Texture and neural network for road segmentation

Road recognition is a main component in autonomous vehicle guidance system. We aim to determine whether an image element is part of the road or not, by using information from the element´s proximity. The wide range of situations have to be processed is a problem, and they impose on the system a capacity of self-adaptation to new cases. A neural network based architecture has been chosen to implement the texture analysis. By applying a local texture recognition process to the whole image, we obtain a segmented image of the road