New lane detection algorithm for autonomous vehicles using computer vision

Abstract

For navigation tasks it is necessary to determine position of the ego vehicle relative to the road. One of the principal approaches is to detect road boundaries and lanes using a vision system in the vehicle. This paper presents a simple and robust method designed to detect and estimate the curvature of road lane boundaries from images provided by a monocular camera. First, we use vector-lane-concept and non-uniform B-spline (NUBS) interpolation method to construct the boundaries road lane. Based on lane detection result, we estimate the curvature of left and right lane boundaries for autonomous guided vehicle systems application. Some experimental results based on real world road images are presented. These simulation results show the efficiency, feasibility and robustness of the algorithm.