Research on Multi-pose Face Recognition based on Block Occlusion

Aiming at the low accuracy of face recognition under occlusion conditions with different poses, the paper proposes a block occlusion multi-pose face recognition method based on feature point extraction. First, the face is divided into blocks according to the occluded area. Different weights are given to different face blocks by adaptive weighting method. The occluded face recognition is completed by using convolution neural network, which can effectively reduce the impact of pose changes and occlusion on face recognition. The research shows that this method has higher recognition rate and better robustness for multi-pose occluded faces.