Design And Implementation Of Online Learning Aided Detection System Based On Face Recognition

In recent years,the rapid development of the Internet,artificial intelligence,big data and other technologies has helped to open the era of Education Informatization 2.0,and smart education has emerged as the times require.At present,online learning has been widely used in distance education,assisted classroom teaching,etc.,which provides a strong guarantee for promoting modern education reform.Online learning breaks the constraints of time and space in the traditional teaching process,and provides a new form of learning for learners.However,compared with the traditional teaching methods,its disadvantages cannot be ignored.There is a lack of interaction between the learners and learners.The learners cannot perceive the learning status and emotions of the learners,there is a serious "emotional deficiency" phenomenon in online learning.In addition,in the process of online learning,how to effectively verify the identity of learners and strengthen the supervision of the effectiveness of online learning is also an urgent problem.At present,face recognition technology has achieved remarkable results in areas such as attendance and security,which also provides new ideas for smart online education.Aiming at the shortcomings of the current online learning systems,this thesis assists in detecting learners' online learning behaviors from two main aspects: face recognition and emotion recognition.This thesis provides a new perspective and method for online learning behavior assessment,and provides effective technical support for the constraints and supervision of learners in online learning.The main tasks completed in this thesis include:The algorithm design and implementation of face detection,face recognition and expression recognition are completed.In terms of face detection,face detection and key point location based on the MTCNN network,and face alignment using the OpenCV affine transformation algorithm.For face recognition,the ResNet18 residual network is used as the backbone for facial feature extraction,and the InsightFace loss function is used to train on the CASIA-WebFace dataset.In the aspect of facial expression recognition,for the problems of facial expressions with large intra-class expression differences and high similarity between facial expressions,this paper improves the facial expression recognition algorithm and redesign A weighted fusion loss function of Center Loss and Softmax,combined with the ResNet18 network,Algorithm proposed in this paper achieved an accuracy rate of 73.447% on the FER2013 test dataset,surpassing other algorithms under a single model and single task.Combining face detection,face recognition,and expression recognition algorithms,a facial information deep perception model for online learning is constructed.The model mainly solves two problems: complete learner identity verification through face recognition and detect learners' emotions through facial expression recognition.This model enhances the supervision of the effectiveness of online learning,and to a certain extent,solves the problem of " emotion absence " in the process of online learning.An online learning aided detection system based on B / S architecture is designed and implemented,and the facial information deep perception model of this paper is applied to the online learning process.Based on the realization of the traditional online learning function,the system regularly captures the learner's face image through the camera,performs face detection,face recognition and expression analysis,and visualizes the results.When learners have more abnormal behaviors and negative emotions,the system promptly gives reminders and interventions.Finally,functional testing,non-functional testing,and performance testing were performed on the online learning aided detection system.The test results show that the system performance is stable,the functions are running normally,the accuracy and speed of face recognition and expression recognition have reached the expected goals.