Reconstructing urban height maps from single SAR intensity images via deep convolutional neural networks is becoming an alternative to conventional SAR-based 3D reconstruction techniques such as SAR interferometry or SAR tomography. However, even with this new technology, data gaps caused by radar shadowing remain an unsolved problem. In this paper, we investigate the potential of a fusion of single-image-based height maps derived from SAR data acquired from both ascending and descending orbits in the context of the generation of a simple LOD1 city model. Experimental results confirm that adding a second image from the opposing orbit leads to a significant boost in the quality of the city model.     «

Reconstructing urban height maps from single SAR intensity images via deep convolutional neural networks is becoming an alternative to conventional SAR-based 3D reconstruction techniques such as SAR interferometry or SAR tomography. However, even with this new technology, data gaps caused by radar shadowing remain an unsolved problem. In this paper, we investigate the potential of a fusion of single-image-based height maps derived from SAR data acquired from both ascending and descending orbits...    »