Synthetic optical tracking of objects in Low Earth Orbit in real-time

Kurzfassung

Since the first artificial satellite, Sputnik 1, was launched to Earth Orbit in 1957 the amount of non-functional artificial satellites in Orbit has been increasing. Current estimates speak of around 700,000 objects with a diameter larger than 1 cm, of which only a small fraction (<20,000) are actively tracked, and most remain unknown. These objects, orbiting Earth several times a day, have the potential to impact each other and functioning satellites at speeds up to 16km/s, creating more debris, which poses a significant threat to Orbit infrastructure. Therefore the German Aerospace Center (DLR), among others, make the effort to find and track more debris by employing passive optical methods. Such methods have been used for decades by astronomers to detect the movement of celestial object. This is traditionally performed by taking single long exposure images where moving objects register as streaks. These streaking methods are very refined and exploit many image processing techniques and according hardware, but they are not without flaws. Their sensitivity limits to object brightness, movement estimation inaccuracy and direction ambiguity, as well as their high processing demand are disadvantages that are hard to overcome. Consequentially, radar telescope based methods are still the dominant but expensive way to detect and track objects in Earth Orbit. The recently developed Synthetic Tracking approach shows promise to excel streaking in performance, accuracy, optical hardware economy and real-time capability, and fuels hopes to develop a passive optical alternative similarly potent as the established radar tracking methods. The goal of this thesis is to develop an algorithm on the basis of Synthetic Tracking that is capable of detecting and tracking moving objects in data obtained by a 'staring camera'. This is a camera system that observes a fixed part of the sky and provides up to 100 images per second. A simulator is designed to serve as configurable, reliable input data basis. The algorithm is implemented to run on graphics processors and exploit their parallel computation capabilities to achieve high performance. The algorithm is evaluated with respect to detection quality and real-time capability.