Registration and matching method for directed point set with orientation attributes and local information

Abstract Point set registration and matching are the basic problems of pattern recognition and computer vision. The key to solve these problems is to determine the correspondences between the two point sets and to describe the deformation between them. In this paper, we propose an optimization model for the registration and matching of 2D directed point sets with coordinate and orientation attributes ( x , y , θ ) . First, the thin-plate spline (TPS) function with coordinate and orientation attributes is derived by variational method to describe the directed point set deformation. Second, an optimization objective function with coordinates and angles is constructed for point set registration and matching. Finally, the objective function is solved by alternately obtaining the correspondences and describing the deformation between the point sets. In the algorithm, the initial solution of correspondences is obtained by utilizing the neighborhood information of the point set to make the algorithm more robust. Several registration and matching experiments were performed on the artificial point sets and FVC fingerprint image databases to verify the robustness, effectiveness, and accuracy of the algorithm. Compared with the current popular algorithm, the proposed algorithm shows higher precision and robustness.