Increases in space traffic have caused an ever-growing need to recognize, track, and determine the operational state of both known and unknown satellites in-situ. Imagery from ground and/or space-based observation platforms provide a convenient and versatile data source for such characterization but is often limited in both quantity and quality. Rapid and automated 3D reconstruction of satellite models from limited 2D imagery would significantly reduce the burden of manual satellite characterization for both analysts and decision makers. VSI proposes Sat3D, a 3D satellite modeling capability from limited input data based on the innovative and proven probabilistic volumetric representation. The VSI team is well suited to program challenges, with expertise in remote sensing, 3D modeling, computer vision, and machine learning.3D modeling from limited input,Bayesian inference of spacecraft geometry,multi-source fusion of ground and space-based imagery