Learning, Fusion and Remote Sensing (AFuTé)

Publié le mar 1 Déc 2020

French version 

Group facilitators

Issue and perspectives

Artificial intelligence is at the heart of technological, scientific, economic, societal and environmental transformations. In this context, data modeling for description, decision, prediction and/or forecasting purposes becomes an essential issue. Our workgroup is dedicated to this issue and brings together methodological works dealing with machine learning (deep learning, data mining), the fusion of uncertain data (probabilities, possibilities, belief functions, fuzzy sets, intervals) and signal processing (wavelets, statistical learning, differential geometry). These works are mainly applied to/triggered by the analysis of remote sensing data with a temporal dimension; most often for environmental monitoring purposes (crustal deformation, erosion, deforestation, glacial retreat, marine pollution). Remote sensing work is also being carried out to produce such data, with the main objectives of measuring displacements, detecting changes and inverting models. Our scientific perspectives are multiple and concern 1) the consideration of the volume, uncertainty and complexity (spatial, temporal, physical properties) of data, 2) the fusion of data and/or models, and 3) the interpretability of the obtained results.

Keywords

Deep learning, data mining, data fusion, uncertainties, remote sensing, time series, environmental monitoring.

News

Dissertation defense: E. Amri, Automatic Offshore Oil Detection Based On Deep Learning Approaches Using Heterogeneous Data Fusion, June 27, 2022.

Best Student Paper Award 2020 awarded to Guilhem Marsy by La Revue Française de Photogrammétrie et de Télédétection for his publication « Automatic detection of moving areas in unaligned image series: Application à la surveillance des mouvements gravitaires » published in the special issue 217-218 of the 2018 CFPT Colloquium.

HDR defense: Y. Yan, Towards operational Earth deformation monitoring and natural hazard prediction, September 6, 2021.

PhD defense: M. Jacquemont, Cherenkov image analysis with deep multi-task learning from single-telescope data, 26 novembre 2020.

PhD defense: A. Hippert-Ferrer, Missing data reconstruction in remotely sensed displacement measurement time series, October 16, 2020.

PhD defense: G. Marsy,  Contribution of stereoscopic « time-lapse » optical imagery for the quantification at high spatio-temporal resolution (4D) of slope dynamics in mountains, September 22, 2020.

Talk on pattern discovery in satellite image time series and displacement field time series, MIAI Grenoble Alpes meeting, May 14, 2020.

PhD defense: C. Lesniewska-Choquet, Possibilistic stochastic modeling and application to change detection in remote sensing images, January 23, 2020.

PhD defense: M. Jauvin, Measurement of surface deformations by satellite radar interferometry – Application to the monitoring of mountain territories and the impact of large construction sites, December 18, 2019.

HDR defense: Alexandre Benoit, Automatic image analysis, from expert to apprentice, December 11, 2019.

Demonstration of DTFS-P2miner at ICDM 2019 (19th IEEE International Conference on Data Mining), Beijing, China, November 8, 2019.

Tutorial on the mining of displacement fields and their confidence measures using DTFS-P2miner, in collaboration with the DM2L and Imagine teams of UMR LIRIS. Autumn school of the MDIS 2019 national conference, Strasbourg, France, October 15, 2019.

Statistical Learning for Signal Processing workshop, Annecy, France, July 15-16, 2019.

SAR & Cryosphere workshop, Annecy, France, June 11, 2019.

Publications

Referenced in HAL.

Ongoing projects

ANR REPED-SARIX: Estimation and recursive prediction of Earth deformation from SAR image time series. 2022-2024.

CNES/TOCSA project SITS Deep : Monitoring of natural territories by satellite optical imaging and deep learning. In collaboration with UMR TETIS and UMR IMS. 2022.

Project iXblue: use of robust statistical methods for object detection and classification in sonar data. iXblue/LISTIC PhD subject. 2020-2024.

CNES/PNTS project SHARE: chronological series of sentinel-1 SAR images in mountainous terrains. Change detection over snow-covered surfaces (dry snow/moist snow) using automatic learning methods. In collaboration with UMR ISTerre, UMR LJK, CNRM, Magellium. 2021-2023.

Members

Professors and associate professors: A. AttoA. BenoitPh. BolonR. BoukezzoulaD. CoquinY. DumondS. GalichetM-P. HugetP. LambertG. Mauris, A. Mian, E. TrouvéL.ValetY. Yan.

Doctoral students: A. Bralet, L. CharrierA. Collas,  M. Gallet, S. Kaushik,  Viet-Hoa Vu Phan.

Completed projects

Project SMGA: use of AI techniques to detect cavities on mountain slopes. In collaboration with Géolithe. 2021.

Project RINA: creation of a demonstrator using AI methods for an operational management of geological Natural Hazards. In collaboration with CEREMA, BRGM and Géolithe. 2021.

Project Heliocity: classification of solar installation monitoring data using machine learning methods. 2021.

ANR MARGARITA: Modern Adaptive Radar: Great Advances in Robust and Inference Techniques and Application. 2019-2021.

Project TOTAL: oil slick detection, large ocean surface SAR data volumes, heterogeneous data fusion, deep learning. 2018-2021.

ANR ReVeRIES: recreational, interactive and educational plant recognition on smartphones. 2016-2021.

Project LDI I-TURN: real-time monitoring and control of a bar-turning machine using fuzzy systems (fuzzy rules, rule aggregation). 2020.

Project SmarterPlan/Linksium: detection and inventory of business objects by deep learning in 360° images of commercial building interiors. Funded by SATT Linksium. 2020.

CNES/TOSCA project START Deep: monitoring of vegetated territories using remote sensing and deep learning techniques. In collaboration with UMR TETIS and UMR IMS. 2020.

Project Géolithe : AI methods for processing data produced by airborne geological radars. 2020.

CNES/PNTS project: missing data restoration in displacement time series obtained from SAR images by statistical learning. 2019-2020.

GammaLearn: characterization of gamma radiationS by deep learning approaches applied to Cherenkov images provided by a single telescope. In cooperation with UMR LAPP, jointly funded by the European project ASTERICS and the Savoie Mont Blanc Foundation. 2017-2020.

ANR PHOENIX: parsimony, Huge Observations of Earth Non-stationarities from Images Time Series. 2015-2019.

ANR VIP-Mont Blanc: understanding and predicting environmental changes: a research project on the morphological evolution of the Mont Blanc massif. 2014-2018.

FUI G4M: multi-profession and multi-material geodetection. 2014-2017.

FUI MISAC: multi-functional Intelligent Surface for Automative and Aeronautics Cockpits. 2012-2015.

European project INTERREG GLARISKALP : risque glaciaire. 2011-2013.

ANR FOSTER: spatiotemporal data mining: application to the understanding and monitoring of erosion. 2011-2013.

ANR REVES: plant recognition for smartphone interfaces. 2010-2013.

ANR EFIDIR: information extraction and fusion for the measurement of displacements using radar imaging. 2008-2012.

Project ADIXEN: event forecasting in datastreams for predictive maintenance. 2007-2010.

ACI MEGATOR: measurement of the evolution of alpine glaciers by optical and radar remote sensing. 2004-2007.