PinPoint Astrometry Failures

Discussion in 'Image Processing' started by Richard Tabbutt, Oct 17, 2018.

Tags:
  1. Richard Tabbutt

    Richard Tabbutt Cyanogen Customer

    Joined:
    Dec 11, 2017
    Messages:
    16
    I am having PinPoint Astrometry failures on "perfect" images of M42 with known good coordinates and pixel scale. PinPoint solution parameters are all default. Solution times out. Same result on all 20 images.

    I just discovered that PinPoint Astrometry does solve the R,G, and B images shot through filters. So my guess is that the Luminance images on which it fails are overexposed. Is there any workaround that can be suggested on these L images? The moon will prevent any more imaging for a couple of weeks.
     
    Last edited: Oct 17, 2018
  2. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    6,273
    Please upload one of your images in FITS format.
     
  3. Richard Tabbutt

    Richard Tabbutt Cyanogen Customer

    Joined:
    Dec 11, 2017
    Messages:
    16
    Here is my latest discovery: I used a low-pass mild FFT filter in a batch process on the Luminance images, which lowered the star saturation by a minuscule amount (from 65,504 to 64,500). My tests showed that PinPoint Astronomy would then solve these filtered images. I then proceeded to align all 20 of them using the astrometric technique, and all but one were solved and aligned. Curiously, the one image that failed is, on quick inspection, indistinguishable from those that were solved. But after a couple of additional failed attempts, I cut it loose.
     
  4. Richard Tabbutt

    Richard Tabbutt Cyanogen Customer

    Joined:
    Dec 11, 2017
    Messages:
    16
  5. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    6,273
    I think the problem is simply that you have too many stars to solve.

    I tried increasing Standard Deviation to 8 to limit the number of image stars, and the catalog magnitude range to 12 to limit the number of catalog stars. It solved quickly when I did that.
     
  6. Richard Tabbutt

    Richard Tabbutt Cyanogen Customer

    Joined:
    Dec 11, 2017
    Messages:
    16
    Thanks. I used those parameters, and the plates were solved. However, the resulting alignment of the 20 images was what I would call poor. Is there a way to extend the timeout period so that the plates can be solved using the default parameters?
     
  7. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    6,273
    It looked like a good solve... 0.5" average residual. Are you getting consistently good residuals for the other images?
     
  8. Richard Tabbutt

    Richard Tabbutt Cyanogen Customer

    Joined:
    Dec 11, 2017
    Messages:
    16
    Yes, and I am also getting good sub-arcsecond residuals. But the stack simply will not align, although it purportedly has done so. The stack alignment errors are in the multi-arcsecond range, and the resultant Combine looks a bit like a scattergram. I have used PinPoint Astrometric alignment successfully on images with many more stars. For example, the "Mexican Nebula" in the North American Nebula yielded nearly 20,000 stars per image, each of which was solved in seconds with, well, pinpoint alignment accuracy.

    My guess is that the first problem, failure to solve with the default parameters, has to do with the saturation levels of the stars. And I also guess that the second problem, failure to align, is due to the Standard Deviation multiplier being set too high (which is necessary to achieve the astrometric solution). The only solution method I've found is to apply the mild filter that I mentioned earlier. But I am quite curious as to why the PinPoint Astrometric process is failing on what I consider to be fairly good images.

    By the way, I also installed the full PinPoint engine (trial), and it yielded the same failures. I now have little hope that these images can be solved as-is, but it would be very useful to understand why they weren't solved so that this is not repeated on future imaging projects.
     
    Last edited: Oct 19, 2018
  9. Bob Denny

    Bob Denny Cyanogen Customer

    Joined:
    Oct 12, 2014
    Messages:
    746
    Location:
    DC-3 Dreams, SP, Mesa, Arizona +1 480 396 9700
    I was just alerted about this over the weekend. I will look at your data ASAP. Meanwhile it is not matter of “too many stars”. PinPoint mitigates this automatically by selecting a few hundred of the brightest stars from the image and from the catalog then solves with this subset. So even if you had 30,000 image stars and 30,000 catalog stars, the solver would be working with typically 300 of each. The actual subset size is programmable property.

    I will get back with an answer.
     
  10. Bob Denny

    Bob Denny Cyanogen Customer

    Joined:
    Oct 12, 2014
    Messages:
    746
    Location:
    DC-3 Dreams, SP, Mesa, Arizona +1 480 396 9700
    Do you have the unprocessed raw FITS? What star catalog are you using? And do you have more than a "sample of one"?

    I had no problem solving your image in Visual PinPoint with the Guide Star/1.1 (w/enhanced accuracy) catalog.

    C:\Users\Robert B. Denny\Desktop\M_42LightLuminance00002216_DarkCalibrated.fit:
    4448 image stars found.
    1161 catalog stars found.
    Solved using 260 of max 500, RMS residual is 0.59 arcsec, order =4
    Solution took 25.5 seconds
    Centerpoint RA = 05h 35m 09.836s Dec = -05d 22' 03.30"
    Pointing error 4.4 arc minutes
    WCS: Roll = 88.67 HScale = -1.385 VScale = -1.385
    PA = 271.332d
    FWHM = 6.89 arcsec
    ZeroPoint = 23.17 (1 sec.)


    And in MaxIm as well, same catalog:

    Locating stars in image
    Solving using 4448 image and 876 catalog stars
    Looking up stars in catalog
    Solving using 4448 image and 876 catalog stars
    Matched 259 of 4448 image and 876 catalog stars;
    Average residual 0.5 arcsec; order 4
    RA 05h 35m 09.8s, Dec -05° 22' 03.3"
    Pos Angle +271° 20.0', FL 565.7 mm, 1.39"/Pixel


    Smaller Search Area in MaxIm (200% vs 300%) so fewer catalog stars. Still, plenty. One issue is the star catalog. Some of them have big "holes" where star clusters, bright nebulae and other extended objects are located. In the area of these objects, there are often relatively few reference stars. Thus I'd like to know what your catalog is. The A2.0 has the biggest of all holes... huge... The GSC is not nearly as bad in that regard.

    That's as much as I can tell you at present.

    Snap1.jpg
     
  11. Richard Tabbutt

    Richard Tabbutt Cyanogen Customer

    Joined:
    Dec 11, 2017
    Messages:
    16
    I cannot achieve these results in MaxIm on any of the 20 very similar images in this group. I am using GSC-1.1 downloaded from your site. Exactly the same image scale. The solution always times out. So I am stymied to see that you have no trouble, and would love to know more about the differences in my system versus yours. Even Doug had to change the Standard Deviation multiplier to 8, and limit the catalog star magnitude to 12 to achieve a solution. But those parameters do not produce solutions that can be aligned. I do have the entire set of 20 images that I can put into my drop box and send you a link. BTW, what parameters are you using under the Advanced tab?
     
  12. Richard Tabbutt

    Richard Tabbutt Cyanogen Customer

    Joined:
    Dec 11, 2017
    Messages:
    16
    Hi Bob, and thank you much for looking into this. Good News! I decided to enter exactly the parameters you used in the solution you posted, and the one that made the difference is the Search Area (as % of image). The MaxIm DL default is 256, but the value that you used is 200. When I made this change to 200%, every image solved with over 200 stars matched in each. When I tried going back to the 256% number, all images failed to solve.
     
  13. Richard Tabbutt

    Richard Tabbutt Cyanogen Customer

    Joined:
    Dec 11, 2017
    Messages:
    16
    I just experimented with reducing the Search Area to 150%, and the images were solved in half the former time, again with over 200 stars each. They also aligned and stacked beautifully. Thanks again.
     

Share This Page