Research On Face Recognition Based On Local Occlusion

Identification and authentication of identity is an important research topic in the Information Age,Face recognition technology has great advantages in the field of identity authentication.Besides,face recognition plays an important role in the fields of human-computer interaction,network technology,public entertainment,video surveillance and so on.However,it faces many difficulties to be overcome.This paper focuses on the face occlusion in the face recognition problem in depth.From the defects of the traditional algorithm in the face of the occlusion problem,the new algorithm is fused to deal with the occlusion.The main contributions and innovations of this paper are as follows:(1)This paper mainly studies the traditional feature extraction method and carries on the related experiment.In the experiment,the recognition ability of the traditional algorithm under different eigenvector is analyzed,and it is proved that the feature extraction of PCA and SVD algorithm is based on the global feature of the image,and the difference information between the samples is not fully utilized,which leads to the defect of the algorithm in the local feature expression ability.The experimental analysis of two sets of local features shows that Gabor features can accurately represent the local feature information of the image.(2)A sparse representation face recognition algorithm based on Gabor transform is proposed to solve the problem of face image occlusion,and a better experimental result is obtained by using Gabor Multiscale filtering algorithm.In this algorithm,the image layering is used to make Gabor filter on each layer of image,and then the sparse representation of the base function of discrete wavelet transform is applied to face image recognition.Finally,by setting up two groups of experiments on AR face database,this paper analyzes the performance of algorithm recognition,and verifies the robustness and effectiveness of the algorithm under real face occlusion.(3)an improved face recognition algorithm based on PCA and HOG segmentation is proposed.The proposed algorithm firstly compensates the global features by using the double-attribute fusion model,and then extracts the local HOG feature information uniformly on the target image.Then,a global classifier is used to weight the fused global feature information and local HOG feature information.performance experiments and random occlusion experiments on ORL and Yale-B face database show that the algorithm is effective in face occlusion with face occlusion.