Face Recognition with Occlusion using a wireframe model and Support Vector Machine

This paper proposes a face recognition system with the capacity of achieving high recognition rates, even if the face image presents significant occlusion, using a wireframe model and a Facefit tool that acquires a coordinate system with fiducial points that characterize the facial parts. Thereby, the occlusions are emulated by the selection of points that represent four cases of partial blockage of the face such as: occlusion in eyes area, occlusion in the area of mouth and occlusion in left and right part of face in a separately way. Computer simulations show that the proposed system provides recognition rates higher than 95% in each case, providing also stability to the identification system. Thus, the wireframe model presents a robust coordinate system that even in complex cases, as data loss, is able to obtain a fairly good recognition rate.