| With the development of automotive industry, the number of automobiles grows rapidly, which causes traffic safety more and more terrible. So intelligent automobiles and automatic driving technology are in fresh demand. There are ten key systems in automatic driving technology and video-vision system is one of them. The author learn about it and do further research on traffic signs recognition which is part of it. Different from recognizing traffic signs in static images, the article plans to do the same thing in video stream. For 1080p@60Hz 24bit color deep video, the volume of a static image is about 47.5Mb, but the volume of a one-minute video is about 171Gb, which is 3600 times of the static image. So it is hard to handle with a large amout of video data.The article researches and analyzes currently popular theory and algorithms of traffic sign recognition firstly. On this basis, the article selects appropriate algorithms and do some improvements according to the hardware system. There are no hardware platform on the market which is in accordance with the design requirements. So a ten-layer printed circuit board (PCB) is designed, which takes Spartan-6 FPGA as video processing core and is also equipped with high-speed memory DDR3 (data rate is 800Mbps) and microcontroller ARM9. Signal integrity and power integrity of the hardware plateform are design difficulties.Software system is also designed to implement the function of recognizing traffic signs from video stream. The software system is mainly including FPGA programming and ARM programming, so two kinds of programming language are used——Verilog HDL and C. The video processing in traffic sign recognition and data storage of DDR3 are tough points in FPGA programming.After final test, the result indicates that both hardware design and software design meet the design requirements and implement the expected function of extracting and recognizing traffic signs from video stream. |
PDF Full Text Request