/Satellite_Detection

This is the repository for the analysis of satellite trails on Hubble images (raw images) in the 2021 ESA summer project "The impact of megaconstellations on space astronomy".

Primary LanguageJupyter Notebook

Satellite_Detection

This is the repository for the analysis of satellite trails in Hubble images in the ESA summer project "The impact of megaconstellations on space astronomy".

  1. Zooniverse_Data.ipynb: Transforms the results from Zooniverse into training data for the machine learning algorithms.
  2. Hubble_Image_Download.ipynb: Downloads all the raw Hubble Archive Images that make up the composite images for the three instruments ACS/WFC, WFC3/UVIS and WFC3/IR and saves them in the format 600x600px.
  3. Image_Classifier.ipynb (Machine Learning algorithm 1): Makes binary predictions "satellite"/"no_satellite" on a given Hubble image.
  4. Mask_R-CNN_Satellites.ipynb (Machine Learning algorithm 2): Detects start/end point and the angle of a satellite trail on a given Hubble image with a satellite trail.
  5. Analysis.ipynb: Resulting analysis script for the raw images. Creates plots showing the time evolution of the fraction and the chance for a satellite trail as well as the fraction for the different filters.