Research On The Algorithm Of Lane Detection And Warning Strategy In Lane Departure Warning System

With the increasing quantity of vehicles, the safety and comfortable of vehicle comes to our focus. A rapid development of intelligent transportation system in recent years has improved the safety of driving in some sense, as a part of intelligent transportation system,lane departure warning system(LDWS) was mostly used in highway, it's purpose is to get the driver who is to o tired to get departure from the lane they are driving back to the right lane. In some sense, traffic accident get a rapid decent and transportation safety is improved as means of it.The main content of this thesis is to do research on the algorithm of t wo main technologies of lane departure warning system which can be listed as lane detection and the tragedy of warning. At first, the history and the condition of the development of the technologies in abroad and domestic of lane departure of warning system is introduced, and the problem of this system in lane detection and control strateg y is proposed. Then, each frame of the image which is get from the camera will be pre-processed, which purpose is to amplify the difference between the lane and the background, getting the lane out from the background. Then the lane and the background will be parted by the threshold. Pre-proposed get many algorithm, including image gray, image binaryzation, algorithm of threshold of segmentation, denoise, image morphology, edge detection, the introduction of proposed priority pixel. Then, the classical approach of Hough transformation in line detection and Sobel operator in edge detection are abandoned, the position of the lane will be fitted by the least square method. Then, the following region of interest and lane will be tracked by Kalman Filter method based on the position of the lane in the former image. In the tracking model, the efficiency and the precision of lane detection is improved. Based on the weakness of Kalman Filter in tracking the target whose variance is big, the Kalman Filter is improved. Through the experiment, the effectiveness of the improved is proved.In view of the disadvantages of classic warning model can be only used in inverse perspective image, through the way of simulation with the software of Prescan/simulink, the classic TLC(Time to lane crossing)warning model can be translate into a new warning model which is based on the angle of the two lanes and the distance from the gravity center of the car to the center of the lane. Different velocity and yaw angle can be set in Prescan software, based on the algorithm of warning model the warning time is easy to be seized, then through the lane detection algorithm the angle of the two lanes and the distance is extracted. At the end, the 3d model of warning is constructed based on velocity, angle and distance. Through a lot of experiment, the robustness and effectiveness of the warning strategy which is based on the angle between the two lanes an d the distance from the vehicle to the middle of the lane is proved.In the system of Windows, based on the platform of Opencv and Vc6.0,the algorithm that is proposed in this article is tested through the way of loading the video which is transcribed in the car, the real-timing and effectiveness is proved through the experiment which is utilized in lane departure warning system,and th e effectiveness of warning strategy is also proved,and the same time the odds of wrong warning is decreased.