Abstract
This paper discusses recent advances in data fusion of remotely sensed data for urban area characterisation. It focuses mostly on feature fusion, as a way to combine information available from multiple sensors, multiple spatial resolutions and multi-temporal acquisitions. The application on urban areas reveals the trend towards application-driven approaches, i.e. techniques aimed at using specific data sets and extracting peculiar features for well-defined tasks. This is also one of the reasons of the reduced importance of pixel-based fusion in this research area.