基于深度差的离群点识别与修正方法

离群点的存在会给点云后续处理与应用带来干扰,文中针对扫描过程中快速通过的车辆或行人造成的离群点提出基于深度差的识别与修正方法.分析此类离群点的产生机理,并运用基于球面投影的点云构网方法构建点云的拓扑关系,在此基础上提出根据深度差识别离群点的方法以及兼顾不同点云质量的阈值自动选择方法;设计了基于二次曲面拟合的点云修复方法;利用真实隧道数据进行实验,实验结果验证了文中方法的有效性和稳定性.

Algorithm of recognizing and correcting the outlier based on depth difference

The existence of outlier will bring interference to the following processing and application of the point cloud. For the outlier caused by the fast passing of car or pedestrian during the scans the algorithm of recognizing and correcting is proposed based on the depth difference. First, the mechanism is analyzed by the use of triangulation method based on the spherical projection to build the topology of the point cloud; then an algorithm using the depth difference to identify the outlier is proposed which can select corresponding threshold from different point cloud automatically~ finally, a method is designed to＇repair the point cloud based on the quadratic surface fitting. Using the real tunnel data in the experiment, the results show that the algorithm has higher stability.