Signal in the Shadows

Digitally Reconstructing the Yu Son

Om Gothi, James Byrne
14 July 2026

How do you tell how deep a vessel is in the water from satellite imagery? We asked ourselves this question a year ago and set out to solve it.

In November 2025, we published an investigation into the clandestine payment channels used to facilitate the movement of Russian oil to North Korea in violation of UN Security Council Resolutions.

Drawing on financial and corporate records, AIS data, high-resolution satellite imagery and three-dimensional reconstructions, this investigation traced the movement of oil tankers and the flow of funds through sanctioned Russian banks to the Vostochny oil terminal in the country’s far east.

As part of this investigation, we sought to determine, using advanced geospatial techniques, whether the vessels visiting Russian and North Korean ports were laden. As a vessel takes on cargo, its hull sits lower in the water, something often visible in satellite imagery. However, to increase the precision of these assessments, we developed a hybrid technique relying on high-resolution imagery, the construction of a high-fidelity digital twin and trigonometry. This methodology enabled us to determine, with a high-degree of confidence, when the vessels were heavily laden and therefore likely loading oil for delivery to North Korea.

The vessel chosen was the DPRK-flagged Yu Son (IMO 8691702), an oil products tanker built by Zhe Jiang Shun Hang Ship Manufacture Co Ltd (浙江顺航船舶制造有限公司) in 2003. As one of North Korea’s most active oil tankers, the Yu Son was designated by the UN Security Council in March 2018, following Pyongyang’s nuclear and ballistic missile tests in 2016 and 2017, respectively.

This piece explains the process by which this digital reconstruction was achieved, and the steps we took to make these determinations.

the control room of a ship with multiple monitors

Building a Digital Twin

Cargo ships at a bustling port under cloudy skies, cranes ready for loading.

Verifying a vessel’s laden state -- calculating precisely how far the hull is submerged -- remains a core challenge for maritime analysts. A vessel sitting low in the water shows a vessel is laden. For an oil tanker, this indicates the ship is loaded with petroleum.

Because a top-down view cannot always account for a ship’s draft, traditional two-dimensional satellite imagery often fails to provide a complete profile. To resolve this issue, we built a detailed Digital Twin Reconstruction (DTR). This methodology allowed us to derive the exact displacement of the Yu Son, providing the necessary evidence of sanction-breaking ship-to-shore transfers.

In cases where imagery and video are readily available, the DTR process can use techniques such as Gaussian splatting and traditional photogrammetry, a process that requires hundreds of high-quality angles. However, in the absence of these resources, we first used a process called Multiple Camera Projection (MCP).

_{Source: Open Source Centre}

Multiple Camera Projection

A professional analyzing data on multiple monitors in a dark room, highlighting cybersecurity themes.

The MCP process is a rudimentary form of photogrammetry, itself the science of extracting information about physical objects from images. As such, we used 3D software to manually align high-resolution satellite images and other ground-level images from a variety of angles.

These images were projected from multiple virtual cameras onto a single 3D ground plane containing the ship. Using this technique, the images were aligned with the cameras to show a complete 360-degree view of the vessel.

_{Source: Open Source Centre}

_{Source: Open Source Centre}

Once this process was complete, we applied a technique known as rotoscoping. Originally developed in the early 20th century and now widely used in animation, film and visual effects, rotoscoping involves tracing or aligning visual forms against reference imagery. Using this approach, we progressively matched wireframe views of the 3D model to visible silhouettes in satellite imagery and other reference images.

We then isolated and identified physical features of the Yu Son, including oil piping and transfer valves, lifeboats, gangways and railings, communications equipment and a wide range of other micro and macro objects. Lacking a large library of ground-level images of the vessel, we collected images of the Yu Son’s sister vessels, which include other ships built by the same shipyard and oil tankers of the same specification. These images were used to recreate micro-level details such as firefighting equipment, mooring winches and fuel piping.

At the end of this process, the digital twin of the Yu Son contained 32 million polygons. This polygon density is equivalent to a film-grade asset used across major Hollywood movies and productions.

Measuring the Shadows

Once this process was completed, the digital twin of the Yu Son could be placed within a veridical model of the real world captured by high-resolution imagery. While traditionally displayed digitally or on paper, satellite imagery files contain a wealth of georeferenced data captured by the sensor. These data packages contain details about the azimuth of the satellite, sun elevation, the precise time and date of capture, and a range of other metadata that are often used in geospatial intelligence.

The content of the image and this data can therefore be used to measure objects and shadows using simple trigonometry. Using ESRI’s ArcGIS platform, we measured the shadow cast by the Yu Son at its highest point, then using a simple formula to calculate the height of the visible portion of the ship, which includes the freeboard and superstructure. It is calculated as:

H = Shadow Length x tan(Sun Elevation°)

Using an image collected by Airbus Defence and Space on 17 October 2024 with the company’s Pléiades Neo sensor - one of the highest resolution commercial satellites in the world - the calculated visible height of the Yu Son was 11.3 metres, based on a 37 degree sun elevation and a 15 metre long shadow.

Keel-to-Waterline Calculation

While this process can be used to accurately measure the height of the vessel above the water, the digital twin is critical for making determinations of the underwater section of the vessel.

As such, we superimposed the digital twin of the Yu Son with the georeferenced satellite image. The 3D model is then submerged into a virtual water plane until the bridge-to-waterline distance matches the calculated height derived from the shadow measurement process.

Because the model includes the hull, not visible in satellite imagery, the OSC was able to measure the distance from the virtual waterline down to the keel. By comparing this draft measurement against the vessel’s lightship draft and load line -- when the vessel is empty and at maximum capacity, respectively -- this analysis can be used to determine the exact cargo volume of a vessel.

Gaussian Splatting

Once the modelling process was complete, we then also used a Gaussian splatting process to convert the Yu Son’s model into a fully interactive 3D scene.

Gaussian splatting is a graphic display discipline where the three-dimensional model is converted into millions of ellipses which can accurately capture the colour, specular and transparency values of the original mesh, essentially enabling the conversion of the high-fidelity model into a fully interactive, photorealistic scene. The outcome of this process can be viewed here:

GAUSSIAN SPLAT INTERACTIVE MODEL

From reconstructing a digital twin in an informationally opaque environment, the OSC has transformed static data into a photorealistic, interactive scene. This transition from two-dimensional imagery to immersive three-dimensional scenes represents a fundamental leap in the capabilities of geospatial intelligence.

Beyond the Frame

The reconstruction of the Yu Son drew on established geospatial methods, but also on techniques more commonly associated with fields such as 3D modelling, animation and visual effects. Bringing these disciplines together allowed us to approach the source material differently, extending both the analytical process and the ways in which its findings could be communicated.

For us, this is an important direction for open source intelligence. New methods create new ways of exploring, testing and engaging with source material, both as practitioners and as consumers of intelligence.

At OSC, we believe the boundaries of our tradecraft should be continuously tested and refreshed. That means looking beyond traditional intelligence disciplines, borrowing techniques from elsewhere and building new ways to analyse and communicate complex subjects. The Yu Son reconstruction is one example of that approach. For us, the next frontier of open source intelligence will be shaped not only by what we can find, but by how far we are willing to rethink the ways we see, analyse and tell the story of it.

POWERED BY

Disclaimer
This document has been prepared by OSC for informational purposes only (the ‘Permitted Purpose’). While all reasonable care has been taken by OSC to ensure the accuracy of material in this report (the ‘Information’), it has been obtained primarily from open sources and OSC makes no representations or warranties of any kind with respect to the Information. You should not use, reproduce or rely on the Information for any purpose other than the Permitted Purpose. Any reliance you place on the Information is strictly at your own risk. If you intend to use the Information for any other purpose (including, without limitation, to commence legal proceedings, take steps or decline to take steps or otherwise deal with any named person or entity), you must first undertake and rely on your own independent research to verify the Information. To the fullest extent permitted by law, OSC shall not be liable for any loss or damage of any nature whether foreseeable or unforeseeable (including, without limitation, in defamation) arising from or in connection with the reproduction, reliance on or use of any of the Information by you or any third party. References to OSC include its directors and employees. For this report, the authors have processed company, entity and individual names recorded in Russian and Chinese. In some instances, names of companies, entities and individuals have had to be translated or transliterated. Every effort has been made to ensure accuracy in translation/ transliteration, and the authors do not accept liability for any unintentional errors made in this regard.

Identification Of Individuals, Companies And Governments In This Report
The purpose of this report is to understand and explain how Russia is circumventing sanctions to continue to fund its ongoing war in Ukraine. To achieve this purpose, it identifies a number of individuals/companies/governments who are believed to be involved in operations to this end. For the avoidance of doubt, OSC does not impute any allegations of wrongdoing on the part of these individuals/companies/governments and makes no representations or assertion that these individuals/companies/governments have any involvement in any sanctions evasion-related activity or are involved in directly or indirectly facilitating sanctions breaches, supplying the Russian defence industry, Russian military and/or Russian military customers in breach of any international (or their own domestic) laws or regulations restricting or prohibiting such action, unless expressly stated in the report.