Seeking advice on lowering the odds-to-solve value

[Prajwel Joseph]

Dear Dustin,

   I am running a local installation of astrometry.net software with custom Gaia index files. The intention is to solve astrometry for many far-ultraviolet images (with ~1'' FWHM). While we get successful WCS solutions for all our test cases using your fantastic software, I would like to make further improvements.

I went through the Astrometry.net article. After reading the section "Verification of Hypotheses", I think modifying the odds-to-solve could be possible for the following reasons.

1. We know the (RA, Dec) pointing information. Our FOV is half a degree in diameter. For the solve-field run, a 1-degree search radius centred on the known pointing information is used. 
2. We also know the pixel scale very well.  

This being the case, how low can be a conservative value for odds-to-solve? I have presently lowered it to 1e6 from 1e9.

Thank you for any help.

With regards,

Prajwel

[Dustin Lang]

Hi Prajwel,

You might find that the "6100-series" index files are perfect for your needs.  They are built from a GALEX catalog, and the scales (4-6) should be well matched to a half-degree field.

http://data.astrometry.net/

As for your question, one thing you can do is turn on the "--verbose" option to solve-field, and then it will print lines like

object 1 of 1000: 0 quads tried, 0 matched.
object 2 of 1000: 0 quads tried, 0 matched.
Got a new best match: logodds 2.99069.
object 3 of 1000: 5 quads tried, 37 matched.
object 4 of 1000: 16 quads tried, 149 matched.
object 5 of 1000: 28 quads tried, 189 matched.
Got a new best match: logodds 3.12527.
Got a new best match: logodds 3.42611.
object 6 of 1000: 64 quads tried, 441 matched.
Got a new best match: logodds 9.32062.
object 7 of 1000: 133 quads tried, 1155 matched.
object 8 of 1000: 201 quads tried, 1596 matched.
Got a new best match: logodds 65.8437.

the "object"s here are stars, so this is saying that, for example, after looking at the 8th star ("object 8"), solve-field formed 201 quadrangles of stars in the image, and found 1596 matches.  I would suggest looking at what those numbers are like for your example images.  If it matches 1,000,000 quadrangles, then a 10^6 odds ratio can easily give you a false positive result.  The other thing you can do is look at the lines saying "Got a new best match: logodds 9.32" etc... this is telling you the best matches found so far.  Usually these numbers will climb slowly as more-and-more convincing false positives are found.  It should give you an idea of what the best non-match is.

cheers,

dustin

--
You received this message because you are subscribed to the Google Groups "astrometry" group.
To unsubscribe from this group and stop receiving emails from it, send an email to astrometry+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/db5558ea-f3b5-4eb7-940c-5bec4667a462n%40googlegroups.com.