Fusion of high-resolution InSAR data and optical imagery for building detection using Conditional Random Fields

Konferenz: EUSAR 2010 - 8th European Conference on Synthetic Aperture Radar
07.06.2010 - 10.06.2010 in Aachen, Germany

Tagungsband: EUSAR 2010

Seiten: 4Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

Wegner, Jan Dirk; Soergel, Uwe (Institute of Photogrammetry and GeoInformation, Leibniz Universität Hannover, Germany)
Haensch, Ronny (Computer Vision and Remote Sensing Department, Berlin Institute of Technology, Germany)
Thiele, Antje (Fraunhofer Institute of Optronics, System Technologies and Image Exploitation, Germany)

State-of-the-art satellite SAR sensors (e.g., TerraSAR-X, CosmoSkyMed) provide imagery of one meter resolution and airborne SAR sensors achieve even higher resolutions. In those data objects in urban areas become visible in high detail. However, SAR-typical effects like layover, shadowing, and interfering backscatter of multiple objects complicate interpretability. Thus, additional information about those objects may be obtained from optical imagery. In this work we combine features of high-resolution airborne interferometric SAR (InSAR) data with features of an orthophoto in order to detect buildings. A Conditional Random Field (CRF) is set up in order to integrate context-knowledge. We show that CRFs are a suitable method for integration of both context-knowledge and multi-sensor features for building extraction.