Research And Implementation Of Image Anomaly Diagnose System Based On Image Analysis For Video Surveillance

In modern society, video monitoring system exists in everywhere. For large-scale video surveillance system, the monitoring points tend to be scattered, resulted in inadequate supervision of its video quality. A lot of video surveillance systems are face with the problem of how to ensure video quality and maintain the video quality with gradually increasing difficulty. The first problem is that how to detect and judge the image anomaly taken play in video surveillance.Usually, the task of inspection for video quality was done by manager to view the screen, and then recorded the results manually. This method lack of automated handling capacity for the surveillance system as managers were not likely keep recording video quality status all the time, thus resulting in lengthen the system’s maintenance time and more complicated. This operation is not conducive to reduce costs and risks. Therefore, this makes sense to design automatic diagnose process of detect the occurrence of video anomaly.Based on image analysis technology, this thesis design and implement a system which can make a diagnosis about several interference phenomena in the video, such as the sharpness anomaly, brightness anomaly, video shake, screen freeze, and automatically generate the statistical results. The processing object designed in the thesis is a real-time video stream or video files saved from stream.In this thesis, the emphasis and difficulty part are several video’s image anomaly diagnostic algorithms based on image analysis. The method of this thesis is that, analysis the frame image extracted from video. Through calculating the inter-frame movement of displacement, moving regions size and mean value of brightness, focus-square value of average gradient, combine with the calculation amount of the normal reference frame, we diagnose the status of the video image about sharpness, brightness, screen, etc. In addition, this thesis also research about the preprocessing of the video frame, the set of diagnostic mechanism, the save and query of diagnosis results, diagnostic statistics and result analysis.The system in this thesis divided into two parts, video stream processing client and local client. The video stream client receives the stream from the front-end devices and network server, convert it into video files after the decoding operation and preprocessing. In local client, analysis image about the quality of the video, then judge the quality conditions of video’s image by using the above several video’s image anomaly diagnostic algorithm. The operations to results, diagnostic settings and database operations are also completed in local client. Those function of automatically generate statistical reports and to analyze the results greatly improving the automated processes and intelligent level of the system.Although, sometimes environmental factors affect the video’s anomaly image diagnostic algorithm largely. But experimental result verify that by adapt to a parameter compatible with the environment which was obtained after numerous experiments, this system has certain degree of accuracy and practical value.