邻域自适应的三维点云滤波算法

针对现有的点云滤波算法存在的精度丢失和收缩的不足,提出邻域自适应选择的算法,有效地改善了点云滤波中丢失精度的问题.算法首先针对原始点和均值点滤波出现的收缩问题,提出混合增采样策略.其次采用邻域自适应选择保持特征部分的滤波精度.最后定义每个采样点以对应的似然函数,并按照其梯度方向进行迭代,通过最大似然估计得到最优滤波结果.实验部分表明,本文三维点云滤波算法对点云滤波精度的保持具有更好的效果.适用于工业生产与检测领域的三维扫描.

3D Point Cloud Filtering by Adaptive Neighborhood

Existing point cloud filtering algorithms will shrink,the filtering result affect the reconstruction precision.In this paper,a new hybrid point cloud filtering algorithm is proposed to improve the filtering accuracy.Firstly,we define a hybrid point cloud by combing mean point and original point.Secondly,the neighborhood of filtering point is adaptively selected using normal and curvature information to preserve local features.Finally,a likelihood function is solved,which obtain the filtering position of each point.The comparison experiments show that this algorithm can improve the accuracy of the existing point cloud filtering algorithm.