Vision-based 3D motion estimation for on-orbit proximity satellite tracking and navigation

Abstract

The main challenge addressed in this work is to develop and validate an algorithm able to track and estimate the relative position and motion of on-orbit, un-modeled targets by using only passive vision. The algorithm developed is based on well-known image processing techniques. To achieve this goal, a number of different approaches were analyzed and compared to assess their performance for a satisfactory design. The code also has a modular general structure in order to be more flexible to changes during the implementation until best performance is reached. Artificially rendered high quality, animated videos of satellites in space and real footage provided by NASA have been used as a benchmark for the calibration and test of the main algorithm modules. The final purpose of this work is the validation of the algorithm through a hardware-in-the-loop ground experiment campaign. The development of the Floating Spacecraft Simulation Test-bed used in this work for the validation of the algorithm on real-time acquisition images was also documented in this thesis. The test-bed provides space-like illumination, stereovision and simulated weightlessness frictionless conditions. Insight on the validity of this approach, describing the performance demonstrated by the experiments, the limits of the algorithm and the main advantages and challenges related to possible future implementations in space applications, were provided by this research.