Target Tracking In Intelligent Video Analysis Research And Implementation Of The Algorithm

Today, the world is in the age of information explosion. It becomes a hugechallenge to extract useful information from the mass of data, especially from videodata, thus intelligent video analysis emerges as the times require. As one of the basictechnologies of video analysis, object tracking has caused a wide attention form theacademia and industry. This thiesis focused on the research and implementation ofobject tracking algorithm, introducing the the application of object tracking algorithm inintelligent video analysis and the basic tasks and categories of object tracking algorithm.The genenel steps of object tracking were summaried and the latest develompments ofobject tracking algorithm was also introduced.Intelligent analysis technology calls for efficient video port and more power of dataprocessing. These needs are just met by DavinciTMdigital signal processors (DSPs)launched by TI. The author participated in the development of two generation of videoanalysis systems based on TI DavinciTMseries chips. Based on DM6437DSP andDSP/BIOS, a framework of video capturing, processing, object tracking and the GUI onPC was constructed on the frist genernetion of system. The secend one was based onDM8168SoC whose video framework was constructed by Linux, SYS/BIOS,DVRRDK and McFW. The ARM cores were responsible for video capture display,coding/encoding, storage and control while the DSP ran the video analysis algorithm,thus the collaborative working of ARMs and DSP on DM8168was achieved. This thesisintroduced the two genenration of systems form the aspect of realization of hardwareplatform and the constructing of software function.Based on the above work, the author choose two object tracking algorithms ofMean Shift and particle filter which were implemented and optimized on the C64x+DSP to take full advantage of C64x+hardware. Without lowering the object trackingaccuracy, the time consumption of the algorithms cuted to one in six. The algorithmswere embedded in actual hardware platform and an intelligent analysis system withfunction of object tracking realized. This thesis detailed in the implementation andoptimization procedure of the two algorithms and summarized the four steps of algorithm research and implementation, introducing the principle, tools and methods tomake the algorithm implemented and optimized in specific hardware platform from itstheory.