一种改进的跨源点云全局配准算法

An Improved Cross Source Cloud Global Registration Algorithm

为了实现具有不同分辨率和精度等特性的跨源点云高精度高效配准,提出了一种改进主成分分析跨源点云配准算法。点云局部数据模式不同,基于局部特征的全局算法配准效果较差,提出基于整体形状分析解决跨源点云配准问题。采用Householder矩阵性质保持PCA主轴方向为右手坐标系,筛去点云变换中的非刚性变换,完成跨源点云的快速全局配准。使用跨源点云库的chair2、lab2、pillow、dustbin对改进算法及一些全局算法进行对比实验,时间减少94%、96%、91%、96%,均方根误差减少60%、99%、70%、63%。对配准结果进行ICP精配准,误差和时间进行对比,改进算法具有良好的配准效果及鲁棒性。

An improved principal component analysis cross-source point cloud registration algorithm was proposed to solve the problem of cross-source point cloud registration with different resolution and precision.The registration effect of the global algorithm based on local features is poor due to the different local data modes of point cloud.The global shape analysis is proposed to solve the cross-source point cloud registration problem.Householder matrix property is used to keep the chirality of PCA coordinate system unchanged,and non-rigid transformation is screened out in point cloud transformation to complete fast global registration across source point clouds.chair2,lab2,pillow and dustbin of cross-source cloud library were used to compare the improved algorithm and some global algorithms.The time was reduced by 94%,96%,91%and 96%,and the root mean square error was reduced by 60%,99%,70%and 63%.ICP precise registration was performed on the registration results,and the error and time were compared.The improved algorithm still has good registration effect and robustness.