Lane Departure Warning System Design Based On The FPGA

With the growing economic strength, the people’s quality of life has been greatly improved.At present our country auto possession ranks second in the world, only behind the United States.The popularity of cars brings great convenient to people’s travelling, but meanwhile, producesnegative effects such as traffic accidents. In the world scope, traffic accidents caused a largenumber of casualties and property losses every year. In order to reduce the number of trafficaccidents, all kinds of auto safety technology has become the focus of the research of eachcountry. As an important automotive active safety technology, lane departure warning systemhas a wide range of research significance.Compared with foreign countries, the domestic research for related technology started late;so far a lot of academic achievements are concentrated on the research of algorithm based onthe PC environment. Be worth what carry is, in the last two or three years, related productsbased on DSP embedded lane departure warning system have appeared. Currently, domesticresearch based on FPGA in the lane departure warning system is less, and this paper putsforward a method based on FPGA and machine vision design. Due to the parallel processingand assembly line structure characteristics of the FPGA, so we can get faster processing speedon FPGA design. In this paper, CMOS light sensor collects the image information, then througha series of image preprocessing process, such as median filter and edge detection, finallyidentifies the lane line in the image of the position through the Hough transform method. In thepaper, we conducted former simulation validation and gate level simulation validation for thesystem design. Paper mainly works as follows:1) In the FPGA design, we completed OV7620light sensor driving interface modules, andrealized the function simulation and gate level simulation. Finally we set up the test platform,and verified the driving interface.2) On the comparison and analysis of several common image filtering method, andultimately choose the median filtering method, which has smaller computation and saves moreimage edge information. And some improvements have been proposed on the existing filteringmethod. Edge detection operator adopted Sobel operator to realize the image edge detection.Each function module has been designed, and also we realized the function simulation and gatelevel simulation. 3) Straight line identification adopted a robust Hough transform method, through theanalysis of the characteristics of Hough transform; the design gives the reasonable structure ofHough transform.