[AstroPy] Alt-Az Splitting a Fisheye Image

Mohammad Shameoni Niaei m.shemuni at gmail.com
Thu Dec 12 03:47:56 EST 2019


Hello the Astropy community

I have a project to detect cloud coverage from ground-based data (All-Sky
camera data).
I'm trying to train a classifier using SVM or CNN.

The problem:
I want to train the machine for parts of the sky. And for this, I want to
chop the image data into pieces. These pieces can be polygons(This part is
not the problemn) or Alt-Az(This is the problem) areas. And I want to use
the Numpy masks (The problem can be solved using a nested loop. But that
would be inefficient I think).
[image: part.png]
At-Az mask example(What I want to achieve.* Image created using gimp*)

I can create a full radius mask which I call pizza mask.
[image: pizza.png]
Full radius (Pizza) mask example

The plan is to have 3 pizza masks for different radii and apply logical
comparison to achieve At-Az mask. (I'm open to better ideas. With this
method I'll need to calculate 3 pizza masks and it takes time.)

Here comes the catch: I need to know what is the linear radius of the given
altitude angle.(Sounds like *someone* didn't attend to
trigonometry/spherical astronomy courses)
The fisheye projection looks like I need to use Sky2Pix_ZenithalEqualArea.
But how?

-- 
Mohammad SHAMEONI NIAEI
Astronomer / Software Specialist
Astrofizik Araştırma ve Uygulama Merkezi (ATASAM),
Yakutiye, ERZURUM/TÜRKİYE
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/astropy/attachments/20191212/ca53aacc/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: part.png
Type: image/png
Size: 141879 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/astropy/attachments/20191212/ca53aacc/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pizza.png
Type: image/png
Size: 129012 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/astropy/attachments/20191212/ca53aacc/attachment-0003.png>


More information about the AstroPy mailing list