Has Star Detection Changed in the new Star Alignment Tool?

[Juan Conejero]

Despite this problem seems already fixed (or close to), I want to stress an important point. Whenever a user has a problem with some tool or script, posting something like "X does not work with my images" does not help too much to solve the problem, in general.

The developer's life is made much easier if he or she can see a sample of the data that is causing the problem. Without data to play with, a developer can only try to figure out what happens, which is a time consuming effort that often doesn't help to understand the problem. So the point I want to stress is:

Unless the issue is really trivial, *always* try to upload a minimal data set that reproduces the problem consistently. This will make our work much easier and more efficient.

Thank you all for your support.

[Andres.Pozo]

Understood and duly noted. 

(although I have to admit that with the previous version the solver always worked for me - but never mind, if it is only on stretched images it is minor).

I'll send you a PM with an example.

I have been working with Ioannis by PM to diagnose the problem.
The latest version of ImageSolver  has a small bug in the initialization of the parameter "Align Algorithm". It always shows the value "Triangles" when the configuration window starts although the selected algorithm was "Polygons".

The images of Ioannis are mirrored so he needs to use the "Triangles" algorithm but he had selected "Polygons" without knowing it.

I already have a fix for this but until it is published you can use a workaround:

• The log shows the algorithm used just before the StarAlignment process. It should say "Using Triangle similarity algorithm".
• If the log says "Using Polygon algorithm with N sides", you have to rerun the algorithm, change the value of "Align Algorithm" to "Polygons" and then back to "Triangles".

[Sean]

Juan,

I've uploaded my series of Star Alignment images to the Endor server. You can see them at http://endor.uv.es/files/data/public/135b44.php. I've also uploaded a screenshot of my StarAlignment settings.

I just reprocessed them this morning with the latest download of StarAlignment in Ripley RC7, Mac OSX Mountain Lion. Of the 14 images, 5 aligned successfully, images R007, R008, R010 (the reference), R013, and R015. Of the 9 that failed, most had an "Unable to find an initial set of putative star pair matches" message, but several had Unknown Exception or Singular Matrix errors. All 14 successfully aligned with the older version of StarAlignment.

Again, I know that these are really marginal images for alignment, and I realize the issues in trying to reliably align them. However, if you have any suggestions for me, they of course are welcome.

Thanks,

Sean

[mschuster]

Sean,

I took the liberty of downloading your images and trying them with both StarAlignment and my scripts.

With StarAlignment I see similar problems on Win7. Multiple PCL Win32 System Exceptions seen with Pentagons, Triangle Similarity and Triangles options.

With my scripts, StarDetector appears to be detecting randomly positioned, small structures that may not be in fact stars.

This is a guess I could be wrong!

Mike

[Sean]

Mike,

Thanks for taking a look. As you saw, there's not much in the way of star structure in these images. I'm trying to add a little Ha to M101 - here's a first try with the aligned images from the prior version of StarAlignment merged in:

http://www.drakevisual.com/galaxies/h6fe5e0a1#h6fe5e0a1

There are only a few smaller regions of Ha, but it does add something to the overall galaxy, so I think I'll keep trying. I may grab some 20 minute Ha tonight while the moon is full, and see if the longer exposure helps. I'm amazed that you were able to work with 20s Ha images!

BTW, Pinpoint inside of ACP also failed to plate solve these images.

Sean

Sean

[Juan Conejero]

Sean, Mike,

Thank you Sean for uploading the data. The problem is with the extremely low SNR of these images. The new star detection algorithm is not well suited for these cases.

Of course, doing worse than previous versions is unacceptable, so this has to be fixed. Good news: I have a new version of the star detector that can work with these images, so StarAlignment registers them without problems. I'll release an update later today (including also a new StarDetector.jsh).

[Sean]

Juan,

Thank you very much for looking into this - great support as always.

I look forward to the new version.

Sean

[Juan Conejero]

Hi Sean,

Thank you for your support. I have just released an update to StarAlignment with a new parameter: Noise reduction. You'll find it in the Star Detection section. By default this parameter is zero. If you set it to 15 all of your Ha images can be registered without problems.

I have also fixed the bug that was causing access violations when less than 8 stars were detected.

[mschuster]

Hi Juan,
Are you planning on adding Noise Reduction to the .jsh? I don't see an update.
Thanks,
Mike

Thank you for upper limit.

[Sean]

Juan,

I just successfully aligned the 14 Ha images, plus another 15 from another night. Everything worked very smoothly with the new Noise Reduction parameter, and quickly - the new StarAlignment is very fast. The Integrated aligned image looks very good.

Thanks for your help with this.

Sean

[Physicist13]

Joan, can you elaborate as to what noise reduction is doing? Thanks!
Patrick

[Sean]

Here's a screenshot from the mouse-over help.

Hope this helps.

Sean

[Physicist13]

Sean, thanks!!...I should have looked!!...Does a bit curious to call it noise reduction though?...more like Gaussian smearing I think it was the name that was confusing me as to what was going on
Patrick