Deep Learning for Matching High-Resolution SAR and Optical Imagery

Abstract

The joint exploitation of SAR and optical data constitute the most important application of data fusion within remote sensing. A key first step in data fusion endeavours is the determination of correspondences between the various data sources. However, due to their vastly different geometric and radiometric properties, the SAR and optical matching problem has few generalizable solutions. Thus the main objective of this thesis is to develop a fully automated, deep learning-based, SAR-optical matching pipeline suitable for use on high-resolution imagery.