First Real-Time System for Detecting Ships in Public Optical Imagery Globally

Analysts and conservationists have long aspired to locate unreported vessels from space using optical imagery. The ability to discern details in color provides the contextual insights necessary to advance how we protect our ocean. Now, that aspiration has become a tangible reality through Skylight’s new optical imagery detection system.

Over the last year, our research and conservation teams have collaborated on building a highly specialized computer vision model to pinpoint vessels in data collected by the European Space Agency’s (ESA) Sentinel-2 satellite. Freely available to governments and NGOs worldwide, Sentinel-2 data empowers the fight against illegal fishing and other maritime crimes. It offers a near real-time view of ships crossing oceans at a remarkable 10-meter resolution, striking a valuable balance between capturing a vast swathe of the ocean and revealing crucial details about vessels on its surface. Since its release, vessel detections from Sentinel-2 have become one of Skylight's most popular data sources – setting a new standard in maritime surveillance that enables swift analysis and response.

“Using AI-identified Sentinel-2 data vessel detections from Skylight, our team can routinely monitor the United Kingdom’s overseas territories for dark vessels,” said Sebastian Jennings, a Marine Management Organisation (MMO) senior data officer “In many remote offshore overseas territories, vessels of interest are typically 20-60 meter fishing vessels, which can be well defined in this new vessel detection tool.”

While there have been many previous efforts to identify vessels in Sentinel-2 imagery, no tool has been regularly processing it globally and with a focus on real-time operational needs. 

“Today, the average delay from when Sentinel-2 imagery is captured to when it is available in Skylight’s database is five hours,” shares Skylight software engineer Linh Tran. “We also detect thousands of “dark” vessels – those without AIS – daily.”

The expanded coverage afforded by the addition of Sentinel-2 to Sentinel-1 data is also particularly notable, with substantial improvements for regions like the Pacific Islands. With its vast expanse and unique challenges, the Pacific Islands stand to benefit immensely from Skylight's comprehensive surveillance capabilities, ensuring enhanced protection of its marine resources and ecosystems. Furthermore, regarding supporting the UN High Seas Treaty and the 30x30 efforts, the information these data provide is crucial. These models enable precise and timely monitoring of IUU fishing even in the most remote areas, significantly enhancing enforcement efforts. 

A key strategy in making these data accessible to those who need it the most is through Skylight’s Application Programming Interface (API). APIs enable Skylight’s data and efforts to be integrated into various platforms, making combining intelligence from multiple sources easier while reducing the hassle of managing numerous screens. Several technologies that serve government agencies globally already utilize data from Skylight’s API. One of those technology partners is SeaVision – a platform developed and provided by the U.S. government to hundreds of government agencies globally. For the last several years, SeaVision has employed Sentinel-2 detections through Skylight’s API, on top of Sentinel-1 and Night Lights data.

“Skylight’s vessel detection data from ESA Sentinel-2 imagery has been a game-changer,” said Charlie Brown, Maritime Advisor. “It is a significant time-saver to have vessel detections automatically flagged for investigation. Also important to note is that these tools do not replace analysts but help enhance their skills.”

Brown uses Skylight’s new Sentinel-2 vessel detections to look for potential illicit activity, such as ship-to-ship transfers between oil tankers. This information helps him identify vessels that might be trading sanctioned oil. Additionally, Skylight allows him to monitor large groups of fishing fleets that operate and anchor in various regular spots at certain times that may suddenly move out at other times.

“Access to Sentinel-2 imagery and the assistance of vessel detections, along with additional contextual data from Skylight, including VIIRS, AIS, and other sources, is necessary for maintaining awareness of activities at sea,” adds Brown.

Skylight’s artificial intelligence model for detecting vessels in Sentinel-2 is the best-in-class in computer vision. It deals with many types of unique artifacts and attributes of these data. The Skylight system links each detection to the corresponding raw image in Sentinel Hub, so analysts can always confirm the AI model's work for themselves. 

As for what’s next, we plan to introduce vessel detections from the U.S. Geological Survey Landsat satellites later this year. The model is openly available for others to provide feedback and deploy in their own systems. This is part of AI2 and Skylight’s broader commitments to foster openness and accessibility to AI models, as well as promote collaboration in the maritime surveillance and marine conservation community. If you are interested in accessing this data or collaborating with Skylight on future efforts, please don’t hesitate to contact us.

‍