The Study On The Method Of Liveness Recognition Combined With Face Detection

In the field of biometrics,face recognition technology is widely used in identity authentication systems as it features high efficiency,stability and convenience.However,with the continuous development of technology,facial information has become more easily accessible and producible.In this case,non-bio face spoofing attacks have taken a heavy blow to the security of the identity authentication system.As the digital authentication is unable to verify a person through biometrics and the staff should not be disrupted when accessing the important platform-enabled website via personal computer in their daily routines,this paper focuses on the biological recognition method based on the face detection under the deep learning framework of Caffe.The paper briefly states the background,significance and status of the research both at home and abroad,looking in detail at the content of each chapter,introducing the current study on the technology of face detection,and giving an introduction of the deep learning platform.Given the deep learning framework,face data,bio and non-bio data are collected and calibrated,along with the analysis of models of face detection and biological recognition,in a way that helps improve the system,making it more real-time and accurate.And face recognition is realized through the training of SVM classifier.Last,establishing a biological recognition system with facial recognition to realize related functions of facial recognition and biological detection.When it comes to the problem of speed in the verification process and the prevention of over-fitting during the training process,the network parameters of the MTCNN are adjusted and cascaded with the convolutional neural network of AlexNet which is simultaneously upgraded.Designing an appropriate 3D convolutional neural network structure,which includes the volume of input,the number of layers,the convolution kernel size and the number of the network,etc.,so as to improve the speed and stability of recognition through the optimization of the network.The system,combined with Caffe deep learning framework and OpenCV to build a biological detection system for facial recognition of human,uses VC++,a strongly-typed language,for programming.The system mainly consists of image data acquisition,algorithm analysis processing,and basic interface control template of software.The results suggest that the system has made significant improvements in instantaneity and accuracy and can detect silently without being occluded.