Detection System, Based On The Dm642 Embedded Vehicle

With the development of the computer vision technology and the embedded system, a vision-based embedded vehicle detection system has become possible. This thesis focuses on developing an embedded remote traffic monitoring system. Firstly, the vehicle detection algorithm is designed and implemented based on PC, which is transplanted to the DM642 embedded platform. The system transmits video and traffic parameter information by Ethernet and implements remote surveillance.During the vechile detection algorithm design, the thesis presents an automatic vehicle detection method based on difference accumulation information. Effective background is modeled by accumulative frame-to-frame difference and could be real-time updated. Then, the moving object can be distilled by background difference method. The vehicle is segmented accurately by the Enhanced Otsu adaptive threshold method. The results show that the proposed algorithm is efficient, adaptive and robust.We adopt the multimedia processing TMS320DM642 from TI Corp. as the core processor for video processing. The system is capable of capturing the video signals from four channels simultaneously, displaying, video processing and Ethernet transmission in a certain condition. During the software design, we adopt DSP/BIOS development tool and RF5 from TI Corp.The Network Developer's Kit provided by TI facilitates us in sending the compressed video packages and traffic parameters to the Surveillance Center via Ethernet.Druing the client software design, we adopt Socket, multithreading and Directshow technologies under the Visual C++ development environment. The software implements surveillanceUI design, data transmission and video play . we use TCP protocol to carry out network communications with DM642, adopt Directshow to finish video decoder and play, the client can obtain the traffic information and choose video channel by surveillance UI.