Ran the first cuda parallel code on the tegra stack last night. Impressive. I wish Mathematica was on those boards. The articles I read for the image processing are these.

http://blog.wolfram.com/2014/08/14/fixing-bad-astrophotography-ii-imaging-mars-with-mathematica/ http://blog.wolfram.com/2010/12/27/fixing-bad-astrophotography-using-mathematica-8-and-advanced-image-deconvolution/

You will have to play around with it a bit to get it working. I'll have to do it again myself as I've managed to loose what I had working. I keep forgetting to do a shutdown on the pi's and ruining the SD image. Now I'm trying to do it with OpenCV which is much easier for me to get my head around. It has a function where you click on a star and it will cut out what ever size you want. That is going to be more useful for what I want to do.

Henrik I've taken your idea and ran with it. Put together a cluster of Jetson Tegra K1 dev boards and I'm using the cuFFT library that comes with it to do the FFT's. Its quick. On a single board a 1024x1024 convolution takes 22ms. On a 5in square board that happily runs on a 12v battery. The OpenCV lib provides all the mechanisms to get the fits file into the matrix for the cuFFT routines to operate on. Eventually I'll use one of the Tegras on my scope and one or more on a vision robot I'm working on. Actually giving my mobility scooter the ability to self drive is the goal. I started with a cluster of Pi 2 boards but the Tegras are much better for math. The Raspi cluster is happily running Univ of Hamburgs climate model now but will eventually have a Hadoop cluster running on it. So thanks for that idea its turning out to be an excellent way to get this accomplished.

Dan

Thanks for the info. I missed these in the deluge of daily e-mail. I'll have to try it out.

Dan

Dan, I am not sure if you have seen this, but there is a nice blog post by Tom Sherlock, perhaps interesting to you:

Serial Interface Control of Astronomical Telescopes

Here is one way:

In[1]:= (* make some data *)
data = Table[
   rSquared = x^2 + y^2;
   3. Exp[-rSquared/5^2],
   {x, -10, 10}, {y, -10, 10}
   ];

In[2]:= (* generate an {x,y,z} list *)
indexedData = 
  Flatten[Table[{x, y, data[[x, y]]}, {x, Length[data]}, {y, 
     Length[data[[1]]]}], 1];

In[3]:= (* fit to Gaussian to recover parameters *)
FindFit[indexedData, 
 i0 Exp[-((x - x0)^2 + (y - y0)^2)/a^2], {i0, x0, y0, a}, {x, y}]

Out[3]= {i0 -> 3., x0 -> 11., y0 -> 11., a -> 5.}