COLLEGE PARK, Md. — September 24, 2026— IonQ, the world's leading quantum full-stack platform and foundry, released research demonstrating how quantum generative models can improve change detection on complex high-resolution Synthetic Aperture Radar (SAR) and interferomic SAR (InSAR) data. Using real satellite sensor data, IonQ researchers compared the quantum approach with classical change-detection methods and tested it on IonQ hardware. The ability to detect ground changes from space has many applications including disaster response, infrastructure monitoring, defense, and land-use enforcement.

"Satellites are exceptional at collecting imagery of the Earth. The value is in knowing what changed and whether it matters," said Jordan Shapiro, President, Quantum Platform at IonQ. "This research shows quantum generative models can take on a statistical challenge that becomes harder as radar imagery gets more granular. It's also a good example of what happens when our space and quantum teams work on the same problem."

SAR sees through clouds, fog, and darkness, which is why defense agencies, disaster responders, and infrastructure operators depend on it. Change detection analytics usually compare an estimate of what a scene should look like if nothing had changed with an actual image of the scene to determine the areas of change. With high-resolution SAR data, the available pixel statistics can become complex, making classical change detection less reliable. Conventional approaches often use additional preprocessing to make SAR easier to analyze, but at the cost of reducing fine spatial detail.

To overcome this challenge, IonQ researchers tested a Quantum Circuit Born Machine (QCBM), a type of quantum generative model, to estimate what each scene's background should look like. The QCBM learns the joint statistical relationship between the before and after images. It then generates reference samples that help estimate the expected background and distinguish meaningful change from normal variation.

The team ran the quantum model on IonQ’s own high-resolution SAR datasets. On challenging non-Gaussian SAR data from an airfield, a simulated quantum model scored higher on filtered F1 than both classical baselines tested. On a challenging non-Gaussian SAR test case from Marine Corps Air Station Miramar, the QCBM achieved a filtered F1 score of 0.41, compared with 0.24 and 0.16 for the two classical baselines evaluated. When preprocessing made the underlying pixel distributions approximately Gaussian, the performance gap largely disappeared. The results indicate that the quantum model's strongest relative performance occurred where conventional statistics were sparsest. The advantage remained when the simulation model was run on IonQ quantum hardware.

InSAR adds even more complexity to the analysis task. This is a technique that uses phase differences across SAR images to measure surface deformation at millimeter-to-centimeter scale. In a separate InSAR experiment involving volcanic lava flows, the quantum and classical approaches achieved comparable peak performance. The results suggest quantum generative models may be especially useful when high-resolution sensing data produces sparse or difficult-to-model statistics. IonQ is building on these results to advance analytics in defense, critical infrastructure, and environmental monitoring.