iraf.net

[Mod Note: Moved from story comment on behalf of khanna]
Hi!I am trying to use identify to trace along horizontal lines in a lunar spectra to unbend them.Can I use pixel coordinates to do this? What should my line list look like? How do I know that Identify has run properly?Thanks,
Kerri

Hello Kerri,Yes it is possible to use pixel coordinates. This is typically done for handling the "S-distortion" of a long slit spectrum. You don't need a line list to do this. You should set the line list to "" (an empty string with no space) and then when you mark a feature (emission or absorption) it will prompt you for a coordinate with the default being the pixel coordinate. Accept the pixel coordinate or give anything you want. Because there is no line list it will just accept what you type. During reidentify it will measure the pixel coordinate at other points and assign the pixel coordinate initially selected interactively as the desired final coordinate. Later the FITCOORDS/TRANSFORM tasks can be used to determine a transformation and to resample the image to align the traced feature(s).Yours,
Frank Valdes

From: Kerri
Date: May 31, 2006 4:37:35 PM MST
To: Frank Valdes <valdes@noao.edu>
Subject: Re: identify and reidentifyHi Frank,Thank you for the timely response. I was able to use your ideas to get the identify portion to work correctly. Unfortunately it seems that reidentify is not doing quite what I need it to.The lunar spectra that I have is 320 x 240 pixels and the spectra fills the entire image. There are lighter and darker lines or rows due to albedo differences. I am using three of these dark lines to correct the S-distortion and I am identifying them in the first step. Although I am using three lines to correct for the distortion, I would like to correct the distortion for every line or row in the image. In my current reduction process, the final image coming out of transform does not have these lighter and darker rows. Not even the three dark lines I identified are there.Do you have any clues where my reduction might be amiss??
Thanks,
Kerri
Begin forwarded message:From: Frank Valdes <valdes@noao.edu>
Date: June 1, 2006 3:46:27 PM MST
To: Kerri
Subject: Re: identify and reidentify
Hi Kerri,I don't know enough about what you are doing. One thing to send is parameters (e.g. lpar transform) and command you are using. So
the steps you should be doing are idenitfy, reidentify, fitcoords, and transform. If you just do the tracing of the dark lines and then apply
the results of fitcoords to the same image with transform you should get something that looks like the untransformed image but with the
lines aligned.Yours,
Frank
From: Kerri
Date: June 2, 2006 12:02:36 PM MST
To: valdes@noao.edu
Subject: Re: identify and reidentifyHi Frank!Again thanks for getting back with me so quickly. I have attached a ZIP file that has all of the par files that I have been using for identify, reidentify, fitcoords and transform. Also included are the input file (m20060415_0111.sub.fits), transform output file (moon_test.fits), and the files generated from each reduction step.My reduction steps have been as follows:(1) Run Idenitfy and interactively choose 3 darker absorption features at lines 39, 142 and 225.
(2) Run Reidentify
(3) Run Fitcoords
(4) Run TransformI am not interactively doing anything except for the Identify step.Thanks Again!!
Kerri
From: Kerri
Date: June 5, 2006 4:26:29 PM MST
To: valdes@noao.edu
Subject: Re: identify and reidentifyHi Frank,I just wanted to check in and see if you have had any luck with my lunar spectrum. I have been trying to change various parameters and keep running them through and still haven't had any luck.Thanks,
Kerri

Hi Kerri,I've looked at your data. The most likely problem is that you are overfitting the data in FITCOORDS. For your data recommend xo=4 and yo=3. I also recommend you delete the traced points at high x because they become noise and would also cause the overfitted surface function to do funny things. If you do come by I can show you how to use FITCOORDS to delete points and figure out the proper fitting orders to use. I also noticed that you had retraced the data several times and so you had many more points to fit than are needed though that should not really cause any problems. I think the symptom of the problem is that the output from FITCOORDS shows:[code:1:084dbe3e41]
NOAO/IRAF V2.12.2-EXPORT xxxx Mon 11:44:03 22-May-2006
Longslit coordinate fit name is m20060415_0111.sub.
Longslit database is database.
Features from images:
m20060415_0111.sub
Map User coordinates for axis 2 using image features:
Number of feature coordnates = 936
Mapping function = chebyshev
X order = 6
Y order = 6
Fitted coordinates at the corners of the images:
(1, 1) = 244.5766 (320, 1) = 246.9409
(1, 240) = 196.7175 (320, 240) = 196.7847
[/code:1:084dbe3e41]Note the last two lines. The right behavior should be:[code:1:084dbe3e41]
NOAO/IRAF V2.12.2a-EXPORT valdes Wed 10:58:28 14-Jun-2006
Longslit coordinate fit name is m20060415_0111.sub.
Longslit database is database.
Features from images:
m20060415_0111.sub
Map User coordinates for axis 2 using image features:
Number of feature coordnates = 184
Mapping function = chebyshev
X order = 4
Y order = 3
Fitted coordinates at the corners of the images:
(1, 1) = 3.460186 (320, 1) = 2.56799
(1, 240) = 241.8973 (320, 240) = 240.1608
[/code:1:084dbe3e41]There is another possible problem which would only apply if you have a version of IRAF labeled V2.13 which is only recently been made available from iraf.net (as opposed to distributions from NOAO which are V2.12a). In that case there is a bug in the version of FITCOORDS which I found when trying to reproduce your problem.Yours,
Frank