移动最小二乘增量式多视点云数据融合算法

Moving Least Squares Based Incremental Multi-View Range Images Integration Algorithm

在散乱数据点移动最小二乘曲面拟合的基础上,提出了一种增量式多视点云数据融合算法.将算法中多视点云数据作为对同一物体表面二维流形的一次采样,采样数据中包含匹配误差、冗余和畸变,把多视点云数据融合问题转换为由包含误差的散乱数据点恢复二维流形的过程.对每一幅当前处理的点云,寻找当前点云与已增量式融合的点云数据的重叠部分,在重叠部分数据集上构造移动最小二乘曲面,将重叠部分的每一个在移动最小二乘曲面上的对应点合并到当前已增量式融合的点云数据集中,从而实现了增量式多视点云数据的融合.实验证明,该算法是一种有效的多视点云数据融合算法,并且可从较大匹配误差、噪声、畸变的多视点云数据中获得较好的融合效果.

An incremental multi-view range images integration algorithm is proposed based on scattered points moving least squares surface fitting. Muhi-view range images are considered as the once samples of 2D manifold, the sampled data may contain registration errors, deformation, and redundant, etc. Whole multi-view range images integration process is considered as a process of finding the optimized fitting smooth 2D surface from those scattered point sets with errors, deformation and redundant. For every current range images, the overlapped section among current range images and the range images set that has been incrementally integrated together is detected. A moving least squares （MLS） fitting surface is constructed over the overlapped section. The corresponding points of the overlapped section on the moving least squares surfaces are merged to the integrated range images set. The Implementation on real scanning 3D data indicates that this MLS based incremental multi-view range images integration algorithm is effective, especially for the range images with large registration error, deformation and redundant.