Method And Implementation Of Abnormal Behavior Recognition In Examination Room Based On Video Surveillance

As the number of candidates increases sharply year by year,the competition for modern examinations becomes more intense,and the fairness of examinations becomes more and more important.At present,the traditional electronic invigilation method that relies only on on-site invigilator and manually checking the video data of the examination room to find abnormal behavior has a high work intensity and is prone to missed detection and misdetection.With the rapid development of information technology,it is of great practical significance to apply artificial intelligence technology to modern invigilation technology.This thesis mainly studies the method and implementation of abnormal behavior recognition based on video surveillance in college environment,which uses the computer automation technology to complete the real-time judgment of abnormal behavior based on the video of the examination room to assist the invigilator in decision-making.In the research of this subject,the above system function design will be completed by combining hardware and software technology.The main tasks completed are as follows:1.By analyzing the characteristics of the examination room environment and the behavior characteristics of the candidates under normal and abnormal behaviors,This thesis put forward the system requirements of the abnormal behavior recognition technology in examination room based on video surveillance,and carry out the corresponding solutions according to the demand analysis.The system is composed of image data acquisition module,system control module and abnormal behavior alarm module.The image data acquisition module is mainly composed of network camera,which transmits image data to the control module through wired network.The system control module is a system hub,which is implemented by the server and the display device.It mainly completes image data processing and other related tasks,including functions such as achieving target detection and abnormal behavior recognition.The abnormal behavior alarm module is composed of an embedded system.When abnormal behavior is found,the system control terminal will transmit alarm information to the embedded processor control board through the wired network to complete the abnormal behavior alarm.In this research,the design and implementation of the system control module for examinee target detection and abnormal behavior recognition as well as the system abnormal behavior alarm module are the core research contents.2.Candidate target detection in the system control module is the basis of the candidate's behavior recognition.The position information of each candidate should be automatically detected before the system performs behavior recognition.Based on the completion of face detection,this thesis combines the skin color and hair color characteristics of candidates,and proposes a new comprehensive method for target detection based on face combined with examinee's skin color and hair color characteristics.When building face detection model,extracting cascade features of gradient histogram and local binary pattern features of candidates' faces as input features,selecting support vector machine training classifier.In order to ensure a higher target detection rate,based on the completion of the candidate face detection,the candidate target detection based on the skin coloring feature is performed in other areas of the image.The above method was used to verify the test data,results show that the target detection accuracy is 95.89% and the average time is 1.69 seconds,which proves that the method meets the system design requirements both in accuracy and time consumed.3.In the system control module,the abnormal behavior of candidates is identified as the ultimate goal of the system.In the system design,the abnormal behavior of candidates is divided into suspected abnormal behavior and determined abnormal behavior.After the completion of the candidate's target test,the subject first proposes a candidate numbering method and a method for determining the candidate's activity range.After the completion of the candidate's target test,the subject first proposes a candidate numbering method and a method for determining the candidate's activity range.On the basis of the completion of the candidate number and the scope of the candidate's activities,On the basis of the completion of the candidate number and the scope of the candidate's activities,by studying the change of hair color and the foreground information within introducing the mixture Gauss model in the candidate's activity range,this thesis proposes a method of identifying the abnormal behavior of candidates based on the hair color-mixing Gauss model.The experimental results show that the method can identify more than 90% of the abnormal behavior of the examination room,and in the different test environment,in the face of complex and variable abnormal behavior of the examination room,even if the behavior category is not limited,the method can still judge the abnormal behavior.4.In addition to detecting abnormal behavior of candidates,a complete examination room monitoring system should also have an alarm function for abnormal behavior.The abnormal behavior alarm module of the system consists of the alarm of the main control room and the alarm of the classroom.The alarm of the main control room is mainly realized by the system software.The main interface of the system displays alarm information and stores abnormal behavior pictures of candidates to the server.The alarm of the classroom is completed by embedded processor.When the system finds abnormal behavior,the system control terminal transmits alarm information to embedded processor control board through wired network,and the alarm lamp module and display module on board realize alarm function together.It is proved by experiments that when the system finds abnormal behavior,the abnormal behavior alarm module can complete the alarm work in real time and meet the system design requirements.