Research On Lane Departure Warning System For Structured Roads

With the rapid development of road traffic and increasing number of car ownership, the incidence of traffic accidents is also increasing, and traffic safety problem is getting more attention. To solve the problem, with the current rapid development of technology, people began to seek a way using computer technology to reduce traffic accident, and computer vision-based driver assistance systems(DAS) is one of the pop topics. Lane departure warning system (LDWS) is also one of the most important parts of DAS, which is designed to remind the driver to avoid the danger of driving from the right route that caused by fatigue driving, so that the driver can timely make measures to keep the vehicle in the right way.The foundation of a LDWS is lane detection and tracking, that is, first to get the lane information through the camera, and then to judge whether the car is departure, last to make warning to the driver. This paper first introduces the basic algorithms of lane detection, studies how to get the lane information of an image in a fast and effective way, puts forward some own ideas and makes some experimental verification, and finally the whole lane departure warning system has been achieved. The article is divided into the following sections:(1) Introduces the background and significance of computer vision-based lane departure warning systems, describes the current development at home and abroad, and analyzes the whole system.(2) Describes and analyzes relevant methods of each module in image preprocessing. Considering the specific requirements of road image processing, selects the right and has targeted improved some of the algorithms(using median filter for image enhancement, improved Sobel operator of edge detection, Otsu of image binarization).(3) Discusses lane recognition algorithm in detail, select the lane recognition model. By partitioning the road image to get the region of interest of lane detection, chooses different methods in the initial lane detection and lane tracking phase(improved Hough transform for initial detection and least square fitting for tracking), which not only ensures the robustness but also improves the real-time performance.(4) Introduces several LDWS models and selects the appropriate one(TLC model). Put forward valid deviation criteria.(5) Summary and Outlook. Make summary of the whole article, analyze the advantages and disadvantages of the methods used in the article, and the future work is expected.