e-mail: belarte (at) unistra (dot) fr phone: +333 68 85 45 78
Bruno Belarte ICube - UMR 7357 Pôle API 300, boulevard Sébastien Brant CS 10413 F - 67412 ILLKIRCH Cedex Bureau C331
Thesis: Extraction et analyse de relations spatiales entre objets d'intérêt dans les images de télédétection guidées par des connaissances du domaine.
Directors: Cedric Wemmert (MCF HDR ICube) and Christiane Weber (DR Géographie, LIVE, UdS)
Co-advisors: Germain Forestier (MIPS) and Manuel Grizonnet (CNES, DCT/SI/AP)
Funding: CNES (french space agency) and Région ALsace
Summary: The new satellite sensors allow the acquisition of images of a very high level of detail at high speeds, thus producing a large amount of data. The manual processing of these data has become impossible, new tools are needed to process them automatically. In this context, effective segmentation algorithms are required to extract objects of interest of these images. However, the segments produced by the algorithm do not match the objects of interest. We propose to change the level of abstraction in order to interpret the objects of interest as objects composed by segments. For this, we have implemented a multi-level learning process, based on expert knowledge, in order to learn the rules of compositions defining objects of interest. To manage the imprecision of the analysis of remote sensing images we propose to use fuzzy logic to model the rules of composition. The proposed method is validated on very high spatial resolution images acquired by QuickBird and Pléiades sensors.
- IUT Computer Science S1: Computer architecture (20h TD / 48h TP)
- IUT Computer Science S1: Computer architecture (28h TD / 52h TP)
- IUT Computer Science S1: Computer architecture (14h TD / 28h TP)