PTZ Research Based On Intelligent Surveillance Technology

As the development of intelligent visual surveillance system, the characteristic message and behavior of objects becomes more and more important for supervisor. How to get the characteristic message of objects we wondered is one of the important functions we cared of visual surveillance system. The usage of PTZ (Pan, Tile, Zoom) camera can shoot face character moving trace behavior of moving object continuously. It can make this monitoring more continuously and accurately.Based on the skill of detecting, distilling and tailing, this article introduces an automatically tailing system by using PTZ camera. This system used the mixed model of static camera and PTZ camera. According to these image lists gained by the static camera, it can develop the image background and distill foreground. Then the using of Kalman filter will make these isolated signals which have supervised the target to be constant trail. On the basis, design the controlling strategy of PTZ camera which can dominate the rolling of the camera and make the camera tail automatically its interesting objects, which will enrich the moving information of the object. The supervising strategy and method of PTZ camera is the core; and the skill of detecting, distilling and tailing of the mobile target is the base of the intelligent visual surveillance system.Aimed at the feature of visual surveillance, this paper develops the mixed tailing model of static camera and PTZ camera. By using static camera and PTZ camera, we build up a real supervising system and discuss the theoretic method of trail model by only using PTZ camera. The mixed tailing model of static camera and PTZ camera divided the supervising scene into several sub-supervising area by using scene cutting method. In addition, give specific arranged place for each sub-supervising area. According to the place reflecting on these specific arranged place gained by the static camera, we can control PTZ camera and let it read the corresponding arranged place.Then we can achieve the purpose of supervising randomly and automatically of the moving object.The trail model of using PTZ camera is the best way of scene surveillance.As the counting way of background mapping can improve the effect of background reconstruction. However, the way of counting is too complicated. It has a high requirement for the accuracy of leaning angle and elevation gained by PTZ camera. The experiment mixed the counting way of mixture Gaussian model, Kalman filter, the model of static camera and PTZ camera.It set up a minor intelligent visual surveillance system on the base of the original experimental condition.The system based on the skill of detecting, distilling and tailing of the mobile target and studied much more deeply on the counting way of supervising the moving object.It established the experimental method based on the counting way of mixture Gaussian model and Kalman filter trail model.The counting way of mixture Gaussian model use the brightness and color of each pixel dot in the image to set up a clear background model which is close with the real one. In the process of supervising the foreground area, the effect of gaining the object is great.The mixed Gaussian background model can effectively adjust to the changing of the scene and keep the system to be stable.By supervising the shade and disposing the noise, the influence of shade and noise can be banished. Therefore, the accuracy of surveillance can be improved. Tailing these moving objects which were supervised in the image list by moving surveillance counting method and evaluating the movement by Kalman filter can decrease the arrangement of getting the feature. Therefore, it improves the effect of counting method. When the object is matching, using the cooperating method of center and space can consider the place and shape of the object.