LANDSAT Image classification and open source solution(s)

Question

I have some experience with ERDAS IMAGINE in doing some supervised/unsupervised LANDSAT imagery classifications and now find myself in the position of needing to do the same but without the access to ERDAS IMAGINE. I've been looking at the GRASS literature and found this here and see that it has similar functionality/tools such as i.class/cluster/gensig as well as PCA and Tassled Cap analysis, which also hold some interest to me. I plan on looking at SAGA as well when I get the chance to see what it can do in respect to classification. I also found this thread as well here but much of what is there dates back a few years.

My question is: are there other Open Source alternatives similar to GRASS or SAGA that I should look into as well that you may have used? Just doing my homework before I commit to something...and I am really only interested in the image classification abilities, or lack thereof. I will be using LANDSAT data for an AOI of 700 sq/mi and would like to utilize the ASTER spectral library. The class number would probably be fairly low as it is not a very diversified landscape. Operating system is Windows 7, 32bit(!) and no budget to speak of. Time frame will be mid-summer, full leaf on.

Answer 1

Orfeo Toolbox is a good option, it now works inside QGIS via the Processing toolbox. It's available via OSGeo4W.

http://orfeo-toolbox.org/otb/otbqgis.html

https://trac.osgeo.org/osgeo4w/wiki/PackageListing

http://www.orfeo-toolbox.org/SoftwareGuide/SoftwareGuidech2.html#x14-210002.1.1

Answer 2

Grass is very powerful, you just have to get used to the location / dataset organization. You can't go wrong using a mix of Grass, qgis and orfeo.

There is also a brazilian software very good called Spring, but it takes some time to get used to the database model it uses.