Can ants build urban street networks?

Our paper deals with the detection and automatic extraction of a hierarchical network of urban streets from maps containing only building footprint data. We develop a new approach for extracting, locating and labelling plausible street networks in a given city, based on geometrical an functional considerations. Using some basic tools from the "Mathematical Morphology" field, we propose simple, robust and efficient techniques for extracting homotopic skeletons of the ground zones and a street-width map. Our method leads to the construction of open or closed connex graphs that we encode and save. Finally, we develop some ant-based techniques to identify plausible elements (streets, boulevards, avenues, lanes, water streams) in this graph.