The rapidly growing population of satellites and space debris, collectively referred to as RSOs, poses major challenges for space operations, astronomical observations, and the preservation of space and the night sky as shared resources. Assessing the impact of this increasing orbital population, as well as developing informed mitigation strategies, requires precise knowledge of both the past and current situation and, wherever possible, detailed information on the characteristics and evolution of individual objects. The archival data gathered by astronomical telescopes represent an invaluable resource for such assessments. These datasets capture serendipitous streaks from RSOs in exposures, offering insights into the orbital population.

In this context, the present work aims to automatically and reliably extract streaks from astronomical images acquired with VST. This large and highly sensitive facility, located at the Paranal Observatory in Chile, records dozens of RSO crossings per night. These detections allow the analysis of the impact of streaks and its evolution over time on the telescope's operations. Through photometric analysis, it is also possible to derive brightness measurements from the same streaks, opening the door to a deeper understanding of the physical properties of the observed objects.

A CNN is employed to detect streaks in raw images. Its training, validation, and testing datasets are carefully designed to capture streaks of varying intensity and brightness fluctuation patterns, across different filters and under diverse background conditions. Because false positives can occur too frequently for reliable analysis, a second CNN classifier was developed to effectively filter and curate the detections. Once identified, images containing streaks are plate-solved and calibrated for precise astrometric and photometric measurements. Detected streaks are then cross-referenced with the space-track catalogue, and lightcurves are extracted by placing a series of adaptive rectangular apertures along each streak.

When applied to one month of data, representing 2357 256-megapixel images, the detection and classification pipeline reach an overall precision of 0.990. This performance is achieved while retaining 97% of true positives from the detection algorithm and discarding more than 96% of false detections. Lightcurves are successfully extracted using adaptive apertures placed along the streaks. Applying these methods to five years of data (2019-2023) resulted in 80564 detected streaks. From these, temporal trends are identified, and the impact of streaks on VST observations is carefully evaluated. The photometric analysis of selected streaks from 2020 and 2021 provides valuable insights into object brightness, allowing the characterisation of RSOs beyond simple counts.

As the rapid growth of the RSO population becomes a growing concern for space operators, astronomers, and society at large, the insights gained in this work can serve as a foundation for dialogue between the various actors involved in or affected by space activities. The photometric analysis demonstrates that astronomical databases are an excellent resource for the physical characterisation of space objects. Finally, the techniques developed in this work can be adapted for use with other telescope data.