This paper reports results on the use of color vision for road estimation. A main technique is the use of a transformation of the color video signal to a hue-saturation-intensity (HSI) representation. The hue component then becomes relatively insensitive to shadows and varying road surfaces, facts that often deteriorate the performance of road estimation schemes on unstructured roads. The transformation to a birds-eye perspective by the inverse perspective mapping (IPM) technique brings further clarity to the results. The methods have been tested in schemes for lane change aid and road departure alarm with good results, using live recordings. The test material includes a wide variety of road types, traffic densities and illumination conditions. For the covering abstract see ITRD E114174.
Abstract