A Robust Vision-based Lane Detection using RANSAC Algorithm

In a pursuit to reduce ever increasing road accidents by developing an advanced driver assistance system (ADAS), this paper proposes a vision-based lane detection algorithm. This paper incorporates a framework constituting of color space conversion, region of interest (ROI), adaptive histogram equalization, clustering of lane pixels, and RANdom SAmple Consensus (RANSAC) to develop a lane detection algorithm. The advantage of adaptive histogram equalization is to adjust the pixel intensity of Shadow and illumination regions in the road image using a contrast limit function. Further, clustering of a lane pixels is used to count and accumulate lane pixels above certain threshold. Finally, a RANSAC algorithm is applied to remove outliers and fit the lane lines model. The advantage of proposed framework is to detect the ego-lane and also all the lane boundaries in the image plane. Moreover, based on visual analysis, algorithm reveals a superior lane detection performance suitable for illumination variation, shadow, and lane variant width.