Vision-based road lane detection for autonomous vehicles

Abstract

Issues related to saving lives by increasing vehicle safety on the road and reducing of road accident have been of great interest to researchers in the context of developing Advanced Driver Assistance Systems. Among the complex and challenging tasks of future road vehicles is road lane detection or road boundaries detection. This is based on lane detection, which includes the localization of the road, the determination of the relative position between the vehicle and road and the analysis of the vehicle’s intended direction. One of the principal approaches to detect road boundaries and lanes is using computer vision system. The purpose of this research is to present a vision based lane detection approach that has the capability to detect lanes in varying road conditions with robustness to lighting change and shadows. In this system a camera is first mounted on the vehicle to get the front view. In the preprocessing phase data is subjected to grayscale conversion, noise removal and edge detection with automatic thresholding to obtain edges. Lines are extracted by using Hough Transform method with the restricted search area. Lane boundary scan uses the information from the previous phases to return a series of points on the right and left side which represent the lines or boundaries of the road. Finally, pair of hyperbolas are fitted to these data points to represent the lane boundaries. The proposed lane detection system can be applied on painted road with curved and straight road structure under different weather conditions. This system has been implemented by MATLAB7.1.The experimental results show that the proposed scheme is able to detect the road lane marking efficiently under different conditions achieving a 96.6% during the day and 92.6% at night time. Ultimately, the system has given superior and robust results compared to other existing systems.