Hi Nina,
The help for ERRMAT says:
In essence, ERRMAT is identical to CROSSTAB with the exception that no
tabulations are made for cells marked with a zero on the ground truth
map. Output also includes column and row marginal totals, errors of
omission and comission, an overall error measure, confidence intervals
for that figure, and a Kappa Index of Agreement (KIA), both for all
classes and on a per category basis.
>From this I assume that CROSSTAB actually includes the background pixels
in the calculation, which would explain why you get a higher KIA
(obviously all the background pixels agree)... the help for CROSSTAB
does not mention the background, but I remember reading in other
documentation that not all Idrisi modules recognize the background value
and ignore it. I guess the only way to know for sure is to try to run
the two modules on a subset of the image where you have no background
pixels, or to assign another value to the background and compare (ERRMAT
apparently will discard zeros, not backgrounds, but in your case they
are the same).
Nina Snyman wrote:
>
> --- Start of Idrisi List message
> I have done a land use classification from a LANDSAT TM satellite image,
> with two different techniques: a supervised classification and an
> unsupervised classification method.
>
> I have extracted the area of interest (which is a river's catchment),
> and gave the "background value" a value of zero. Both sets of land use
> classes were reclassed to one set of common land uses classes.
>
> I have compared the two classifications by means of CROSSTAB, which
> calculated the Kappa Index of Aggreement as 0.68 for the areas inside
> the river's catchment. I also used ERRMAT to compare same the two land
> use images with each other, giving the one as the classified image, and
> the other as the ground truth image. Here the Kappa Index of Aggreement
> were calculated as 0.40. Why the big difference in Kappa Index of
> Agreement values?
>
> Thanks
> Nina
>
> --
> Nina Snyman
> Hydrological Research Unit, University of Zululand
> PO Box 13 Tel. +27 35 4600655
> Nkwaleni Fax +27 35 4600634
> 3816
> South Africa
>
> --- End of Idrisi List message
>
> IDRISI-L is an unmoderated list. As a service to list users, please post a
> summary of useful responses to your questions.
> To post a message to IDRISI-L, send it to [EMAIL PROTECTED] To
> subscribe to the list, send an email from the address to be subscribed to
> [EMAIL PROTECTED] with "subscribe" (without quotes) as the first line
> of the message body. To unsubscribe, do the same, only with "unsubscribe"
> as the first line of the message body.
--
Marco Albani - PhD Candidate, Quantitative Landscape Ecology
Department of Forest Sciences, University of British Columbia
3041 - 2424 Main Mall - Vancouver, BC V6T 1Z4 Canada
Phone: (604) 822 8295 Fax: (604) 822 9102
