In 2017, humanity had its first encounter with an interstellar object named 1I/”Oumuamua. This distant traveler piqued the curiosity of astronomers due to its unique characteristics that differed from anything previously observed. Speculation abounded about its origins, even entertaining the controversial notion that it might be an extraterrestrial probe or a fragment of a discarded spacecraft passing through our solar system.
Public interest in the possibility of “alien visitors” was further heightened in 2021 with the release of the UFO Report by the Office of the Director of National Intelligence (ODNI). This report shifted the study of unidentified aerial phenomena (UAP) from a clandestine affair overseen by government agencies to a legitimate scientific pursuit.
With a keen focus on both celestial phenomena and orbital objects, scientists have begun proposing innovative ways to leverage recent advancements in computing, artificial intelligence (AI), and instrumentation to aid in the detection and analysis of potential “visitors” from beyond. Notably, a recent study conducted by a team from the University of Strathclyde explored how hyperspectral imaging, coupled with machine learning techniques, could revolutionize the process of detecting and understanding these enigmatic phenomena.
The research team, led by Massimiliano Vasile, a professor of mechanical and aerospace engineering, comprised researchers from various disciplines, including Mechanical and Aerospace Engineering and Electronic and Electrical Engineering at the University of Strathclyde, as well as the Fraunhofer Center for Applied Photonics in Glasgow.
Their findings, detailed in a preprint titled “Space Object Identification and Classification from Hyperspectral Material Analysis,” have been made available on the pre-print server arXiv and are currently under review for publication in Scientific Reports.
This investigation represents the latest installment in a series of studies that explore the potential applications of hyperspectral imaging within space-related activities. The first paper, titled “Intelligent Characterization of Space Objects with Hyperspectral Imaging,” was published in Acta Astronautica in February 2023 and was associated with the Hyperspectral Imager for Space Surveillance and Tracking (HyperSST) project. Notably, this project was selected for funding by the UK Space Agency (UKSA) and laid the foundation for the European Space Agency’s (ESA) Hyperspectral space debris Classification (HyperClass) project.
The most recent paper delved into the prospect of applying hyperspectral imaging techniques to the emerging field of UAP identification. This entails the collection and analysis of data spanning the electromagnetic spectrum from individual pixels. Such analysis is typically conducted to identify different objects or materials present within images.
Vasile explained that the fusion of hyperspectral imaging and machine learning has the potential to narrow down the search for possible technosignatures by minimizing false positives arising from human-made debris objects like spent rocket stages and defunct satellites. He elaborated via email to Universe Today that by scrutinizing the spectral data, it becomes possible to ascertain material composition even from a single pixel. Furthermore, understanding the temporal variations of spectra allows insights into the motion characteristics of these objects.
Vasile and his team proposed the establishment of a data processing pipeline tailored for UAP image analysis, employing machine learning algorithms. The initial step involves assembling a dataset of time-series spectra belonging to various space objects, encompassing both satellites and objects in orbit. This dataset should encompass diverse scenarios, trajectories, lighting conditions, and precise information about the geometric properties, material distribution, and attitude motion of every orbiting object at all times.
In essence, the researchers require a comprehensive database of all human-made space objects for comparative analysis to effectively filter out false positives. Given the scarcity of such data, Vasile’s team developed physics simulation software to generate training data for their machine learning models. Their approach then incorporated both machine learning-based classification and traditional mathematical regression analysis to link spectra with the materials producing them.
By using machine learning-based classification, the researchers could associate the probability of detecting specific material combinations with corresponding classes. Upon completing the pipeline, Vasile shared that they conducted a series of tests, yielding promising results. These tests encompassed a laboratory setting with a satellite mockup composed of known materials, high-fidelity simulations mimicking real-space observations, and telescope observations of satellites and the International Space Station. While some tests produced favorable outcomes, others were less successful due to the limited size of the material database.
In their forthcoming paper, Vasile and his colleagues will present the aspect of their pipeline pertaining to attitude reconstruction. This presentation is anticipated to occur at the AIAA Science and Technology Forum and Exposition (2024 SciTech), scheduled from January 8th to 12th in Orlando, Florida.
Source: Universe Today