I run into problems with live stacking. Some details of my setup:
Lens/Telescope: DIY refractor, 50mm f/3.6, poor optics taken from a low cost Celestron Binocular
Mount: Simple EQ mount with motorized RA axis
Camera: Raspberry Pi HQ Camera, where the raw format is a jpg+RAW
Computer: Raspberry Pi 3B+, handling image capture and live stacking
As the Pi 3B+ is not powerful and that the RAW image format from Pi HQ camera need processing to become a DNG, I restrict to PNG format with 2x2 binning with resolution 2028x1540. The image file size is about 2-3M. The pixel size is 1.55um.
With rather high light pollution and no CLS/UHC filters, I limited the exposure time to 4s. Discounting those subs with clouds, I manage only stack less than 20% of the subs. I take out 2 sample images here (img01 and img02) for analysis. I test the quads pattern and found they are different, so cannot stack. I guess most other images share this issue.
I cannot figure out what can be the cause: cromative abbrevation? tracking error? hot pixels? Would like to get some advices. Attached files for reference.
I conduct more study today and get even more puzzled. I input an extreme condition in the alignment tab to limit the max number of stars for tetrahedron detection to 30, hoping to force the engine to use the brightest stars. To my surprise, the quad diagram ignores almost all the bright stars but use the dimmer ones. In the log it does day "selecting the 30 brighest stars only"! So apparently those bright stars in the image are not considered as stars by the program. I believe it might be related to the fact that these brightest stars are not round so they are being ignored? Attached images for reference. Having such limitation in my equipment, is there a way in the software parameter setting that can be helpful?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
If the star HFD is larger then maybe 10, ASTAP could ignore the stars. Looking to you images, the stars have a pretty large halo making the star big for the detection. So I suppose ASTAP ignore them because they are bigger then the detection window of 25x25 pixels.
The best thing to do it to bin more for alignment. Try to force bin 2x or 3x. See attached screenshot. Keep the star detection on 500 stars max.
Thanks a lot. I tried 2x and 3x binning per your advice using the images I captured last time. For 2x the success rate is 89% and for 3x it is 70%. It works! Really great.
With binning it seems that the image lost the color information, the Bayer pattern is wiped out? Any method to get the color back or they are lost permantly? Frankly for live stacking color is not so important. Attached image is the resultant jpg for 2x binning.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
More info. The last post was related to a simulated live stacking on my PC using bin down 2x. I tried to do the simulation again in the Raspberry Pi using bin down 2x with the same sets of image. The success rate is even better slightly over 90%! And the color is "preserved"? The asinh stretch factor in Raspberry Pi is 4, while last time on the PC is was 10. Anyhow this is a great step forward. Thanks.
Binning is only for alignment. The stack will not be binned!
The colour will be preserved if you have the full resolution image. Binning prior to stacking in 2x2 and save to PNG (I restrict to PNG format with 2x2 binning ) will kill any colour information. So for colour binning 1x1 is required but most likely not possible with your RaspberryPI3. But in case you use a more powerful computer you can force debayer/demosaic. See attached. screenshot This all uses CPU power. Even my RPI4 will struggle.
You image show big halos most likely due to colour failure. This doesn't help star detection. So 90% success is a plus.
Thanks for the info. I believe I will live with the Pi 3B for a while.
Last night the sky was quite stable and clear and I tested in the field. I did live stack for 3 DSOs, each 40x4s exposures. All 120 frames stacked successfully in the Pi, 100%. ASTAP is great.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I run into problems with live stacking. Some details of my setup:
Lens/Telescope: DIY refractor, 50mm f/3.6, poor optics taken from a low cost Celestron Binocular
Mount: Simple EQ mount with motorized RA axis
Camera: Raspberry Pi HQ Camera, where the raw format is a jpg+RAW
Computer: Raspberry Pi 3B+, handling image capture and live stacking
As the Pi 3B+ is not powerful and that the RAW image format from Pi HQ camera need processing to become a DNG, I restrict to PNG format with 2x2 binning with resolution 2028x1540. The image file size is about 2-3M. The pixel size is 1.55um.
With rather high light pollution and no CLS/UHC filters, I limited the exposure time to 4s. Discounting those subs with clouds, I manage only stack less than 20% of the subs. I take out 2 sample images here (img01 and img02) for analysis. I test the quads pattern and found they are different, so cannot stack. I guess most other images share this issue.
I cannot figure out what can be the cause: cromative abbrevation? tracking error? hot pixels? Would like to get some advices. Attached files for reference.
I conduct more study today and get even more puzzled. I input an extreme condition in the alignment tab to limit the max number of stars for tetrahedron detection to 30, hoping to force the engine to use the brightest stars. To my surprise, the quad diagram ignores almost all the bright stars but use the dimmer ones. In the log it does day "selecting the 30 brighest stars only"! So apparently those bright stars in the image are not considered as stars by the program. I believe it might be related to the fact that these brightest stars are not round so they are being ignored? Attached images for reference. Having such limitation in my equipment, is there a way in the software parameter setting that can be helpful?
The attachements
If the star HFD is larger then maybe 10, ASTAP could ignore the stars. Looking to you images, the stars have a pretty large halo making the star big for the detection. So I suppose ASTAP ignore them because they are bigger then the detection window of 25x25 pixels.
The best thing to do it to bin more for alignment. Try to force bin 2x or 3x. See attached screenshot. Keep the star detection on 500 stars max.
Han
Thanks a lot. I tried 2x and 3x binning per your advice using the images I captured last time. For 2x the success rate is 89% and for 3x it is 70%. It works! Really great.
With binning it seems that the image lost the color information, the Bayer pattern is wiped out? Any method to get the color back or they are lost permantly? Frankly for live stacking color is not so important. Attached image is the resultant jpg for 2x binning.
Missing attachment from last post.
More info. The last post was related to a simulated live stacking on my PC using bin down 2x. I tried to do the simulation again in the Raspberry Pi using bin down 2x with the same sets of image. The success rate is even better slightly over 90%! And the color is "preserved"? The asinh stretch factor in Raspberry Pi is 4, while last time on the PC is was 10. Anyhow this is a great step forward. Thanks.
Binning is only for alignment. The stack will not be binned!
The colour will be preserved if you have the full resolution image. Binning prior to stacking in 2x2 and save to PNG (I restrict to PNG format with 2x2 binning ) will kill any colour information. So for colour binning 1x1 is required but most likely not possible with your RaspberryPI3. But in case you use a more powerful computer you can force debayer/demosaic. See attached. screenshot This all uses CPU power. Even my RPI4 will struggle.
You image show big halos most likely due to colour failure. This doesn't help star detection. So 90% success is a plus.
Han
Thanks for the info. I believe I will live with the Pi 3B for a while.
Last night the sky was quite stable and clear and I tested in the field. I did live stack for 3 DSOs, each 40x4s exposures. All 120 frames stacked successfully in the Pi, 100%. ASTAP is great.