| Re: Clustering x,y coordinates with IDL? [message #70311 is a reply to message #70220] |
Mon, 29 March 2010 13:20   |
Maxwell Peck
Messages: 61
Registered: February 2010
|
Member |
|
|
On Mar 30, 2:38 am, Mort Canty <m.ca...@fz-juelich.de> wrote:
> Am 29.03.2010 11:44, schrieb Maxwell Peck:
>
>
>
>> Hi All, This is not strictly an IDL question but I'm hoping it's
>> something simple to do and I'm not missing something painfully
>> obvious!
>
>> I have a set of x,y values e.g.
>
>> x = [34.5,36.7,35.6,80.5,81.2,79.3]
>> y = [50.6,51.2,53.2,48.5,50.3,51.2]
>
>> The values will be sparsely clustered overall but tightly within an
>> approximate 10x10 box. What I'd like to do is replace each clusters
>> values with the average x,y value. This doesn't have to be 'perfect',
>> the edges aren't that important.
>
>> My initial thought was to use hist_2d or hist_nd to try and calculate
>> the 2d histogram with a binsize of 10 and use this to average the
>> values but I can't get it to work. My other thought was using griddata
>> or something similar, or at worst generating an approximate binary
>> image and running a window over it.
>
>> Any advice or suggestions would be appreciated.
>
>> Regards
>
>> Max
>
> The IDL function CLUSTER() implements k-means clustering. I think that's
> what you want.
>
> Mort
Thanks Mort. I had seen this I am concerned though about how it will
deal with single/a few points as opposed to real 'clusters'. Also the
distance with which points are considered to be a cluster as well is
concerning. I will give it a bash anyway and see how it looks. I had
thought there must be a nice way to do it with histogram but I just
can't get it to work.
Regards
Nax
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
