Robust road line color recognition based on 2-dimensional S-color space

Abstract

In this paper, we propose an illumination invariant lane color recognition method. Most of the conventional lane color recognition methods suffer from various illumination changes. In the past, the HSV color space has been commonly used to tell white and the yellow road lines, because the HSV color space is a range of specific colors. However, it is known that accurate road line recognition is difficult using the HSV space, because the road illumination is not static but always dynamic. In this paper, we propose a robust road line color recognition method by introducing a 2-dimensional S-color space. The white and yellow color features are clustered in the 2-D S-color space. The centroid of the feature samples in S-space is tracked continuously for real-time lane tracking.