Road detection from high-resolution satellite images using artificial neural networks

This article treats the possibility of using artificial neural networks for road detection from high-resolution satellite images on a part of RGB Ikonos and Quick-Bird images from Kish Island and Bushehr Harbor, respectively. Attempts are also made to verify the impacts of different input parameters on network's ability to find out optimum input vector for the problem. A variety of network structures with different iteration times are used to determine the best network structure and termination condition in training stage.It was found that when the input parameters are made up of spectral information and distances of pixels to road mean vector in a 3 × 3 window, the network's ability in both road and background detection can be improved in comparison with simple networks that simply use spectral information of a single pixel in their input vector.