Moving Target Detection In Video Surveillance System Of Bank

Video surveillance system is a widely used security and protection system at present, which hasthe advantages of real-time, intuitive visibility and the ability to monitor and record the emergency.It is widely used in some important areas like banks, airports, traffic junctions, etc. With thedeveloping of computer technology, video compression technology, computer vision, artificialintelligence technology, video surveillance becomes so intelligent that it will need fewer people oreven no people to monitor. Moving target detection technology in computer vision can be a goodpromotion to the intelligent video surveillance system.The article describes several common methods of moving target detection and analyzes theadvantages and disadvantages of these methods. For the real-time characteristics of videosurveillance system, we have screened the methods. We tested the effect of the three framedifference method and Gaussian Mixture Model method and summarized the applicable scenes andfeatures of these two methods. Based on these experiments, we put forward an improved movingtarget detection algorithm.The improved moving target detection algorithm is based on GMM and uses the result of thethree frame difference method to help the update of the background model. Firstly, the OTSUmethod is used to improve the three frame difference method in the threshold part. The improvedmethod uses adaptive threshold to separate the foreground from the background, so the result isoptimized. Secondly, the result of the frame difference method is applied to the GMM, and thealgorithm will adjust the Background update rate of the GMM according to the relationship betweenthe result of the frame difference method and the final result. The background update of thetraditional GMM uses the whole image of the current frame to update the background model, so themoving object can be updated into the background and affect the accuracy of the background model.In this article the improved algorithm uses the update method with mask to update the backgroundselectively, so it will make the update of the background model more correctly. The article alsoanalyzes the background update under the condition of light mutation and proposes a simple andeffective method to adapt to the new scene after the light changed. Finally, the new method alsoprocessed the shadows which existed in the process of moving target detection and eventually made a better detection result. Through repeated experiments, the improved algorithm is lesstime-consuming, with a good real-time, can be applied to video surveillance system.In the application part of the article, it designed how to report the results of the moving targetdetection to the monitoring center and how the monitoring center display the result after receivedthe report detailed, including the reported frequency after detected moving target, the treatment ofthe same moving target, how the monitoring center display the reported video on the big screen andthe update policies. The article designed an intelligent video surveillance system, which is complete,effective and consisted of detection, reporting, and display of the moving targets.The moving target detection algorithm in this article is designed and developed on the platformof VC++6.0combining with OpenCV computer vision library. The system is able to detect themoving target, display the reported video.