Face Recognition Based On Gabor Local Feature And Deep Convolution Neural Network

Relying on these core technologies,the technology of the face recognition has reached a new level.It has become an important way to verify personal identity information.In particular,the research of face recognition technology based on deep learning neural network has made great progress in recent years.In this paper,a new face recognition method is proposed on the basis of deep learning neural network.In the construction of network models and the design of algorithms,there are three main improvements in this paper.1.Optimiz the input image.In the traditional face recognition algorithm,the input of the neural network is the face target image,which not only makes too much redundant information in the network,but also greatly increases the computation of the network.Before the face target image is input,the image is divided into blocks,and the local features of the face are extracted by the two-dimensional Gabor transform.2.Multitask training.Compared with the classical method,the model training function of the existing methods is relatively simple,and can not perform different recognition tasks at the same time.This paper combines the detection,recognition and classification tasks of the target image,and trains the network model.It is close to the performance of the existing system in recognition rate performance,and has a certain practical application value.3.Optimize the network structure.In face recognition,the structural design of the network model is the key.At present,the structure design of the mainstream network model is more complex,and the requirement of the hardware performance is high.Therefore,in the experiment,the time consuming of the network training and the actual detection effect can not be coordinated.Based on the traditional CNN structure,this paper further increases the number of network layers.The model tests were carried out based on the public face database ORL,Yale and Extended Yale B.Compared with the related algorithm.It can be seen from the experimental resultst that the recognition speed and recognition accuracy in this method had been significantly improved.And the algorithm was better than the classical algorithm such as LBP-DCNN,Gabor Faster R-CNN,Gabor-BP classification methods.The average recognition rate can reach 98.33%when the network model was used for face recognition.