Research And Implementation Of Detection Method For Abnormal Behavior Of Examination Based On Video Surveillance

The enrollment test has become an important measure for selecting talents in China.The fairness and transparency of recruiting are highly concerned by the whole society.At present,all the test sites are equipped with electronic video equipment for the entire process of the test,in preparation for later review.This kind of monitoring method is not smart enough,it is simply a video recording in the examination room,lacking real-time and initiative.In view of the shortcomings of video surveillance in the existing examination room,this topic combines cutting-edge computer vision technology to research and implement the detection method of abnormal behavior based on video surveillance.This topic innovatively uses the target detection technology based on deep learning to test the target of the candidate.At the same time,the combination of moving target detection and skin color detection is used to detect the abnormal behavior of the examination field,and the effect is remarkable.The research content of this paper mainly includes three parts:the target detection of the candidates,the abnormal behavior detection of the examination room,and the intelligent monitoring system design of the examination room.Firstly,in view of the lack of robustness of traditional target detection algorithms such as HOG,this thesis adopts the target detection algorithm YOLOv2 based on deep learning and proposes an improved YOLO-D network.By improving the fine-grained features of different convolutional layers,the improved network can increase the robustness of target detection and ernhance feature transfer,making this network structure more suitable for the candidate test tasks of this topic.Secondly,in the abnormal behavior detection part of the examination room,a detection method based on the combination of moving target detection and skin color detection is proposed.The method calculates the contour and region of the movement through the motion history image of the candidate,and combines the positioning of the candidate's target detection method to determine whether it exceeds the normal range of motion,and can detect the abnormal behavior such as the sneak peek and pass the note.The skin color detection method detects the color of the hand of the candidate,and if the candidate's hand leaves the desktop for a certain period of time,it can detect the abnormal behavior of the mobile phone.Thirdly,in view of the shortcomings of the existing examination and control system,this project has designed a set of intelligent monitoring system for the examination room.The system has the functions of attendant statistics,abnormal behavior test,abnormal behavior warning,abnormal behavior history query and examination room order evaluation.