No, it’s not a racy post. It’s a play on words some funny programmer chose for the name of this StarExtractor. Actually, it’s called sextractor without the space, but I digress.
This represents a big step forward in my pursuit of a clear shot of the night sky. The process that I have worked out requires automatic alignment of each shot from my nocturnal photo-shoots, but the algorithms I have been using keep failing. My suspicion is that they fail because they are looking for features and the stars are just a bunch of points – featureless.
This, however, changes everything. This is a line drawn among the fifty brightest stars in the central area of the image and I was able to produce it automatically with the help of sextractor, which identified and classified the stars in the image. These lines will hopefully be enough of a set of features for the other algorithms to match.
Onward and forward!
the afterburner 
Almost there!
Would the null-space/dark sky within each “point” triad suffice as necessary features?
LikeLike
Probably so, and others have done similar work with these triads. My problem so far has been detecting them. In other words, I didn’t have any way yet to algorithmically find the points where the stars lie.
sextractor is awesome because it does sub-pixel estimation for the center, size, and magnitude, even rotation (for galaxies) of the stars in the image. I can do practically anything I want now with it.
Thanks for the question. I’m thinking of trying to create a Voronoi diagram based off of the results,which could be a piece of art in its own!
LikeLike
I’d like to see that diagram, and some of the funky transforms you apply.
You may find nVidia’s CUDA project helpful in the long run for optimizing processing time, matrices, and particle sims/analyses (and neural networks, in case you want side-side-side project)
LikeLike
Here’s a taste: http://www.inf.ufpr.br/vri/alumni/2013-PeterPerroni/GPU-PSF-Estimation-with-CPSO-dissertation.pdf
So much research, so little easy-for-the-consumer-to-use software.
LikeLike