基于改进DBSCAN和双边滤波算法的点云去噪

Point Cloud Denoising Based on Improved DBSCAN and Bilateral Filtering

为了提高激光雷达点云去噪效果,提出了改进DBSCAN和双边滤波算法。首先通过点云数据空间欧氏距离划分栅格,从而删除无效点云;接着基于距离-密度方法在密度较大的段内找到DBSCAN的eps值,避免错误值出现,使用自适应方法获得DBSCAN的min pts值;然后利用点云邻域的曲率以及法向夹角优化双边滤波权值因子,便于保持点云的细节特征;最后给出了算法流程。实验结果显示所提算法能够去除接近点云模型的噪声以及混杂在点云模型中的噪声,评价指标较优。

In order to improve the denoising effect of LiDAR point cloud,a point cloud denoising method based on improved DBSCAN and bilateral filtering is proposed.Firstly,the grid is divided by the Euclidean distance of the point cloud data space,so that the invalid points are deleted.Secondly,distance density method is used to find eps value of DBSCAN in the section with high density to avoid error values,and the adaptive method is used to obtain the min pts value of DBSCAN.Thirdly,bilateral filter weight factor is optimized by the curvature and normal angle of the neighborhood,thus maintaining the detailed characteristics.Finally,the process is given.Experimental results show that the proposed method can remove the close and mix noises,and the evaluation indices of the proposed method are better than those of other algorithms.