Non-Rigid Point Set Registration Based on Global Prior and Local Structural Constraint

Coherent point drift (CPD) is a classic non-rigid point set registration algorithm. Inspired by the CPD idea, an improved CPD method is proposed in this paper. Firstly, we establish a global prior based on the graph feature to dynamically allocate Gaussian components. Secondly, a new neighborhood is defined to flexibly adjust the range of unevenly distributed points. Finally, a local structure constraint based on local neighborhood is proposed, which ensures the structure stability of the point sets. Experimental results on synthetic and real data sets show that the proposed method achieves good performance in degraded data, such as deformation, rotation, and noise.