The camera that does not shoot frames: a technology designed to catch UAP without missing the moment
Sandia researchers propose event-based sensors in which every pixel reports only when local light changes. The result is very fast response, little redundant data and motion signatures that may distinguish birds, bugs, planes, satellites and drones.
The paper appeared in the Spring 2026 Naval Postgraduate School issue and was written by Kaylin Hagopian and Dr Joshua Shank of Sandia National Laboratories. It reports typical 100-microsecond response and real-world data volumes of 1 to 5 percent of a framing camera.
Processing algorithms are less mature than conventional-camera algorithms. The work presents a promising detection and classification method, not proof of an exotic UAP and not a replacement for radar, telemetry and additional sensors.
Record change, not an entire empty sky
A conventional camera records complete frames at a fixed rate even when most pixels contain unchanging sky. In an event-based sensor each pixel acts independently and reports only brightening or dimming, concentrating bandwidth on motion and change.
One hundred microseconds and a fraction of the data
The researchers report a typical response time near 100 microseconds. In real-world Sandia scenes, the sensors usually generated only 1 to 5 percent of the data volume produced by framing cameras. That combination can support persistent wide-area monitoring without streaming hundreds of megabytes each second.
Birds change speed and direction fluidly, nearby bugs may make wave-like patterns, aircraft lights flash periodically and satellites move relatively straight without the same beacon rhythm. Stars also appear through atmospheric scintillation, but their slow movement becomes distinguishable over time.
The WATCHER processing pipeline groups events, filters noise and first separates periodic from non-periodic signals. It then examines motion, flashing, shape and detail. Every cluster begins unidentified, and the objective is to remove familiar possibilities methodically rather than assume anomaly.
Noisy pixels, lens choice, focus, lighting and sensor settings can still mislead. The authors say the algorithms remain immature. A credible system should combine event sensing with conventional imagery, radar, accurate time and position so that an unusual event is preserved as a data package rather than a lone point of light.