Lane Mark Detection with Active Contour for Gradient Domain

Abstract

The extraction of lane mark is an essential step for several vision-based systems in intelligent transportation applications. In this paper, we propose a novel method for lane detection using a new active contour model. The new model is a modified version of the classical C-V model in gradient domain. The initial level set curve can be anywhere in the image without changing the segmentation result. To select the lane mark contour, moment invariant is utilized to make shape analysis. The experiment results on real scene image show that the proposed algorithm could detect lane mark accurately.