Robust iterative closest point algorithm with bounded rotation angle for 2D registration

The iterative closest point (ICP) algorithm is efficient to register point sets, but it is easily trapped into a local minimum. The difficulties of obtaining an optimal minimum are the variety of the transformation and finding a suitable initial value. This paper introduces an inequality constraint of the rotation angle into the least square model for 2D point set registration problem and then solves the new model by a more robust ICP approach which bounds the rotation angle of the transformation. In each iteration, a closed-form solution of the transformation is obtained according to the monotonicity of the objective function with respect to the rotation angle. The boundary of rotation angle and initial value are estimated by the principle component analysis. A series of experiments validate that the proposed method is much more robust.