André Stumpf soutiendra sa thèse intitulée "Landslide recognition and monitoring with remotely sensed data from passive optical sensors" le Mercredi 18 décembre 2013 à 14h00, dans l'amphithéâtre Rothé (5 rue René Descartes) de l'Université de Strasbourg.
Jury :
Rémi Michel, UPMC/IPGP, Paris
Sébastien Lefèvre, UBS/IRISA, Vannes
Jean-Michel Dischler, UNISTRA/ICube, Strasbourg
Michel Jaboyedoff, UNIL/CRET, Lausanne, Suisse
Norman Kerle, UT/ITC, Twente, Pays-Bas
Jean-Philippe Malet, UNISTRA/EOST-IPGS, Strasbourg
Anne Puissant, UNISTRA/LIVE, Strasbourg
Christiane Weber, UNISTRA/LIVE, Strasbourg
Résumé :
Summary: Landslide detection, inventory mapping and monitoring are indispensable for disaster management, landslide hazard assessment and the quantification of erosion rates in mountain environments. The enhanced availability of VHR satellites, UAVs and consumer grade digital cameras offers a great potential to support those tasks at regional and local scales whereas there is still a lack of processing tools for the efficient extraction relevant information from different types of optical imagery.
This PhD research is dedicated to the development and evaluation of state?of?the?art machine learning and computer vision algorithms to support the mapping and monitoring of landslides with optical remote sensing data. The main research topics are: (A) The development of image analysis workflows for VHR satellite optical sensors combining object?oriented, supervised learning, and active learning algorithms to construct landslide catalogues after intense triggering events; (B) The evaluation of stereo? and multi?view image matching pipelines to measure surface deformation with both terrestrial and VHR satellite images at the local scale; (C) The adaptation of Gaussian matched filters for the detection of surface fissures in time?series of UAV images as geo?indicators of landslide activity; (D) The proposition of criteria and guidelines to select the most appropriate combination of sensors ? platforms ? image analysis techniques for landslide investigations according to different user needs.
The research has been established in the context of the European Project Safeland (Living with Landslide Risk in Europe, 2009?2012), the French?funded ANR Project Foster (Spatio?temporal data mining: application to the understanding and monitoring of soil erosion, 2010?2014) and the French Landslide Observatory ? OMIV (CNRS/INSU).