Adaptive SAR Image Processing Techniques to Support Flood Monitoring from Earth Observation Data

Abstract

This chapter addresses the exploitation of Earth Observation (EO) data in the operational chains for flood monitoring and post-event damage assessment, focusing specifically on the task of post-flood mapping. In this context, this work provides a general review of our research into image processing techniques with an emphasis on adaptive methods applied to synthetic aperture radar (SAR) images. These procedures involve no restrictions on SAR acquisition parameters (frequency band, polarization, spatial resolution, and observation angle). Depending on the data availability, different maps can be produced. When multi-temporal images are available, two different products can be generated: fast-ready flood maps; and detailed flood maps;. The former is a color composite image that enhances the visualization of changes that have occurred after an event. The latter is a more detailed map obtained after a segmentation process. In contrast, when only an image acquired on a single date is available, a water body map can be generated. All these maps are intended as support for institutional interventions. Since only methods of segmentation and numerical data fusion are applied, such results are not final classification products. They are symbolic and not semantic maps, generated using fast and simple procedures that can be used as input for a classification purpose or employed by the user in other application tasks. The experiments described here were performed on real SAR images related to different datasets. The images were acquired from COSMO-SkyMed (CSK) and RADARSAT satellites.