Research On Face Recognition Based On Convolution Neural Network

In biometric technology,face recognition plays an important role because of its unique physiological characteristics.Human face features,as the most characteristic physiological characteristics of human beings,can be used as a standard for biometric recognition.Even twins,the similarity of the human face can not reach 100%.There are differences in any face.Face recognition can also quickly identify the user's identity under the condition of non user cooperation.This technology is friendly to users,and it is fast and accurate.Therefore,in the field of biometric recognition,face recognition technology has always been a very important recognition method.This paper mainly explores and studies face recognition technology from deep learning,and combines the basic face recognition algorithm process,through face detection and face key point localization to complete the overall face recognition algorithm.In the last stage of face image matching,we use convolutional neural network structure to train and recognize face images,so that we can get high accuracy and speed of face recognition.The main contents of the paper are as follows:(1)This paper first describes the main composition of the image based face recognition technology and the extraction method of the required features.The algorithm of face region detection,the algorithm of face region feature extraction and the algorithm of face recognition are introduced in detail.The three algorithms and the experimental process of the corresponding algorithm are obtained,and the experimental results are obtained for the following comparative experimental analysis.(2)This paper uses the FuSt structure detection algorithm to detect the face image in the database,and on this basis,identifies the face size.In the known face regions,five key points(nose tip location,mouth corner location and pupil location)were determined based on the key points location algorithm of CFAN structure,and the key point feature was used to test the subsequent face matching algorithm.(3)Convolution neural network structure using VGG-19 to realize the face matching algorithm,the first network structure of VGG-19 and the basic principle and implementation of description,then according to the feature vector extraction to the configuration parameters,so as to train the whole network structure,the facial image recognition process,and then complete the whole experiment process of face recognition,experiment results.(4)The algorithm proposed in this paper is tested in ORL database and AFW database.Compared with the above classical algorithms,the recognition accuracy and recognition speed of the algorithm are verified by cross validation.