| Bank ATM video surveillance is important to guard customer’s safety when using the ATM. But the traditional ATM video surveillance system can’t detect illegal actions on the ATM automatically. Because the number of ATM cameras is huge, the monitor staff may not find the actions immediately, and take effective actions to deal with it, which may cause heavy losses. To fix this problem, we have designed Intelligent Video Analysis algorithms which are aimed at ATM, and these algorithms are used for abandon detection and camera cover detection. In the abandon detection part, we applied modified Single Gauss Background Model, which is implemented easily and efficiently, and the result of which is good. In the camera cover detection part, we used the rate of non-zero pixels in the frame to detective, which has good result with dark cover. Meanwhile, this method is really simple. When implementing, the two methods above come out based on OpenCV, which is an open source computer vision library from Intel.On the basis of ATM Intelligent Video Analysis algorithms, we also designed the whoie surveillance system, which is consisted of three parts:server, gateway and client. Thanks to the use of H.264SVC, this system has two advantages. One of which is supporting various resolutions. So, in the normal case, we can transfer video in low resolution to save the bandwidth. If the illegal is detected, higher resolution video is transferred. The other advantage is that this system can easily adapt the quality change of the network bandwidth. As a result, we can use low resolution video when the network bandwidth became worse. To improve the efficiency of the system, I choose the DVR boar VRC6008of Stretch Company to play the role of codec. What’s more, the SDK of which is also used for our development. In the end, I develop the client application program successfully, and high recognition rate, quick response speed and solid running state are proved by the test. |
PDF Full Text Request