Application Of Face Recognition Technology Based On CNN In Classroom Attendance System

With the rapid development of artificial neural networks,its researchs and applications in the field of computer vision has made an important breakthroughs.Face recognition is an important branch of computer vision.In recent years,more and more research results have adopted the method of convolutional neural network to improve the accuracy of face recognition and achieved good results.Class attendance plays an important role in the current teaching management of colleges and universities.The traditional method of attendance is time-consuming and inefficient,which has a certain impact on normal classroom teaching.In this regard,this thesis proposes an effective classroom attendance method based on face recognition.The method adopts the face detection and recognition technology based on convolutional neural network to automatically detect and identify the student face in several pictures acquired by the teacher in the classroom,thereby achieving fast and reliable classroom attendance and effectively making up for the lack of traditional methods.The main research work is as follows:1.The Faster R-CNN algorithm for target detection is to perform global search on all possible candidate regions in the image,while the candidate regions of the face target have their distinctive features.If these features can be acquired and used to generate high-quality candidate regions,on the one hand,the search space can be narrowed,and on the other hand,the accuracy can be improved.In this thesis,K-Means clustering method is used to learn some typical values of face candidate regions from a large number of annotation data,and based on this,the Faster R-CNN face detection algorithm is optimized for RPN network candidate region search strategy.The detection speed and accuracy of the Faster R-CNN face detection algorithm are improved.The experimental data show that the eight typical anchor size bounding boxes obtained by K-Means clustering can achieve the best face detection effect.2.There are many research results in face recognition using CNN.However,if these results are directly applied to classroom attendance,they are easily affected by problems such as image quality and light,resulting in low recognition accuracy.Based on the existing face recognition algorithm of CNN,this thesis optimizes its network structure and parameters,and obtains an efficient face recognition model in this application scenario.The experimental results show that the recognition accuracy of the face recognition model proposed in this thesis reaches 98.33%.3.The classroom attendance system based on the above-mentioned Faster R-CNN face detection algorithm and CNN face recognition model has a good practicability.The system is divided into three functional modules according to different user roles,including sub-function modules such as management of basic information and statistical analysis of attendance.The basic information mainly includes students,teachers,courses,classes,etc.The experimental results show that the classroom attendance system designed and implemented based on face recognition technology can meet the classroom attendance requirements and has practical application value.