Geocoding of Synthetic Aperture Radar Images Using Digital Vector Maps

In this letter, a possible approach to automate synthetic aperture radar (SAR) geocoding (or to automatically refine the existing one) is presented. The procedure is based on edge detection, where ground control lines obtained from digital vector maps are located in the SAR images. In geocoding, affine, projective, and rigorous SAR sensor transformation models can be used. According to the experiments, the procedure seems to be accurate and reliable as long as the initial values of the geocoding transformation model are accurate enough. The procedure copes well with displacements of about ten pixels on SAR images, but good results have been obtained in cases of even larger displacements. According to the well-defined checkpoints, the resulting geocoding accuracies for test images were around a few pixels. When the rigorous SAR sensor model was used, the accuracies were less than two pixels in most cases.