×

INDI Library v2.0.6 is Released (02 Feb 2024)

Bi-monthly release with minor bug fixes and improvements

Speeding up plate solving on RPi 3

  • Posts: 174
  • Thank you received: 27
It currently takes about 30s to plate solve a full size image 3500x2500 on my RPi3. It seems the most time is taken by image prep whereas plate solving itself is very fast. I am using sextractor because it makes solving more reliable. Was anyone able to speed up solving on RPi? How did you do this?

Alex
5 years 5 months ago #30222

Please Log in or Create an account to join the conversation.

Take a 4x4 image?
5 years 5 months ago #30225

Please Log in or Create an account to join the conversation.

  • Posts: 174
  • Thank you received: 27
It is for reference images. I normally image same objects several nights so I use previous night's image as a reference.
5 years 5 months ago #30228

Please Log in or Create an account to join the conversation.

  • Posts: 20
  • Thank you received: 9

Creating a flag in kstars is very useful, then use those RA/DEC values (copy/paste) in your scheduler. I do this to go back night after night, when the weather allows.

Are you offline plate solving? I find it is quicker and I can't really tell the difference in accuracy. Maybe other's have, but in my opinion it works well.
5 years 5 months ago #30231

Please Log in or Create an account to join the conversation.

  • Posts: 6
  • Thank you received: 0
Are you using --downsample 2 or 4 ?
If you know roughly where you are, you can use the --ra and --dec flags. Otherwise --scale-units --scale-low --scale-high help
I think I am down to 20s on a Rpi3. I have not used sextractor yet.
5 years 5 months ago #30690

Please Log in or Create an account to join the conversation.

  • Posts: 1029
  • Thank you received: 301
If the image is already solved, you could patch the astrometry driver to use WCS information directly?
Most of the preparatory work is spent in loading the indexes. There is an option in astrometry.cfg to load those indexes in parallel, but it requires much memory.
There was a branch of the solver discussing the possibility to keep the indexes in ram back in 2017, but I don't know the status, and in any case it would require too much memory for the Raspberry.
You'd better be moving the solving process to a stronger computer. Well, in any case 30s is still fast compared to other procedures.

-Eric
5 years 5 months ago #30707

Please Log in or Create an account to join the conversation.

Time to create page: 0.477 seconds