Research On Moving Object Segmentation And Its Implementation In Embedded System

The research on moving object segmentation algorithm is always one of the hottest topics in the fields of image processing and the most important tasks in actual video processing systems, as well as the key step of completing object tracking and recognizing. Meanwhile, it directly influences the accomplishment of computer vision and image processing, etc. It improves the development of the relevant applications including image processing, pattern recognition, artificial intelligence, computer vision, etc. It is widely applicable in the fields of video surveillance, video compression, robots vision, missle guidence, medical diagnosis.TMS320DM642 is a 32-byte fixed-point Digtal Signal Processor (DSP) with the high performance designed for the application of multimedia processing fields by Texas Instruments Corporation. DM642 is based on C64x-core framework and integrates many perpherals including video I/O ports and so on. It is cost-effective in embedded video processing system. In this thesis, the moving object segmentation algorithm is developed on DM642 hardware target board.In the thesis, the algorithm of moving object segmentation in the complex background is implemented in the embedded system. Firstly, based on the given scenario the right algorithm is selected by Matlab simulations from the moving object segmentation algorithms, including frame difference, background difference and optical flow. Secondly, the algorithm of moving object segmentation is accomplished in the DSP system. Its development consists of the research on the methods of OpenCV migration into the DSP system, the allocation of resources of DSP and the design of software modules and the coding of the algorithms. Thirdly, the embedded system of moving object segmentation algorithm based on SEEDVPM642 is accomplished. Fourthly the algorithm optimization, the debugging of its programmes and the relevant solutions are summarized. Finally, the segmentation of the results and the analysis of the proposed system are illustrated.