基于多重特征匹配的点云配准算法

针对最近点迭代（ICP）算法搜索匹配点对规则单一、准确度低的问题，提出一种基于多重特征匹配的点云配准算法。首先采用改进自适应八叉树算法分割点云，通过移动最小二乘法（MLS）对其叶节点进行局部拟合后，计算点的多重特征；然后提出了基于多重特征的点对相似度，选取满足相似度约束的点对作为匹配点对，进而求取旋转矩阵和平移矩阵实现点云配准。实验表明，该算法能在保持点云配准速度较高的基础上，有效提升配准的准确度，且准确度的提升幅度随着点集数量的增大呈升高趋势。

Point cloud registration algorithm based on multiple-feature matching

To solve the problem that iterative closest point(ICP) algorithm has a single feature for searching and low accuracy for registration, this paper proposed a point cloud registration algorithm that based on multiple-feature matching. It chose the improved adaptive octree algorithm to segment the point cloud. Then calculated the multiple features of the points after performed moving least squares(MLS) algorithm to fit the leaf nodes. Next, this algorithm introduced the point pairs similarity that based on multiple features to establish the matching points. Lastly, it computed the rotation matrix and translation matrix to achieve registration. Experiments show that this algorithm can effectively improve the accuracy of registration on the basis of keeping the point cloud registration speed high. And with the number of point sets increasing, the trend of accuracy for this method is increasing.