Research Of Face Recognition Based On Sparse Representation

In recent years,face recognition technology has been developed by leaps and bounds.Face recognition is widely used in many fields such as finance and security as a means of identity authentication.In practice,the recognition effect of face is affected to illumination,posture,expression changes and occlusion easily,resulting in a decrease in recognition rate and making a certification error.In the existing face recognition algorithm,the face recognition method based on sparse representation has certain advantages to solve the above problems.It does not need feature selection and has simple model with a certain robustness to complex environment.But the method still has some shortcomings,such as can not make full use of some local information and in the complex environment,the robustness has still to be improved.Aiming at the above problems,this paper based on the purpose of improving the robustness of face recognition,and the face recognition algorithm based on sparse representation is improved with a face recognition algorithm based on sparse representation is proposed.The main research work in this paper is as follows:(1)In order to improve the robustness of the algorithm under illumination change,this paper proposes a sparse representation of face recognition algorithm based on LBHOG feature dictionary initialization.In this method,the gradient calculation formula is improved by using B-spline filtering to obtain the BHOG gradient feature which is not sensitive to illumination.And then combine the BHOG feature with the local LBP feature to overcome the problem that the sparse representation classification model lacks local information to a certain extent.(2)In order to improve the sparse representation of the face recognition algorithm in complex environment robustness,combining the LBHOG local dictionary and 2DPCA global dictionary,the paper proposes a multi dictionary sparse representation of face recognition algorithm based on hierarchical discriminant.The algorithm overcomes the problem that the traditional sparse representation is influenced by gesture,expression change and occlusion,and improves the robustness of the algorithm in complex environment.(3)In this paper,the improved algorithm is applied to the field of payment authentication,and the Web-based face recognition payment authentication system is designed to achieve the basic face payment function.This paper experimented on multiple face databases such as CAS-PEAL,Yale B,ORL and Yale.the experimental results prove the effectiveness of the proposed algorithm.