Advanced Techniques for Estimating and Re ning Orientation Vectors of Space Object Imagery

We describe three advanced techniques incorporated into the design of a model-based image analysis system which automatically estimates the orientation vector of satellites and their sub-components. The system, implemented in Khoros, operates on images obtained from a ground-based optical surveillance system. Features of each satellite image are rst extracted by partitioning the image and constructing a model representation. Second, pose estimates are obtained from model-matching across a model database. Finally, pose reenements are derived from photogrammetric or geometric information. We discuss three advanced techniques: eigen-indexing, robust aane point matching, and random edge sampling. Eigen-indexing is a novel indexing method which constrains the number of model candidates and signiicantly improves the overall system speed by providing an eecient way to access the model database. We include a theoretical analysis of the eigen-indexing technique. Robust aane point matching signiicantly improves pose-reenement performance on degraded imagery. Finally, random edge sampling is a novel pose estimation technique used on those satellites which are diicult to partition and therefore violate nominal requirements for aane point matching. We demonstrate these techniques on estimating orientation vectors from challenging satellite imagery.