基于K4PCS与ICP算法在点云配准中的应用

Application of K4PCS and ICP algorithm in point cloud registration

针对生产后端产品互换性的精度要求,采用基于关键点的4点全等集合(Keypoint-based 4-Points Congruent Sets,K4PCS)算法和迭代最近点(Iterative Closest Point,ICP)算法。提出一种K4PCS-ICP点云配准算法,由K4PCS粗配准和ICP精配准部分组成,将K4PCS算法得出的坐标状态转换矩阵作为ICP算法坐标状态初始值,再根据前后2次的迭代误差和迭代次数等条件,输出最适合的转换矩阵。实验结果说明,K4PCS-ICP算法与独立的K4PCS算法和ICP算法相比较,有效提高了坐标状态矩阵的准配性,点云配准的精度得到明显改善。

In view of the accuracy requirements of production back-end product interchangeability,the Keypoint-based 4-Point Congruent Sets(K4PCS)algorithm and Iterative Closest Point(ICP)algorithm are adopted.This paper proposes a K4PCS-ICP point cloud registration algorithm,which consists of K4PCS coarse registration and ICP fine registration parts,takes the coordinate state conversion matrix obtained by K4PCS algorithm as the initial value of the coordinate state of ICP algorithm,and then outputs the most suitable conversion matrix according to the two iteration errors and iteration times.The experimental results show that compared with the independent K4PCS algorithm and ICP algorithm,the K4PCS-ICP algorithm successfully improves the accuracy of the coordinate state matrix,and the accuracy of point cloud registration is significantly improved.