SUSO: Scene Understanding by SAR-Optical Data Fusion

• Applicant: Professor Dr.-Ing. Michael Schmitt
• Subject Area: Geodesy, Photogrammetry, Remote Sensing, Geoinformatics, Cartography
• Term: since 2020
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 455432852

The aim of the project is to develop a methodology for the semantic 3D reconstruction of urban scenes from SAR and optical remote sensing data. This methodology consists of both modern machine learning approaches and multi-sensor data fusion. Specifically, a method will be developed that combines multi-task learning for simultaneous estimation of land cover and urban topography from individual satellite images with classical SAR-optical stereo reconstruction in an iterative way. This iterative approach has several advantages: On the one hand, the topography reconstructed from single images of both sensor modalities is regularized by means of precise stereo measurements. On the other hand, the sensor-specific elevation maps are precisely co-registered to each other and georeferenced by stereo matching. The combination of SAR and optical data also has the advantage that, in the case of urban structures, both vertical and horizontal surfaces can be mapped, resulting in a true semantic 3D representation of the urban target area. The transferability of the approaches developed within the project will be evaluated experimentally on the basis of selected study areas and datasets. It is expected that the project will contribute to a more flexible use of multi-sensor satellite data for the holistic analysis of urban areas.

• DFG Programme: Research Grants
• Co-Investigator: Dr.-Ing. Stefan Auer, Ph.D.