Research And System Design Of Deep Face Recognition

With the rapid development of Internet technology,deep learning technology which based on big data drive has gradually become one of the mainstream technologies in the field of artificial intelligence.The demand for intelligent monitoring systems has made face recognition technology widely recognized by academics and industry.The identity verification system based on face recognition technology is also gradually applied in our daily life.However,due to the large amount of computation,the system based on the deep neural network is mainly deployed on the remote server.The client needs to communicate with the server to obtain the recognition result.The computational complexity is high,and the time consumption is further increased during the transmission process.At the same time,most face recognition systems are short of face liveness detection or just use a simple liveness detection algorithm,which are easily broken by illegal users,and greatly increase the security risks of the system.In order to develop a safe and efficient face recognition system,this paper focuses on two key technologies in the face recognition system-face livingness detection and face recognition.Firstly,based on deep learning technology and traditional image processing technology,this paper proposes a multi-classifier fusion face liveness detection algorithm.The algorithm can effectively resist photo and video attacks in actual application scenarios by combining three types of detetors,eyeblink detection,border detection and moire pattern detection.Secondly,based on the model compression technique and the deep convolutional neural network,a face recognition model with less parameter and high recognition rate in face verification is trained.This model only takes about 128ms to extracts the deep feature of a face image when using PC(Intel(?)Core i5-4590 CPU@3.30GHz).At the same time,we also did some experiments on the problem of small sample face recognition,which improved the recognition rate of the model on small face database.Finally,this paper deploys the above-mentioned face liveness detection algorithm and deep face recognition network into the forward computing framework based on C++.At the same time,we use Qt to write the graphical user interface and use USB camera as the image acquisition hardware to construct a practical face recognition system.The system can implement face verification and face recognition,with high recognition accuracy and good user experience.