The Analysis And Research Of Video Monitoring System Based On Image Analysis In The Test Scene

In recent years, with the computer, camera equipment, network transmission technology greatly enhance and prices fall, video monitoring is widely applied in all walks of life. And in all kinds of examination, to put an end to lax invigilation, collective fraud and other issues, video monitoring technology is the most effective means to realize the test of openness, fairness and justice. However, the traditional "electronic monitors" simply by cameras record the test process, then use the artificial way to supervise. It cannot effectively find the problems existing in the examination, and would increase the cost of the exam and the pressure of the examiners in monitor room.Based on this, this paper applied Image analysis to the video monitoring in the Test Scene, to design and implement a test monitoring system which can automatically locate the candidate targets and found that the examinee fraud behavior. This paper researches and implements the examinee detection and cheating detection two major functions. And this paper designed the cheating storage alarm forensics, the examinee markers, cheating order statistics and the test evaluation standard, etc.In this paper, we studied and learned the traditional HOG detection algorithm, and applied it to the examinee tests. Aimed at the problems that don’t adapt to the test scene, this paper proposes a HOG detection algorithm based on multi-window improvement. The algorithm has a set of different sizes of detection window. And divided the detected image into different areas,different detection applicated different window for maintaining a higher detection rate and reducing the available at the same time. According to the examinee target characteristics in the test scene, improve traditional HOG detection algorithm of the unnecessary computation problems during reducing the image. Experiment after careful validation of the SVM classifier, this paper completed the examinee identification with the head and shoulders detection, And proved the algorithm accuracy is higher than85%.In this paper, analyzing the characteristics of the students cheating, using the cheating analysis method which is combined the mixed gaussian background modeling method combined with skin color segmentation location method. First of all, the system used the mixed Gaussian background modeling method to found the activities of the examinee. on the basis of the activity examinee, using skin color segmentation, selection and location method to located the examinee’s position of the head and hands. Finally, according to the study of motion and position judging relationships, found the suspected cheating targets, out of normal behaviour examinee, complete the abnormal behavior of the candidate’s judgment.Apart from the above two main functions, this paper also according to the system requirements, design the cheating storage alarm forensics, the examinee markers, cheating order statistics and the test evaluation standard function, etc. The experiment proved this system has good examinee detection and cheating detection ability, by test the southwest jiaotong university in2012"computer basis" final exam video.