Research And Implementation Of Face Recognition Technology-based Registration System For University Students

At present,the traditional manual registration method is still used in the registration process of college students returning to school.This method is time-consuming,waste of manpower,low efficiency,data statistics are not timely,easy to cause personnel gathering,items cross contact,unable to carry out effective statistics and management of student registration information.Face recognition is an important biometric authentication technology,which has been widely used in military,financial,public security,daily life and other fields.Face recognition has the characteristics of non-contact,high stability,high accuracy,and not easy to replicate.Under the background of the normalization of epidemic prevention and control,face recognition can effectively reduce the risk of contact transmission.This thesis designs and implements a student return registration system based on face recognition technology,which can directly complete the registration work in real time under the unrestricted environment,and can timely and effectively count the student registration data,but also effectively avoid the hidden trouble of the agent.The main innovative research results obtained in this thesis are summarized as follows:1.The lightweight face detection model MA＿Retina Face based on improved Retina Face is proposed.In this model,the Retina Face backbone network is replaced by the lightweight network Mobile Netv2,which effectively reduces the number of parameters and computing resources of the model,and improves the detection speed.In order to compensate for the information loss caused by the inconsistent feature scales of FPN(Feature Pyramid Networks)in Retina Face,adaptive spatial feature fusion(ASFF)is performed on the output feature maps of FPN.The experimental results show that the AP values of MA＿Retina Face in Easy and Medium grades are better than those of other detection models,and have good real-time performance.Compared with other face detection models,MA＿Retina Face has the best comprehensive performance.2.A face recognition model CA＿RFB＿Mobile Netv2 based on improved Mobile Netv2 is proposed.In this model,Mobile Netv2 is used as the basic network model,RFB(Receptive Field Block)module is introduced to improve the feature extraction ability of lightweight network,and CA attention mechanism is used to autonomously enhance and suppress the features in RFB module,which improves the robustness of the algorithm.Experiments show that the model has a higher recognition accuracy than Mobile Netv2.Compared with other face recognition models,the model recognition speed is improved while the recognition accuracy is guaranteed,and it can meet the needs of real-time recognition.3.Designed and implemented a set of college students’ return registration system based on face recognition technology.The system combines the above improved face detection and recognition model,and uses Python programming language and Py Qt5 to complete the GUI interface construction and system function development,combined with SQLite database to complete the storage of student registration information.