Regularized Clustering Based Double Manifold Discriminant Dnalysis For Low-resolution Face Recognition

SSPP(single sample per person)face recognition technology aims to use one face image for identity confirmation.This technology is widely used in judicial law enforcement,monitoring and security,remote authentication,mobile payment,attendance system and other fields.However,in practical applications,the face images used for identity verification usually come from the ID card,which makes the sample available for training very limited,and leads to a significant decline in the recognition rate of traditional face recognition methods.Therefore,the recognition technology using a single sample has very important research significance and broad application prospects.Firstly,this paper introduces the research background and significance of face recognition with the condition of single sample.Then,the classical face recognition algorithms are analyzed in detail,such as principal component analysis,linear discriminant analysis,local binary pattern and related algorithms.By analyzing and summarizing the advantages and disadvantages of single sample face recognition method,this paper proposes a face recognition algorithm based on single training sample.When using a single sample image for training,due to too few training samples,the effective information for recognition is greatly reduced,which leads to the reduction of the recognition rate of the algorithm.In order to solve the problem of face recognition with a single sample,this paper establishes different manifolds based on global and local features to form a double-manifold structure.Finally,the effectiveness of the proposed method is verified by the experimental results on a single sample dataset.In order to solve the problem of face recognition with single face image in low resolution,this paper proposes a regularized clustering method based double manifold discriminant analysis(RC-DMDA).Firstly,regularization algorithm and clustering algorithm are used to effectively alleviate the over fitting problem in the case of single sample,and then two manifolds are established through different features,so as to obtain more available information and help face recognition.Finally,this paper uses ar face data set,FERET face data set,LFW database and other standard face databases to test the algorithm recognition performance under various conditions.Finally,the experimental results show that the proposed method not only has fast calculation speed,but also has higher recognition rate for low resolution.