ASI178MC Plate Solve too many stars

[graems]

When using SharpCap to call the ASTAP plate solver with my ASI178MC camera, it usually fails with the number of stars reported in the 500 range (realistically it should be about 50). If I just capture a FITs file and feed that to ASTAP, ASTAP reports a more realistic value of roughly 50 stars. If I change the camera format from RAW16 to RGB24 and save to a FITs file and then feed that to ASTAP, I can reproduce the 500 star range. This makes me wonder if there is an issue with the way Sharpcap converts the raw data to the PNG format for ASTAP? The color camera has a 14 bit A/D.

Win10, SharpCap 4.0.9538.0-2023-01-23, ASTAP 2023.05.09, 64bit with D80 database

Graem

[admin]

Hi Graem,

yes, I am starting to suspect too that Astap is only reading 8 bits of data from PNG files, even when SharpCap writes them with 16 bit depth images. There is another parallel thread running on here that was heading towards the same conclusion.

I imagine that with the lower bit depth data, Astap is starting to see noise in the pixel data as stars, leading to a lot of 'fake' stars causing the plate solve to fail.

I will be digging into this to see if I can confirm and fix as required.

cheers,

Robin

[graems]

Hi Robin: I did some investigating. SharpCap was set to Raw16. I put the lens cap on my camera and told SharpCap to do a plate solve with ASTAP. I also did a manual save to FITs. I captured the 2 files and did some comparing. Note these captures are at 2 slightly different times (I couldn't get SharpCap to copy the same image to PNG and FITs at the same time). What it shows is that the hot pixels in the FITs files (which appear as single pixels) are bled across multiple pixels in the PNG file. I think this causes ASTAP to get confused as to what are stars and what are hot pixels.

I repeated this test with my B/W camera (ASI432MM) and the PNG file does not show any bleeding. This would explain why my B/W cameras are working for plate solving and my colour camera is not. I would attach the corresponding files, but they are over the 1MiB limit.

Graem

[admin]

Hi Graem,

interesting... The spreading of the pixels is because the image being sent to the plate solving enging is being debayered from a RAW image to colour. This is necessary because otherwise the grid pattern of the raw image can upset some plate solving applications.

I have (as an experiment) set up the latest SharpCap 4.1 beta to send FITS files to Astap rather than PNG, but if the difference is down to the debayering process then that will probably not help, since I am still debayering the image before saving it as FITS.

There are a couple of options in Astap that might help deal with this, including the 'Ignore stars less than' option and possibly the option to downsample the image. It would be worth playing with them.

Feel free to email me the images if they are too large to attach - you can use the 'Contact Us' link at the bottom of this page.

thanks,

Robin

[graems]

Hi Robin - I did some quick testing of ASTAP plate solving with SharpCap 4.0.9538 64bit vs SharpCap 4.1.10745 64 bit. I'm still using the ASI178MC camera with a finder scope of focal length of 222mm. Exposure set to 1.9 seconds.

Using the old version of SharpCap and no binning, ASTAP would report about 400 stars and would always fail plate solving. If I used the "-z 2" option, ASTAP would report about 90 stars and would solve about 3/4 of the time.

Using the new version of SharpCap and no binning, ASTAP would report about 50 stars and solve about half of the time. If I used the "-z 2" option, ASTAP would again report about 50 stars and would solve all of the time.

So in conclusion, the new version of SharpCap cuts down the number of fake stars that ASTAP reports. I still need to use binning to get the best ASTAP performance. I assume that 2x2 binning is almost like converting a bayer matrix to mono (since the 2 green, 1 blue and 1 red pixels are getting converted into 1 pixel?).

Graem

[han59]

Hi Robin, Hi Graem,

The bleeding effect I made visible here:

beta to send FITS files to Astap rather than PNG, but if the difference is down to the debayering process then that will probably not help, since I am still debayering the image before saving it as FITS.

The bleeding transforms a single hot pixel to a 3x3 pixel star makes the exclusion of hot pixels very difficult. Sending either a 16 PNG or 16 bit FITS would not make a difference but avoiding debayer would definitely do.

The order of preference for solving a OSC camera image is the following:

1) Bin 2x2 if the original dimensions are at least 2000x2000 pixels
2) Raw image bin 1x1
3) Debayered image.

It would be very beneficial if debayering could be avoided for solving

Han

[admin]

Hi Han,

thanks for confirming. I have recently made a change in SC 4.1 Beta that will send a RAW bin 1x1 image to Astap instead of a debayered image. This change is in SharpCap 4.1.10736 and later versions. I have not done anything with 2x2 binning in SharpCap at this stage, but I have - I think - noticed that Astap will do the downsampling on its own under some conditions - SharpCap sends -z 0 on the command line, which I think triggers this.

PS. Is there a way to send a XY star list to Astap to plate solve? I have wondered if that might be useful in some cases.

cheers,

Robin