The stained part: new software will help to find the remains of fuel oil in the Black Sea

Russian scientists have developed a simple and effective algorithm for searching for oil product stains using drones. The AI is currently analyzing the photo and video data. However, it needs a lot of power to work, so you have to send snapshots to remote servers. And this requires time and financial resources. The new development performs all calculations directly on board the aircraft without human intervention. Read more about the innovation in the Izvestia article.

Oil slick search software

Specialists from St. Petersburg State University of Aerospace Instrumentation (GUAP) have developed software for automated search of oil product spills on the surface of the seas and oceans using UAVs. The software is installed on the computing power of the aircraft and analyzes data from its cameras, and then transmits frames with detected anomalies to the monitoring post. The technology makes it possible to distinguish oil pollution from accumulations of natural debris or algae with 98% accuracy.

— The development can be used in environmental monitoring systems on board aircraft, in particular for remote sensing of the seas and oceans. The method helps to detect anomalies in the form of oil product spills in photographs. The resulting image is divided into sequentially intersecting frames. The next frame occupies at least 75% of the area of the previous frame. After the algorithm has worked, namely calculating the statistical parameters of the histograms of the brightness distribution during image processing, a decision is made on the absence or presence of a spill of petroleum products. In this way, it is possible to distinguish a spill of petroleum products from algae or the wake of a ship," said the author of the project, senior lecturer at GUAP Dina Vasilyeva.

One of the largest oil product spills in the Black Sea occurred in December 2024 as a result of the collapse of two tankers due to a storm. About 3-4 thousand tons of fuel oil got into the water. The coast in the Anapa region was hardest hit. The elimination of the consequences took several months. On September 19, 2025, the authorities announced that they had completely cleaned the bottom in the area.

On black smoke: residents of Adygea complain massively about producers of burnt coal

Similar problems are being fixed in at least five regions — how artisanal industries pollute the air

However, as Igor Shkradyuk, coordinator of the industrial greening program at the Wildlife Protection Center, who participated in the elimination of the consequences of the accident, explained to Izvestia, the storm season is now beginning in the Black Sea. Strong unrest can lead to the surfacing of new portions of petroleum products from the tanker, which is still at the bottom. Smaller hydrocarbon leaks occur regularly. The last one was on August 29 in Novorossiysk. Therefore, the task of searching for pollutants on large areas of the sea surface remains relevant.

Most of the programs currently used to detect oil spills use artificial intelligence methods. Powerful computers are needed for their operation, so images collected from the drone have to be sent to remote servers for analysis, which is expensive and time-consuming.

Significant computing resources are needed to apply AI deep learning methods. At the same time, neural networks are focused on working with specific images, while the consequences of oil product spills are very diverse, so the required computing resource can increase by two orders of magnitude. There are also limitations on the amount of data processed.

Such systems mainly use low-resolution frames (227x227 pixels), which reduces their accuracy. Neural network training requires a large set of images and databases, which are practically non-existent in Russia. And in order for the AI to identify the spot, it must occupy at least 50% of the frame.

Detection of light fractions

In contrast, the new algorithm can find spots that take up from 7% of the frame. That is, it detects the most minor contamination. Data processing on board a drone saves money and time. The software works with higher quality images (640x480 pixels). This allows it to detect spills with even greater certainty.

As Elena Shuvalova, an ecologist who participated in the liquidation of the accident in the Black Sea, told Izvestia, the development of GUAP can be used to search for light fractions of petroleum products that weigh less than water, float in and form spots.

— This is a necessary development. Only light fractions of fuel oil can be found on the surface. The ability to quickly find these spots will be very useful, because there are already ways to deal with light fractions of hydrocarbons. Booms and sorbents are used for this purpose. They have learned how to work with them both in the world and in our country. Timely destruction of these spots in the sea, before they come ashore, eliminates a lot of problems," she said.

Military reactivity: Russia records an increase in the militarization of the Black Sea due to NATO and the EU

What is the West trying to achieve and how can a direct clash be avoided?

Such systems need to be tested and compared with others. The most important thing for users is their efficiency, cost-effectiveness and ease of use, the specialist emphasized.

As Denis Krivoguz, a senior researcher at SFU, noted, the technology looks promising, but its capabilities can only be assessed with practical application.