No abstract

Abstract-Point sets registration, also known as point matching, is to find the one-to-one correspondence between two point sets as well as the related transformation. In comparison with most state-of-the-arts handling the rigid transformation between two point sets, in this paper, we focus on the more complex nonrigid transformation by considering it as a linear assignmentleast square problem. We design a non-rigid point matching algorithm by adopting the Genetic Algorithm (GA) to find an optimal solution of the liner assignment-least square problem, where we define a series of excellent Genetic operators. A population initialization method and a special genetic operator in this paper significantly improve the performance of the GA. The experimental results using public 2D point sets justify that Genetic Algorithms based point matching algorithm can achieve good performance in terms of large scale deformation and rotation.