Approaches for processing spectral measurements of reflected sunlight for space object detection and identification

The proliferation of small, lightweight, 'micro-' and 'nanosatellite' (largest dimension < 1m) has presented new challenges to the space surveillance community. The small size of these satellites makes them unresolvable by ground-based imaging systems. Moreover, some satellites in geo-orbit are just simply too distant to be resolved by ground systems. The core concept of using Non-Imaging Measurements (NIM) to gather information about these objects comes from the fact that after reflection on a satellite surface, the reflected light contains information about the surface materials of the satellite. This approach of using NIM for satellite evaluation is getting new attention. In this dissertation, the accuracy of using these spectral measurements to match an unknown spectrum to a database containing known spectra and to estimate the fractional composition for materials contained in a synthetic spectrum is discussed. This problem is divided into two parts, a pattern recognition problem and a spectral unmixing problem. Two methods were developed for the pattern recognition problem. The first approach is a distance classifier processing different input features. The second method is an artificial neural network designed to process central moments of real measured spectra. This spectrum database is the Spica database provided by the Maui Space Surveillance Site (MSSS), Hawaii USA and consists in spectra from more than 100 different satellites. For the spectral unmixing part, four different approaches were tested. These approaches are based on the ability of spectral signal processing to estimate fractional composition of materials from the measurement of a single spectrum. Material spectra were provided by the NASA Johnson Space Center (JSC) to create synthetic spectra. A statistical approach based on the Expectation Maximization (EM) algorithm as well as a constrained linear estimator were used to estimate fractional compositions and presence of materials in a synthetic spectrum. The last two unmixing methods are based on inverse matrices, singular value decomposition and constrained pseudoinverse. The results for material identification and relative abundance estimation are presented as a function of signal-to-noise ratio and as a function of the number of material used in the synthetic spectrum. The best results for the satellite classification problem were obtained using an artificial neural network, which has a total classification rate of 84%. The best spectral unmixing method is the linear constrained estimator. The detection rate is on average 94%, the estimation error rate is 1%, and the error detection rate for the detection of material not present in the spectrum is 10%.