Research Of Face Recognition Method Based On Convolution Neural Network

Face recognition technology is a hot research topic in the field of computer vision.it has been applied to criminal investigation,financial service,attendance system and other fields,and plays a very important role.At present,face recognition technology still faces challenges in terms of anti-interference and recognition speed.Convolution neural network(CNN),as one of the representative methods of deep learning,has shown great advantages in the field of image processing.In this paper,the method of face recognition based on convolution neural network was studied.Firstly,this paper analysed the method of face detection using cascaded convolutional neural network(MTCNN).After training the network,a cascaded network which can realize face detection was obtained,and the face image used in face recognition was extracted by the network.In order to reduce the computational complexity of face recognition network and improve the image quality,the extracted face samples were processed by image grayscale,image histogram equalization and image normalization.In order to study the factors that affect the performance of convolution neural network,a face recognition network model based on LeNet-5 model was built by using TensorFlow deep learning framework,and CASIA-FaceV5 face database was used for network training and learning.The network initialization method,convolution kernel size and activation function were studied.The experimental results show that the convergence rate and recognition rate of the network can be improved by using 7 × 7 size convolutional nuclear energy in combination with elu activation function.The optimized parameters obtained in the experiment were applied to the network,and the recognition rate of the optimized network on CASIA-FaceV5 face database can reach 99.5%.Finally,the face recognition method of convolution neural network combined with long-term and short-term neural network(CNN-LSTM)is studied on the basis of convolution neural network.Two face data sets,CASIA-FaceV5 and AR,were used to train and test the network respectively.The results showed that the face recognition method based on CNN-LSTM was compared with the face recognition method based on CNN.The former method can improve the convergence speed of the network and reduce the training time of the network.At the same time,the recognition rate is as high as 99.8%on the AR face database.The method is feasible.