Generalized Iterative Closet Point Algorithm Based on Lie Algebra Parameterization

摘要

The Generalized Iterative Closest Point(G-ICP) algorithm is a high-precision point cloud registration algorithm, which uses Euler angles to parameterize a rotation matrix. However, Euler angle parameterization has singularity at a specific angle, which leads to application limitations. In this paper, a new G-ICP algorithm has been proposed, in which the Lie algebra on so(3) is used to parameterize a rotation matrix and the Gauss-Newton method is employed to optimize the objective function of the G-ICP algorithm. The experimental results have demonstrated that our method performs well compared with the state-of-the-art G-ICP alorithm [1] and simultaneously holds better estimation accuracy of pose with fast convergence speed in the open RGB-D TUM datasets.