The Designing Of HD Vehicle Detection System Based On Fpga

With the dramatic increase of the number of cars, urban traffic situation attracts people's attention more and more. How to improve traffic management in the urban has become the focus of public attention. Not like electromagnetic induction coil based system, digital image processing based traffic control and surveillance system has won a continuing expansion programme of civil authority, because of its high processing speed, low cost, convenient installation and maintenance, versatility, and so on.This paper implements a vehicle detection system based on image processing. After discussing the disadvantages of PC-based and DPS based vehicle detection system, an embedded vehicle detection system based on FPGA is proposed, which has high reliability, stability. The traffic scene is acquired by the CMOS or CCD camera, and processed by the FPGA. Due to the highly parallel processing ability and re-configurable property of FPGA, the system can analyze the vehicle information more real-timely and flexibly.A new generation of embedded video vehicle detection system is realized,which is based on modular hardware system design procedures. The system is comprised of DDR2 SDRAM frame buffer control module, image pre-processing module, and image detection modules. The problem of image sampling, storage, processing and transmission are well resolved by the system, and good foundation is laid in terms of system functionality expansion. In the algorithm, the video-based vehicle detection algorithm is studied. The background subtraction method, image binarization and the threshold of using histogram analysis method are discussed. The image is processed using histogram equalization, and the image detection accuracy is improved. In the realization, hardware structure of the algorithm on the FPGA system is designed which utilizes embedded systems processing technology.The system uses the Verilog HDL language. Functional simulation and performance analysis are processed in each module and algorithm. Experimental results show that the system has high detection accuracy and good noise immunity. The system can effectively improve the detection rate, with good real-time.