Algorithms Study On Object Detecting And Tracking Of Multi-Azimuth

Nowadays digital surveillance system has got a rapid development with the improvement of its stability and reliability and the decrease of cost for surveillance equipments (for example the camera, the Pan-and Tile Drive and so on). For its increasing importance, the digital surveillance system has been used widely in public area such as bank, hotel, supermarket, airport and station.The research work of this paper was mainly focused on the visual object detecting and tracking for surveillance system. An algorithm of multi-azimuth object detecting and tracking was proposed for supervision and measurement of the object.Considering the variability of the moving object condition and change of the environment light, an advanced color filter was applied to obtain color character of the object. And then boundary detection and color character segmentation were used to realize the detection of moving object.As to the requirement of real-time, the multiscale matching of moving object was realized with the application of wavelet transformation. Furthermore, Kalman filter was used for prediction of trace and the tracking of moving object was obtained. This algorithm not only could reduce tracing time, but also had adaptability to the change of moving object' s shape.The method presented in this paper of Multi-azimuth object detecting and tracking provided a solution to detect and track moving object for fhe camera on the Pan-and Tile Drive. And with the focus-changing function of the camera, the full appearance of moving object could be obtained to provide detailed information of aimed object for the following work.