Computing the pixel locations of satellites as efficiently as possible for large number of fits files

[astro-dyl]

Hi,

I have a problem where I have tens of millions of FITS files that I need to calculate the pixel positions of all satellites overhead in.

I currently do this as follows (stripped down example):

sat_objects = np.array(load.tle_file(tle_file))

for fits_file in fits_files: #len(fits_files) is >10,000,000
    # extract wcs and datetime (dt) from fits file

  for sat_obj in sat_objects:
  
      difference = sat_obj - telescope_position
      
      topocentric = difference.at(dt)
      
      elev_tle_deg = topocentric.altaz()[0].degrees
      
      if elev_tle_deg < 0:
          continue
      
      ra_tle, dec_tle, dist_tle = topocentric.radec()
      
      x_loc_tle, y_loc_tle, _, _ = wcs.all_world2pix(ra_tle._degrees, dec_tle._degrees, 0, 0, 0)
      
     # if the pixel locations are not NANs, the satellite is overhead and i process it further

Looping over every satellite in every image is necessary but inefficient. I have images of the full sky, meaning that if a satellite has an elevation greater than 0 degrees then I want its pixel position in the image. Only if a satellite has an elevation greater than 0 degrees do I calculate its pixel position (maybe this isn't the most efficient way to do it).

I have two questions:

• Am I able to vectorise the satellite objects through the pixel position calculation?
• Am I able to cheaply calculate if a satellite is overhead at a given time so that I don't need to waste computation on satellites which aren't visible? (most satellites in the TLE list won't be visible at a given time).

Thanks for your help. I really appreciate the skyfield library and community allowing me to do my research well!

Cheers,
Dylan