Abstract
Improving the geo-localization of optical satellite images is an important pre-processing step for many remote sensing tasks like monitoring by image time series or scene analysis after sudden events. These tasks require geo-referenced and precisely co-registered multi-sensor data. Images captured by the high resolution synthetic aperture radar (SAR) satellite TerraSAR-X exhibit an absolute geo-location accuracy within a few decimeters. These images represent therefore a reliable source to improve the geo-location accuracy of optical images, which is in the order of tens of meters. In this paper, a deep learning-based approach for the geo-localization accuracy improvement of optical satellite images through SAR reference data is investigated. Image registration between SAR and optical images requires few, but accurate and reliable matching points. These are derived from a Siamese neural network. The network is trained using TerraSAR-X and PRISM image pairs covering greater urban areas spread over Europe, in order to learn the two-dimensional spatial shifts between optical and SAR image patches. Results confirm that accurate and reliable matching points can be generated with higher matching accuracy and precision with respect to state-of-the-art approaches.
Highlights
We generated a dataset out of 46 orthorectified optical (PRISM) and radar (TerraSAR-X acquired in stripmap mode) satellite image pairs acquired over 13 city areas in Europe
To minimize the impact of the different acquisition modes of PRISM and TerraSAR-X, we focus on flat surfaces where only the radiometry between the synthetic aperture radar (SAR) and optical images is different
For a better evaluation of our results, we compare our method with three available baseline methods: the similarity measure normalized cross-correlation (NCC) [55], the similarity measure mutual information (MI) [56], and a MI-based method (CAMRI) which is tailored to the problem of optical and SAR matching [10]
Summary
Data fusion is important for several applications in the fields of medical imaging, computer vision or remote sensing, allowing the collection of complementary information from different sensors or sources to characterize a specific object or an image. Satellites like TerraSAR-X [1] exhibit an absolute geo-localization accuracy in the order of a few decimeters or centimeter for specific targets [2]. Such accuracy is mainly due to the availability of precise orbit information and the SAR imaging principle. Both images have a ground sampling distance of 1.25 m
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
