Automatic Fusion of SAR and Optical Imagery based on Line Features

Abstract:
A generic image processing chain for the fusion of very high resolution optical and SAR imagery is developed. It is fast to operate and provides reasonably good accuracy. This method is capable of rapidly registering optical and SAR images based on extracted features rather than on pixels. Indeed, at very high resolution, pixel approaches fail on above ground structures. An approach on a higher semantic level is required. The fusion is generally applicable to many kinds of remote sensing imagery due to it being generic. Its goal is to enable feature based registration of unrectified optical and SAR images with sub metric resolution in urban areas. Important aspects of this fusion approach are the very high resolution of the images, the introduction of a classification, rapid processing and the modular construction of the entire processing chain based on the open source software library ORFEO Toolbox (OTB).