Road boundaries detection using color saturation

Our aim is to automatically detect the road borders in a road scene image. This is useful to many road scenes analysis applications, in both fields of vehicle guidance and civil engineering. Difficulties arise because pavements are often heterogeneous and because illumination variations often occur in outdoor scenes. Some vehicle navigation projects use colour images for road borders detection [1, 3], but most of the time they consider RGB features, which are sensitive to shadows and pavement defects. Besides, the proposed classification methods are often computationally intensive [7]. In contrast, chromatic saturation is a discriminant one-band feature, so a simple threshold can be used, making the classification quite fast. Moreover, chromatic saturation is quite insensitive to shadows and pavement variability.