Face Recognition Based On Fusing Multiple Bayesian Classifier

Face recognition is a hot research topic in the area of image processing,computer pattern recognition and biometric recognition.It is attracting widely attentions on account of its great academic and practical application value.In recent decades,face recognition technic has achieved great advance under the efforts of many scientific research works.However,face recognition still face serious challenges under the conditions of the users don't cooperate with the system,the sampling is unsatisfactory and the amount of face image is very huge.According to our research,for the database which has a large amount of face samples the distributed parallel processing mechanism of pattern classifier is propitious to settle the problem of matching face accurately in it.During this process,we should solve a few of key issuses,such as the training,designing and realizing of classifier.In this paper,in order to meet the requirements of matching complex facial image sample high-performance and realizing distributed parallel processing,we combined the trend of the development of the classifier in recent years in the field of face recognition,and then proposed a novel face recognition method.The detailed implementation process of this method is as follows.Firstly,27 landmarks were located based on a CLM(Constrain Local Model)model.Then,face patches centered on each landmark were extracted and further split into non-overlapping cells.These face patches'LBP(Local Binary Pattern)feature can be used for creating local Bayesian classifiers by doing Joint Bayesian training.And the local classifiers were integrated in the framework of logistic regression.Finally,a face verification model was taken shape.The specific research works includes:(1)we have proposed a method of extracting face feature through combining CLM with LBP.Compared with the traditional methods,such as the way to extracting global LBP texture feature and extracting Low-dimension structural feature based on the landmarks,our method could extracts the structural feature and texture feature in the meantime.And at the aspect of feature extraction it also could realize the reasonable compromise.(2)An original method of face recognition based on fusing local Joint Bayesian classifiers is put forward.In this method,distributed approach has been adopted during the period of training and testing local Bayesian classifiers.So,compared with using traditional global Bayesian classifier,the classifiers' accuracy of training and Real-time identification(testing)is improved.That is to say,we could use High-dimension LBP feature to train and test local classifiers in order to achieve the purpose of improving system performance.(3)we have presented a method to fusing local Joint Bayesian classifier based on a logistic regression model.Aming at the concrete application of face recognition,the way to modeling this model is given.