Modern Earth Observation systems provide remote sensing data at different temporal and spatial resolutions. As regards optical sensors, nowadays, the Sentinel-2 program supplies images with high temporal resolution (every 5 days) and high spatial resolution (10m) that can be useful to monitor land cover dynamics. On the other hand, Very High Spatial Resolution images (VHSR) are still an essential data to figure out land cover mapping characterized by fine spatial patterns. Understand how to effectively leverage together these complementary sources of information to deal with land cover mapping is still challenging. With the aim of dealing with land cover mapping through the fusion of multi-temporal High Spatial Resolution and Very High Spatial Resolution satellite images, we propose an End-to-End Deep Learning framework, named M³Fusion, able to leverage simultaneously the temporal knowledge contained into time series data as well as the fine spatial information available in VHSR images. Experiments carried out on the Reunion Island study area asses the quality of our proposal considering both quantitative and qualitative aspects. Finally, the performances of our method are also compared to a standard machine learning approach (Random Forest) used as baseline.

M³Fusion : Un modèle d’apprentissage profond pour la fusion de données satellitaires Multi-{Echelles/Modalités/Temporelles}

BENEDETTI, PAOLA;Ruggero Pensa;Dino Ienco
2018-01-01

Abstract

Modern Earth Observation systems provide remote sensing data at different temporal and spatial resolutions. As regards optical sensors, nowadays, the Sentinel-2 program supplies images with high temporal resolution (every 5 days) and high spatial resolution (10m) that can be useful to monitor land cover dynamics. On the other hand, Very High Spatial Resolution images (VHSR) are still an essential data to figure out land cover mapping characterized by fine spatial patterns. Understand how to effectively leverage together these complementary sources of information to deal with land cover mapping is still challenging. With the aim of dealing with land cover mapping through the fusion of multi-temporal High Spatial Resolution and Very High Spatial Resolution satellite images, we propose an End-to-End Deep Learning framework, named M³Fusion, able to leverage simultaneously the temporal knowledge contained into time series data as well as the fine spatial information available in VHSR images. Experiments carried out on the Reunion Island study area asses the quality of our proposal considering both quantitative and qualitative aspects. Finally, the performances of our method are also compared to a standard machine learning approach (Random Forest) used as baseline.
2018
Conférence Française de Photogrammétrie et de Télédétection CFPT 2018
Marne-la-Vallée, France
25-28 June 2018
Conférence Française de Photogrammétrie et de Télédétection CFPT 2018
1
8
https://rfiap2018.ign.fr/sites/default/files/ARTICLES/CFPT2018/Oraux/CFPT2018_paper_benedetti.pdf
Land Cover Mapping, Data Fusion, Deep Learning, Satellite Image Time series, Very High Spatial Resolution
Paola Benedetti, Raffaele Gaetano, Kenji Osé, Ruggero Pensa, Stephane Dupuy, Dino Ienco
File in questo prodotto:
File Dimensione Formato  
CFPT2018_paper_benedetti.pdf

Accesso aperto

Descrizione: paper
Tipo di file: POSTPRINT (VERSIONE FINALE DELL’AUTORE)
Dimensione 2.59 MB
Formato Adobe PDF
2.59 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2318/1670094
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact