Discrimination between roofing materials and streets within urban areas based on hyperspectral, shape, and context information

In the context of automizing the process of urban mapping, hyperspectral imagery allows a detailed differentiation of characteristic surface cover types. Due to the spectral similarity of surface materials used for different surface cover types (e.g. roofing bitumen and asphalt), the spectral information alone cannot solve the ambiguities in the class decision process. Additional knowledge, such as context information, is necessary to improve the mapping of urban surface cover types. In this paper, an existing approach for the combination of hyperspectral data and shape knowledge is extended and improved for further automation of the image analysis. The technique is tested on hyperspectral data of the HyMap sensor. The results demonstrate the potential of this method.