Event-Based Vision for Space Navigation

Industrial Problems Seminar

Connor Hashemi
Kitware, Inc.

Abstract

Star trackers determine the precise orientation of satellites and spacecraft by mapping and tracking stars in the night sky, providing attitude estimates with arcsecond-level accuracy. This makes them critical for space navigation, but conventional star trackers tend to be bulky, power-intensive, and limited in update rate, all which conflict with the growing demand for small, inexpensive satellites in low-Earth orbit (LEO). Event-based sensors (EBS), also known as neuromorphic sensors, represent a paradigm shift in imaging: instead of recording full image frames, each pixel independently reports brightness change “events” with microsecond precision. This enables continuous-time sensing with high dynamic range and low power consumption—all which make it an ideal fit for star tracking. However, conventional tracking algorithms designed for synchronous frame-based data cannot be directly applied to these asynchronous measurements. In this talk, I will present our recent advances in using EBS for star tracking, including a novel asynchronous tracking algorithm and a high-precision centroiding method. I will conclude with a discussion on future work and the broader implications of this technology for space-domain awareness (SDA).