PinPoint distortion coefficients

Hello,
I need to understand the data PinPoint has added to the image FITS headers and above all the distortion coefficients (TR1_0, TR1_1, ... TR1_14, TR2_0, ... TR2_14). I can make an educated guess of what they might mean, but I truly need to know - guessing is not an acceptable option. Reference Guide (a companion to ACP Observatory Control Software) states: "These terms are described in more detail in the document called FITS World Coordinate Systems (WCS).". I have spent several days going through all the material and scientific articles within that NASA web site and I have never ever seen any reference to TR1 and TR2 coefficients. I may just be blind, or they are not there.

Any suggestions on where to find the polynomial those coefficients describe?

I promise to be tremendously grateful for the answer,
// Rami Rekola

Please see this message series on this issue. The main point is:

Quote:

---

Ah, since the details of that are something I developed while a "starving programmer" (and well before the SIP stuff) and now I make my living doing this, I'd rather not open my book of tricks. The biggest trick is how to do the analysis, but seeing the algorithm for the corrections would give a fair bit of it away.

What is stopping you from using the engine? It's just a few lines to get the full accuracy full frame. Or are you doing some analysis on Linux or something?

---

Hi Rami,

I was wrong about "TR2_6" item in the PinPoint discussion in the Reference Guide -- I'm sorry that statement was misleading. I am in the midst of updating the entire Reference Guide, and I will clarify that. The publically available documents I refer to, talking about WCS coordinates or FITS header elements, do not describe any of the "TR" coefficients Bob Denny developed for PinPoint.

Thank you for the answer. I could not see the message series on distortion terms in the header, though, as it says I do not have a permission to access the page.

I'm sorry, I didn't know the distortion coefficients were your invention. I thought they were something generic, described in the FITS reference manuals.

I shall give a wee background to my request. I am working with Prof. Bo Reipurth from the University of Hawai'i. He has two VYSOS telescopes, which are both running on your software. We have acquired some 100 000 images over the past several years and are beginning to reduce them now. We are trying to make a pipeline that takes the images in and spits the astrometry and photometry out. Even though the telescopes are using Windows computers to run your software, all of our data reduction will (have to) be done in our Linux servers. This is partly because of how the data storage has been arranged and partly because all our existing analysis tools, including IRAF, run only in unix environments. We don't even have access to Windows computers powerful and fast enough to process 100 000 images in any meaningful time frame. If you are unable to help us, I guess we just have to reinvent the wheel.

Cheers, // Rami

Thanks Dick! No worries. It is good to know though, as I did think initially that the answers would be available from the link on that Reference Guide page.

Rami - - PinPoint could be used with a relatively small JavaScript script to produce science grade astrometry of all of the detected objects in an image. It would take nominally 5 seconds per image on a modest Windows system. Extending the script to loop through 100,000 images would be trivial. That would take about 140 hours. Josh/VYSOS has ACP licenses which include full PinPoint functions. However producing millimag / science photometry is not possible with PinPoint alone. It could extract accurate Flux measurements, and with the APASS stars from the UCAC4 catalog, could produce decent photometry depending on the filters used to acquire your data. Since the APASS reference stars are going to be in each field, again this should be straightforward. Reinventing the wheel on Linux will likely take more time, and will need testing and validation.

Hello, I am really interested to know how to use the distortion coefficients TRi_j for a project in order to convert pixel coordinates (from an image of the sky) into equatorial coordinates. Therefore, for last year, have you got anything interesting about these coefficients ?
Thank you for your help,
Alice

I thought I had answered this, somewhere... The coefficients are private but there are PinPoint API calls XYtoSky() and SkyToXY() that will transform between image and equatorial coordinates with the distortion corrections included. You get full accuracy across the entire frame. The API is described in the PinPoint user guide. Are you a customer? If so you can post to the PinPoint support section.

Yes indeed you did answer this somewhere else. Thank you I understand better now. Would you know exactly the accuracy we can get with the distortion coefficients included, as a function of the input image ? Is it more accuracy than using only the SIP coefficients ?

Moreover does the "average residual" computed after plate solving take into account distortion correction ?

Besides in your function XyToSky(x,y) a pixel is defined as (x,y). Would you know exactly where are the x, y located ? Are they located in the pixel center or in the upper left side of the square pixel ?

Thank you for your answer
Regards

There are no SIP coefficients in PinPoint, instead the system we have pre-dates SIP and does essentially the same thing. I can't guarantee anything about the accuracy, you will need to test with your data and decide whether PinPoint is "good enough" or not. The average residual does take into account the distortion. Everything takes into account the distortion And note that "distortion" is the departure from the planar projection you select (usually TAN) not the total "distortion". In other words PinPoint produces the best fit WCS then applies additional corrections to the best of its ability. The XY origin is the standard FITS conventions, of the upper left. You will learn these things if you decide to use the API.