Sonar is usually used to map the ocean floor, and seabed composition (e.g., sand, sand, or rock) affects how the sound is reflected.
This usually means that sonar measurements at different depths and distances can give accurate soundings of the sea ‘s possessions. For example, how submerged currents propagate, how the warmer sea changes with the climate, or where best to listen to whales.
The research was part of a job contracted by The Defence and Security Accelerator (DASA), a part of the Ministry of Defence, to improve monitoring of the UK’s vast marine territories using high tech sonar.
The technology may also be potentially used for sea tomography across entire ocean basins, like the Arctic, to examine the consequences of climate change on the oceans and better empower the sustainability of individual activities in fragile environments and ecosystems.
Dr. Blondel explained: “There are many different variables that influence how noise waves are propagated in water, as some frequencies of sound can travel farther than others.
“If you think of an orchestra’s sound, as you go further away, you may lose the high-frequency noise of the violins but continue to be able to hear the lower frequency notes of the cellos. The beating of drums could be felt even further.
“This is identical to sea sounds, which come from the weather, like rain and storms, the animals, such as fish and snakes, and people, with ships and offshore activities.
“For this project, we wanted to model how sonar echoes were changed with depth, salinity, and temperature so that we could use sound to quantify these factors in the ocean.”
The investigators first analyzed the many features of underwater environments and classified them into different types.
They used Probabilistic Generative Modelling to develop several AI algorithms for identifying submerged surroundings.
After creating the AI algorithm, the investigators tested their performance on a wide selection of simulated acoustic data representing a broad spectrum of underwater environments.
The tests demonstrated that their Probabilistic Principal Component Analysis (PPCA) algorithm could classify submerged environments from simulated sonar measurements with an average accuracy of 93%.
An alternative Latent Variable Gaussian Process (LVGP) model also showed a strong performance and enabled them to attain an even higher classification accuracy of 96%.
The simulations showed that accurate classification could happen even with sonar measurements over short spatial periods, making it appropriate for practical usage, e.g., slow-moving autonomous vehicles.
Marcus Donnelly, Technical Lead in Environmental Data Science in SEA Ltd, stated: “This project surpassed all our hopes for AI algorithms applied to the complexity of sonar in the underwater environment.
“We anticipate continuing our cooperation with the IMI following favorable comments from the MoD.”
The researchers expect the technique could be utilized in the long run to monitor the effects of climate change.
Dr. Blondel stated: “Climate scientists monitor sound propagation in the ocean around the rods to observe temperature changes over time. Our techniques could help determine where to find monitoring stations to provide the most comprehensive data employing the optimal variety of measurements.”