Design And Development Of Class Attendance Program Based On Face Recognition Technology

Class attendance is an essential part of teaching process.Nowadays,it is still done manually by teachers,which is time-consuming and ineffective.On the other hand,face recognition technology is developing rapidly,which is widely used in secure and attendance area.Thus,in this paper,we research face detection,face pretreatment and face recognition algorithms in classroom environment,and design an automatic class attendance program based on face recognition technology.Light varies in different location of classroom,and student's gesture in same camera is different.These differences are challenges to face recognition technology.Thus,it is necessary to research and use appropriate algorithms.In this paper,we firstly use popular and well-effected adaboost to detect faces from classroom images quickly and accurately.Then we use a series of methods to pretreat the detected faces,including DoG and size normalization.As for feather extraction,we use a texture description operator named LBP(Local Binary Pattern),which is simple and have strong ability of classification,then we use PCA+LDA algorithm to decrease the dimension of LBP feature.Traditional average-blocked LBP algorithm performs weak in non-standard situation such as multiple poses,so we propose several image block methods to deal with non-standard face images in classroom situation,including multi-scaled block,traversing image block and key points block.In multi-scale block method,we estimate the facial key points by experience,then we divide face image by 12 different scaled patches according to the location of facial key points.It receives recognition rate of 82.3%in View2 protocol of LFW database,which improved a lot compared with 68%of average-block method.Traversing image block is the improved method of multi-scale block,since the latter divides images by experience.It uses multi-scale windows to traverse the image and then gets 9 patches which characterize face better by training.It receives recognition rate of 83.25%.Key points block divides face image by 12 patches based on 68 facial key points extracted by AAM algorithm,which is proposed to solve facial key points alignment problems.But limited by key point extraction accuracy and LBP accuracy,it reached recognition rate of 84.1%on LFW database.Finally,we design an automatic classroom attendance program completely based on user's needs,including the overall system architecture,camera module,multi-threaded module,data analysis and management module,database design an so on.With application of above algorithm,this program realize the whole process of classroom environment image capture,face detection,light pretreatment,size normalization,feature extraction,feature classification and get attendance information.In the end of this paper,we demonstrate the program.