Nicole Gagnier, The Boeing Company; Trent Kyono, The Boeing Company; Jacob Lucas, The Boeing Company; Michael Werth, The Boeing Company; Justin Fletcher, SMC/DirSP-G; Ian McQuaid, Air Force Maui Optics and Supercomputing Site; Michael Brannon, Air Force Research Laboratory
Keywords: machine learning, data generator, imaging, LEO
Abstract:
The numerous applications of machine learning for space domain awareness (SDA) have underscored the need for robust datasets of space objects, particularly Low Earth Orbit (LEO) satellites. In this paper, we present the Scored Images of LEO Objects Generator (SILO-G), a lightweight data generator that uses satellite models and wave-optics simulations to create quality scored images of LEO satellites as if observed from a ground-based optical observatory with varied turbulence conditions. In addition to the quality scores, SILO-G includes metadata such as segmentation maps, feature locations, and pose to increase utility across the SDA field. SILO-G utilizes a GPU-based wave optics simulation to greatly increase wave optics simulation speed relative to CPU-bound approaches. The data generator framework of SILO-G enables a lightweight implementation in which the user can quickly edit parameters and generate large scale datasets without large storage or processing overhead. Finally, we discuss some of the many novel applications that the SILO-G dataset enables.
Date of Conference: September 15-18, 2020
Track: Machine Learning for SSA Applications