The Classroom Human Behavior Detection Based On Deep Learning

Along with the Internet technology gradually mature,"Internet plus"era has arrived.To Internet plus education as an example,the Internet is changing many aspects of teaching and learning,evaluation mode and curriculum structure,curriculum form etc.Using computer vision technology to improve the quality of Internet education courses plays a vital role in the further development of Internet education.It is of great significance to improve the quality of course education by using deep learning technology to detect and judge the behavior of objects in video.In this context,this paper proposes the following research objectives:Based on the deep learning target detection and behavior recognition algorithm,to achieve a set of classroom scene moving target behavior detection system,real-time detection of the target,and mobile camera recording video to the target.In order to achieve this goal,this paper mainly carries out three parts of work.The first part is the study of the principle of motion detection algorithm.This part includes the research of moving target detection algorithm and behavior detection algorithm,including the classical moving target detection,the use of artificial design features for behavior detection and the use of deep learning network for behavior detection.The moving object detection algorithm includes CodeBook,ViBe and GMM.The artificial detection algorithm based on artificial design features includes HOG features,SIFT features and other traditional manual operators and classifiers And the combination of optical flow method;based on the depth of the learning algorithm based on region proposed target detection algorithm and regression-based target detection algorithm.This part laid a theoretical foundation for further research.The second part is the realization of the behavior detection algorithm in the classroom scene.This part includes collecting the behavior detection data set in the classroom scene,implementing different algorithms to detect the target behavior in the classroom scenario,analyzing and comparing the advantages and disadvantages of different algorithms,and providing the core technical support for the next system implementation.The third part is the concrete realization of the whole testing system.By comprehensively analyzing the performance of different algorithm models,a solution based on depth learning in the classroom scene motion detection system was designed:a dual camera with the program to achieve a good video recording effect.This paper describes the system principle and implementation process,and then through the C + + language on the Windows platform to achieve a complete system to meet the real-time,to achieve the desired results.