融合曲率信息的点云配准方法

Point cloud registration method integrating curvature information

针对点云配准时受原始位姿局限及配准效率、鲁棒性低的问题,提出一种融合曲率信息的点云配准方法。首先,将海量点云进行重心邻近点体素下采样,采用融合曲率信息的提取算法提取特征点;其次,通过三维形状上下描述符进行特征描述并利用改进的随机抽样一致性算法进行粗配准;最后,在具有较好位姿的情况下采用Symm-ICP进行精配准。试验结果表明,该算法对于不同位姿的点云数据均保持较高的配准精度。本文提出的算法配准效率优于其他算法,并具有较好的鲁棒性。

In this paper,a point cloud registration method integrating curvature information is proposed to address the problems of the original posture limitations,low registration efficiency and robustness of point cloud registration.Firstly,the massive point cloud is downsampled from the voxels adjacent to the center of gravity,and the feature points are extracted by the extraction algorithm that incorporates curvature information;secondly,the feature description with 3D shape context and coarse registration using an improved RANSAC;finally,the Symm-ICP is used for fine alignment under the condition of good posture.The experimental results show that the proposed algorithm maintains high alignment accuracy for point cloud data with different positions.The alignment efficiency of the proposed algorithm is better than other algorithms and has better robustness.