Vision-based Pose Estimation for Space Exploration

Abstract

Autonomous navigation lies at the heart of space exploration, empowering spacecraft to traverse uncharted realms, conduct scientific investigations, and execute intricate manoeuvres in close proximity with precision. Central to autonomous navigation is pose estimation,which revolves around the precise determination of an object's spatial position and orientation based on the analysis of captured images. In this particular research context, the emphasis is on pose estimation executed from an external vantage point, primarily involving imagery in the context of a two-spacecraft scenario. Traditional methods of pose estimation have long relied on geometric and analytical techniques, which an be susceptible to lighting conditions or poor images. By creating a system architecture that looks at a time sequence of images to better understand more than a single image, it is possible to overcome these limitations. This thesis introduces a novel approach to pose estimation, with a specific focus on the dynamic application for non-cooperative spacecraft.