一种基于尺度变化的点云并行去噪方法

Point cloud parallel de-noising algorithms based on scale change

三维扫描设备获得的点云数据不可避免地存在噪声,为去除不同尺度的噪声、较好地恢复出点云数据并保持模型的几何特征,采用了一种基于噪声尺度变化的点云并行去噪方法。该方法对于大尺度噪声,采用半径滤波与改进的具有噪声基于密度的聚类(density-based spatial clustering of applications with noise,DBSCAN)算法去除;对于小尺度噪声,采用改进的双边滤波算法滤除,去噪后能使模型特征不被破坏。并且,采用八叉树并行化提高双边滤波的速度,对比传统双边滤波,去噪效果更好且去噪速度提高至120%。

Noise inevitably exists in point cloud data acquired by three-dimensional scanning equipment.In order to remove noise at different scales,recover point cloud data better and maintain geometric characteristics of the model,a point cloud parallel de-noising method based on scale variation of noise is adopted.For large scale noise,the radius filtering and the improved density-based spatial clustering of applications with noise(DBSCAN)algorithm are used to remove it;for small scale noise,the improved bilateral filtering algorithm is used to remove it,so that the model features are not destroyed after de-noising.Octree parallelization is used to improve the speed of bilateral filtering.Compared with the traditional bilateral filtering,the de-noising effect is better and the de-noising speed is increased to 120%.