Ponente
Descripción
Since the launch of Sputnik 1 in 1957, the number of satellites in orbit has increased exponentially. The recent deployment of constellations such as Starlink has intensified the impact on astronomical observations, especially due to satellites in Low Earth Orbit (LEO). Although these systems provide major benefits in communications, meteorology, remote sensing, and defense, they also degrade the quality of the night sky by adding noise and obscuring targets in astronomical images. This interference affects both ground-based and space telescopes and has become a major challenge for modern astronomical surveys, which generate massive datasets requiring efficient automated processing.
This work proposes the use of Artificial Intelligence to automate the detection of artificial objects through their trails in astronomical images. The system identifies the celestial position of detected objects, compares them with satellite databases, and analyzes their light curves to determine their operational status. Neural networks are employed to process large volumes of data without requiring complex preprocessing steps.
Data acquisition is carried out using a 107 mm refractor telescope located at an observatory in Pillarno, Asturias, equipped with an ATIK APX 60 CMOS sensor. Around 100,000 images with 2-second exposures have been collected at random sky coordinates. Once a trail is detected, the system searches for matches in artificial object databases. If no match is found, the object is automatically classified, for example as a satellite, rocket stage, or space debris, and its operational state is estimated through light curve analysis