A robust lane detection method based on hyperbolic model

Lane detection is an essential part of safety assurance in intelligent vehicle and advanced driver assistance systems. Despite many methods having been proposed, there still remain challenges such as complex road surface and large curvature. In this paper, we present a robust lane detection method under structured roads to solve these issues. The method contains two parts: straight line detection in near field and curve matching in far field. Instead of generating top-view image by inverse perspective mapping (IPM), we propose a new form of IPM application to reduce noise that we only take advantage of sub-pixel-level spatial relations and project line segments obtained by line segments detector to top-view image. Then, we apply density-based spatial clustering of applications with noise to clustering segments and design a fusion method to extract the optimal lines combination from clusters. Finally, a weighted hyperbolic model is proposed to finish curve fitting. The results of experiment indicate that the method has robust performance in complex environment.