Where are kd trees generated?

[wgwe...@telus.net]

 I have recently downloaded the precompiled Ubuntu version using Ubuntu's software manager, and have tried to solve one of the example jpg's.

The program complains that it cannot find the kdtree header in an unrelated image file, and subsequently failed to read kdtree from the same file. Finally it tried and failed to read an index from the same unrelated file.

1- Where are the kd tree files generated?
2- is there a directory in which they are meant to go?
3- Is there a directory that indices and kd tree files  *must* go ?

By the way, I tried to capture the error messages with > and | tee to a file, but they are apparently not written to stderr.

[Dustin Lang]

Hi,

I believe most error messages are written to stdout.  You can capture them all via

solve-field .... > out 2>&1

(in bash; not sure about csh)

I am confused about the error messages you are getting.  The index files (usually name index-*.fits) normally live in /usr/local/astrometry/data or something like that.  There is a config file (normally /usr/local/astrometry/etc/astrometry.cfg) that tells solve-field where to look for index files.  (Solve-field looks for that config file relative to where the "astrometry-engine" executable is located; no paths are hard-coded).  I am sure that the Ubuntu packages install things self-consistently, probably in /usr rather than /usr/local.  Don't move anything.

I can't say much more without seeing the logs, but I *believe* the ubuntu package is set up so that just installing it and running "solve-field test.jpg" (for your image "test.jpg") should just work.  You do not need to move image files anywhere to solve them, or anything like that.

The index files contain (multiple) kd-tree structures.  In the olden days, each index was in three separate files; the code still checks for that structure, so even if everything is installed correctly you may see messages that look like error messages about not finding files named *.quad.fits, *.ckdt.fits, and *.skdt.fits.; ignore those.

cheers,
--dustin

[wgwe...@telus.net]

Hi, Dustin.
I found out why I got the kd tree messages. I had a copy of my image in the same directory as the indices, and the program was looking for the trees in the image file. It seems to look into any file that is not a .cfg file. ( I changed the extension on the image, same result,) The good news is that when I removed the file I got a reasonable looking solve.

*but*... There is always a but.  When I tried to display the (image).new file in ds9, to view the WCS info, it crashed the program. I'm still working on that one, as well as trying to repeat the solve with a different image. No luck so far. The program got as far as giving me a reasonable plate centre and pixel scale, but no solve. When I re ran the program with the pos'n and scale as parameters, I got the same result, no solve.

How does s/n affect the solve? Would it be useful to bin down (not resample) large images to improve s/n.

Thanks for your help,
Bill

[Dustin Lang]

Hi,

If you want to post an example image (eg, submit to nova.astrometry.net and post a link to it here), I can try to provide specific advice.

We typically need about the 20-30 brightest stars to be reasonable centroided to get a solution, so as long as your images have enough S/N for that, it's fine.  We often find that "--downsample 2" helps, especially for larger images; it's the default on the nova.astrometry.net web service.

cheers,
--dustin