Frigate: A Novel Dataset Comprising Wide Field-of-View Astronomical FITS Images of Low Earth Orbit Scenes for Machine...

Area of Science:

• Astronomy and astrophysics
• Computer vision
• Space situational awareness (SSA)

Background:

• Publicly available space surveillance datasets are scarce.
• Commercial providers possess valuable but inaccessible data.
• Machine learning and computer vision can advance SSA.

Purpose of the Study:

• Introduce the Frigate dataset, a novel collection of Low Earth Orbit (LEO) images.
• Provide processed images with enhanced quality for SSA applications.
• Facilitate research in satellite tracking and orbital debris detection.

Main Methods:

• Collected still-frame images of LEO using ground-based optical systems.
• Developed image processing techniques for artefact removal, background subtraction, and sharpening.
• Documented optional downstream processing methods like stacking and streak detection.

Main Results:

• A novel astronomical dataset (Frigate) of LEO images is now available.
• Processed images mitigate common artefacts like lens glare and vignetting.
• The dataset supports machine learning tasks for space surveillance.

Conclusions:

• The Frigate dataset is a valuable, openly accessible resource for the SSA domain.
• Processed imagery improves the utility of astronomical data for computer vision tasks.
• This initiative addresses the need for public space surveillance data.