RANSAC: Unable to find a valid set of star pair matches.

[jerrymacon]

Some of my sets of images give RANSAC errors for all the images.  Example:  36 images, all good focus, all nicely aligned.  Typical error:
Detecting stars: done
3366 stars found.
* Limiting to 500 brightest stars.
Matching stars: done
99 putative star pair matches.
Performing RANSAC: done
*** Error: RANSAC: Unable to find a valid set of star pair matches.
Using:  DSLR Nikon D5200, 6036x4020 pixels, ISO 12800 or 25600, exposure 60 sec.  Very very dark site. No moon.
Out of 5 targets with ISO 12800, only 1 worked (perfectly), all others gave RANSAC errors on all images.
Out of 3 targets with ISO 25600, none worked, all gave RANSAC errors on all images.
Is the RANSAC code not able to cope with the high ISO?
Thanks for any help on this.

[cdesselles]

It's not just your Nikon and the high ISO.  I get the same issue all the time with a Canon T2i and ISO 1600 or even 800.  I thought it was something I was doing wrong somewhere but haven't been able to identify  the problem.  I am certainly listening with both ears to see what helpful answers you attract here concerning this issue. 

Thanks for asking the question.  If you hadn't beaten me to it, I certainly would have asked it soon myself.   

[pfile]

a good way to debug this is to use StarAlignment in "detected stars" mode and then take a look at what SA thinks the stars are. if it's picking up hot pixels or other noise, you can try changing the star detection threshold or increase the Hot Pixel removal or Noise Reduction parameters in StarAlignment.

rob

[cdesselles]

Thanks, Rob.  I'll certainly give that a try.

[jerrymacon]

Problem for me appears to be sub-quality images, aggravated by high ISO's and/or long focal lengths.
The following allowed all the ISO 12800 sets to align roughly half or more of the images in a set:
Using 'Batch PreProcessing Script', Lights tab, 'Image Registration', set Noise Reduction from <Disabled> to 1.
The 2 out of 3 ISO 25600 sets still would not align.  Tried numbers up to 10, still no luck.
These images were not nearly my best images.  They all had some combination of deficiencies in focus, tracking, light clouds.
Further evidence of this: 
My rig has a 90mm refractor at 630mm, and 200mm RC at 1624mm, both taking images at the same time.
All sets at 630mm PreProcessed perfectly.  Only had the problems with the 1624mm sets.  On other imaging nights with better quality images, similar 1624mm sets have all PreProcessed nicely.
Hope this helps.