Georeference of High Resolution TerraSAR Images with RPFs

The first civilian satellite SAR sensor was launched in 70s. Optical sensors, launched in the same decade, attracted more interest than SAR sensors, as they have the advantage of true optical color, while the complex geometry of SAR sensors usually makes radar images look odd, especially in ragged terrain and urban areas. During the last 40 years, optical systems were actively developed to offer products of 1 or even 0.4 m resolution. During the same period SAR products were of very low resolution of 10 or even 25 m.

But this has recently changed. During the last two years, there are commercially available high resolution SAR images by the German satellite TerraSAR-X and by the Italian satellite constellation CosmoSkyMed. The resolution of 1 m that is available for these systems, can be compared to optical systems. Although modern SAR products are still very expensive and they suffer from speckle noise and distortions inherited from their geometry, SAR is an all-weather, day and night sensor, offering information about the properties of the targets, their 3D geometry and their evolution through time. As a result, there is a growing demand for processing methods on par with optical processing methods.

In this paper, Rational Polynomial Functions (RPFs) are investigated for the georeference of high resolution TerraSAR images. A whole scene is georeferenced using real, measured Ground Control Points (GCPs). The scene covers a typical area in Greece: ragged terrain covered with sparse vegetation and scattered buildings. The results are tested with the use of independent Check Points (CPs).