Automated Techniques for Extracting Four Dimensional Space-Time Data from Ground-Based Imagery of Space Objects

This project will develop techniques for processing sequences of images of a space object, taken from ground-based telescopes, to create a three dimensional model of the object and to calculate the object's attitude motion as a function time. This Shape from Motion problem has been studied since the 1980's, but the approaches have not been subject to rigorous performance testing, and the algorithms typically assume much higher image clarity and spatial resolution than ground-based space surveillance systems can deliver. First, we will create a prototype system that applies the most promising of the current shape from motion techniques, and evaluate its strengths, limitations, and feasibility using simulated and real imagery of space objects. This system will be delivered as a working tool for research and special mission analyses. Second, we will examine novel approaches to the shape from motion problem that are tailored to the characteristics of the space surveillance problem. One approach integrates feature identification with feature tracking to efficiently accommodate noisy data and features that move in and out of view. Another replaces the point cloud representation of object shape with a meshed surface or a set of voxels, making results more useful to other analysis tools.