An Adaptive Method of Color Road Segmentation

This paper presents a new adaptive and robust method for color road segmentation. The method shows good performance in segmentation of road with shadows. The color features of road are obtained in a sample region and are expressed as two scatter diagrams of red-green and blue-green. A fitting and predicting approach is used to extend the features to the whole image. The pixels in the image are then classified as road pixels or non-road ones based on the extended characters.