| Re: yet another 2d matching question [message #72039 is a reply to message #71975] |
Tue, 03 August 2010 13:53  |
JDS
Messages: 94
Registered: March 2009
|
Member |
|
|
On Jul 31, 7:47 am, Gray <grayliketheco...@gmail.com> wrote:
> On Jul 30, 6:23 pm, JD Smith <jdtsmith.nos...@yahoo.com> wrote:
>
>> Paulo spotted the issue. What determines whether a given point in the
>> search list "is not matched to a closer point"? Your 1-to-1 match
>> will be sensitive to the input ordering of the target list. The
>> intention of match_radius is to specify the maximum separation beneath
>> which all matches are "equally good". For example, the statistical
>> uncertainty in the position itself. Multiple matches would then imply
>> either is an equally good match. If you still wanted to do this (for
>> example if you are conducting a match for which sub-match_distance
>> separations are still meaningful), it will have to be a pre- or post-
>> processing step, since all matches are performed in parallel (which is
>> what gives MATCH_2D its speed).
>
>> JD
>
> Hmm... if all matches are equally good within the match_distance, then
> how does match_2d prioritize matches when there is more than one
> source in list b within the match radius of list a? This could happen
> when, for example, the positional accuracy of the sources in each list
> is low, but there is a possible shift (translation+rotation+etc.)
> between the members of the two lists which necessitates a larger match
> radius.
This is only true if your match_distance represents some positional
uncertainty; i.e. it's not meaningful to say a given star is 50
milliarseconds closer when the precision with which you know your
search list is 2 arcseconds. It does simply return the closest point
within match_distance if there are multiple matches. You could
certainly alter this to return *all* matches within match_radius, then
use post-processing to enforce a 1-to-1 matching. IDL 8's new LIST
type would make this much easier than before (when I've used
REVERSE_INDEX style arrays for the same purpose).
JD
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
