Research On Application Of Convolutional Neural Network Face Recognition Technology In Online Exam

In recent years,with the development of information technology and the impact of the new crown epidemic,online examinations have been used in many places,such as classroom testing,postgraduate re-examination and so on.The online test not only presents a novel test method for candidates,but also makes up for the lack of time and space in the test process.However,there are still deficiencies in this model.Under the current development trend,face recognition technology has been widely used in attendance,security and other fields,but there are still problems and deficiencies that are difficult to identify such as substitute examinations and plagiarism in online examinations.Aiming at the problem of online detection of students’ behavior and status in exams,this paper provides a new perspective for online exam evaluation by effectively sensing the emotions and expressions of candidates,face recognition as the entry point,and preventing candidates from cheating in the exam stage.Behavior and perceived emotional changes provide technical support.The work that needs to be done in this thesis is as follows:This thesis first analyzes and expounds the research progress of face recognition at home and abroad,and introduces the convolutional neural network from four aspects,which are divided into four parts: convolution,neural network,convolutional neural network and Py Torch deep learning framework.In order to better complete the face detection,face recognition and emotion recognition,the convolutional neural network is introduced,and the appropriate loss function is explored,and the face detection algorithm,face recognition algorithm and emotion recognition algorithm are mainly improved and tested.Therefore,the MTCNN network is selected to carry out face detection work,and the facial feature points are located,and the Open CV affine transformation is used to correct and align the face to ensure consistent image specifications,thereby improving the face detection accuracy;Res Net18 residual network is used.As the backbone network,the face feature extraction in face recognition is completed,and the Insight Face loss function is selected to combine with the backbone network to achieve the purpose of improving the accuracy of face recognition.In view of the large difference in intra-class expressions and the extremely high similarity of inter-class expressions,which are the key issues to be solved in emotion recognition,this thesis weights Center Loss and Softmax on the basis of the Res Net18 network to improve the emotion recognition algorithm.The training on the FER2013 test set The accuracy rate reached73.45%,the highest accuracy rate compared with other algorithms under the same training conditions.Combine face detection,face recognition and emotion recognition algorithms to create a facial information perception model and apply it to the online examination system,use face recognition to verify candidate information,and record the emotional changes of candidates in the examination through facial expression recognition.The application of online examinations improves the effectiveness of online examination supervision,reduces the phenomenon of substitute examinations,and solves the problem of not being able to perceive the emotional changes of candidates during the examination.have important meaning.The online examination auxiliary detection system in this thesis is based on B/S,combined with the facial information depth perception model,to complete the face detection,face recognition and expression analysis in the online examination,and the statistical results are displayed in a visual way and can be automatically warned.It is presented in the form of a bar chart or a line chart.If there are many negative emotions and behaviors in a certain test,the invigilator will remind you.Finally,the online examination system is tested.The test results show that the face recognition accuracy and speed of face recognition using convolutional neural network have exceeded the expected target,and the face recognition technology of convolutional neural network has been realized online.application in exams.