The Research Of Moving Object Detection And Tracking System And Its Implementation On DSP

At present, the video surveillance system has been widely used in various aspects, and moving target detection and tracking is an important part of it. Object detection involves verifying the presence of an object and possibly locating it precisely. Object tracking is to monitor an object’s spatial and temporal changes. It is of great theory value and project meaning to realize a real-time,stable and effective Object detection and tracking system.Object detection and tracking is the first step of the intelligent video surveillance system, which is the basis of Intelligent Recognition and other high-level behavior understanding of motion objects. To meet the real-time requirement and the large volume of data processing in monitoring tasks, we make a in-depth research and discussion in object detection and tracing of video surveillance system, and design a Blackfin DSP-based moving object detection and tracing system.First of all, several commonly used moving object detection and tracking algorithms are discussed and compared, focusing on the frame difference method and the Meanshift method. Secondly, a complete hardware platform for the Blackfin DSP-based video processing system is designed, based on the overall hardware structure and the function of the system, including the communication and control module, video input and output module and external memory module, etc. Thirdly, the hardware platform is configurated and tested, incuding the the clock circuit, the DMA transmition and the I2C bus, focusing on how to use the PPI port to communicate the DSP board with PC and the FPGA subsystem, the test results are given. Then, using the tool of Visual DSP++, a model of frame difference method and Meanshift method is built and tested, including how to optimize the program. The results of the test show that the corresponding functions can be achieved, and the design requirements can be met.