Rapid Land Mapping by TERRASAR-X VHR data

According to many application requirements, especially in the security field of applications of Earth Observation (EO) data, rapid mapping is one of the more important tasks to be performed. It may be useful for damage assessment, treaty monitoring, emergency management, and it’s mainly based, due to time constraints, on manual data analysis and digitization of this visual analysis results. As such, it is based on optical data only, since SAR data require radar experts to be understood and correctly interpreted. SAR data are those more likely to be available in the immediate aftermath of any event, however, due to the all-weather capabilities of the radar sensor, coupled with the very high spatial resolution (mandatory for man-made structure monitoring) and the short revisit time of low earth orbit systems like TerraSAR-X There is therefore the need for rapid mapping and possibly automatic tools able to exploit these data and provide usable maps to decision makers and final users. This work is devoted to present a processing chain suitable to analyze TerraSAR-X VHR data and provide within hours detailed maps of the areas of interest. A common methodology for scene interpretation is based on knowledge-based segmentation of the image into simpler elements, exploiting the relationships between objects and features. This approach, usually labeled as “top-down” analysis is also implemented in this work. The novelty of the analysis proposed in this work is the contemporaneous exploitation of spectral and spatial features. Spatial feature are here referring to both texture analysis and linear element extraction and recombination, which allows a better characterization of the elements in the scene than each of the two spatial analysis taken alone. Moreover, specific approaches are introduced for different parts of the scene, and associated spatial features are chosen accordingly.