Abstract: Aiming to solve the problem that the high-precision map data collected at different times or different devices cannot be matched well, this paper proposes an algorithm of realizing high-precision map registration using the regularized local-weighted thin-plate spline algorithm. The K-D tree is used to search the local neighborhood control points to generate a regularized local thin plate spline transformation field. The elements to be registered are locally weighted to generate the post-registration elements. The experimental results show that this algorithm is superior to the traditional linear least squares algorithm and improves the registration success rate.

Key words: high precision map, registration, local weighted thin-plate spline, regularization, K-D tree