Abstract
Because of the increased number of unmanned space missions, and a growing complexity in these missions, the ability to manoeuvre around nearby objects is becoming increasingly crucial. These manoeuvres can include avoidance, observation or docking but each are dependent upon the initial approach. Such capabilities exist, but are inadequate in the general scenario, when the rendezvous with a non-cooperative target must be autonomous and restricted to using passive imagery. A critical issue to address is the ability to estimate the relative position and attitude (or pose) of the target object in an appropriately short time, which is dictated by the mission environment. This research introduces a novel method of obtaining a fast estimate of a target’s pose, by describing it with a degenerate ellipsoid, or spheroid. From the captured images, the bounding ellipse of the projected target is used to reconstruct the descriptive spheroid and, by extension, the target itself. This concept is then applied both to satellite and rocket imagery using simulated images, where the pose of the target is determined up to the spheroid’s axis of symmetry. The focus of this research has been the derivation of the spheroid’s projection onto an image plane - and sphere - and through inversion, the analytic reconstruction of that spheroid. Additionally, algorithms for incorporating spheroid modelling with range information, stereo cameras and feature detection methods are investigated, yielding novel approaches to target identification and tracking. The errors incurred from approximating the target as a spheroid are also evaluated, imparting a method of determining the descriptive spheroid - the spheroid that incurs the least error upon reconstruction. Finally, an accuracy comparison between a global target description and a feature-detection method is presented, which quantifies which approach is more suitable in differing circumstances. The fundamental objective of this PhD is to enable the capability of fast, autonomous pose estimation of a known passive target using spheroid modelling.