| This paper designs an automatic attendance system for the classroom scene,which is mainly based on the classroom monitoring video stream and realized by face recognition.The main work is as follows:(1)Firstly,this paper studies face recognition technology,including face detection algorithm and face recognition algorithm.In the aspect of face detection,Retina Face algorithm based on deep learning shows excellent effect in the monitoring scene.Compared with the classic MTCNN algorithm,both the effect and the detection speed are greatly improved.Therefore,Retina Face is selected as the face detection algorithm of this system.In the face recognition part,Face Net method is selected.The experiment shows that Face Net has good effect in unconstrained face recognition.(2)In the experiment,it is found that the blur and angle of the face image will cause certain interference to the face recognition and affect the accuracy of the face recognition.Therefore,this paper attempts to add a human face image quality evaluation link in the face recognition process.In this paper,we try to build a face image quality evaluation data set,combined with Face Net face recognition algorithm to find cosine similarity to mark the data.The training model shows a good effect on the test set.After the experiment,it is found that after the image quality evaluation part is added,the low-quality image can be effectively screened,to a certain extent,the phenomenon of false recognition is reduced,and the accuracy of the system has been improved to a certain extent.(3)In view of the low resolution of the face images of the students in the back of the classroom,it is proposed to use a zoom digital camera to collect images.The camera's preset function is used to divide the classroom into multiple areas.The camera automatically patrols each preset position,so as to collect high-resolution images of each location.(4)Finally,an attendance management system is designed by using the above research,including client,background and image acquisition module.Video data is collected from the classroom for face recognition,and then attendance is determined according to multi frame voting.The final results are stored in the database for teachers and students to view. |
