Researchers at the Alan Turing Institute’s DARe centre and the University of Birmingham have built an AI model that identifies parts of satellites in orbital radar images. The work could help track what is happening to objects in increasingly busy orbits, though it remains research, not an operational service.
Watch Desk analysis
What happened
The model labels components including solar panels, antennas, thrusters and robotic arms in high-resolution radar images. Resultsense says the team presented the work at the European Radar Conference on 9 October.
The system processes images in sequence, carrying information about a component from earlier frames so it can keep tracking it when a satellite turns and the part disappears from view. The images come from a sub-terahertz inverse synthetic aperture radar developed by Professor Marina Gashinova’s group at Birmingham. Resultsense’s report says the radar’s resolution makes component-level identification possible.
Why it matters
Knowing that an object is in orbit is not the same as knowing what it is made of or how it is configured. Identifying parts could support space-domain awareness, including work on satellite rendezvous, servicing and debris removal. Resultsense reports that the researchers see growing demand for monitoring satellites and debris as orbital traffic increases.
The specific development here is an AI system interpreting radar imagery to identify and track satellite components. That makes this a computing and space-monitoring story, not simply a general warning about crowded orbits. The report does not give accuracy figures or say when the system might be used operationally, so claims about practical performance or deployment would be premature.
Our read
Component-level tracking is a useful step towards understanding what is actually moving overhead, rather than treating every satellite as a featureless dot. The clever part is keeping track of a component across changing views; the next test is whether that works reliably on real-world imagery, beyond the research demonstration. Resultsense also places the work in the context of UK sovereign capability, but that is an aim, not proof of an operational advantage already delivered.
What to watch
- Whether the researchers publish accuracy results or further details of the system’s evaluation.
- How the model performs as satellites turn, components become obscured, or imagery varies.
- Whether the work progresses towards operational use for satellite monitoring, servicing or debris removal.
Discussion spark: Should investment in space-domain awareness prioritise identifying what satellites are made of, or tracking their movements and detecting debris?
Sources and evidence
- Turing model maps satellite parts from orbital radar images – Resultsense (9 October 2026, 08:10 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.