Skip to content

Typical asteroid detection workflow

mostanes edited this page Apr 13, 2019 · 2 revisions

A typical workflow for detecting asteroids using Umbrella2 would be similar to the following:

  1. Pre-process input images with a noise removing filter (ex. RestrictedMean or CoreFilter).
  2. Create a master image by combining the others, removing as much as possible the moving objects (ex. applying MultiImageMedian).
  3. Subtract the master image from the other images (ex. MaskByMedian).
  4. (if attempting a fully automated pipeline) Attempt to infer the relevant parameters for the detection algorithms.
  5. Apply the detection algorithms (DotDetector, RLHT).
  6. Filter the resulting light sources.
  7. Pair the light sources into tracklets (currently there is no good pairing algorithm, but PoolMDMerger can be used, albeit with some difficulty).
  8. Filter and output the tracklets.
Clone this wiki locally