Research On Road Detection And Lane Departure Warning Algorithm For Autonomous Vehicle

With the rapid development of transportation and the rapid increase of vehicle in China, traffic safety problem is increasing seriously; people pay more and more attention to the problem of traffic operation safety. In order to improve the overall safety performance of the vehicle, people begin to seek for an assistant driving system to improve overall safety performance of the vehicle, and the lane departure warning system based on machine vision technology is a key part in the assistant driving system. Lane departure warning system can give a signal alarm when the driver is in a fatigue and distraction state which leads to unconscious lane departure, so that the driver, so that the driver can correct the vehicle driving state in time to avoid traffic accident. Based on machine vision, this paper mainly studies and achieves the lane line detection and tracking and warning decision algorithm s for lane departure warning system.First, this paper introduces the research background and the meaning of this project and summarizes the research status at home and abroad. We mainly study on the structured road environment. In order to preferably acquire the lane line feature, we make experimental comparison and analysis on each step of commonly used algorithm for road image preprocessing, to select a set of suitable algorithm for this project. First of all, set the lower part of road image as static interested area, gray scaling by weighted average method; Then, de-noises the noise by median filter, and use OTSU for image binarization; At last, the improved Sobel operator is adopted to improve the edge enhancement.Then, set up lane line model, using an improved Hough transform based on prior knowledge to extract the global lane line; Improve the lane line tracking method, use the classical Kalman filter algorithm to predict the lane line parameters, establish lane line dynamic local interested search area, with the least square method to fit lane line in the local interested region and extract the lane line parameters. Chose different lane line recognition methods to extract line in initial detection stage and tracking stage, which not only guarantee the robustness and real-time of lane line identification, but also improve the real-time of algorithm. The experimental results show that the algorithm has a good recognition effect on each scenario.Finally, combining the road pavement information with the vehicle current location information, establish a lane departure warning model without camera parameter calibration. The algorithm is based on the distance which is between the centre of vehicle and lane boundary to judge whether the vehicle lane departure or not. The algorithm is simple to calculate and the alarm is accurate.In this paper, we test this algorithm of lane line detection and tracking on several videos of city roads. And make several lane departure tests to verify the effective of the lane departure warning decision method.