A 3D face recognition method based on corresponding point

Against the problem of low recognition rate caused by facial expression influences and the problem of time-consuming in on-line recognition, a 3D face recognition method based on corresponding point is presented. First, applying CPD (coherent point drift) warp algorithm to pretreated template face and human model, the relationship of corresponding point between two models is established. Then, try to discard the instable part of point cloud in the face model caused by the change of expressions. Finally, calculate the similarity according to the matrix obtained from the corresponding point projecting to a particular sphere, which can be used to identify different faces. The simulation experiment on FRGC(face recognition grand challenge) V2.0 database demonstrates that the proposed method has a good online real-time performance and is robust to deal with expression changes.