Automated Detection of Satellite Trails in Ground-Based Observations Using U-Net and Hough Transform
arXiv:2407.19461 · doi:10.1051/0004-6361/202451663
Abstract
The expansion of satellite constellations poses a significant challenge to optical ground-based astronomical observations, as satellite trails degrade observational data and compromise research quality. Addressing these challenges requires developing robust detection methods to enhance data processing pipelines, creating a reliable approach for detecting and analyzing satellite trails that can be easily reproduced and applied by other observatories and data processing groups. Our method, called ASTA (Automated Satellite Tracking for Astronomy), combines deep learning and computer vision techniques for effective satellite trail detection. It employs a U-Net based deep learning network to initially detect trails, followed by a Probabilistic Hough Transform to refine the output. ASTA's U-Net model was trained on a dataset with manually labelled full-field MeerLICHT images prepared using the LABKIT annotation tool, ensuring high-quality and precise annotations. This annotation process was crucial for the model to learn and generalize the characteristics of satellite trails effectively. Furthermore, the user-friendly LABKIT tool facilitated quick and efficient data refinements, streamlining the overall model development process. ASTA's performance was evaluated on a test set of 20,000 image patches, both with and without satellite trails, to rigorously assess its precision and recall. Additionally, ASTA was applied to approximately 200,000 full-field MeerLICHT images, demonstrating its effectiveness in identifying and characterizing satellite trails. The software's results were validated by cross-referencing detected trails with known public satellite catalogs, confirming its reliability and showcasing its ability to uncover previously untracked objects.
References in corpus (21)
- LSST: from Science Drivers to Reference Design and Anticipated Data Products
- The Zwicky Transient Facility: System Overview, Performance, and First Results
- ATLAS: A High-Cadence All-Sky Survey System
- Astrometry.net: Blind astrometric calibration of arbitrary astronomical images
- The Low Earth Orbit Satellite Population and Impacts of the SpaceX Starlink Constellation
- Proper image subtraction - optimal transient detection, photometry and hypothesis testing
- On the Impact of Satellite Constellations on Astronomical Observations with ESO telescopes in the Visible and Infrared Domains
- Mitigation of LEO Satellite Brightness and Trail Effects on the Rubin Observatory LSST
- Optimal and Efficient Streak Detection in Astronomical Images
- Small near-Earth asteroids in the Palomar Transient Factory survey: A real-time streak-detection system
- MaxiMask and MaxiTrack: two new tools for identifying contaminants in astronomical images using convolutional neural networks
- The BlackGEM telescope array I: Overview
- MeerCRAB: MeerLICHT Classification of Real and Bogus Transients using Deep Learning
- Linear feature detection algorithm for astronomical surveys - I. Algorithm description
- AutoSourceID-Classifier. Star-Galaxy Classification using a Convolutional Neural Network with Spatial Information
- AutoSourceID-Light. Fast Optical Source Localization via U-Net and Laplacian of Gaussian
- Concerns about ground based astronomical observations: quantifying satellites' constellations damages
- Consistency Tests for Comparing Astrophysical Models and Observations
- AutoSourceID-FeatureExtractor. Optical image analysis using a two-step mean variance estimation network for feature estimation and uncertainty characterisation
- Satellite shadows through stellar occultations
- OneWeb Satellite Brightness -- Characterized From 80,000 Visible Light Magnitudes