Key Technology Research Of Lane Departure Warning System

Traffic accidents have some similarities, seize the common accident can avoid accidents effectively. According to the statistical data from the US Federal Highway Administration, in the United States, all the terrible traffic accidents in 2006, there were forty percent accidents resulted of lane departure, and a rising trend year by year. As you can see, the lane departure phenomenon is the first problem for driving security. Lane departure warning system is a kind of device, witch can give notice in advance when the vehicle will deviate from the lane, to remind the driver to correct deviation error and avoid deviation phenomenon. Therefore, the introduction of lane departure warning system, which can reduce the incidence of traffic accidents and ensure the driving safety, is of great significance to the whole world.How to segment the image and extract the lane-marking information from the road images is the key technology of lane departure warning system. In the running process of a vehicle, it can cause much interference on the scene image due to the bad whether and obstacles. This will lead to extract the wrong feature points and let the lane detection unfaithful. In this paper, the road image is divided into blocks of information extraction, using dynamic threshold for the road image segmentation, it can remove the light uneven interference and reserve the lane marking information greatly. Besides, it contact Shenjun operators and Prewitt operators to extract the lane gradient map, for the purpose of selecting the lane marking points. To extraction the lane markings by Hough transform, according to the two lane geometry after perspective for the further filtering out of the lane detection. Finally, it will verificate the lane based on the statistics of the posterior algorithm to add the reliability, and provide accurate data support for departure warning system.The bent lane detection algorithm is also the key technology of lane departure warning system. It is a big challenge to fit and reconstruct the bent lane accurately due to the complicated road alignment and camera perspective effect. This paper adopts to linear-hyperbolic model to match the complex road alignment. In the condition of ensuring model precision, reducing the model to avoid complicated correlative coefficient calculation. Tracking the parameters of the model by using the Kalman filtering method on the assumption that video images are continuous. It can reduce the processing time of a single image and improve the real-time performance of the system. According to the change of angle between the two lanes on the two successive images or near scene and far scene on a single image as the switching condition between two models. It make the whole system more flexible and convenient.