[R-sig-Geo] Modules for using geostatistics for image classification

[Roger Bivand]

On Mon, 3 Jan 2011, Chethan S wrote:

> Hello everyone!
>
> I am using GRASS with spgrass6 for my work. I will be using variograms in
> the process of landsat image classification. I am quite ok with GRASS but am
> finding R really tough. I understand that spgrass6 is a link between GRASS
> and R which can read and write raster/vector layers. Out of really many
> packages in R, for generating variograms out of landsat images which
> packages of R can be used? I want to go for image segmentation after that -
> i.e., for further work I will be using object oriented classification.

You broke several basic list rules in this posting. You cross-posted to 
R-help - never cross-post, simply post on one list and wait for replies. 
If this doesn't satisfy you, or if so directed, post to another list. 
Cross-posting is bad because helpers on one list are probably not 
subscribed to others, so no-one can see all the replies to a thread as 
they are split among different lists. In addition, you stress that you are 
not willing to take the effort to try to find out for yourself - as a 
reply on R-help has already pointed out by suggesting a Google search.

Had you taken the trouble to visit the Spatial task view on CRAN, you 
would have found plenty of links to R packages for geostatistics. However, 
how that might fit into classification of satellite images is a completely 
different, and arguably wrong-headed question. Variograms and kriging are 
used for interpolation, not classification. So your work flow is seriously 
muddled too.

You seem to want to resample your images to a different resolution using 
geostatistical methods (why?) before classifying the new images; possibly 
you want to fill in missing pixels in your images before classification 
rather than resampling, but you don't say this. An area in which 
geostatistical methods are used in image classification is in the analysis 
of image texture, but you do not mention this as your goal. You would 
anyway find it easier to use the GRASS module r.texture instead of 
geostatics, I think.

In any case, from your spate of postings on the GRASS users list, you do 
seem to rely on other people to do your work for you, don't you? Your 
posting on GRASS-stats of 6 December is very like the above, and went 
unanswered, because you were asking for an example study to follow.

You need to design your own study (with your supervisor?), and post again 
only once you have clear and specific questions showing where the 
functions in the packages you are using are causing problems.

Roger Bivand
(list admin.)

>
> Regards,
>
> Chethan S.
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no