Using Deep Learning to Sweep for Underwater Mines

using-deep-learning-to-sweep-for-underwater-mines

NATO’s Centre for Maritime Research and Experimentation (CMRE) is a scientific research facility located in La Spezia, Italy. Formerly the NATO Undersea Research Centre (NURC), the center focuses on technology research in the defense of maritime forces against terrorism and piracy. It works to develop secure networks, mine countermeasure systems, protection of ports and harbors, anti-submarine warfare, modeling and simulation, and marine mammal risk mitigation. One CMRE research priority is to remove the need for Navy personnel to risk their lives clearing mines. The CMRE is refining the use of high-performance, efficient information processing systems installed in underwater robots that work in real time as a new way to sweep for mines.

Why There Is a Need

In the Adriatic and other European seas, the seafloor is littered with tens of thousands of mines, bombs, and other munitions that were lost or abandoned after World Wars I and II. Because of the risk of explosion or the release of toxic chemicals such as mustard gas, these mines pose a constant danger to commercial shipping fleets and fishermen, as well as the environment and the food chain.

Clearing the mines is a big problem. When World War II ended, more than 25,000 U.S.-laid mines were still in place off the shores of Japan. The U.S. Navy was unable to sweep them all, limiting its efforts to only critical areas. After sweeping for almost a year, the Navy abandoned its efforts in 1946, leaving 13,000 mines still in place. Over the next 30 years, more than 500 minesweepers were damaged or sunk while trying to clearing the remaining mines. Given the fact that 80% of all commercial trade travels by sea, removing mines remains imperative decades after the two wars ended.

How Deep Learning Can Help

Traditional automated underwater vehicles (AUVs) work along preplanned survey routes and record all the data for offline processing. They do not independently adapt to environmental conditions or sonar performance. To develop a better way to detect mines, the CMRE has developed a new autonomous underwater vehicle called MUSCLE (Mine-hunting UUV for Shallow-water Covert Littoral Expeditions) and included deep learning technology to give MUSCLE adaptive capabilities. Using high-resolution, high-frequency synthetic aperture sonar (SAS) married with the deep learning convolutional neural net (CNN), MUSCLE features onboard object pattern recognition capability to locate, identify, and classify mines in real time. MUSCLE is well suited for operations in shallow coastal waters where deeper draft vessels cannot go. It can produce high-quality images of objects on the seafloor, is self-navigating, and has the ability to cover a large area sailing on its own.

The Path Forward

Perhaps the most exciting aspect of this technology is the research being conducted to modify it to send a warning to ships that are approaching underwater mines. The value of saving lives and ships cannot be overstated and the information collected by one ship could be shared with others and lead to an eventual solution of the mine problem. The lives saved would not be limited to humans; technology may also someday be adapted to help protect marine mammals, fish, and other sea creatures.

MUSCLE is also being used to collect data that will allow NATO researchers to achieve state-of-the-art advancements in the area of seabed mapping. Improved capabilities in seabed mapping could also lead to advancements in other fields, such as environmental and archaeological surveys, as well as mine sweeping.

Applications like MUSCLE are only one of numerous practical uses for artificial intelligence (AI), and Tractica anticipates that the defense sector will be one of the largest markets for AI during the coming decade. In Tractica’s recently published Artificial Intelligence Market Forecasts report, we forecast that global AI revenue in the defense sector will reach $2.3 billion by 2025.

md-aimf-3q16-chart-defense

Comments are closed.