Robust registration of surfaces using a refined iterative closest point algorithm with a trust region approach (Q513664)

Consider the problem of finding a rigid body transformation, consisting of a rotation and translation, such that a set of data points fits to a given surface under the transformation. The authors propose a trust region method combined with an M-estimation technique for updating the rigid body transformation which combines the advantages of the different convergence behaviors for both the point-to-point distance minimization and point-to-plane distance minimization iterative closest point algorithms using a robust criterion. Two different ways of bounding the proposed trust region are discussed and the convergence is analysed. Numerical results on some test problems are presented.