Last modified 3 years ago
Source Detect Pipeline
This describes the pipeline that transforms images into x,y lists. It feeds into the BlindPipeline and/or the TweakPipeline.
what we do
We run simplexy via fits2xy.
Simplexy ought to be described in a LaTeX file in the codebase. Hogg is ticketed with that.
how we test it
Here is a random assemblage of possibly non-trivial images for simplexy to properly handle:
- http://oven.cosmo.fas.nyu.edu/blind/status.php?job=tor-200704-71466048
- http://oven.cosmo.fas.nyu.edu/blind2/status.php?job=tor-200705-17309242
- http://live.astrometry.net/status.php?job=alpha-200704-56850152
- http://oven.cosmo.fas.nyu.edu/blind/status.php?job=e4e535e224
- http://oven.cosmo.fas.nyu.edu/blind/status.php?job=5e6694392a
- http://oven.cosmo.fas.nyu.edu/blind/status.php?job=tor-200704-78019335
- http://oven.cosmo.fas.nyu.edu/blind/status.php?job=30c25330ed
- http://live.astrometry.net/status.php?job=alpha-200704-93693449
what we would do in an ideal world
- Obtain or determine a cosmic-ray mask, or more generally, a bad-pixel mask, either by user input or by runing la_cosmic or equivalent.
- If the later steps require it, interpolate over bad pixels.
- Obtain or determine an approximate inverse variance image, either by user input or by cheap statistics, modified by the bad-pixel mask.
- Obtain or determine an approximate seeing or resolution or point-source response for the image.
- Detect peaks by some sensible algorithm / code (using, possibly, the point-source response as some kind of filter), producing a brightness-ranked list of x,y positions, down to some plausible "plate limit".
- Group extremely nearby peaks by friends-of-friends (using, possibly, the point-source response size as the linking length) and keep in the list of image sources only the brightest peak in each group.
- Save these image sources in some well-defined format in some well-defined directory structure with some well-defined file name in the astrometry.net archive.
