2023_programme: Emulation of Optical Images for Underwater Target Classification using Sonar Imagery

• Session: 15. Towards Automatic Target Recognition. Detection, Classification and Modelling
  Organiser(s): Johannes Groen, Yan Pailhas, Roy Edgar Hansen, Jessica Topple and Narada Warakagoda
• Lecture: Emulation of Optical Images for Underwater Target Classification using Sonar Imagery [invited]
  Paper ID: 2020
  Author(s): Albert Thomas, Berthomier Thibaud, d'Alès Benoît, Pailhas Yan, Furfaro Thomas
  Presenter: Berthomier Thibaud
  Abstract: The task of detecting, classifying, identifying and localizing underwater mines with sonar imagery is a difficult process due to the challenging nature of the environment. Furthermore, the creation of comprehensive, representative datasets, involving data collection in real environments, can be expensive and time consuming. This project proposes a novel approach to tackle this issue by using optical images to generate emulated sonar data. In a dark room, representative dioramas were staged using target models laid in a sandbox. The scene was illuminated with a strong offset light to create highlights and shadows as in a sonar image. An optical camera then captured a diverse set of data, with varying light source, orientation, range, altitude, objects’ shape, burial, type, and environment complexity. The optical images were then modified to mimic forward looking sonar images. This preliminary dataset enables us to quickly develop an automated target recognition framework for particular sensors without requiring a large collection of “real” data. The resulting algorithm was then tested on real sonar images to validate the methodology and its accuracy. This approach has the potential to lessen the costs and time associated with traditional data collection methods, while providing a more comprehensive set of data.
    Download the full paper
• Corresponding author: Mr Thibaud Berthomier
  Affiliation: STO CMRE
  Country: Italy
  e-mail: