Design And Implementation Of Auxiliary System For AI Smart Classroom Based On Raspberry Pi

With the advent of the “Internet +” era,the blueprint of the “2.0 Action Plan of Educational Informationization” has been on paper.The application of artificial intelligence technology to education and teaching has also become a hot topic today.“Smart campus” comes into being in this context.It combines advanced educational concepts with the new generation of information technology,which effectively promotes the progress and development of educational information.As the first scene of teaching,the core position of “Smart Classroom” is self-evident.The traditional classroom teaching mode is often inseparable from the teachers' high-intensity double-line work in and out of class,and it also faces problems such as the difficulty of reasonable allocation of class teaching time due to the unbalanced development level of students.How to use emerging information technology to promote the efficient development of classroom teaching,reduce the burden on teachers,and get rid of the teaching dilemma of lack of interaction between teachers and students has become an important research topic.Therefore,a mobile auxiliary system for AI Smart Classroom is studied and designed in this thesis,which is divided into two modules: classroom attendance and teaching situation reference.It is composed of software and hardware parts,and the following work has been carried out respectively:Firstly,in the classroom attendance module,the online student face database is first established.Then,the Dlib classroom and affine transformation are combined to preprocess the real-time multi face images taken by the system in the classroom.After that,the 1:N face comparison function of Baidu face recognition API is used to compare the detected,corrected and segmented student faces with the faces in the face database,so as to realize the students' classroom attendance and ensure the accuracy of recognition.Moreover,the Linux version of face offline recognition SDK is deployed on the ground side,which can realize the localization operation of the system without network.Among them,the design of the image acquisition module and the preprocessing of the original image are the key and difficult points in the research process.Secondly,in the teaching situation reference module,FERPlus is used as the training data set,which has been processed by data enhancement,and the number of data sets has been expanded.At the same time,the lightweight neural network model is built by using the paddlepaddle framework,and the open GPU resources provided by Baidu AI studio platform are used for high-speed model training.At the same time,the effects of different optimizers and neural network structures on the model training are compared,four types of classroom expressions are redefined,and a quantifiable teaching situation reference module is designed.Finally,the hardware part of this system is composed of Raspberry Pi and its external camera and touch screen.The corresponding environment construction and module transplantation are mainly done for Raspberry Pi.The test of the classroom environment shows that the auxiliary system for smart classroom designed in this thesis can achieve the expected functions,with the advantages of cost-controllable,manageable,easy to expand,efficient,convenient,and mobile.It can effectively help teachers check attendance in class and provide a platform for teacherstudent interaction.