Enhancing Bathymetric Data Processing with Artificial Intelligence: A 2D U-Net Approach for Hydrographic Applications

Accurate bathymetry is critical for navigation safety, oceanographic modeling, military operations, and sustainable seafloor management. The CHOF (Future Hydro-Oceanographic Capacity) armament program, led by the DGA (French MoD Procurement Agency), aims to modernize the acquisition, processing, and dissemination of hydro-oceanographic information. Under this program framework, a joint initiative by Shom (French Hydrographic Office) and AMIAD (Ministerial Agency for Defense AI) explores artificial intelligence (AI) in order to accelerate bathymetric data processing. Inspired by previous research on Multi Beam Echo Sounder (MBES) outlier detection, this work focuses on developing sovereign deep learning models. It introduces criteria for selecting representative data for the training phase and specific evaluation methods tailored to the operational requirements of hydrographic services. This work highlights the potential of AI to enhance bathymetric data processing, ensuring consistency, scalability, and operational efficiency while maintaining compliance with hydrographic standards.