Learning astrometry for making cloud monitor

[Dhruv Paranjpye]

Dear all,

I am an Electronics and Telecommunication engineering undergraduate student and I want to build a night sky cloud monitor
which would be useful at astronomical observatories. For doing so, my plan is as follows:

1) Take periodic images of the sky (wide field) using a digital camera mounted on a microcontroller controlled mount.

2) Perform automatic astrometry on the images so that if the number of stars is less than a threshold, we can infer it is
cloudy in that area of the sky.

3) Trigger an alarm/message to the users indicating that it is cloudy. (We assume that there was no moisture on the lens when the image doesn't show any star.

I specifically need help in (2) because I am new to performing astrometry and I need help in learning it. Would someone please help me with how to start off?

[Dustin Lang]

Hi,

Is this a project for a class, or just for fun?

Why do you need to do astrometry?  You're just counting stars, right?  Astrometry means figuring out where on the sky the camera is pointed.

By the way, many large observatories have an infrared cloud camera -- for example, at Cerro Tololo there is RASICAM:
https://www.youtube.com/watch?v=9bMXew_hSwU
Clouds glow in the infrared, so they're very easy to see.

Have fun,
--dustin

[Hans]

Hi,

I tried to build just that, using astrometry.net's image2xy to count stars. I have it running but it is not reliable at all as lots of cloud edges also get counted as stars.
So I need to improve star detection and would welcome ideas !

I use this ugly oneliner to count stars from my raspberry pi allsky camera :

cat 20161004_222135-rpi3.jpg |jpegtopnm 2>/dev/null|ppmtopgm|convert -crop 1750x1944 - - |convert - crop.fits && image2xy -O -g 9 crop-0.fits|awk '{ print $3 }'
54

where I also cut off the right side of the image as some rolloff roof is visible there. The whole thing runs in a loop that also controls exposure length and the counted stars number is sent to rrdtool from which I make plots like attached.


-- Hans

[Dustin Lang]

(replying to a DM),

Okay, so you want to know what stars are supposed to be visible so that you can compare your star count to the number you would expect in that position.

You can either run the astrometry.net code on your own machine -- you'll use the 'solve-field' program;
http://astrometry.net/doc/readme.html

Or there is a web service -- nova.astrometry.net (also with an API).

One thing though: we usually don't do well on very wide-field images; search this forum for details.

cheers,
--dustin

[Dustin Lang]

Two more thoughts:

- starting with trying to create something like Hans has done, counting stars, would be a good start

- if you mount the camera pointing straight up (not on a mount), then once you've figured out the Dec center at a given time, you can then compute the pointing at any future time.

cheers,
--dustin