Benthic habitat mapping using UAV photogrammetry and machine learning algorithms

Traditional methods for marine habitat mapping are often time-consuming and rely on sparse sampling. To overcome these limitations, a new method has been developed for large-scale mapping of benthic marine habitats. By deploying several synchronized autonomous underwater drones and using AI, this new approach enables large-scale, continuous, and high-precision seabed mapping and monitoring at depths down to 200 meters.

Several survey areas covering up to 10 hectares were surveyed focusing on ecologically significant habitats such as Posidonia seagrass meadows, kelp forests, and sensitive areas like harbor access zones. The aim was to precisely detect, map marine habitats and individual species, thus contributing to environmental impact assessment for a coastal civil engineering project and the exploration of relevant metrics for long-term monitoring. This solution has also been used for fiber optic cable localization and coupling estimation for Distributed Acoustic Sensing (DAS) experiments and enables the precise detection of subsea infrastructure and obstacles, such as unexploded ordnance (UXOs).

As a result, this solution enhances the understanding and the ability to better monitor benthic habitats while also supporting more effective risk management and operational planning. Its capacity to conduct large-scale mapping and long-term monitoring of benthic ecosystems opens new perspectives for the monitoring and sustainable management of marine resources.