Digital tools give better picture for fishers of quantity and species of catch

Researchers are combining AI, cameras and DNA traces in seawater to build a transparent picture of what is being caught, and support more sustainable fisheries.

Since 2024, Torreele has led OptiFish, an EU-funded initiative developing technology that can automatically recognise fish species and improve how catches are monitored.

The four-year collaboration involves researchers from eight countries working to improve fisheries monitoring and management across Europe. They combine AI, computer vision, electronic monitoring, genetic analysis and robotic systems to build an integrated digital monitoring system.

“The better your data is, the clearer your story is and the more complete your picture becomes of what is happening in your marine environment,” Torreele said. 

The timing matters. The EU introduced rules to close enforcement gaps that let illegal and unreported fishing go undetected, on the logic that quotas only protect fish stocks if they are being followed. Under the revised Fisheries Control Regulation, electronic reporting is already mandatory for most vessels, with tracking extending to smaller boats by 2028 and digital traceability rules reaching processed fish by 2029.

But digital reporting alone will not close the information gap. Without detailed data on which species are being caught, where and in what quantities, fisheries managers cannot make sound decisions about sustainable catch limits, trace seafood through the supply chain, or help fishers plan fishing trips.

“Everything starts with your data,” said Sander Delacauw, a marine biologist at ILVO. “If you want to manage sustainable fisheries, you need to make sure that all the data you collect is high quality and sufficient for stock assessments that feed into quotas and management decisions.” https://projects.research-and-innovation.ec.europa.eu/en/horizon-magazine/casting-wider-net-digital-tools-bring-europes-fisheries-sharper-focus

“But the problem is that a lot of sampling is currently manual, and you cannot send observers to sea 24/7,” he added. “When cameras are used, people often need to watch footage manually, which takes enormous amounts of time.”

OptiFish is tackling this by developing a suite of technologies, such as cameras paired with AI algorithms that can recognise and count fish species.

“We start by collecting a sufficient amount of training data,” Delacauw said. “It’s like teaching a child to recognise something – you need to show them enough pictures, so they actually understand.”

The team is turning this training material into high-quality image datasets that other researchers can use to teach their own AI models.

Across Europe, these AI camera systems are being tested on pumping vessels, sorting tables, beam trawlers and small-scale fishing operations. The system is trained to identify species, even when fish overlap or are partially hidden, using quality checks that flag anomalies and help assess the reliability of the result.

One early result comes from OptiFish partner DTU Aqua, which built an AI pipeline to track individual fish as they move along conveyor belts under electronic monitoring, so the same fish isn’t mistakenly counted twice. Tested on six similar-looking species, it correctly matched fish 90.43% of the time

The team is also developing a robotic arm that separates overlapping fish on sorting belts and tables. This will improve image quality for the cameras.

In the sea, fish shed DNA particles as they swim, and the project uses this environmental DNA (eDNA) to sample the water and estimate the biomass of a particular stock.

“It’s like humans shedding hair,” Delacauw said. “By collecting water samples or deploying special probes attached to fishing nets, we can identify which species are present in the water.”

GPS sensors, water current monitors and fuel consumption data add further context, such as location and sea conditions.

Delacauw explained the logic of combining several technologies.

“Manual sampling gives very high-quality data on individual fish, but only occasionally. Cameras can measure 24/7, but are limited to what’s visible. eDNA can indicate what species are present in an area. By bringing all these methods together, we take the best of each approach to create an overall picture.”

OptiFish is now testing and refining these systems across five pilot sites spanning different fisheries and European regions, from the North Sea and the Mediterranean to the Black Sea.

For the full article, please click here.