Bayesian Inference of Spacecraft Geometry and Pose

Space platforms are complex dynamical systems and there are many reasons to model them from ground telescope imagery as well as from space-based sensor platforms. One key motivation is to determine the operational state and configuration of the vehicle. Key questions such as did the solar panels fail to deploy, did the object begin to tumble unexpectedly may be addressed by a sufficiently detailed and accurate 3D reconstruction resulting from inspection imagery. It is proposed to model the formation of imagery of orbiting spacecraft (satellites) observed by ground telescopes based on a probabilistic framework. The unknown parameters of the inference process include spacecraft pose and geometry as well as unknown atmospheric aberration properties. The posterior distribution of these parameters is obtained by Bayesian inference using particle filtering. The probabilistic framework supports straightforward fusion of available information, such as derived from photometric stereo or shape from shading. A range of complementary image feature measurements are proposed to provide redundancy to occlusion and to overcome ambiguity in any one information channel. The VSI team provides the breath of technical skills to carry out the proposed research including 3D reconstruction and pose estimation, space vehicle technology and state estimation for space platforms.