DEM-assisted neural network for SAR-to-optical image translation - Laboratoire d’Informatique, Systèmes, Traitement de l’Information et de la Connaissance
Communication Dans Un Congrès Année : 2024

DEM-assisted neural network for SAR-to-optical image translation

Résumé

SAR-to-optical remote sensing modality translator neural networks are mainly trained on flat areas preventing their use to detect gravitational movements as landslides in steep sloped areas. In this paper, we first propose a new SAR-DEM-optical dataset in mountainous regions to improve performances of SAR-to-optical image translators under these extreme conditions. Then we upgrade SARDINet (SAR Distorted Image translator Network) model previously developped for urban areas, to take a Digital Elevation Model (DEM) together with the SAR image as input and perform translation in natural mountainous environment. Multiple fusion strategies are explored to merge efficiently SAR and DEM images: late fusion, early fusion but also an intermediate fusion based on balanced separable convolutions. These approaches are compared to the original SARDINet and two standard adversarial networks -Pix2pix and CycleGAN -showing improvements in distorted regions.
Fichier principal
Vignette du fichier
Article_IGARSS24_final.pdf (5.31 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04720021 , version 1 (03-10-2024)

Licence

Identifiants

Citer

Antoine Bralet, Trong Nghia Ngo, Emmanuel Trouvé, Jocelyn Chanussot, Abdourrahmane Atto. DEM-assisted neural network for SAR-to-optical image translation. IGARSS 2024 - IEEE International Geoscience and Remote Sensing Symposium, Jul 2024, Athens, Greece. pp.7649-7652, ⟨10.1109/IGARSS53475.2024.10641788⟩. ⟨hal-04720021⟩
8 Consultations
3 Téléchargements

Altmetric

Partager

More