Research On Object Detection And Tracking Algorithms Based On Multiple Cameras

As the development of technology and society, using multi-camera network and video processing technology to achieve the intelligence of video surveillance is a research hotspot in the field of computer vision. There are two critical problems have to be solved in multi-camera intelligence video surveillance system. The first one is the object detection and tracing using single camera. And the problems of objects handover in multi-camera system is the other one we have to deal with. Based on the application background of the intelligence video surveillance, the objects detection and tracking algorithms of the multi-camera system with overlapping areas is studied in this paper. The research includes following aspects:Analyzing the performance and application occasions of the objects detection and tracking algorithms in single-camera tracking system. Researching on the key techniques and existing algorithms in multi-camera system. Summarizing the technical difficulty and algorithms studied currently.A new method based on the camera calibration and combined with the environment model is proposed to calculate the relationship of homography between two images. The corresponding coordinate's relationship of the two images that got from different cameras is calculated by this method. It is accurate and request only one calculation. The relationship will be unchanged once the camera system is finished.Studying the algorithms of objects handoff in multi-camera system and realizing the experiment of multi-objects handoff by the method proposed in this paper. Also a method to solve the problem of conglutination in dual-camera system is proposed in this paper. The experimental result shows its efficiency.Developing a tracking system of multi-camera to detect and. track objects in indoor environment. Two cameras are adopted in this system to enlarge field of view. The object detection and tracking is achieved in each single-camera when objects are not in the overlapped regional and the object handoff is realized by homograph relations when objects are in the overlapped regional. The experimental result shows the robustness and efficiency of the system.