Research On Face Recognition Access Control System Based On Double Cameras

With the improvement of people's safety awareness,the face recognition access control system has been widely used in train station entrances,airport terminal gates,government agencies,banks and other places with a high security level,however,for these places with a high security level,if only the traditional color camera is used for face recognition,then photofaces and other fake faces are easily mistaken for living faces to identify.This will lead to incorrect authentication,which poses a security risk.Therefore,in this paper the method combining thermal imaging with color imaging is proposed.In order to improve the security defense system of the face recognition access control system and solve the problem that the face recognition system is vulnerable to being attacked and deceived,based on a large number of research literatures at home and abroad,this paper focuses on the design of face liveness detection module for face recognition access control system.Also,in this paper two methods are proposed so as to realize face liveness detection,one is to use traditional methods for contour matching,and the other is to adopt a kind of deep learning algorithm to identify face liveness,and then implement face liveness detection by programming.The novelty of face recognition access control system designed in this paper is that the image acquisition device of the access control system combines color camera with thermal imaging camera.Moreover,application of the thermal imaging camera also endows face recognition system with strong defense capabilities against the three existing spoofing attacks(i.e.photo,video,and facial mask).Aiming at the design of face recognition access control system,the main tasks completed in this paper are as follows:1.According to the application occasion and actual requirements of the face recognition access control system,overall design of the system is carried out.The hardware design consists of the image acquisition module and the access control system switch module.The software design consists of face detection module,face liveness detection module and face matching module.2.The face liveness detection method based on contour matching is elaborated,i.e.Firstly,by means of the camera model theory,the positioning model from thermal imaging camera to color camera is derived.Secondly,by means of image preprocessing operations and the improved multi-motion lines detection convex hull algorithm,face outline is extracted from the image of the thermal imaging camera.Thirdly,by means of YCb Cr ellipse clustering skin color model,face area is segmented from complex background in the image of the color camera;and then adopting the morphological processing and the improved multi-motion lines detection convex hull algorithm,face outline is extracted from the image of the color camera.Finally,by means of the image moment theory,the difference function is established so as to identify face liveness.3.The structure of VGG16 deep convolutional network model is elaborated,and the model is used to realize face liveness detection.The detection method is divided into four steps,inlcuding data acquisition,data enhancement,model training and liveness detection.4.Using C++ programming language and software Qt5.8 to complete programming,face liveness detection method based on contour matching is validated;using Python3.5 programming language and Spyder software to complete programming,face liveness detection method based on deep learning is validated;and the experimental results are analyzed briefly.