Satellite and Underwater Sonar Image Matching Using Deep Learning

  • Matheus Santos FURG
  • Giovanni Giacomo FURG
  • Paulo Drews FURG
  • Silvia Botelho FURG

Resumo


Aerial images provide rich information about the Earth surface, while underwater perception is severely limited due to the physical characteristic of water. This work proposes a new problem domain where a matching process fuses aerial and underwater images. The proposal is designed to aid underwater navigation in partially structured environments such as marinas and harbors. A pipeline combining image processing techniques and Convolutional Neural Networks is presented. The method is validated in a real dataset with quantitative and qualitative results.
Palavras-chave: Satellites, Image segmentation, Acoustics, Sonar, Geology, Semantics, Image color analysis
Publicado
23/10/2019
SANTOS, Matheus; GIACOMO, Giovanni; DREWS, Paulo; BOTELHO, Silvia. Satellite and Underwater Sonar Image Matching Using Deep Learning. In: SIMPÓSIO BRASILEIRO DE ROBÓTICA E SIMPÓSIO LATINO AMERICANO DE ROBÓTICA (SBR/LARS), 16. , 2019, Rio Grande. Anais [...]. Porto Alegre: Sociedade Brasileira de Computação, 2019 . p. 108-113.