保留灰度特征的配准点云数据精简算法

Refinement with Gray Features Preserving for Registered Point Cloud Data

论文提出了一种有效的点云精简算法,用于处理配准后的点云数据,解决其存在的数据量过大、点密度不均等问题,同时可以保留点云的灰度和几何特征。对于给定点云,首先基于体素栅格生成一个粗点云。然后,通过不动点迭代算法,将粗点云投影到给定的点云上,用每点邻域内的近似重心点来代替原点位置。同时,在正则化约束条件下,将局部排斥力掺入到距离过近的点用于增加惩罚项,保证了点云的均匀分布。最后,利用KDTree查找重采样后点的最邻近点,将最邻近点的灰度信息映射至重采样后的点。实验结果表明,该算法在充分保留点云数据几何特征和灰度特征的前提下,能有效滤除配准重叠区域的冗余数据,且保持点云的均匀分布。

In this paper,an effective point cloud refinement method that allows is proposed to preserve gray features to pro⁃cess the multiple registered point cloud with redundant data and non-uniform densities.For a given point cloud,a coarse point cloud is first generated based on a voxel grid.By using fixed point iterative algorithm,the coarse point cloud is then projected onto the given point cloud and the origin point is replaced with the approximate center of gravity point in neighbor field of each point.Meanwhile,under the regularization constraint,local repulsive forces are incorporated to the points whose are too close to the point for increasing penalty item so as to ensure the uniform distribution of point clouds.Finally,KDTree searching is used to find the closest point to the resampled point and the gray level information of these closest points are then assigned to the resampled points.Experiment results have shown that the proposed method not only effectively filters the redundant data in the registration overlap re⁃gion,but also maintains the uniform distribution of point cloud under the condition of preserving the geometric and gray features of the point cloud data.