Studies On Detection And Tracking Of Multiple Objects In Intelligent Video Surveillance

Intelligent video surveillance system can watch behaviors of object intelligently and detect abnormal events, which has significance for theories and applications. This paper focuses on the technologies of detection and tracking of multiple objects. Firstly, we refine the method of detecting shadows in HSV color space by analyzing the illumination model and color characteristics of the moving shadow. Secondly, a tracking method of multiple objects is proposed based on the complicated moving states of multiple objects. This method tracks different objects in different way depending on the object moving states determined by its characteristics. Thirdly, on the basis of analyzing color characteristics of the object and establishing color models, we put forward a method to solve multiple objects occlusion problem by color blocks of objects. Finally, by studying the function of multi-camera surveillance system, we design a kind of structure for the system and propose a tracking method in multi-viewpoint by combining homography with FOV lines between two different views. This method can solve the problem of handing on and corresponding objects in different views effectively.