Multiscale multisensor decision level data fusion for urban mapping

The combination of multiple maps of urban area extents is becoming more and more common, as the availability of multiple EO data sets is increasing constantly, and different teams develop efficient techniques to extract human settlements from SAR and optical data. The next challenge is the need to combine all of these maps using an approach suitable to multi-source but also multi-resolution data sets. This work provides a first step towards this aim, by considering a “pixel aggregation” technique which proves to be more effective than standard techniques to resample all data sets on a common grid at the intended spatial resolution. Results on Beijing using SAR data from three different sensors and multispectral data from Landsat are offered to prove the effectiveness of the proposed procedure.