[R-sig-Geo] Image Classification - Wish List?

[Ned Horning]

Eric,

I strongly recommend "image segmentation algorithms and tools" even if 
it's not necessarily a classification algorithm. Image segmentation and 
tools to derive features from the segments are, for the most part, 
lacking in open source software. There are some segmentation tools out 
there (Spring, OTB/Monteverdi) but they have limitations that make them 
less than ideal. Access to more classification algorithms is less of an 
issue especially with the wealth of options available in R, a reasonable 
selection available in GRASS, and the minimal gain in classification 
performance realized when using "new and improved" classification 
algorithms.

I don't have much to offer on the programming side but am happy to help 
with development ideas and testing/feedback for segmentation or image 
classification algorithms.

Ned

On 04/04/2012 08:16 AM, Eric Momsen wrote:
> Hello all,
>
> I'm looking for some feedback and ideas regarding the need to improve
> image classification abilities in GRASS GIS.  My understanding is that
> since the required classification techniques aren't available in
> GRASS, the researchers doing that work are found elsewhere.  Hopefully
> this is a good place to find those experts?
>
> For those that aren't already aware of it, Google Summer of Code
> provides support for students to spend the summer programming for Open
> Source software projects.  The GRASS GIS idea list included the
> following items that caught my eye:
>
> 4. Implement image segmentation algorithms and tools
> 5. Implement region-based classification
> 6. Implement hierarchical classification tools (e.g. being able to
> create a large class "forest", with subclasses of different types of
> forests)
>
> The methods currently implemented in GRASS are at:
> http://grass.osgeo.org/wiki/Image_classification
>
> > From discussions on the grass-dev mail list, I understand that
> some/many of the new algorithms to meet these needs are implemented in
> R and other software.
>
> I'm wondering:
>
> What tools/workflow are used for this type of work?
> Are the current methods OK, or would some integration in GRASS be better?
> Are any algorithms not available in other software that couldshould be
> implemented in GRASS?
> Are any experts in image classification willing to mentor or provide
> informal advice about the subject matter this summer?  (There is one
> volunteer to mentor from the software side, it would be nice to have
> someone to answer questions about the classification techniques.)
>
> Thanks for any and all input!
> Eric
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo