[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Minimum spanning circles and Voronoi diagrams




> -----Original Message-----
> From: Chris Albertson [mailto:chrisalbertson90278@yahoo.com]
> 

You are saying just use the stddev number for the set?  Yeah, but how do you
intelligently combine both the ra and dec error bars into one number?  This
why I thought of the MEC, as using the radius to determine the amount of
scatter.  And, if there is a significant difference between the circle
center and the mean position, then there must be outliers.

> I think you want to simply define a minimum angle threshold
> based on the error bars on the astrometry.  15 arcseconds
> worked for the Mk III.

Actually, I am.  But, there may end up being lots of points, and I like
doing unusual stuff in code.  Besides, with 100 points, you have 4950
computations.  By computing the convex hull, I get small set of points.  My
ulterior motive is learning.  Which is why I joined the Tass group in the
first place :-)

> 
> Near the equator why not just use SQRT(ra^2 + dec^2) with
> ra and dec in degrees to compute the distance.  It's fast.
> You can cut down the search by first applying a simple box
> function
> 

I'll have to take a look at this.  Thankfully, there is an SLA perl module
Astro::SLA, so this would fit nicely with the rest of my stuff.  Thanks for
the pointer.

> 
> Or you could simply link to a good library like "SLALIB"
> and call the "DESEP" function and get an debugged, exact
> answer.  This is in FORTRAN but I have a C wrapper library
> that makes using it easy from C.  Works for me on both
> Linux and Solaris.
> 
> http://www.starlink.rl.ac.uk/star/docs/sun67.htx/sun67.html#xref_
> 
> the SLA library was recommended to me by people on this list.
> I agree now.  You really need this if you are going to do
> anything connected with astronomical coordinates.  It is the
> "standard" library that is in wide use.  This problem is
> not hard but others are
> For example, pretty soon you may want to match catalog data
> that is in a different referance system or you need to precess
> the data to a diffent epoc.  
>