A Study Of Face Recogniyion Based On Collaborative Representation-based Classification

Face recognition has always been a hot topic in scientific research and an important topic in biometrics.In recent years,many researchers in the tireless efforts,put forward a lot of efficient algorithms,made a lot of achievements,and the face recognition developed rapidly.Collaborative representation classification is a fast and effective classification algorithm based on sparse representation classification.The cooperative representation method is classified by an l2-norm minimization method.Compared with the sparse representation algorithm,the cooperative representation method greatly reduces the computational complexity.The research content of this paper is mainly based on the identification of collaborative representation classification,The main research work is as follows:Firstly,this paper introduce the background,development history and current situation of face recognition,and the methods of several main technologies in face recognition such as subspace algorithm,support vector machine,neural network classification and sparse coding classification.In addition,this part also introduces commonly some database used to the face recognition experiment.Secondly,the collaborative representation algorithm based on dictionary extension and the face recognition algorithm based on PCA_LDA and cooperative representation are introduced respectively.The collaborative representation algorithm based on dictionary extension adds a feature matrix to the traditional training dictionary,which increases the proportion of the feature information in the dictionary matrix,thus improving the accuracy of recognition.The face recognition algorithm based on PCA_LDA and cooperative representation combined the feature subspace of the PCA algorithm and the LDA algorithm.The feature information of the face is compressed into a smaller subspace,and then the cooperative representation algorithm is used to classify.In the experiment part,ORL,AR,YALE,FERET and other face database are selected as the experimental objects,and the algorithm is compared with some traditional classical algorithms,and the recognition rate and recognition speed are very high.At the same time,Gaussian randomnoise and occlusion are added to the test images of some databases,and the influence of the two algorithms is analyzed.The algorithm has strong anti-interference ability to noise and occlusion.Finally,this paper summarized the algorithm in this paper,and expound the achievements of this paper and the shortcomings of this study.