Research And Implementation Of Classroom Attendance System Based On Face Recognition

Class attendance is one of the important means to ensure student attendance,and it is also an important part of the student's course performance.Classroom attendance can supervise students well,thus ensuring the quality of classroom teaching.At present,the main method of attendance is still manual or random checking by teachers.This method of manual naming not only takes up the class time but also can not solve the phenomenon of early leave,signing and absenteeism of students,and can not supervise the student's attendance.There are also a variety of attendance machines on the market,but due to the high cost of hardware and deployment,the attendance problem cannot be solved well.With the continuous improvement of computer vision and image processing theory and technology,as well as the deep development of deep learning and artificial intelligence in recent years,it provides new ideas for automated classroom attendance.In order to solve the above problems,this paper will adopt face recognition technology to realize automatic attendance.Due to factors such as the movement and posture shift of the human body in the classroom environment,the image of the face region in the sample frame will be blurred,thus affecting the performance of face recognition algorithm.In order to solve the influence of the distortion of the face image in the sampling frame on the recognition result,this paper proposes a method for evaluating the quality of the face image.We first fine-tune the pre-trained VGG19 deep network and then use the features extracted by the VGG19 neural network to train the SVM model,and finally obtain the quality evaluation score of the image.The quality evaluation score is used to filter the images to be compared,which ensures the reliability of the image quality in the recognition process.In the real environment,we carried out practical experiments.The experimental results show that the image quality assessment model can effectively filter low-quality face images and improve the accuracy of the system identification.It proves the effectiveness of our proposed method and its effect on improving the accuracy of recognition.How to select high-quality face area image in the video stream,how to ensure the accuracy of the recognition result will be the most important problem in the attendance system designed in this paper.According to the above functional requirements and the mentioned problems,this paper will specifically carry out the following research work:(1)The system obtains the information of the location of the face region through the face detection method of MTCNN,and trims and normalizes the acquired face image to reduce the influence of environmental factors such as illumination on the recognition.(2)The image quality evaluation model is trained to evaluate the quality of the face image before recognition,and to screen out high-quality face images and use the high-quality face images for recognition to improve the effectiveness of face recognition.The model is improved in the self-built face image library,which makes the trained face image quality evaluation model more suitable for face quality assessment.(3)Using FaceNet to extract features from high-quality face regions and obtain single-recognition results through SVM.In order to avoid the contingency caused by the single recognition result,the voting algorithm is used to ensure the reliability of the recognition result to obtain the final recognition result.(4)Using the face recognition technology and the face image quality assessment model to complete the attendance system,and testing the system in a real environment,to ensure the accuracy of the system.