Automated Techniques for Extracting Four Dimensional Space-time Data from Ground-based Imagery of Space Objects

Stellar Science has developed innovative techniques for the automated processing of sequences of images into three dimensional models in near real time, and demonstrated those techniques on space surveillance imagery. Using feasibility studies that rigorously characterize the performance of several well-known shape-from-motion techniques, in Phase II we will develop a fully-operational system that incorporates algorithms for tracking over very large (> 45 degrees) rotations between individual image frames, accommodates articulating components, utilizes a priori sensor knowledge when available, and creates robust shapes to more closely approach true “models” of satellites. We have two unique strategies to achieve these objectives. This first is the integration of the feature tracking and shape reconstruction stages to effectively accommodate both noisy data and the occlusion and reappearance of features as the object rotates. The second strategy builds an object shape with a meshed surface or a set of voxels derived using the “space carving” technique. This two-front approach will produce a robust, well-engineered system capable of building 3D models of satellites from real-world space surveillance imagery. Stellar Science brings to this solution unique operational experience in both space surveillance imagery and shape-from-motion techniques and a proven record of delivering robust scientific algorithms and software.