Research And Analysis Of Classroom Students’ Learning Status With Video Analysis

The classroom listening status of students is an important reference indicator for evaluating students’ class learning and teachers’ teaching conditions.Usually,these works are completed by the school’s teaching supervisors through attending lectures.However,for every teacher,supervision cannot be done every time.Listen to all classes.Therefore,the evaluation obtained is not necessarily comprehensive and objective.This paper uses computer vision technology to analyze the video images of students in the classroom scene,automatically detect the state of students in class,and build a judging system of students’ classroom learning state.The main research contents of this paper are as follows:(1)Established a head detection model of students in class based on Yolov4＿DB algorithm.The state of students’ classroom listening mainly depends on the state and movement of the students’ heads,so the detection of the students’ heads in class has become an important step in the evaluation system of students’ classroom learning status.Based on Yolov4＿DB algorithm,this paper establishes a classroom student head detection model.The model is based on Yolov4 and introduces the Drop Block layer to extract features of the image,which improves the generalization ability of the model.(2)Established a research model on the classification of classroom students’ learning status based on the Mobilenetv2＿S algorithm.The state and expression of the student’s head are an important indicator to judge whether the student is listening carefully.In this study,this state is divided into two states: listen and not listen.The state is classified through the classification model of Mobilenetv2＿S.The model uses the inverted residual network to extract features of the image.In the activation layer,the Swish activation function is used to replace the Re LU activation function in Mobile Netv2,which solves the problem of network degradation and improves the detection accuracy of the model.(3)Use classroom surveillance video images to establish an image data set containing 15,466 images for system testing.Part of the data set is used for model training and the other part is used for experimental detection,The head detection model and head state classification model proposed in this paper have obtained detection results of 99.09% and 99.00% respectively.which have well completed the status analysis of students’ learning in class.