Autonomous shape modeling of small bodies using infrared image silhouettes

Missions to small bodies require accurate shape characterization for navigation and scientific purposes. In this paper, a Shape from Silhouette (SfS) algorithm is presented which uses the visual extents of the body (silhouettes) in a set of infrared images to infer its three-dimensional shape. The algorithm leverages intersections between rays (each corresponding to a silhouette pixel) from different images to extract the small body shape. A novel sampling method is introduced to optimally extract the relevant silhouette pixels from each image that provide salient information to the shape modeling process. This reduces the computational burden of the algorithm while preserving the salient shape information from each image. The use of infrared images ensures the robustness of the algorithm to lighting geometries. The SfS algorithm is applied to synthetic images of four small bodies at two different approach latitudes of 0° and 45°. At a 0° approach latitude, the resultant shape models have an RMS error of 1.333 m, 4.056 m, 53.99 m, 341.6 m for Bennu, Itokawa, 67P/C-G, and Eros respectively. At a 45° approach latitude, the resultant shape models have an RMS error of 3.159 m, 4.672 m, 54.27 m, 217.1 m for Bennu, Itokawa, 67P/C-G, and Eros respectively. The algorithm’s sensitivity to pole errors (of $2-10^{\circ }$) is characterized for the four bodies which shows an expected degradation in performance as the error increases. The degradation is worse for Itokawa, 67P/C-G, and Eros compared to Bennu due to their more irregular shape. The algorithm is also applied to real infrared images taken by the TIR camera on-board the Hayabusa2 spacecraft to generate a shape model of asteroid Ryugu. The RMS error of this shape model is 8.652 m. This algorithm can be used in future missions such as HERA and Hayabusa2$♯$ to obtain initial shape models of the target objects using on-board infrared images.