density-based spatial clustering of applications with noise（DBSCAN）algorithm by applying voxel-grid partitioning to the three-dimensional point-cloud to create a set of grid cells,which greatly reduces the search scope of each object′s neighborhood in the data-space range.The improved algorithm can quickly find each cluster,which separates the target point-cloud from the outliers,thus eliminating the outlier noise in the point-cloud.The experimental results show that the proposed algorithm can process point-cloud data in real-time,ensure threedimensional geometric features of point-cloud,effectively recognize and filter out outlier noise,reduce the scale of point-cloud,and speed up the subsequent processing efficiency of the point-cloud.Using this algorithm,the accuracy of matching between the frames is doubled,and the matching time is only one-third of the time before denoising.