Space Object Identification and Classification from Hyperspectral Material Analysis

TL;DR

This work targets space-object identification from hyperspectral data by first extracting material composition through two unmixing strategies (ML-based ANN and library-based least squares) to produce Material Abundance Curves, followed by probabilistic classification of objects into predefined classes. The authors implement a simulation-driven pipeline with time-series, single-pixel hyperspectral signals, and a color-index normalization to ensure geometry-agnostic material inference, then couple MACs with synthetic-data trained classifiers to achieve probabilistic object labeling, including an option to label uncertain cases as UFOs. Key contributions include a head-to-head comparison of ML and non-ML unmixing under aging and incomplete libraries, a probabilistic framework for material presence, and robust classification performance across 13 object types, emphasizing explainability and generalization to unseen objects. The approach enables practical, transparent space-surveillance capabilities by linking spectral features to material composition and object class with quantified uncertainty, suitable for ESA OSIP HyperClass and related applications.

Abstract

This paper presents a data processing pipeline designed to extract information from the hyperspectral signature of unknown space objects. The methodology proposed in this paper determines the material composition of space objects from single pixel images. Two techniques are used for material identification and classification: one based on machine learning and the other based on a least square match with a library of known spectra. From this information, a supervised machine learning algorithm is used to classify the object into one of several categories based on the detection of materials on the object. The behaviour of the material classification methods is investigated under non-ideal circumstances, to determine the effect of weathered materials, and the behaviour when the training library is missing a material that is present in the object being observed. Finally the paper will present some preliminary results on the identification and classification of space objects.

Figures (24)

• Figure 1: Classification pipeline and probabilistic material identification.
• Figure 2: Plot of multiple spectral measurements of the same aluminium foil at different orientations
• Figure 3: Example of three simulated satellites: a) Iridium-NEXT, b) Starlink, c) DubaiSat-2. Colours correspond to regions with different materials. The reflectivity spectra for each region were obtained from laboratory experiments using mock-ups.
• Figure 4: Example of ANN predictions for material abundances for an Iridium-NEXT satellite. Over 200 simulations with differing orbital and rotational conditions, the mean squared error in the predictions was 5.32 $\times$ 10$^{-3}$
• Figure 5: Schematic of library unmixing algorithm