Miguel Velez-Reyes, The University of Texas at El Paso; Jiarui Yi, The University of Texas at El Paso
Keywords: hyperspectral remote sensing, unresolved objects, unmixing
Abstract:
USA depends economically and militarily in space assets. Satellites provide a multitude of critical services, which are critical for US defense, security and economic wealth. Space situational awareness (SSA) is needed to have a clear picture of the environment surrounding US space assets to detect any changes or potential threats. Radar ground assets are primarily used for observing targets in Low Earth Orbit (up to ~2000 km) while optical ground assets are used to assess the environment at higher altitudes; both of which do not routinely use imaging sensor technology. Current ground-based space telescope technology cannot spatially resolve objects in space that are distant (orbits beyond 1,000 km altitude, e.g. GEO) or that are small (e.g. nanosatellites). These objects are denoted as unresolved objects. An approach that can potentially extract quantitative information about unresolved objects is hyperspectral remote sensing. The high spectral resolution of hyperspectral sensors allows extraction of information about the material composition of the unresolved object from their contribution to the measured spectra. Thus, hyperspectral remote sensing can provide a quantitative approach to assessing/extracting orbiting objects material information. Even though the object cannot be spatially resolved, it can be spectrally resolved.
In this paper, we study the application of unmixing to hyperspectral remote sensing of unresolved objects for space situational awareness. We first review library-based unmixing and its application in the context of hyperspectral remote sensing of unresolved objects. We introduce the concept of spectro-temporal signatures of unresolved objects and how can they be used to characterize the object and in the unmixing process. In particular, we study how the spectro-temporal signature may be used by unsupervised hyperspectral unmixing methods to extract material composition of unresolved objects. Simple simulations are used to illustrate these ideas. Our main goal is to understand the limitations of unsupervised hyperspectral methods for remote sensing of unresolved objects and how the prior information for existing spectral libraries can be used to develop semi-supervised unmixing methods. We also want to explore the idea of how “new” signatures extracted from semi-supervised methods can be used to expand existing libraries.
Date of Conference: September 15-18, 2020
Track: Non-Resolved Object Characterization