Discovering 3-D Structure of LEO Obects

Jacob Lucas, The Boeing Company; Trent Kyono, The Boeing Company; Julia Yang, The Boeing Company; Justin Fletcher, Odyssey Systems Consulting

Keywords: Machine Learning, LEO, resolved imagery

Abstract:

A 3-D object model provides invaluable information for space domain awareness.  However, in practice, there are countless reasons why a 3-D object model may not exist or is unattainable.  In this work, we aim to investigate the discoverability of 3-D object structure from passive observations of objects.  Specifically, we aim to investigate the feasibility of using Neural Networks via Neural Radiance Fields for uncovering 3-D object models.  Experimentally, we demonstrate feasibility on a simulated set of satellite images, where we can compare object structure to ground-truth.

Date of Conference: September 14-17, 2021

Track: Machine Learning for SSA Applications

View Paper