Systemetic Research And Implementation Of Student’classroom Situation Analysis Based On Deep Learning

At present,most classroom recording methods still use traditional manual check-in and recording methods,which is tedious and inefficient.With the in-depth research and application of deep learning,big data and other technologies,it has provided an important impetus for the development of education informatization.Attendance,head-up rate detection and other systems that analyze the classroom situation of students using cameras came into being.However,using information technology to deal with this type of problem,there are still the following challenges:the resolution of the face collected by the high-definition camera in a densely populated environment is not high and the face area is small;the existing face recognition algorithms are designed for small faces The accuracy of the detection is not high;the head-up detection during the learning process cannot be combined with face recognition,and the head-up situation of each student cannot be accurately analyzed in real time.To this end,this article builds a prototype of a student classroom situation recognition system that includes attendance and head-up rate detection,and studies related algorithms to solve the above technical problems.The main research contents are as follows:(1)Tiny face detection algorithm: Due to the large number of students in the classroom,the face size of the back row students captured in the classroom environment is small and difficult to detect.This paper proposes a tiny face detection algorithm.The algorithm is tested on the FDDB public dataset and the face detection Class Detection dataset under self-built real environment.The test results show that the algorithm has a good face detection effect in the classroom environment.(2)Face recognition algorithm based on super-resolution reconstruction:Because small-sized faces have fewer facial features,it is difficult to perform accurate face recognition.Therefore,a super-resolution reconstruction algorithm is designed in this paper to reconstruct low-resolution faces into high-resolution faces,and then perform face recognition.Experiments prove that the algorithm can improve the accuracy of face recognition in the classroom environment,and The problem of low accuracy of small face recognition in limited scenes is solved.(3)Student classroom head pose estimation algorithm: Based on face recognition,the student’s head state is identified to determine whether the student is in a state of head-up listening to the lesson.This paper designs a recognition algorithm based on deep learning networks for head-up pose And the self-built students raised their heads to recognize the Class Head dataset for training and testing.The test results show that the algorithm has a good classification effect.(4)Build a classroom situation analysis system prototype: Based on the above algorithm,a set of student classroom situation analysis system is designed and implemented.The system analyzes classroom videos taken by the camera to generate classrooms that include students’ late arrival,early departure,absence,and head-up rate.Situation analysis statistics.The comprehensive condition query method can be used to query students’ attendance information and headings for a class or semester.The research results in this paper can improve the accuracy of face detection and recognition in dense environments,and can effectively recognize the head posture of students during the learning process.The developed application system can not only save time costs such as manual sign-in and recording,but also help teachers understand the classroom situation and assist teachers in classroom management.It has important research significance and promotion value in the field of smart education.