Self-supervised Methodologies for Orbital Object Characterization

Space situational awareness (SSA) demands rapid, accurate analysis of orbital objects to mitigate potential hazards to satellites. A significant challenge in this field involves characterizing objects from ground- based observations that often yield unresolved imagery. This research introduces a novel self-supervised learning approach for characterizing orbital objects using telescope-based spectral data. The self-supervised methodology offers particular value for identifying previously uncharacterized objects-a critical capability for comprehensive space situational awareness. Our approach leverages an extensive dataset of simu- lated spectral signatures from orbital objects with diverse physical properties and orbital parameters, ensuring robust and reproducible classification results. We evaluate our algorithm's effectiveness through comparative analysis against the raw spectral data. The comparison employs a novel method that detects the amount of material-composition signal within the sample. This research advances the technical foundation for improved autonomous characterization of objects in Earth orbit.