53 replies to this topic

[UB_Astro]

Please keep discussion (or a copy) here.

Thanks!

Yes, it's great to see and learn from experts discussing issues.

[artik]

Beta6.

 

Per @han.k request added support of additional controls related to cooler. 

And in general added more controls (like AE and AWB)

 

https://github.com/a...eases/tag/beta6

[han.k]

I will test the new version tomorrow.

 

We had some brief clear moments:

 

1 image of 5 seconds stacked:

 

9 images of 5 seconds stacked:

 

25 images of 5 seconds stacked:

 

 

Solving

 

 

 

 

 

 

 

 

 

[han.k]

Below the stack of 116 images of 5 seconds exposure time. This using my S7 android phone directly connected to the ASI1600MM. Telescope a 100 mm APO. I stretched the resulting tiff a little in ASTAP and compressed it heavily to get under the 500 kb limit. Moon was shining and the sky was not very clear but is worked fine.  No darks or flats where used. Cooling of the ASI1600MM was not functional. 

 

Before the stacking starts, you have to be a little patient but it started this time reliable. Stream view didn't work but that is already reported as an issue.

 

So it is a very nice app with a lot of potential

 

a) Two minor problems. When unlocking the phone, sometimes the stack view is lost. The red buttons are still there but the stack view is missing.

b) Satellites tracks are becoming a problem. So having each images saved (on SD) for stacking later using sigma clip stacking is desirable

 

Han

 

The compressed version:

 

The original stack image you can download as new-image.fits from here:

https://nova.astrome...783277#original

 

Han

Edited by han.k, 26 April 2023 - 05:55 PM.

[artik]

Two minor problems. When unlocking the phone, sometimes the stack view is lost. The red buttons are still there but the stack view is missing.

 

 

It is likely that browser went into sleep mode and disconnected from live stacking video stream. Just refresh the browser and video will be restored.

 

(I still need to make sure I load various values like pixel size etc)

 

Satellites tracks are becoming a problem. So having each images saved (on SD) for stacking later using sigma clip stacking is desirable

 

Can you please send me the raw data recorded?

 

I have a simple and efficient idea to remove satellite tracks in live stack.

 

Basically for each stacked pixel after registration calculate sum minus max

 

Basically remove max pixel value for N images, thus if there is a frame with bright satellite it should be removed.

 

It should work very efficiently but I need data to validate the approach.

 

 

I think my next things to add based on feature requests with high priority:

 

- Save data to SD instead of standard location (It is clear how useful it can be)

- Add alternative location for ASTAP data (so you can move to SD, from SD etc)

- Restore values (like pixel size and other form settings)

- Add satellite removing (pending data)

- Bugs/issues that may come from cooling/temperature support

- I'll probably reduce minimal video framerate (currently around 10FPS) since it can be highly problematic for huge frames.

 

Currently pending for next + 1:

 

The presets - i.e. change camera settings by preset selection

 

Correct me if I missed something important

Edited by artik, 27 April 2023 - 12:53 AM.

[han.k]

Here are the files for 30 days. This site is a little tricky with a lot of advertisement but after a few tries you can free download the file phone.7z (7zip):

 

https://ufile.io/tthb1b7s

 

 

Some files are 8 bytes because I run out of disk space. No app error was given.

 

 

And here the stacked files:

 

https://ufile.io/sascyw1q

Edited by han.k, 27 April 2023 - 04:45 AM.

[han.k]

Removing satellite tracks will be difficult with live stacking.

 

About stacking algorithm

 

So what your doing is as follows:

 

Assuming the images are A,B,C,D...

sum:=A

result:=A/1

 

sum:=sum+B

result:=sum/2

 

sum:=sum+C

result:=sum/3

.....

 

 

 

For ASTAP live stacking I apply what I call serial stacking:

 

Assuming the images are A,B,C,D... then serial stacking goes as follows:

 

result:=A
result:=(result+B )/2
result:=(result*2+C)/3
result:=(result*3+D)/4
result:=(result*4+E)/5
....

 

Maybe this is a little more efficient. In ASTAP I keep the images as floating point value arrays (4 bytes=single per pixel). You can't do this with integer values.

Edited by han.k, 27 April 2023 - 05:20 AM.

[artik]

Ok I did the satellite removal based on your data. Worked like a charm :-)

 

 

Based on your data cropped centre 1/4 of the image (I only little bit adjusted stretch parameters).

 

BTW amazing camera and optical setup. And this is without cooling!

 

In ASTAP I keep the images as floating point value arrays (4 bytes=single per pixel). You can't do this with integer values.

 

I use floating point as well. Much simpler and yet very efficient.

 

If you stack using 16bit integers you start loosing accuracy for initial frames quite fast. I think better optimisation may be using half float maybe on GPU. But this is future.

 

In any case for registration I'm using FFT based phase correlation algorithm that is very fast. So I need these floats.

[han.k]

Tested the beta6 version. Cooling works! Two things:

 

1) Cooling power is a "get".  This one is not reported. This one is important because if you try to cool too deep it goes to 100% cooling but the efficiency becomes less and temperature rises due to the electrical power put in. Normally you can cool down about -35 degrees below ambient max.

2) I would move the measured temperature below the temperature setpoint.

 

Having these value in the TIFF description would be nice (Astro-Tiff)

 

 

 

[artik]

Cooling power is a "get".

 

Ooops. This kind of stuff you need to know from experience. I have my ASI224MC for about 3 month and obviouslt never had cooled camera.

 

I'll fix it as well.

 

I would move the measured temperature below the temperature setpoint.

 

I report them in order ASI gives them. But I assume I can move some values back and forth

 

Having these value in the TIFF description would be nice (Astro-Tiff)

 

 

In to do list (you opened a ticket)

[artik]

BTW can you open the UI in different browser - chrome or edge I want to see if it is related to browser compatibility or to screen dpi/resolution.

 

It looks horrible. 

[han.k]

Ok I did the satellite removal based on your data. Worked like a charm :-)

 

 

BTW amazing camera and optical setup. And this is without cooling!

 

 

In any case for registration I'm using FFT based phase correlation algorithm that is very fast. So I need these floats.

Your satellite removal algorithm works good. It just removes (or ignores) the brightest pixel from the series. I will try to add it to ASTAP . The problem is that is will add extra step in the stacking which will add processing time. But I can make it optional. The problem I see it adds processing time. The image download time and processing time can take a great portion of the total time. So maybe in practise the effective exposure time is only 50% of the total time. I haven't looked to it but it will be interesting to see what the image cycle time is using 5 seconds exposure time. Something you could report in Openlivestacker.

 

The telescope is quad APO with a build-in flattener.

 

FFT is interesting. I assume it only compensates for shift and not for rotation? Rotation correction will be required if mount polar alignment is not good. In ASTAP the star detections are used for alignment. This will work even for rotation and meridian flip.  Note that the quality of the stack could be measured by testing the median HFD value of the stars. Good alignment will give the lowest median HFD or FWHM value.

Edited by han.k, 27 April 2023 - 08:52 AM.

[han.k]

Wikipedia tell me that FFT can also be used to compensate for rotation.

[artik]

Your satellite removal algorithm works good. It just removes (or ignores) the brightest pixel from the series. I will try to add it to ASTAP .

 

 

The problem is that is will add extra step in the stacking which will add processing time. But I can make it optional. The problem I see it adds processing time.

 

 

I made it optional. About processing time - I think it is relative small portion in comparison to FFT, iFFT, auto-WB, auto-stretch etc. I did quick measurements it added ~30% - but I assume I'll get better result when I'll use intrinsic as usual instead of naive OpenCV function - most of time critical code I run using vector register operations.

 

In any case I still have quite lots of room for optimization of performance:

 

(a) put post processing and registration+stacking in different threads

(b) do actual registration in several parallel threads

© do proper mono - stacking (I currently convert mono to RGB which need to go at some point)

 

And BTW honestly I liked image with trails more - it gives some kind of authenticity to live stack 

 

FFT is interesting. I assume it only compensates for shift and not for rotation? Rotation correction will be required if mount polar alignment is not good. In ASTAP the star detections are used for alignment. This will work even for rotation and meridian flip.

 

Yes in FFT I do shift only and for Alt-AZ I compute derotation and it works very well.

 

Even with poor polar alignment that isn't best the rotation error should be smaller by order of magnitude in comparison to shift. 

 

Finally it is tool for Live-Stacking/EAA - so in general user is not expected to run in for hours like in AP. But performance is critical. I don't know how good ASTAP's performance but when I did some stacking with Siril the registration speed on PC was irrelevant for any live stacking on handheld device and even PC. Also frequently registration errors were significant.

 

Consider that I also aim to support cameras like sv105 has have 0.5s max exposure or guys who do non-tracking EAA. So frame-rate is fast. Currently OpenLiveStacker handles sv105 with 0.5s exposure (frame-rate is around 1/0.8 actually) and probably 0.25s quite well.

 

Another point I found (at least with Siril) that FFT gave much more stable results (even with sub-pixel accuracy) in comparison to star registration especially on poor quality images given by sv105 (which is BTW quite Ok for bright objects like globular clusters).

 

Once I see rotation issues I may add some algorithm for small angle error correction (maybe Michal Irani's algorithm). But I'll see how it goes. Feature after Feature 

 

 

Wikipedia tell me that FFT can also be used to compensate for rotation.

Yes. For this you do what is called log-polar transform and convert rotation to shift problem.

I tried it first before went for manual derotation compensation and the results were poor at least on low quality images of M31 from SVBony sv105.

 

Bottom line currently it seems that what I have Ok. And I'll add features as needed.

[artik]

Ok yet another improvements and bug fixes release: https://github.com/a...eases/tag/beta7

 

• Added support of true mono stacking instead of converting mono-to-rgb, note you need to rebuild your calibration frames
• Plate solving parameters pixel size, focal length and the object are now saved and recovered per restart
• UI fixes related to width of the field - adaptations for small and narrow screens
• Fixed cooler power to be readonly
• Added support of satellite removal in live stack - by removing brightest pixel in stack
• Various performance improvements
• Added option to move application to SD card does not work

 

 

1) Cooling power is a "get". 

 

 

@han.k

 

Can you please test following:

 

- UI fixes related to width of the field - adaptations for small and narrow screens - how it looks now on your browser?
- Fixed cooler power to be readonly - check if it fixed - I don't have cooled camera

- I also did code rewriting to make stacking Mono internally if input it mono instead of previous converting mono to RGB. It should vastly improve the performance. But be aware that you'll need to take calibration frames once again.''

Edited by artik, 29 April 2023 - 09:58 AM.

[artik]

One more I fund a cool stuff I can run two apps visible on Android one above other - keyword Android split screen,

 

You open app manager (the left button usually then press at the top on the title and select split screen)

 

 

This is OpenLiveStacker and SynScan - so I can control the mount via wifi and see live video

Edited by artik, 29 April 2023 - 10:05 AM.

[han.k]

Probably it is better it is better to put latitude and longitude on two rows. Not one.

 

 

Same for pixel size and focal length:

 

[han.k]

The cool capacity is working.

 

The three SET values are better together.

 

A timed loop to read automatic get the GET values each second would be nicer. Users could have problems to understand you have to click on GET to refresh the get values.  I would remove the GET's and make getting them fully automatic with an interval of one second.

 

The check mark for cooler is working. but the "off" text is confusing.

 

 

The default setpoint for cooling is 0. You have to remove the zero before you can enter a minus value. A default of -15C would be better. 

 

 

Default for exposure I would set 5000 ms / 5 sec.

 

 

 

 

Edited by han.k, 29 April 2023 - 05:53 PM.

[han.k]

This is the result of this night after 100 x 5 seconds cooled with ASI1600, cooled at -15C and Moon at 69%

 

The 100 separate images:

https://ufile.io/24qf3c5r

 

The result unmodified:

 

Edited by han.k, 29 April 2023 - 05:21 PM.

[han.k]

After start the stack is too low and you have to scroll too find it.

 

 

Sometimes after unlocking the stack is gone. After some shifting it is suddenly there

[artik]

Probably it is better it is better to put latitude and longitude on two rows. Not one.

 

Generally I optimise for landscape layout since most cameras have wide frames. In such a case it is much better to have them in one row. But I want to make sure it is usable in portrait mode and narrow screen. So free text like flow arranges it automatically.

 

A timed loop to read automatic get the GET values each second would be nicer. Users could have problems to understand you have to click on GET to refresh the get values.  I would remove the GET's and make getting them fully automatic with an interval of one second.

 

I got your point. I can probably make an option to update values every few seconds when settings are open.

 

The check mark for cooler is working. but the "off" text is confusing.

 

I see... Maybe just not show default for on/off options. It indeed can be confusing even for WB and AE.

 

The default setpoint for cooling is 0. You have to remove the zero before you can enter a minus value. A default of -15C would be better.

Default for exposure I would set 5000 ms / 5 sec.

 

 

Defaults are given by SDK/camera and this what I show.

 

On my 224MC camera AFAIR auto exposure is on by default. On sv105 it is also on. And when you start camera you do want short exposures - for example for aligning you want move camera fast - have high gain short exposure to align on a star.

 

So I don't want to change defaults.

 

I want to create presets so user can set his own values according to his/her needs

 

The three SET values are better together.

 

Good point. I think I need to "sort them" differently. I'll see what I can do.

 

After start the stack is too low and you have to scroll too find it.

 

Can you please check it with Chrome or Edge browsers? Just close Samsung browser and open 127.0.0.1:8080 in Chrome or Edge.

 

Because I don't see such a behaviour with Samsung on Android 10 or Chrome or Edge. This kind of issue is hard to debug because there is no Samsung browser in simulator and built-in WebView works fine as well even on Android 7 simulator.

Edited by artik, 30 April 2023 - 07:50 AM.

[han.k]

Note the solvers for Android X86 and X86_64 are now available.

 

Landscape mode does improve the arrangement. It is not so important but I would lock the text description and the corresponding edit box together. Maybe in a groupbox. Then when the screen width is too short they scroll to the same row making it more readable. Now the edit box scroll but the text not. They should be locked together.

 

Landscape:

 

I'm pretty sure for the Samsung browser in stream mode the image is moved sideways every few seconds. I managed to capture it after the first move. After the second move it disappears.

 

This shows that for the Samsung browser the image is placed random each time?

 

For the Chrome browser the image is fixed at one place.

 

 

But for the Samsung browser you can zoom out. The Chrome browser can't.

 

Chrome browser left, Samsung browser right.

 

Chrome browser left, Samsung browser right.

 

The Samsung browser image positioning problem does not happen for the smaller M44 simulation image. So the problem happens only for large images like 2328x1760 pixels

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

[artik]

I would lock the text description and the corresponding edit box together.

 

Got you. Fixed.

 

The Samsung browser image positioning problem does not happen for the smaller M44 simulation image. So the problem happens only for large images like 2328x1760 pixels

 

Interesting. I'll try different image size 

[artik]

Ok I released 8th beta: https://github.com/a...eases/tag/beta8

 

• Fixed warning of app is build for old Android version
  
  Now this is big change because I changed target build SDK and need to change quite lot of things in permissions policies and other stuff Android changes/breaks each and every time between releases.
   
• Added support of framerate limiting to prevent overload in case of observing bright objects like Sun or Moon - I noticed problems when trying to run with small exposures (for Sun for example) I had the app overloaded easily. So I added limiter so you can prevent app overloading by dropping frames.
• Fixed: label and input box aren't wrapped separably
• Updated ASTAP is included and now build for all supported platforms: x86, x86_64, arm64-v8a, armeabi-v7a thanks to @han.k!

 

I played a bit with SyncScan app and found the best way to use them both

 

You open synscan in Window mode and it is very useful - especially since you can minify one and then bring it back.

 

Here is the screenshot of some capturing of Live Moon video

[han.k]

Installed v8 and had brief look to it. The arrangement of the inputs and text looks good.