Gabriel VASILE

Welcome to Gabriel Vasile's Home Page

Area of Expertise

Data Processing & Analysis
Machine Learning, AI
Speckle, Radar
Remote Sensing, Earth Observation
Polarimetry, Interferometry, Tomography
Acoustics and Vibration

Midjourney AI Artwork

Contact Information

GIPSA-lab, CNRS / Grenoble INP
11 rue des Mathématiques
Domaine universitaire – BP 46
F-38402 Grenoble Cedex, France

gabriel.vasile AT grenoble-inp.fr

Midjourney AI Artwork

Hebbian Hub

Advocating for free software, independent benchmarks and innovative applications in the fields of Machine Learning & Artificial Intelligence (ML & AI). Visit the Hebbian Hub.

Gabriel's Bio – Research Scientist

Gabriel Vasile received the M.Eng. degree in electrical engineering and computer science and the M.S. degree in image, shapes, and artificial intelligence from POLITEHNICA University, Bucharest, Romania, in 2003 and 2004, and the Ph.D. degree in signal and image processing from Savoie University, Annecy, France, in 2007.

Professional Experience

2008 – present

Research Scientist, CNRS – Section 03

GIPSA-lab (UMR 5216), Data Science Department, Grenoble

Co-Head of PICSA team since 2026. Head of SigmaPhy team (2020–2025). Scientific responsible of the Mont Blanc technical platform since 2021.

2007 – 2008

Postdoctoral Researcher (CNES fellowship)

ONERA DEMR, Palaiseau – polarimetric texture analysis of SAR images

2006 – 2007

Development Engineer, CNRS

Laboratoire des Images et des Signaux (LIS – UMR 5083), Grenoble

2002 – 2003

Engineering Assistant, IMT

National Institute for R&D in Microtechnologies, Bucharest

Education

2004 – 2007

Ph.D. in Electronics / Electrotechnics / Control Systems

Savoie University (LISTIC) & POLITEHNICA University (LAPI) – cotutelle

Prix de thèse "Bourse de la Vocation – Prix des Neiges" 2008

2003 – 2004

M.S. (DEA) in Image, Shapes and Artificial Intelligence – Major

POLITEHNICA University of Bucharest

1998 – 2003

M.Eng. in Electrical Engineering & Computer Science

POLITEHNICA University of Bucharest

PhD Supervision

2025 – present

Simone MORICI (co-supervision 50%)

CIFRE with NXP Semiconductors · Dir.: Marco Congedo (GIPSA-lab)

"Geometric and imaging methods for automotive radars"

2023 – present

Abdelazyz RKHISS (supervision 100%)

CIFRE with ENGIE Green & CSTB

"Wind turbine noise extraction using deep learning"

2019 – 2023

Ileana-Madalina CIUCA (co-supervision 30%) – defended June 2023

Cotutelle with UPB · Dir.: M. Gay & S. Ciochina

"PolSAR image processing methods, monostatic and bistatic"

Sep – Nov 2025

Weili TANG (doctoral visit)

Collab. S. Vitale (Parthenope U. Naples) & G. Fornaro (IREA-CNR)

"Deep-Learning Based Detection of Persistent Scatterers on Sentinel-1 data"

Teaching (55h eq. TD / year)

Grenoble INP – Phelma & IUT 1 Grenoble:

Digital Signal Processing (20h)
Metrology and Sensors (5h)
Numerical Methods (30h)

Fellowships & Awards

2009–2013, 2017, 2019

ERASMUS / ERASMUS+ Teaching Fellowships

UPB / Constanta Maritime U. / Valahia U.

2017

Campus France – Découverte Chine

Chinese Academy of Sciences, Beijing

2008

Bourse de la Vocation – Prix des Neiges

Interdisciplinary PhD award, Savoie & Haute Savoie

2005

Bourse du Gouvernement Français

Cotutelle PhD fellowship

2003–2006

EGIDE, ERASMUS & MIRA Fellowships

International Scientific Exchange

Annual visits – Politehnica Univ. of Bucharest (2010–2023) · ETASM school (2022–2025)
U. of Alaska Fairbanks (2023) · Chinese Academy of Sciences (2017) · French Embassy Beijing & Shanghai (2016)
Crna Gora U. (2011, 2012, 2015) · Arizona State U. (2014) · Villanova U. (2011, 2013) · Schlumberger (2013) · U. of Rijeka (2013)

Academic Collaborations

S. Vitale (Parthenope U. Naples) & G. Fornaro (IREA-CNR) – SAR satellite imagery
S. Ciochina & A. Anghel (UPB) – radar satellite imagery, RRSM
E. Trouvé (LISTIC-USMB) & J.-M. Nicolas (LTCI-Télécom Paris) – glacier monitoring

Languages

French (fluent) · English (fluent) · Romanian (mother tongue) · Italian (average)

Media & News

"Ils veulent mesurer 7j/7 les ondes EM", Dauphiné Libéré, 2 Aug 2022
Radio New's FM 101.2 – Surveillance de la pollution EM, 4 Feb 2022
Forum des idées & Budget participatif de Grenoble 2022
Détection du cormoran, l'Ain Agricole, 27 Dec 2019
Envoyé spécial, France 2 (Oct 2007) · Le Monde (Oct 2006)
Fête de la Science: Grenoble (2008), Chambéry (2007), Chamonix (2006)

Publications

Full list from HAL · IdHAL: gabriel-vasile · ORCID: 0000-0002-6458-3634 · IdRef: 120253240

Fetching publications from HAL…

Research Activities

Research themes focus on new signal processing methods for multivariate imagery (very high spatial resolution) for environmental monitoring by remote sensing.

PolSAR Signal Processing — Geometric Classification (PolBaRi)

The PolBaRi algorithm exploits the polar decomposition S = UH of the scattering matrix, where the Hermitian factor H is rotation-invariant. Classification uses Riemannian k-means++ with the AIRM metric on the manifold of positive definite Hermitian matrices. Validated on RAMSES (X-band) and EMISAR Foulum (C-band) datasets, achieving 97% accuracy while preserving polarimetric texture. Riemannian extensions of Davies–Bouldin, Krzanowski–Lai, and Silhouette indices enable fully data-driven class number selection (Vasile & Congedo, IEEE IGARSS 2025).

Automatic class number optimization via the Riemannian Calinski-Harabasz index.

PolSAR — Real Representation Scattering Matrix (RRSM)

For bistatic SAR, our RRSM formulation converts the complex consimilarity problem to a real eigenvalue decomposition. Eigenvalues appear in ±pairs: real for reciprocal targets, complex conjugate for significant asymmetry. Validated on Full-Wave simulations (C-band, 5.405 GHz). Directly applicable to upcoming missions: ESA Harmony, LuTan-1, DLR HRWS.

Wind Turbine Noise Estimation — Deep Learning

In A. Rkhiss's CIFRE thesis (ENGIE Green / CSTB), LSTM networks with Huber loss and constrained activation estimate wind turbine noise L_WTN from total noise L_TN without stopping the machines. MAE of 0.26 dB(A) with environmental data (wind speed + electrical production), vs. 0.72 dB(A) with acoustic data alone.

RNN architecture: two recurrent layers, batch normalization, dense layer, dropout, linear output.

Wind turbine noise predictions with (left) and without (right) environmental data.

Underwater Acoustic Monitoring — Fish Counting (ACOUEAU)

The FEDER ACOUEAU project (2017–2021) prototyped an IoT ultrasound fish counter (300–600 kHz, bistatic). Detection models fish passages as rare events with a Gamma-distribution confidence interval. Classification uses deconvolution + Hilbert transform + matched filtering. Deployed at Breil-sur-Roya: >95% trout detection across 5 target classes.

ACOUEAU: experimental chain and acoustic signature classification (96% trout, 85% leaves, 100% wood).

Glacier Monitoring by SAR Remote Sensing

Combining D-InSAR and amplitude correlation on instrumented glaciers (Mont-Blanc massif), we quantified velocities up to 1.7 m/day on the Bossons glacier and revealed seasonal dynamics from space. TanDEM-X analysis quantified radar penetration bias in dry snow (4–6 m in X-band above 3500 m). Sentinel-1 (6-day revisit) enabled continuous systematic monitoring.

Professional Service

Editorial Activity

Senior member IEEE · Member IEEE GRSS
Associate Editor – IEEE Trans. Geoscience and Remote Sensing since 2022 (~20 papers/month)
Editorial Board – AI section, MDPI Remote Sensing since 2020
IEEE reviewing: TGRS, TAES, GRSL, JSTARS, TIP, JSTSP
Reviewing: AMS MathSciNet, IET, Elsevier, EURASIP JASP, OSA, ISPRS JPRS, Traitement du Signal, Télédétection

Conference Service

IEEE IGARSS TPC & reviewer (2008–2026) · GRETSI reviewer (2022, 2023, 2025)
IGARSS 2025 & 2026: 3 sessions chaired + Session Organizer "SAR Tomography" & "Subsurface Remote Sensing"
Co-chair IGARSS (2010, 2020, 2023) · Co-chair EUSIPCO (2014)
Chair/Co-chair MTS/IEEE OCEANS (2015–2018)
TPC: CoSeRa (2015–2020), COMM (2010, 2012, 2014), ISSPIT, ICA, APACE, CENCON, IEACon, PES APPEEC

Workshop Organization

École de Physique des Houches spring school (2011)
Workshop Analyse multivariée des images radar – GIPSA-lab (2013)
Non-Stationary Signal and Image Processing – COMM 2014, Bucharest

Grant Evaluation

Romanian CNCS: evaluator/rapporteur – PCE-2012, TE-2012, PCCA-2013, PED-2016, PED-2019, PED-2021, PED-2023 (10 proj., 2024), PED-2024 & PCB-RO-MD-2024 (13 proj., 2025) · ICT panel 2022 & 2025
French ANR: ASTRID (2015), AAPG2021 · German DFG (2020) · Croatian Science Found. (2019)
ANRT CIFRE (2020) · UGA IRGA (2021) · OSU THETA (2021) · Reims EXEBIO (2024) · Slovak SRDA (2025)
1 jury de thèse at Univ. de Limoges (examiner, 2020)