Brian McReynolds, U.S. Air Force Academy; Rachel Oliver, Air Force Institute of Technology; Peter McMahon-Crabtree, AFRL Space Vehicles Directorate; Zachry Theis, AFRL Space Vehicles Directorate; Tobi Delbruck, Institute of Neuroinformatics, UZH/ETH Zurich
Keywords: Event-based, Neuromorphic, SDA, Photometric reconstruction
Abstract:
Neuromorphic Event-based Vision Sensors (EVS) have demonstrated unique capabilities in SDA applications by providing excellent temporal resolution with low power and sparse data requirements. EVS achieve these distinct advantages by reporting pixel level change events in response to localized brightness changes, and as a result are well suited for detecting resident space objects (RSOs) with relative motion against the dark sky background. Due to their sparse output, EVS are useful for detecting and localizing RSOs, but have not demonstrated the ability to further characterize celestial objects due to the fact that they ostensibly offer no absolute photometric information. The ability to extract photometric information from the event stream would greatly expand potential use cases of EVS for SDA.
Recently, a few groups have demonstrated the ability to map EVS output to object brightness with limited success, but the estimation error increases dramatically for dimmer objects. Using a custom point source RSO simulator to simulate sub-pixel, point source objects moving against a dark background and recording with a 4th generation Sony IMX636 sensor, we show that EVS response can be reliably mapped to object brightness. We first develop a theoretical model from first principles using both the logarithmic compression and the illumination dependent, finite temporal response bandwidth of pixel’s photoreceptor. Incorporating the pixel’s temporal response facilitates accurate photometric reconstruction even at the lowest detectable light levels. Ultimately, this new ability to extract photometric information from an EVS event stream greatly expands potential use cases of EVS for SDA, providing a future pathway for data efficient photometric, spectral, and polarimetric RSO characterization with high temporal resolution.
Date of Conference: September 16-19, 2025
Track: SDA Systems & Instrumentation