Improved geometric accuracy in SAR sensors implies that more complex models of the Earth may be used not only to geometrically rectify imagery, but also to more robustly calibrate their radiometry. Current beta, sigma, and gamma nought SAR radiometry conventions all assume a simple "flat as Kansas" Earth ellipsoid model. We complement these simple models with improved radiometric calibration that accounts for local terrain variations. In the era of ERS-1 and RADARSAT-1, image geolocation accuracy was in the order of multiple samples, and tiepointfree establishment of the relationship between radar and map geometries was not possible. Newer sensors such as ASAR, PALSAR, and TerraSAR-X all support accurate geolocation based on product annotations alone. We show that high geolocation accuracy, combined with availability of high-resolution accurate elevation models, enables a more robust radiometric calibration standard for modern SAR sensors that is based on gamma nought normalised using an Earth terrain-model.
A REVISED RADIOMETRIC NORMALISATION STANDARD FOR SAR
ABSTRACTImproved geometric accuracy in SAR sensors implies that more complex models of the Earth may be used not only to geometrically rectify imagery, but also to more robustly calibrate their radiometry. Current beta, sigma, and gamma nought SAR radiometry conventions all assume a simple "flat as Kansas" Earth ellipsoid model. We complement these simple models with improved radiometric calibration that accounts for local terrain variations. In the era of ERS-1 and RADARSAT-1, image geolocation accuracy was in the order of multiple samples, and tiepointfree establishment of the relationship between radar and map geometries was not possible. Newer sensors such as ASAR, PALSAR, and TerraSAR-X all support accurate geolocation based on product annotations alone. We show that high geolocation accuracy, combined with availability of high-resolution accurate elevation models, enables a more robust radiometric calibration standard for modern SAR sensors that is based on gamma nought normalised using an Earth terrain-model.