A multi-step curved lane detection algorithm based on hyperbola-pair model

Abstract

In this paper, we propose a multi-step algorithm based on a hyperbola-pair model for lane detection. We represent the lane markings on the road by a modified hyperbola-pair model, which contains two parts. The first one is a parallel straight line model, which is achieved by Hough transform. The second one is a hyperbola-pair line model, which is achieved by a searching strategy with the parameters got in the first stage as initial parameters. Experiment results show a robust performance in a noise condition, such as part of the lane markings are occupied by other vehicles in the lanes. The algorithm can provide accurate information of location and curvature of the lane markings in real-time.