EFIDIR: Extraction and Fusion of Information for Displacement Measurement by Radar Imaging - ANR project Data Mass and Knowledge 2008-2011
- University Savoie Mont Blanc / LISTIC : Laboratory of Computer Science, Systems, Information and Knowledge Processing.
- University of Rennes 1 / IETR : Institute of Electronics and Telecommunications of Rennes.
- CNRS / GIPSA-Lab: Grenoble Image Parole Signal Automatique.
- CNRS / LTCI: Information Processing and Communication Laboratory, Télécom ParisTech.
- CNRS / LGIT: Laboratory of Internal Geophysics and Tectonophysics, Chambéry/Grenoble.
- CNRS / LG: Geology Laboratory, ENS Paris.
Emmanuel TROUVÉ, LISTIC, University Savoie Mont Blanc University.
The data currently provided by SAR (Synthetic Aperture Radar) sensors on remote sensing satellites are particularly voluminous: typically 1 Gigabyte is required to archive 1 interferometric pair of a 100kmx100km ERS scene, and at least 40 pairs are required to archive the data. The data currently provided by the SAR (Synthetic Aperture Radar) sensors of remote sensing satellites are particularly voluminous: typically 1 Gigabyte is required to archive 1 interferometric pair of an ERS scene of 100kmx100 km, and at least 40 pairs are required for a time series. To enable a complete and systematic exploitation of this large amount of data, it is necessary to develop specific processing techniques to access geological, geomorphological or tectonic information. The availability of this large quantity of data, combined with the uncertainties on the parameters of the physical models (ground truth and/or acquisition systems) leads to the use of data mining and information fusion techniques.
The EFIDIR project aims to develop an open archiving and processing platform adapted to the specificities of SAR data and to the large time series used for displacement measurements. In the 1990s, thanks to the data acquired by the first generation of satellite SAR sensors (ERS-1, ERS-2, JERS, Radarsat-1), it was shown that, by judicious use of time series at a given site, it was possible to deduce, over large areas, displacements of the order of a fraction of a wavelength. In particular, on examples of urban subsidences, differential array interferometry (using stable scatterers called PS: permanent scatterers) achieves millimetre-level accuracies on the displacement, but requires large time series (at least 40 images). At present, new generation SAR satellites are - or will be - placed in orbit: ENVISAT, ALOS, Radarsat-2, Terrasar-X and CSK. Most often associated with new modalities (such as polarimetry), these sensors deliver new data, which are even more voluminous in terms of storage space because they are better resolved, opening the way to new applications, but which will also require reformulations of the processing chains currently used.
The need for a " Mass of Dataand Knowledge" project covering the entire information processing chain is explained, on the one hand, by the volume and nature of the data (raw data "RAW data" or complex multi-varied images "SLC data" supplied by the space agencies) and, on the other hand, by the various phenomena that disturb and bury the information sought: speckle, decorrelation, atmospheric disturbances: To obtain thematic information from these multi-temporal interferometric and polarimetric data, a complex processing chain must be implemented (SAR synthesis, interferogram generation, PolSAR/Pol-InSAR decomposition, fringe unwinding, georeferencing, artefact correction, geophysical model inversion,...). At present, this chain exists only in partial form through closed and expensive commercial software where it is very difficult to make adaptations for a new theme or a new type of data (high resolution, polarimetric, etc.) for which they were not initially intended. Also, in this context of specific databases, the design, production and validation of specific codes complementary to the processing elements available in free software are the aim of this project.
The proposed application is characteristic of the expectations of a user community, the "geophysicists", with regard to the "information processing" community to design and make operational original approaches capable of overcoming the current obstacles thanks to the exploitation of the mass of data and the taking into account of the expertise of the field. The project relies on databases linked to several original themes such as
- the low amplitude (but large spatial extension) movements linked to the filling of large dams (Serre Ponçon lake, studied by the LG),
- glacier surface movements (LISTIC, GIPSA, LTCI), both localised and of greater amplitude, for which, following the ACI MEGATOR Data Mass (2004-2007), a vast and diverse database is available (ERS, ENVISAT), in the process of being provided (ENVISAT, ALOS, E-SAR) or to be provided (for satellites not yet launched: Radarsat-2, Terrasar-X),
- volcanic movements (Mexican "laboratory" volcanoes studied by the LGIT). In these different cases, the current treatment chains have shown their limits.
The project is divided into three sub-projects:
- SP1 which aims to develop complementary tools for signal and image processing justified by the proposed themes. To this end, work will be carried out on "coherent targets" and on refocusing, as well as on the introduction in this context of tools dedicated to polarimetric data.
- SP2, which aims to integrate these tools into software platforms that allow the retrieval of ground displacement parameters. It will focus on small displacements (mainly based on "permanent targets"), and on large displacements (with a particular emphasis on polarimetric data).
- SP3, information fusion step, which is intended to allow the transition to geophysical quantities. This last phase will validate the complete processing chain on pilot thematic applications.
In order to achieve these objectives, this multidisciplinary project brings together 4 "ICST" laboratories specialising in SAR image processing and information fusion and 2 earth science laboratories. This grouping guarantees the development and dissemination of methodological tools guided and validated by the applications. These advances will make it possible to transform the mass of data from satellite radar imagery into displacement measurements and thus enrich the knowledge of observed geophysical phenomena.