Detecting and grouping words in topographic maps by means of perceptual concepts

We describe a method for the semi-automated extraction and clustering of characters occurring in scanned topographic maps. The method takes into account the oriented strings that are frequent in topographic maps. We describe the whole application but we emphasize the module devoted to string extraction based on a suitable human vision model related to this particular problem. The experimental results show the capability to exploit this method for semi-automated character extraction to the aim of building or updating data in Geographic Information Systems.