Under Occlusion Research On Face Recognition Algorithm

With the development of the society,face recognition has been widely used in identity authentication,human-computer interaction,video surveillance and so on.Face recognition technology plays a huge role in all walks of life,but it also has many problems to be ironed out.This paper mainly studies the problem of face recognition under occlusion condition,and presents a series of research work from three aspects: processing ability of continuous occlusion,feature extraction of human face,image segmentation of occlusion and non-occluded region.The main contributions and innovations of this paper are as follows:(1).In this paper,the traditional feature extraction method is studied and the effect of the number of eigenvectors on face recognition is analyzed by eigenvector selection experiment.This experiment shows that the PCA algorithm ignores the difference between different samples,and lacks the representation of local feature information.In addition,two-dimensional Gabor function is analyzed by similarity matching experiment and local face feature detecting experiment.The results of these two experiments show that the two-dimensional Gabor,as the kernel function of mathematical transformation,can extracts the local feature information of the face well.(2).A new face recognition algorithm based on the improved PCA algorithm based on the dual attribute model,and then the local 2D Gabor algorithm(Double Attribute Model based Gabor,DAMG)is proposed to solve the face recognition problem from the perspective of linear subspace The algorithm integrates the global feature vector and the error feature vector to generate a double-attribute eigenvector based on the double-attribute model,extracts the segmented feature of the target image according to the two-dimensional Gabor and design a overall classifier to weighted classifing double-attribute feature vector and localized block vectors In this paper,the analysis of DAMG algorithm's performance and recognition errors' s samples show that the algorithm has very good robustness in the occluded face recognition process.(3).In this paper,a dual-weighted error distribution model based on CV model is proposed,which solves the problem of face recognition under occlusion from the perspective of high-dimensional image representation.First of all,the occlusive image is segmented based on CV(Chan-Vese)model to get the error images in different regions.Secondly,the conditional probability error model based on the gradient direction is revealed.Compared with the low-dimensional feature vector,the conditional probability error model.Finally,according to the occluded and non-occluded regions of the image error distribution,the distribution models of the two errors are deduced separately to form a double weighted error distribution model.In this paper,three experiments of face image's random occlusion,facial feature occlusion and real face occlusion are carried out to verify the effectiveness of the proposed algorithm and DAMG algorithm in face recognition under occlusion conditions.