Research And Implementation Of Human Tracking Based On Embedded Systems

Detection and tracking technology of moving human are hot and difficult in computer vision technology. Intelligent security monitoring system based on detection and tracking technology of moving human can not only protect the safety but also save a lot of manpower and resources. So detection and tracking technology of moving human has broad application prospects and enormous social and economic benefits.Currently, most moving human tracking system are PC-based framework. They analysis and computing the video data that obtained by camera. However, PC’s Shortcomings are large bulk, high power dissipation, and high cost. Therefore, Thesis adopted TMS320DM642 which is a Digital Signal Processor produced by Texas Instruments and CCS software system as development platform. On this basis, thesis studied the detection and tracking of moving human based on Embedded Systems.Thesis introduced the development status of the detection and tracking of moving human and the relevant theoretical basis. Thesis introduced embedded video processing system based on DM642 processor and software development environment. All these work provided an effective basis for the detection and tracking of moving human. Considering the computing capable of embedded systems and relatively large changes in speed of human, thesis adopted the multiple frame difference method. Driver configuration programs of related modules were provided. On the basis of the hardware system design, thesis studied the kernel structure and function of DSP/BIOS and designed the application system. They provided foundation for algorithm validation based on embedded systems. In order to extract the moving human target from the moving targets effectively, thesis adopted aspect ratio of moving human’s external rectangle. In view of the human body’s block issue, thesis studied the tracking method based on Kalman motion estimation and Camshift tracking method, and combined the Camshift with Kalman motion estimation method to improve reliability and accuracy of tracking. The experimental results obtained by embedded systems proved the effectiveness of work in thesis.