Research on the Algorithm for Off-line Rectification of Welding Seam Based on Improved ICP

In order to improve the low registration efficiency and poor accuracy caused by the size error and initial position's major difference of the workpiece in the registration process of off-line welding rectification, this paper adopts a coarse registration first and then fine registration's process strategy. On the basis of quaternion algorithm to achieve coarse registration, an iterative closest point (ICP) registration algorithm based on K-D tree nearest neighbor search method is proposed. In the precise registration algorithm, the K-D tree nearest neighbor search method is first used to improve the search speed of corresponding point pairs, and then the bidirectional search method is used to improve the registration accuracy. In order to verify the feasibility and effectiveness of the algorithm, the registration experiments are carried out on the point cloud data of circular weld, square weld and complex weld. The results show that the accuracy of the proposed method is 90% higher than the PCA algorithm when the time is similar, and compared with the traditional ICP algorithm, the registration accuracy and time of this method are improved by 35% and 55%. Therefore, the method presented in this paper has good efficiency, accuracy and stability in solving the problem of weld track point cloud data registration.