一种改进ICP点云配准方法的研究

An Improved ICP Point Cloud Registration Method

点云配准是计算机视觉中一个基本而又重要的研究课题。针对现有配准算法对初值敏感、特征描述符普适性差的问题,提出了一种Harris3D与改进RANSAC结合粗配准和基于新型加权因子与新型特征描述符ICP精配准的两步配准法。改进RANSAC方法不断迭代,为精配准提供良好的位姿初值。点云的法线计算可充分描述点云特征的描述符与加权因子。在精配准中根据特征距离查询最近点,不断计算点云间特征距离,根据3σ准则剔除误匹配点对,从而实现加快收敛和提高精度的效果。结果表明,该算法相比传统ICP算法,收敛时间仅为其20%,使最终的配准误差降低至0.008 mm以下,可对一般点云进行快速坐标系对齐。

Point cloud registration is a basic and important research topic in computer vision.Aiming at the problems of existing registration algorithms that sensitive initial values and poor universality on feature descriptors,this paper proposes a two-step registration method including manual rough registration and ICP fine registration based on new weighted factor and new feature descriptors.The normal calculation of the point cloud can adequately describe the characteristics and weighting factors of point cloud descriptors.In the precision registration,the nearest point is queried according to the feature distance,the feature distance between point clouds is constantly calculated,and the mismatched point pairs are removed according to the 3o criterion,thus achieving the effect of accelerating convergence and improving accuracy.The results show that compared with the traditional ICP algorithm,the convergence time of the proposed algorithm is only 20%,and the final registration error is reduced to 0.008 mm.