Geospatial AI for Maritime Surveillance

Deeper look at the Bay of Bengal and exploring solutions to thwart intrusion cases using artificial intelligence and deep learning

By Sadhli Roomy

Maritime Security is a facet of national security entailing protection of marine environment, marine economy, and human security over regional seas and territorial waters and rivers. Intrusions like IUU fishing, piracy, drug and arms trade, human trafficking etc. are regular occurrences in every regional waters in the world and have the potential to cripple a nation’s maritime resources, and/or affect human lives and the economy at large. For the significance of maritime resources and for the sake of preserving national security of regional and international waters,governments deploy security agencies to patrol and monitor their waters.

To put economic impact into context, it's been reported that a ‘dark fleet’ of hundreds of Chinese fishing vessels illegally caught squids worth 0.5 billion from the North Korean waters since 2017. Dark Fleets equals to ghost ships who turn off their AIS(Automatic Identification System). For communities dependent on fishing as a livelihood option and primary food source, breach like these can work to destabilize entire villages and/or ports – threatening fish stocks, seasonal predictions,and marine ecosystem. And this is just ‘one’ case of maritime intrusion to contend with.

Intrusion Cases in the Bay of Bengal

The Bay of Bengal is a water region spanning 2,600,000 square kilometers and bordered by India, Bangladesh, Myanmar, Andaman and Nicobar Islands, Sangaman Kanda, Sri Lanka, and Sumatra. These nations, along with other peripheral countries and landlocked regions depend on the bay for fishing, tourism, maritime transport, resource extraction, e.g. oil, gas, and hydrocarbon reserves.

Intrusions in the Bay of Bengal consist of IUU fishing, piracy, illegal drug and arms trade, and human trafficking predominantly if we compare multiple sources.

Case-in-point: Ya ba trade in Bangladesh

Illustrating a case of illegal drugs, Bangladesh has a drug epidemic of ‘Ya ba’ abuse, an addictive drug with a mixture of methamphetamine and caffeine. This tablet contraband is manufactured in Myanmar and is imported through illegal distribution channels. There are two ways that these get into the country, one is through Rohingya refugees who are employed as carriers – who smuggles it to Bangladesh. Because of lack of livelihood opportunities and desperate conditions in refugee camps in Cox’s Bazar, Rohingya refugees are often documented to engage in drug and arms trade to earn a living. The second route for this drug to enter Bangladesh, and possibly the more problematic one, is through fishermen in Cox’s Bazar who takes the tablet andbrings in either via Cox’s Bazar or travels all the way to Kuakata situated on the west to bring those in via new route.

The question is: why is this route more problematic than the other? While there is no official record of this, Bangladeshi fishermen travels to international waters to collect drugs from Burmese vessels, travels back to Teknaf or other coastal regions and/or travels to Kuakata directly. It is difficult for security agencies to track this behavior since international waters are beyond the Coast Guard’s jurisdiction and hence monitoring of these waters are limited. Conversely, Bangladeshi fishermen can’t enter the Myanmar EEZ since the Myanmar BGP canarrest and/or, if circumstances demands, shoot at fishermen. There has been multiple documented cases of this happening resulting in capture, injury or fatality. Unlike the land, the seas do not have a chokepoint, hence it is very challenging for the Coast Guard to identify suspicious boats and mobilize resources to reach and capture these boats appropriately.

Focusing on the point of ‘suspicion’, how do security agencies establish the base for suspicion and conversely mobilize operations to effect? AIS (which can be spoofed or turned off)? Gathering radio or cell-based chatter over long distances? Suspicious maneuvering? These are all enormous factors to consider, are highly subjective, and depend heavily on the ‘human element’.

Drug trade is an inherently complex problem to address wherein there are possibilities of informants/insiders of these cartels existing within Bangladesh’s own security agencies. One noteworthy incident was that of Lt. Commander Faizul Islam’s undercover operation to apprehend the leader of the Nayapara Cartel which ended badly as the cartel members, somehow, already knew of when the undercover operatives are to come – bringing about the question on how they knew of a covert intervention. A similar problem can be replicated during operation over waters if informants communicate operational or critical information to supply chain actors.

Therefore, a solution that inhibits timely, automated, and actionable information for high command to deploy resources has become a situational necessity as of today. Perhaps addressing critical paths in the supply chain to disrupt drug trade might merit exploration, i.e. when a trade is going on in international waters and I think SatShipAI is a worthy contender to consider.

What is SatShipAI?

A spin-off of Nodalpoint Systems, SatShipAI uses artificial intelligence along with high-quality satellite imagery to detect and track vessels. They work through data generated from Sentinel-1, an earth observation satellite deployed by the Copernicus Programme of the European Space Agency along with other satellites as per demand. SatShipAI is supported by the Copernicus Accelerator, nVidia Inception Programme, Blue Growth, and EMODnet; and have won the NCI Agency’s Defence Innovation Challenge 2019.

Use-case: Using geospatial AI to assess suspicious ships on the high seas

The image below showcase evidence on a possible interaction between a Burmese fishing trawler with a Bangladeshi fishing boat – a suspicious behaviour.

A suspicious behaviour mapping system using AI can be a possible execution for SatShipAI’s system. Articulating interaction points via spatial imagery between Bangladeshi fishing boats and Burmese fishing trawlers – which are suspected nexuses of the contraband supply chain – is a quick way to categorise suspicious boats and feed this information to security agencies for employing timely and targeted interventions on the seas. Sentinel-2 data acquisition, supported by local Sentinel stations, can produce images in a 10-15 mins intervals – enabling a quasi-real-time monitoring – enough to plot, detect, and track intrusions. This can help identify interaction points e.g. proximity of vessels, pattern of frequenting a specific area on the EEZ and international waters by a subset of vessels, mapping time-intervals of activities, and behavioural pattern of vessel groups moving from/to suspected locations – all of which are actionable information that can be invaluable in the right hands.

What exactly is the problem with existing surveillance systems?

Current surveillance systems and/or mode of operations heavily depend on Automatic Identification System (AIS). While fairly new in implementation in Bangladesh, AIS is the standard monitoring and tracking system throughout the world. AIS works via VHF transceivers on a ship which is used to gather and project information such as unique ID, position, course, and speed. This helps vessel officers and maritime authorities to track and monitor vessels. VesselFinder is a free AIS finder platform to see how responsive the system is; as evidenced by the picture below:

However, it comes with its own set of restrictions. Current AIS models have a range limitation; where AIS-A covers 20 nautical miles and AIS-B covers 10 nautical miles. Not to mention that smaller fishing vessels and many trawlers do not yet have AIS installed, which makes total surveillance impossible. The more concerning aspect of AIS is it being a cooperative system – meaning that vessels can switch off their transponders and become untraceable through the AIS network. Conversely, vessels can also see other vessels which turned on their AIS as per the freely accessible VesselFinder platform which can be problematic. In short, surveillance depending on AIS is proving to be an outdated concept.

What other surveillance systems are available? There are a few. While I can’t speak for other countries, Bangladesh Navy (not Coast Guard) houses two Dornier 228-NG aircrafts equipped with a 360-degree surveillance radar, Telephonics RDR-1700B radar and operator console, HF, VHF/UHF and VHF FM radios, search-and-rescue(SAR) direction finder, 6 observer seats, and 2 bubble windows. However, we are yet to confirm whether these aircrafts are used to patrol for IUU fishing, narcotics and arms trade, human trafficking, and/or piracy which are domains of Bangladesh Coast Guard.

Other intrusion cases where geospatial AI systems can add value

Now this specific example conforms to drugs and arms trade. There are several other intrusion cases in the Bay of Bengal that can be helped with AI-based maritime surveillance systems like the one by SatShipAI such as IUU fishing and marine resource protection – where there has been multiple intrusions from/in Bangladesh reported, piracy and sea-borne robbery – a recurrent case throughout the bay particularly in Cox’s Bazar and the south-western belt of Bangladesh, and human trafficking – where widespread cases of exploitation of vulnerable groups like the Rohingya are recorded.

Isn’t this just another GIS termed as Geospatial AI?

We often come across statements like these. To answer the question, we need to understand what GIS or Geographic Information System is. By Esri’s definition, GIS is a framework for gathering, managing, and analyzing spatial data and organizing these data in layers of information, visualizations using maps, and rendering 3D scenes. This help users to make better decisions based on deeper insights generated from the GIS e.g. patterns, relationships, situations, etc.

Geospatial AI or Geo AI uses the same framework of GIS but enhances it with high-end and automated analytics based on neural networks, able to provide accurate predictions and intelligence when executed correctly. This simple expansion on GIS’s existing capabilities allows for cutting down operational costs and brings about much needed simplicity when consuming spatial information with little human interventions. This can be done through GIS regardless but when processing large datasets involving parallel scenarios, the limitation of GIS comes to light where deep learning and neural network takes the cake. With the enabling infrastructure currently available for use, e.g. Sentinel satellites, solutions like SatShipAI can ideally integrate into almost any situations.

The purpose of this article is to showcase new avenues that security agencies can explore, enabling deep learning to better predict intrusions of nefarious forces and make better decisions. As we are on the cusp of the 4th industrial revolution where AI plays a pivotal role, we would ideally come across game-changing solutions like that of SatShipAI in the near future in an enhanced and contextually appropriate form to solve real-world problems that affect our economy and lives.