Re: 13 "new" objects/matching and overlap

[Chris Albertson <chris@topdog.pas1.logicon.com>]

This message bounced back.  Here is goes once more.  Sorry if you
see it twice.
-------------------

aah@nofs.navy.mil wrote:
> 
> Rereading Chris' message, I see why he only gets 6 hits/star, since Glenn
> was the third data source and his cameras are pointed south.  I thought
> on first reading that Mike G.'s system was the third source.
>   However, I'm a little surprised that he only gets 6 hits even so, since
> there are always a few cases where crowded fields will give more than one
> object within a reasonable match radius.  I would have expected a handful
> of objects with >6 observations.  Also, is there really no overlap at all
> between Glenn and Tom/Nick?
> Arne


The software will not match two objects that appear in the same frame
no matter how close they are.  Also if two frames overlap only the first
detected sighting is kept.  The actual algoritum is:

1) Get observation
2) get all catalog entries within a 15 arc second radius
3) Case 1: No catalog entires found --> make new entry
        2: Exactly one catalog entry --> a match
        3: More than one entry --> pick closest one as match.
4) Check for special cases (above) and reject match if found.

It took something like 40 hours of compute time to merge the
7.5 million TASS observations with the 2.5 million TASSM16 
catalog. 1/2 of this time is just because Postgres is slow.

==========

I don't know where the cameras are aimed but looking at a sample
of data from I band, with 0 < ra < 1.0 deg I see these limits.

 select min(dec)::float/3600000 as min_dec,
        max(dec)::float/3600000 as max_dec, 
        site_id
      from observ_3v01 
      where ra < 3600000 
      group by site_id;

          min_dec|           max_dec|site_id
-----------------+------------------+-------
-4.95119972222222|-0.483899722222222|B      
-1.10669972222222|            1.9178|G      
-1.57999972222222|            1.3322|H      
(3 rows)

So yes there is a small band where all three overlap.  Looking
at more then one degree of data may open this up a little but
much.

-- 
   --Chris Albertson             home: chrisja@jps.net        
     Redondo Beach, California   work: chris@topdog.logicon.com
-- 
--Chris Albertson

  chris@topdog.logicon.com                Voice:  626-351-0089  X127
  Logicon RDA, Pasadena California          Fax:  626-351-0699