StreakMind: AI detection and analysis of satellite streaks in astronomical images with automated database integration

Abstract

Context. Artificial satellites and space debris are increasingly contaminating astronomical images, affecting scientific surveys and producing large volumes of streaked exposures. Manual inspection is no longer feasible at scale, and reliable identification and characterisation of streaks has become essential for both the quality control of data and the monitoring of objects in Earth orbit.

Aims. We present StreakMind, an automated pipeline designed to detect near-Earth objects (NEOs) and satellite streaks in astronomical images, characterise their geometry, and cross-identify them with known orbital objects. The system integrates all inference results into a structured database suitable for large surveys.

Methods. A YOLO-OBB model was trained on a hybrid manual-synthetic dataset of 2335 images and used to detect streaks in processed FITS frames. Geometric refinement, inter-frame association, satellite cross-identification, and Gaussian-based confidence scoring were then applied to produce final identifications, which were stored in a normalised relational database. In this work, images acquired at La Sagra Observatory (L98) with a Celestron C14+Fastar telescope were used to develop and test automated streak detection and characterisation methods.

Results. On the test set, the model achieved a precision of 94% and a recall of 97%. It reliably detected faint streaks, delivered consistent geometric reconstructions across the dataset, and performed robust satellite cross-identification. The Gaussian-based confidence scoring provided stable identification probabilities across consecutive frames.

Conclusions. StreakMind demonstrates strong potential for large-scale automated analyses of linear streaks produced by both NEOs and artificial satellites in ground-based astronomical images. The pipeline offers high detection reliability, robust geometric reconstruction, and reproducible satellite cross-identification within a fully integrated end-to-end framework.