Design And Implementation Of Conference Sign-in System For Intelligent Face Recognition

With the development of society,conference,as a kind of the most communication activities,plays an important role in in all walks of life.What's more,it's scale has been larger and larger in recent years.The traditional conference sign-in is mainly carried out by means of manual statistics.Before the conference,the staff of the conference group holds the roster of the participants,and then signs and confirms each person invovled according to the relevant information.There are sevral problems such as cumbersome summary,common errors,feedback not in time and so on.Recently,the development of information technology has made the conference sign-in system widely used,but all kinds of sign-in systems existed have different degrees of the cost problem and identity theft.In view of the shortcomings of various conference sign-in systems,this thesis designs and implements the smart face recognition conference sign-in system by using face recognition as the sign-off method.This thesis first analyzes the requirements of the system,designs the overall architecture of the system according to the application requirements and divides the system into three functional modules: user registration,on-site check-in and information management.Considering the efficiency of face recognition and time consuming,this thesis implements the face recognition function based on the deep learning method.Related work is as follows:(1)For the application scenario where the conference checks in,this thesis designs a face recognition module based on Face Net improved structure.Firstly,the histogram equalization processing is performed on the acquired image whose average brightness is lower than the threshold,and then the absolute position information of the face and the key point coordinates are obtained by the multi-task Convolutional Neural Networks(MTCNN)face detection method.The method based on Open CV affine transformation is used to realize face extraction and face extraction using Face Net open source model.The test system based on B/S mode is designed and implemented for face 1:1 and 1:N recognition methods,and its test environment is under the camera network.Then,due to the small distance among different face features based on Face Net,it is hard to determine the threshold of the feature distance.This thesis optimizes the Face Net network,replaces the ternary loss function with the Arc Face loss function,and takes the LRes Net50-IR model structure as the training model on the CISIA-Webface dataset.Meanwhile,it uses the pre-training model to achieve the face recognition module and achieves good results.The accuracy and speed of the tested face recognition reached the requirements of system design.(2)According to the business requirements,the system is implemented according to the research of face recognition algorithm.It is deployed in B/S mode.The front end uses Bootstrap and Vue.js framework,the back end uses Flask framework based on Python language,and the database uses My SQL.The data interaction between the client and the server is done based on the HTTP protocol.This thesis first realizes the separation of system permissions through the login function,and then designs and implements each module of the system.The user registration module implements the information uploading and input detection functions,and integrates the face recognition method for the establishment of the conference face feature database.The sign-in module uses the camera network to obtain real-time image data,and integrates the face 1:N recognition method.The information management module implements conference release,conference management,user management,and information statistics functions,and completes the control of the conference check-in process.Finally,the system is tested on the PC side and the mobile side respectively.The results show that the smart face recognition conference sign-in system has stable performance,beautiful interface and perfect functions.The accuracy and speed of face recognition meet the design requirements,and the automatic control of conference attendance is realized.In the future,it can be promoted and applied in the field of conference management.