Simulation and Analysis of Event Camera Data for Non-resolved Objects

Photon counters and exposure-based cameras cannot simultaneously provide exquisite temporal and spatial resolution in one sensor. Event (i.e. neuromorphic) cameras provide both. Their potential space domain awareness (SDA) applications include short-arc orbit determination and high frequency light curve analysis. Until now, no dynamic event camera simulations for non-resolved objects have been conducted. In this work, we develop numerical and analytical event camera models to better understand potential applicability to SDA efforts. We approximate the event pixel dynamics as a first order differential equation driven by logarithmic photon flux. Photon arrivals are modeled as a stochastic Poisson process. We consider static point sources with constant and time-varying brightness as well as moving sources. We then compare simulated and real event data, showing clear consistencies. For moving sources, we find that the event structure is sensitive to near-instantaneous rates. This sensitivity could provide significant astrometric advancements. We also explore event camera simulations for sinusoidal mean brightness variations, illustrating the complex dependence of frequency estimates on brightness amplitude, oscillation frequency, and event circuit parameters. The presented models improve our fundamental understanding of event data for non-resolved objects. Also, the analytical moving point source model facilitates event-based observability studies and matched filter template generation for object detection, tracking, and characterization. The sensitivity to target parameters observed for the simulated and real event data demonstrates the great potential of event cameras for high spatio-temporal resolution astrometry and photometry.

Automated summary

This article introduces the potential applications of event cameras in the field of space domain awareness, including orbit determination and light curve analysis. By developing numerical and analytical event camera models, the understanding of event data is improved, and the potential uses on non-resolved objects are demonstrated.