Research And Implemention Of Object Segmentation And Tracking In Complex Scenes

The technology of video object segmentation and tracking is an important branch being studied in modem science and technology field. The relative research project can be served for the need of national defence and civil applications. The thesis describes the state of automatic target recognition ,track system and the developing trends of the relative technology, deeply systemic research the image sementation,recognition and tracking algorithm. The solution of relative software system and hardware system based on the C6202 high-speed DSP and large-scale PLD is developed. The system design can be divided: hardware platform design,hardware system's program design and DSP software design.The design of hardware platform is mainly discussed in this paper. The method of modern electronic system design is taken into the design of whole hardware system, In the development of hardware system, the function expansibility and stability are emphasized greatly. The modularized structure is used in the design of hardware system .It not only reduces the complexity of design, but also benefits the group work. The function of hardware system includes conberting video signal to digital image deta and digital image processing and data exchange with the host computer.We discuss pretreatment module and edge detection module in the domain of image processing . In pretreatment module, we use morphologic-opening to eliminate edge burrs and island spots.,and morphologic-closing to fill up heaks and splits; In edge detection module, we use wavelet transform to complet edge detection. And they are applied in the DSP system, and the detailed realization steps were expounded in the aspects of programming and hardware characteristic etc.At last, we discuss debugging of hardware system, including debugging of video frequency interface, debugging of reading and writing FIFO, debugging of image frame memorizer,and debugging of interface between DSP and PCI9054.The system has some advanced and practicability because of using of wavelet multi-scale edge detection and morphologic filter arithmetic. While the content too much the task involved, there are many problems have to lucubrate and discuss. So there are many problems need to be improved and consummate.