This paper highlights the potential of combining Synthetic Aperture Radar (SAR) and optical data for operational rapid urban mapping. An algorithm consisting of a completely unsupervised procedure for processing pairs of co-registered SAR/optical images is proposed. In a first stage, a texture analysis is conducted independently on the two images using eight different chain-based Gaussian models. In a second stage, the resulting texture images are partitioned by an unsupervised fuzzy K-means approach. Finally, a fuzzy decision rule is used to aggregate the results provided by the classification of texture images obtained from the pair of SAR and optical images. The method was tested and validated on images of Bucharest (Romania) and Cayenne (French Guiana). These two study areas are of different terrain relief, urban settlement structure and land cover complexity. The data set included Radarsat-1/ENVISAT and SPOT-4/5 images. The developed SAR/optical information fusion scheme improved the capabilities of urban areas extraction when compared with the separate use of SAR and optical sensors. It also proved to be suitable for monitoring urbanization development. The encouraging results thus confirm the potential of combining information from SAR and optical sensors for timely urban area analysis, as required in cases of disaster management and planning in urban sprawl areas.
Keywords: SAR sensors; fuzzy K-means classification; information fusion; optical sensors; rapid urban mapping; texture analysis.