[GRASSLIST:183] Re: imagery classification tools for classification of terrian types

[Dylan Beaudette]

Markus,

Thanks for the tips. I have been able to get these two commands that you 
suggested to work in the past... but this time no luck:

i.gensigset trainingmap=terrain.train group=terrain subgroup=terrain 
signaturefile=terrain.sig

Finding training classes ... 100%
3 classes
Reading image data ... 100%
Clustering class 1, with 103499 pixels
Warning: Removed a singular subsignature; number 1; 9 remain
Warning: Removed a singular subsignature; number 1; 8 remain
Warning: Removed a singular subsignature; number 1; 7 remain
Warning: Removed a singular subsignature; number 1; 6 remain
Warning: Removed a singular subsignature; number 1; 5 remain
Warning: Removed a singular subsignature; number 1; 4 remain
Warning: Removed a singular subsignature; number 1; 3 remain
Warning: Removed a singular subsignature; number 1; 2 remain
Warning: Removed a singular subsignature; number 1; 1 remain

Error: unreliable clustering
try a smaller initial number of clusters
Warning: Removed a singular subsignature; number 1; -1 remain

Error: unreliable clustering
try a smaller initial number of clusters
Subclasses = 0; Rissanen = 0.000000;  Solution: Number of subclasses is 0
Clustering class 2, with 732 pixels
Warning: Removed a singular subsignature; number 1; 9 remain
Warning: Removed a singular subsignature; number 1; 8 remain
Warning: Removed a singular subsignature; number 1; 7 remain
Warning: Removed a singular subsignature; number 1; 6 remain
Warning: Removed a singular subsignature; number 1; 5 remain
Warning: Removed a singular subsignature; number 1; 4 remain
Warning: Removed a singular subsignature; number 1; 3 remain
Warning: Removed a singular subsignature; number 1; 2 remain
Warning: Removed a singular subsignature; number 1; 1 remain

Error: unreliable clustering
try a smaller initial number of clusters
Warning: Removed a singular subsignature; number 1; -1 remain

Error: unreliable clustering
try a smaller initial number of clusters
Subclasses = 0; Rissanen = 0.000000;  Solution: Number of subclasses is 0
Clustering class 3, with 45545 pixels
Warning: Removed a singular subsignature; number 1; 9 remain
Warning: Removed a singular subsignature; number 1; 8 remain
Warning: Removed a singular subsignature; number 1; 7 remain
Warning: Removed a singular subsignature; number 1; 6 remain
Warning: Removed a singular subsignature; number 1; 5 remain
Warning: Removed a singular subsignature; number 1; 4 remain
Warning: Removed a singular subsignature; number 1; 3 remain
Warning: Removed a singular subsignature; number 1; 2 remain
Warning: Removed a singular subsignature; number 1; 1 remain

Error: unreliable clustering
try a smaller initial number of clusters
Warning: Removed a singular subsignature; number 1; -1 remain

Error: unreliable clustering
try a smaller initial number of clusters
Subclasses = 0; Rissanen = 0.000000;  Solution: Number of subclasses is 0


.. could this be a result of my training data being too coarse ?


this results in a segfault in i.smap ....

thoughts?

Dylan


On Thursday 16 March 2006 12:43 pm, Markus Neteler wrote:
> Dylan,
>
> try
>  i.gensigset
>  i.smap
>
> works FAR better IMHO...
> The training areas may not be too small (few pixels).
>
> good luck
>
>  Markus
>
> On Thu, Mar 16, 2006 at 12:29:49PM -0800, Dylan Beaudette wrote:
> > Hi everyone,
> >
> > Attempting to do some simple geomorphologic classification based on
> > parameters calculated from a DEM.
> >
> > I have tried a couple permutations of i.cluster i.gensig, and i.maxlik
> > without success here are the commands that I used in my tests:
> >
> > #1. training data based classification
> >
> > i.group group=terrain subgroup=terrain
> > input=slope_n,er9,profc,maxic,minic,longc
> >
> > r.digit  --->terrain.train
> >
> > i.gensig trainingmap=terrain.train group=terrain subgroup=terrain
> > signaturefile=terrain.sig
> >
> > #results in:
> > Finding training classes ...
> >  100%
> > 3 classes.
> > Calculating class means ...
> >  100%
> > Calculating class covariance matrices...
> >  100%
> > Signature [1] not invertible.
> > Signature [2] not invertible.
> > Signature [3] not invertible.
> > Writing signature file [terrain.sig]
> >
> > #i.maxlik dies with:
> > ERROR: Can't read signature file [terrain.sig].
> >
> >
> > 2. unsupervised classification:
> > i.cluster group=terrain subgroup=terrain sigfile=terrain.sig classes=5
> >
> > #results in:
> > Iteration 1: % Convergence: 100.00 (0s elapsed, 0s left)
> >
> > i.maxlik group=terrain subgroup=terrain sigfile=terrain.sig
> > class=terrain.class
> >
> > #results in a map with all cells set to 1
> >
> > I am using GRASS6.1-CVS and all of my input maps are FCELL or DCELL ....
> > could this be the cause of my problems?
> >
> >
> > As a side note, this little adventure was started after reading a past
> > paper by Roger Bivand on the use of R for a simple cluster-based approach
> > to classifying terrain [1]. In his paper he goes over the aspects of
> > getting data to move between GRASS and R, leaving the interpretation of
> > the results to a geomorphologist. The R-approach seemed ideal, however I
> > thought that I would try it with the GRASS classification commands first.
> >
> > [1] Roger Bivand. Integrating GRASS 5.0 and R: GIS and modern statistics
> > for data analysis. Internal Paper (?)
> >
> > Thanks!
> >
> >
> >
> > --
> > Dylan Beaudette
> > Soils and Biogeochemistry Graduate Group
> > University of California at Davis
> > 530.754.7341

-- 
Dylan Beaudette
Soils and Biogeochemistry Graduate Group
University of California at Davis
530.754.7341