Research And Design Of Intelligent Epidemic Prevention Access Control System Based On Face Recognition

Since 2020,SARS-CoV-2 has been spreading all over the world,which has brought great inconvenience to people’s production and life.Wearing masks is an effective way to prevent it.However,after wearing masks,the recognition accuracy of face recognition system has been greatly reduced.It is often necessary for people to take off masks for identification in some occasions such as company or laboratory attendance,entrance and exit of hospitals or nursing homes and station security inspection.It is not only troublesome,but also brings great challenges to epidemic prevention.This thesis closely focuses on the key engineering and technical problems of intelligent epidemic prevention access control system,in order to realize the rapid and accurate recognition of face identity with and without masks,and carries out research from the perspective of face recognition data set and loss function.The main work of this thesis includes:(1)According to the specific implementation process of face recognition,this thesis defines the necessary steps of face recognition task-face detection,face feature extraction and comparison,analyzes the functional requirements of the system in detail,expounds the working logic and process of system software and hardware,and the overall scheme design of the access control system is completed.(2)For face detection,a data set containing 8000 images is constructed and used to train SSD algorithm.The small target is enhanced by random expansion and random clipping to improve the generalization ability of the network.Due to its high computational complexity,it is lightweight and improved to improve the detection speed.It is combined with BiFPN module and CIoU_Loss improves detection accuracy.For face feature extraction and comparison,the face recognition training set and test set with and without masks are reconstructed,as well as the corresponding test files.Based on ArcFace and inspired by CenterLoss,a between class loss function is proposed to expand the gap between classes to obtain feature vectors with stronger representation ability.The sphere20 recognition network is replaced by CSPNetS and combined with attention mechanism CBAM to further improve the recognition accuracy.(3)The upper computer management software is developed based on Pyqt5 to facilitate the registration,modification and deletion of relevant information of internal personnel.According to the face detection algorithm,face feature extraction and comparison algorithm,system working logic and other parameters,the relevant hardware is selected,the prototype of access control system is actually developed,and the overall test of the system is carried out.The test results show that the upper computer software runs stably and can manage the system information correctly and easily.The face detection and feature extraction and comparison algorithm transplanted into the access control system can accurately and quickly deal with the face recognition task of "wearing mask" or "without wearing mask",and the real-time performance and accuracy can meet the needs of practical application.