Automatic Extraction of Urban Road Networks from Ikonos Images Using a Fuzzy Mathematical Morphology Approach

High-resolution commercial imaging satellite such as IKONOS provides an important new data sources for urban mapping and geographic information systems (GIS) applications. This paper presents a fuzzy mathematical morphology method for automated extraction of urban road networks from IKONOS imagery. In this proposed method, the road networks in a complex urban scene are firstly modeled, followed by modeling other different types of undesirable structures such as buildings and trees, both of which are based on their geometric and radiometric properties in the IKONOS imagery. Then, the shape, size, and gray-scale values of the structuring elements are selected based on their geometric and radiometric properties in order to determine which objects should be retained or eliminated. A series of fuzzy morphological operators combined with the different structuring elements are used in order to retain bright, long rectangular structures such as roads and eliminate other undesirable non-road structures.