ROAD EXTRACTION AIDED BY ADAPTIVE DIRECTIONAL FILTERING AND TEMPLATE MATCHING

In this paper, the problem of the detection of road networks from Optical/SAR images is addressed. Our procedure try to overcome problems due for the noisy surround and improve road extraction by implementing a directional filter. The method takes into account four separate steps. The first consists of an automatic procedure to adapt the direction of filtering on the basis of the predominant direction of the roads present into of the image, the second step try and discard the “blobs” that do not possess the usual characteristics of the roads such as elongation, the third step entails the real road extraction by the Hough transform routine. Then, in order to eliminate redundant segments and to avoid gaps between part of the same road we apply an algorithm able to connect each other the extremities of the segments on the basis of tolerances and related to the spatial resolution. The proposed procedure was tested on a couple of images, one a fine resolution SAR images and a fine resolution optical image. The experiments have shown an increase in the completeness and correctness indexes after the procedure with respect to more standard extraction methods.