Difference in online and installed build solutions

[Alex Priest]

Hello,

I've got a bunch of images I am trying to get the Ra and Dec for using astrometry. I have downloaded and built the software onto Ubuntu on a virtual machine.

The files I am inputting are currently .xyls source lists, but I have also solved with JPEG images as well.

My question is, when I run the files, either the JPEGs or .xyls files, the online service and the build that I have installed give slightly different results.

FYI, my images are quite large, roughly 24 x 16.5 degrees field of view.

For one particular image, when I upload to the online service I get RA ~= 158 deg and Dec ~= 16.5 deg, but when I use my version of the software I get Ra ~=161.5deg and Dec ~= 21 deg.

This happens for both the source lists and the JPEG images.

Any ideas of what could cause this? 

I've tried refining what options I put into the downloaded version, as well as installing some more indexes (which just seemed to make it take longer to solve).

The version of the software I'm using I believe is up to date, I only installed it a month ago.

Thanks! :)

[Dustin Lang]

Hi,

The online version uses a different set of index files (the "200-series", based on the USNO-B catalog) than the currently available ones (the "4200-series", based on 2MASS), so the set of matched stars and the solution you get out will be a little different.

cheers,
--dustin

[Alex Priest]

Hi Dustin,

Thanks for your previous reply, it is interesting to know that in general. However, I'm not sure the different indexes are the problem.

I've tested again with a jpeg file, a .xyls file for the same image both on and offline.

It seems that the jpeg and src list put through the online version do agree with the results of the jpeg when put through the offline build, with only small differences in the right ascension result and near identical declination result.

It also seems that they are all solving with indexes from the 4100 series, namely with 4114, 4117 or 4113. So unless the online versions are made with different catalogues, but named the same, I'm not sure the catalogues are the problem.

However, when I then run the source list through my offline build, I get results that are different by about 4 degrees in Ra AND 5 degrees in Dec which is extremely significant! 

We're talking about differences in Dec of a third the images FOV.

If you have any other suggestions of what could be causing this difference I'd greatly appreciate it, this is for a University undergraduate project :)

I've attached a couple images of the online build's results and what the offline version says. I've got full log files if it helps at all as well.

Thanks heaps.

[Dustin Lang]

Hi,

I'm confused.  In your log, I see:

Field center: (RA,Dec) = (158.021096, 16.508045) deg.

and the snapshot of the JPEG web version says

Center (RA,Dec) = (158.054, 16.513)

so only different by a small amount.

Oh, the log is for the JPEG file, and you're saying that the source list offline yields a different result.  Please send that log.

One issue with X,Y lists is knowing where the center of the image is.  You should use the --width and --height arguments to tell it how big (in pixels) the image is.

cheers,
--dustin

[Alex Priest]

Hi,

Sorry, I realise what I sent wasn't very clear.

The two attached images were screen captures of the online astrometry results, and the txt file contains the command line output from Linux for BOTH the Jpeg first and the Source list second, when asking the offline build to solve them.  

I've attached two log files to this message, which contain the same command line output but are sepereated now for clarity. They are both for solving with my offline build.

I've also included the line which I input to ask astrometry to solve for each. You may notice I have told astrometry the width and height of the original image for the source list.

If you want more details I can run the solver again with the verbose option enabled and show you the output then.

Cheers

[Dustin Lang]

Aha, yes, I see =)  Thanks for the clarification.

Okay, so just to be totally clear: you are finding:

web version with JPEG = web version with XYLS = offline version with JPEG != offline version with XYLS ?

I'm looking at the XY list you uploaded, and its coordinates are between X ~ 0 to 4000 and Y ~ 0 to 2500, yet the JPEG has size 7360x4912, and neither of those match the --width and --height you specified... so what's up with that?

It's important because when you give it a --scale-low, it uses that and the image width to compute an angular scale per pixel, and *that* is the quantity that is used to keep or reject matches.

Meanwhile, could you also please send the --verbose log running with the .xyls file?

Thanks,
--dustin

[Dustin Lang]

oh, and if you could please attach the file so we know we're talking about the same thing, that'd be great.
thx

[Alex Priest]

Hi Dustin,

I'm not doing a good job of explaining am I hahaha, sorry.

Yes, this statement is true: 

'"web version with JPEG = web version with XYLS = offline version with JPEG != offline version with XYLS ?"

However, I think I have just found the problem, and it turns out I am just an idiot!

You mentioned that I told astrometry the source list size was half the jpeg, Thats because the raw file is downsampled by 2 before the sources are extracted to make the xyls.

But I have just discovered that I think I put the height and width dimensions in the wrong way around!!!

I tested what would happen if I entered the pixel width as height and pixel height as width and the results seem to now agree.

Thank you for all your help!

Alex 

[Dustin Lang]

Great news!  Glad to hear it.
cheers,
--dustin