Since the maximum class membership probabilities are equal, I would think
that both results are "correct" (probabilistically speaking). But why does
the classifier truncate the rule images to one decimal place? And why don't
they all sum to unity?
-Mort
Dr. M. Canty
Forschungszentrum Juelich
D-52425 Juelich, Germany
Tel. (0049) (0)2461 614885
Fax. (0049)(0)2461 612496
----- Original Message -----
From: "Jonathan Greenberg" <greenberg@ucdavis.edu>
To: "ENVI Mailing List" <starserv-envi@ucdavis.edu>; <ENVI-L@envi-sw.com>
Sent: Thursday, April 29, 2004 9:55 PM
Subject: Question about classification (and ENVI implementation)
> Recently, I was running the rule classifier on some probability images and
> realized, much to my chagrin, that ENVI has a behavior which I thought is
> incorrect, and I was wondering what is the "real" answer to this
situation.
> It deals with when you have ties in the max probability for two or more
> classes. For example:
>
> Class1,Class2,Class3
> .4,.2,.5
> 0,0,0
> .2,.4,.4
> .4,.4,.4
>
> ENVI will produce the following class map:
> Class3
> Class1
> Class2
> Class1
> -- in the case of a tie, ENVI defaults to the lowest class number. It
> turns out, this behavior IS documented, but this doesn't mean the behavior
> is legitimate. It seems to me that the "correct" output should be:
> Class3
> Class0
> Class0
> Class0
>
> -> where class0 is unclassified. I suppose my question is a) from a
> classification standpoint, which of these two outputs is "legit", b) how
do
> I get envi to produce the 2nd output (what I think is "correct") without
> significant programming, and c) how many of you realized ENVI was doing
> this?
>
> --j
>
> --
> Jonathan Greenberg
> Graduate Group in Ecology, U.C. Davis
> http://www.cstars.ucdavis.edu/~jongreen
> http://www.cstars.ucdavis.edu
> AIM: jgrn307 or jgrn3007
> MSN: jgrn307@msn.com or jgrn3007@msn.com
>
>
This archive was generated by hypermail 2b29 : Thu Apr 29 2004 - 23:59:42 PDT