Application Research Of Intelligent Invigilation System Based On Group Recognition

The traditional manual invigilation method has problems such as cumbersome invigilation process,difficulty in obtaining evidence of cheating behaviour and too subjective judgement.In this paper,we focus on the intelligent detection of abnormal behaviour in examinations,and build an intelligent invigilation system by using face recognition and head pose estimation methods.The specific research content and results are as follows.(1)Automatic identification of candidates.To address the problem that the candidate sign-in process relies too much on manual work,the candidate identity is automatically identified based on the face recognition method to achieve candidate identity verification in a group environment.The quantitative experimental results on the public dataset LFW and the recognition accuracy in the real examination hall show that the model can efficiently complete the candidate check-in and greatly avoid fraudulent behaviours such as test substitution.(2)The head pose estimation model CS-WHENet was constructed to address the peeking behaviour of candidates during the examination,analyse the masking situation in the real examination room,and construct the CS-WHENet model based on the WHENet and CBAM methods to extract facial features as fully as possible by fusing channel information with spatial information to extract facial features to the maximum extent possible to assist in head posture estimation and improve the The accuracy of candidate peeking behaviour discrimination was improved.(3)A behavioural detection method for passing notes in exams is proposed.A fusion of YOLO-v4 and the three-frame difference method is used for the detection of examinee passing behaviour.The relative position of the candidate area and the candidate’s motion connectivity is used as the basis for determining the passing behaviour.Experimental results in a real exam room demonstrate the accuracy and effectiveness of the method.(4)A verification system for intelligent proctoring was developed.Python,My SQL,Vue,Django and other language tools were used to implement the core algorithms of this paper to verify the identity of candidates and detect abnormal behaviour.The results of the application in a real examination hall show that the intelligent invigilation system can greatly simplify the process of invigilation and assist invigilators in their work.