Research Of Face Recognition Algorithm Based On Improved CS＿LBP And LDP Feature Extraction

Face recognition technology refers to a high-tech technology that analyzes face images through the internal logic structure of the computer and extracts effective face description features for identity screening.Face recognition technology has become a research focus in the fields of artificial intelligence,computer vision and pattern recognition due to its wide application prospects.In face recognition technology,how to choose face description features and how to extract face features effectively is an important part of research.In recent years,the feature extraction method based on Local Binary Patterns(LBP)has achieved high performances in the field of face recognition,so it has been deeply studied by scholars.In this paper,the face recognition technology is systematically studied and researched in depth.The paper focuses on the LBP algorithm,the CS_LBP algorithm and LDP algorithm proposed on the basis of LBP algorithm to carry on analysis and research,and does research and improvements in the following aspects.Firstly,the basic definition,operation steps,pattern development,improved methods,advantages and disadvantages of LBP feature description operator are analyzed,and the face recognition process and basic principles based on LBP operator are summarized.The gray invariance,rotation invariance and uniform pattern of the classical LBP operators are mainly studied,as well as the advantages and disadvantages of each pattern.Then,the two improved algorithms of LBP,namely CS_LBP algorithm and LDP algorithm,are studied.Based on their respective advantages and disadvantages,by internally and externally combining CS_LBP operator with LDP operator,Adaptive Weighted Center Symmetric Local Directional Patterns(A_WCSLDP)and feature fusion algorithm based on improved CS_LBP operator and LDP operator are proposed respectively.In addition,the adaptive threshold is defined in the A_WCSLDP description operator,which enhances the robustness of the algorithm.Moreover,for the A_WCSLDP algorithm and feature fusion method,an adaptive weighting processing operation based on image sub-block information entropy is proposed,which makes the texture information of different regions in the face image more accurately described.Finally,comprehensive comparison verification experiments on CMU_PIE face database and YALE face database prove that the A_WCSLDP algorithm and feature fusion method have good superiority in terms of recognition rate and robustness against noise pollution,compared with LBP8,1,LBP8,1u2,CS_LBP and LDP algorithm.