Research And Design Of Face Recognition System Based On Deep Learning

With the continuous development of society,people's identity information becomes more and more important in production and life.Face recognition is currently a more advanced identity verification technology.It is not only a research hotspot of computer vision but also has been widely used in many fields such as security,finance,and e-government.It is extremely popular in the era of rapid development of science and technology.Theoretical and practical value.With the development of computing power of computers,deep learning algorithms have been widely used in face recognition in order to pursue higher recognition rates.Convolutional neural network algorithm is often used in face recognition in deep learning algorithms.Convolutional neural network can automatically learn features with powerful representation ability and improve the recognition rate of face recognition.In practical applications,face recognition also needs to deal with lighting,pose,and other issues in the actual application.It also requires corresponding auxiliary algorithms to deal with these issues.The face recognition algorithm in this paper,based on the convolutional neural network algorithm,is improved in order to deal with the illumination problem in face recognition.In this paper,an improved gamma transform is used to grayscale the different areas of the highlight and shadow regions of the face image.The interference in face recognition is greatly reduced in the image preprocessing.Then,the feature extraction of the face image is performed using the LBP feature that is robust to light,and the robustness of the face recognition system to the illumination is further improved.Finally,the convolutional neural network is used to accurately recognize the face image.After testing,the recognition rate of this algorithm for face recognition system under different lighting conditions has not changed much,and they are all above 96.3%.A face recognition system is implemented on Altera's DE1-SOC embedded development board in this paper.Qt and Opencv were ported on the DE1-SOC development board for software programming and image processing,and Google's Tensorflow machine learning platform was ported to run convolutional neural networks.Afer the tested algorithm is transplanted to the system,the system can accurately recognize the face.Each function module of the system can also operate normally.It is a face recognition system that meets the requirements of face recognition applications.