@constant1462, counting stars is indeed a better or at least a more 2020 solution. If you can do it by software then you should
You triggered me to do some tests with OpenCV. There are plenty of "object count" examples available.
Thanks for the idea.

Hello gerardvs.
So unfortunately I'm terrible at programming.
Otherwise I'll try to do something about it.
That's why I'm asking if anyone has an idea how to do it or knows a way to do it.
And I'm not very good at English.

Just did a quick test with an old image. I saw many examples but used this one to see how it does on my usually too dark images.
I think it is fairly easy to either analyse the guider images or the real images and do a prediction of cloud coverage once the star-count has dropped significantly.

To be continued.....

You can get the MLX90614 (cloud sensor part of the Weather Radio) in narrow field of view version. The standard part has a 90 degrees FOV but the MLX90614 is also available in 10 or even 5 degrees FOV.

What I am coming at is that a specialized/limited version of the Weather Radio could be built with a narrow FOV MLX90614 that would be put on the same axis and in from of the telescope tube. That could give you a good idea of the cloud cover in the FOV of the telescope, and the standard Weather Radio would give you an idea of the general cloud cover.

Ideally though, an all-sky (thermal?) camera with cloud detection/mapping over the KStars window would be preferred but... that's another story.

Hi Gerardvs
That's exactly what I thought of as an idea, so cool !!

The thing, now, it will be finished to integrate into kstars as we do for weather alerts.

But I think that on will have to add an additional camera, but that's not a problem, otherwise.

In all will I will wait impatiently to hear from you.

See you soon
Christophe

So I have the standard version of my weather station, by the way I didn't even know there was another angle.

I am not sure you can get the narrow FOV parts with evaluation boards like the Adafruit or DFRobot ones but they are available as discrete parts that you could use if you have soldering skills. Unless they are surface mounted on the evaluation boards, I am quite certain that one could replace the standard part with the narrow FOV one fairly easily.

If you are interested/curious, you can look at the Melexis datasheet for the MLX90614, www.melexis.com/en/documents/documentati...s/datasheet-mlx90614 .

I just found this interesting paper that could be relevant to what you want to do. It uses an all-sky camera (180 degree FOV) to estimate and map cloud cover for the observation of Cherenkov photons.

s3.cern.ch/inspire-prod-files-9/98f5d2c8...5cadd2963404c8ae5eec

Hello
So indeed it looks good, but I admit not having understood all because my English is not as good as before.
Does he put a simple all sky style camera or is it an IR camera?

So the most, it is if you manage to make a correspondence between what the camera detects and manage to say that in 10min it will be covered in kstars.

But I think it's more complicated to set up than a star-counting camera, right?

I think the all-sky camera is not necessarily IR, a standard camera should be sufficient.

From what I get from the paper, a simple 1/0 detection/counting of stars (detected or not) will not take into account the possibility of light clouds through which some starlight could pass attenuated and be detected, leading to think that there are no clouds, thus giving a wrong estimation of the cloud cover. It may be interesting to evaluate the usefulness of coupling an all-sky camera star counting algorithm to a small FOV MLX90614 sensor to measure the "telescope pointed" sky temperature. You may still need to account for atmospheric effects and other factors but that's making for a more interesting/challenging project.

But don't you think that with an all sky camera you might have a problem with the rotation of the stars?
Whereas with a camera that always looks at the same place we should be more precise on the detection of clouds?

If by "a camera that always looks at the same place" you mean a camera that point in the same direction as the telescope, yes that would be easier and likely more precise but you will likely require a camera FOV that would make detecting incoming clouds possible, before they get in the FOV of the capture camera It should also be independant from the capture camera so it could monitor cloud movements while a capture is going on. That could be an interesting application for a PoleMaster that would be mounted on the optical axis of the scope.

Anyway, that's all food for thought.