Research On Vehicle License Plate Recognition System Based On Zynq

As a core part of the intelligent transportation systems, the technology of vehicle license recognition can play a very important role in the intelligent traffic management. The technology of vehicle license recognition based on video processing merges image processing, computer vision and pattern recognition technology. This kind of recognition technology has the advantages of high recognition rate, good versatility, fast recognition speed etc. It becomes one of the important research directions in transportation field.This paper proposes a new system implementation method of license plate recognition system based on FPGA+ARM on a condition of finishing some researches. This new method takes full advantage of excellent parallel processing ability and rich logic resource of FPGA,coupling with the advantage of ARM in the treatment of complex control algorithm and running an operating system, etc. It has characteristics of high computing speed, good real-time ability, easier development, and so on.This paper chooses an HSV based positioning method to locate the license plate according to the characteristics of texture and color. Then, after using Hough transform to modify the inclined license plate image, a method of the combination of an improved projection method and the prior knowledge is utilized to divide the character of license plate. Finally, an improved template matching method is used to recognize the normalized characters.The system is based on the development platform of ZedBoard produced Xilinx company and use the Zynq-7020chip combining FPGA+ARM as the core. The Linux system running on ZedBoard captures images via USB camera, implements license plate recognition algorithm within Qt and OpenCV functions. The recognition result is outputted to HDMI displayer which is controlled by FPGA. Finally the paper describes that the license plate recognition program is tested the ZedBoard. This system can complete the image acquisition and program realization and has high recognition speed, low consumption and a good human-computer interaction function.