Cloud Sensor

[SilverAstro]

A low light camera (may not have to be hugely sensitive ?) looking at the sky sees dark patches and light or white patches. We convert it to a bit map (if it is a jpg) and count the pixels that are above a threshold and count those that are below, if the below are greater it sounds an alarm ! ( or maybe we dont need to count all the lights, - if the dark pixels are greater than x% of the total then alarm. )**

In another topic (Clear Sky, IR or Thermal) I was contemplating IR/temperature sensing as many others have done but very few appear as completed working projects and no one has popped up to say " I got one working it's brill " !! so is probably a blind ally and maybe IR etc is overthinking the means.

@Gina says (of her All Sky Cam) that dark patches always have stars in them, and cloudy patches are always light. If I am not misunderstanding her, I am supposing that we need not actually detect the stars, just the dark bits,  so perhaps an el-cheapo camera might be sufficient unto the task ! ?

Could it be as simple as this, presumably not else it would be in use all over the interweb, so what am I overlooking, I dont have the gear yet ( might have to go get something ! ) to look for the pitfalls but some of you do and have already tried this ?  and can tell me where I have gone wrong. ?

Any of you keeping an archive of your skycam pics ? could test this theory, save me getting all excited ?

** having more ideas as I type ! Maybe just a histogram package from ImageMagick or similar could do it ??

Thanks for reading / thinking

 

 

[Gina]

I have a number of ASC video recordings in AVI format hence they are huge - several GB.  When we get ultrafast fibre broadband in this area - which might be early next year depending on getting enough people interested and prepared to sign up for a 12 month contract - I plan to provide my full definition DSO images and ASC video recordings on my web site but ATM it would take all night and maybe part of the next day to upload each video or high res image.

[SilverAstro]

4 minutes ago, Gina said:

it would take all night and maybe part of the next day to upload each video or high res image.

Eeek yes, but no need to do that, , , look through a selection of frames to see if a majority of those with histogram peaks to the left coincide with clear(-ish?) sky and viceversa ? Or if there are any false positives where left histo shows lots of dark but NO! stars .

As a starters !   If it looks promising then we can volunteer one of our code experts to write us a routine to ring a bell  

 

[ncjunk]

 

I had a cloud sensor working with the MLX90614, i think, IR sensor and it does work very very. You need a decent rain sensor with inbuilt heater as well and this can provide the heat to keep the dew off of it.

My sensor died because the test case wasn´t waterproof and the electronics got wet. I moved to the AAG cloud watcher as I don´t have time to redo the homemade one or integrate it into ACP and the AAG does it all for 300 euros and was effectively the same hardware that i bought for mine.

The IR sensor is very easy and you are only looking for the temp difference between ambient temperature reading, that the MLX provides with onboard temp sensor, and the sky temp. A normal clear temperature reading varies depending on atmospheric conditions and time of year. So, in the winter the IR sensor can be reading -30 with ambient of 5 giving a difference in temperature of -35. When cloud comes over then the sky temp will rise to above 0 giving a difference of -5 but you'd be able to detect cloud, dust or changes in conditions before that.

At this time of year, summer here in spain, the atmospheric conditions are worse and the sky temp reading can be anywhere from -1 to -12 deg for clear with an ambient temp of 24 deg C.

If you have the time and enjoy playing then the IR sensor is really quite easy but you do have to tweek the alarm trigger points for Clear, Cloudy and partially cloudy depending on the time of year and conditions.

I'll usually check the readings before the start of the night and change the limits as required.

you could always program the Arduino to replicate the comms and replies from a commercially available unit and use their code. I know of a couple of people that use their homemade units with their own software but I don't think you are finding lots of people with homemade one's because of models like the AAG that are "only" 350 euros which in our hobby is relatively cheap compared to everything else.

If you enjoy tinkering then I recommend doing an IR sensor as it's quite interesting, it would be easier than the camera method you're describing here. My gut feeling is that this camera method will work to a certain degree but won't be as sensitive as the IR sensor and more of a challenge to implement - you never know though until you try.

 

Regards,

 

Neil C

 

[skybadger]

The peltier sensor approach is used at some national observatories, essentially measuring sky temp successfully. 

The visual inspection method using a camera also ought to be quite easy but requires a less than dark sky, because for a given exposure and temperature, you need to measure the sky brightness over the frame and that will also vary by time of night. 

Hence the alternative approach that doesn't rely on sky brightness is to count stars and measure holes where there aren't any. Detecting star psf's is relatively easy. Detecting holes might not be. A low number of stars indicates cloud. The base number of stars you might expect would depend on camera and size of field and local zenith. But if you can reach 6th mag, a 10 degree field will be big enough, otherwise you need to go to 30 degrees or so to get enough star coverage in sparse areas of sky.

What about laser scanning the sky, or networking into the local forecast observations sites and polling for a conclusion?

M

[ringz]

Just expect someone to call if you are shining infrared lasers into the air and the police have been flying about looking for a suspect with heat sensing goggles

 

[SilverAstro]

2 hours ago, ncjunk said:

I had a cloud sensor working with the MLX90614, / inbuilt heater as well and this can provide the heat to keep the dew off of it.

AAG cloud watcher 

300 euros

you never know though until you try.

Thanks for all your thoughts, very interesting, I will obviously have to think again/some more about the IR sensor then    You are right, very wise words > "you never know though until you try. "

I have used a hand-held unit ( mech. engineer friend demoed one when they were rel.new,  think he was using it for engine hot-spot tests) and it was quite surprising, of course one knows the theory about outer space being cold and mother earth/clouds being cosy and warm but there is nothing quite like seeing it in action to see the truth ! One of my other hair-brained thoughts is to mount such a device outside with a solenoid or servo pretending to be my thumb, have not yet figured out how to get the reading back indoors --

-- oh, till now, ding,  a Pi and a NoIR maybe , hmmmm,,,, wouldnt do for the wake-up in the night tho'

I was about to ask for link, or go google, for AAG,ACP but then you said 300 eeek! not sure if any Perseid is worth that !

[SilverAstro]

51 minutes ago, skybadger said:

is to count stars and measure holes where there aren't any. Detecting star psf's

What about laser scanning the sky, or networking into the local forecast observations sites and polling for a conclusion?

psf ?

Interesting thoughts thanks, the only reason I was contemplating ignoring stars was thinking about cheap cameras like the NoIR etc. may be good for dark/light patches ? but not for stars. A limited field of view would not be a prob at my site ! I am surrounded quite close by trees !

I would pref. not to think of lasers - you know where that will take the thread - oh ho look out already

[skybadger]

PSF = point spread function, the shape of the intensity  distribution of photons around a star centre on a detector.

I did build the robot electronics scanning thermal imagery array for building an all sky picture with 5degree pixel resolution. Seemed the right size to start worrying about rain to me. The issue I found was pattern noise caused poor images but probably down to my code rather than anything else. Try their website for more info.

M

[SilverAstro]

11 minutes ago, skybadger said:

PSF = point spread function, the shape of the intensity  distribution of photons around a star centre on a detector.

I did build the robot electronics scanning thermal imagery array for building an all sky picture with 5degree pixel resolution. Seemed the right size to start worrying about rain to me. The issue I found was pattern noise caused poor images but probably down to my code rather than anything else. Try their website for more info.

M

Ah yes, thanks, I got as far as star, profile and function and then gave up !

The ? robot electronics scanning thermal imagery array

EDIT Ah ! this one ? https://www.robot-electronics.co.uk/htm/tpa81tech.htm

[skybadger]

http://www.robot-electronics.co.uk/tpa81.html

[Wayne_Ford]

Have a look at  Tektite Skies

http://www.mcdougalltech.com/page1/page1.html

 

[SilverAstro]

6 hours ago, Wayne_Ford said:

Have a look at  Tektite Skies

http://www.mcdougalltech.com/page1/page1.html

Thanks, nice find, looks interesting