I'm wanting to run an unsupervised classification of multiple geospatial layers within Q-GIS or other open source software. I have both vector and raster information: airborne radiometric, DEM, polygonal maps of soil and geology.
I was hoping to undertake Fuzzy-c means type classification by combining radiometric data (pixel), DEM and specific attributes from regional soil and geological survey.
I am hoping to find a simple interface for this analysis, one that also supports appropriate pre-processing.
Any recommendations much appreciated.
Clint
Do a search for "random forest", in particular the work of Leo Breiman. You can find free software available either for Python or for R. There are links for tutorials and examples
Your best shot for QGIS is still the SCP(Semi Automatic Classification Plugin) I think.
Dear Clinton,
Although not a simple interface, but you can do the entire workflow in R statistical software. For rasters, use package 'raster', for vectors, use packages 'sp' or 'sf'. For import/export of spatial data (both vectors and rasters): package 'rgdal'. Package 'ppclust' can do the Fuzzy C-Means classification:
https://cran.r-project.org/web/packages/ppclust/vignettes/fcm.html
As far as I understand, you have spatial data from different sources (and of different types and at different resolution). Hence, you may use functions extract(), focal(), over(), spatial_sync_raster(), rasterize(), aggregate() and disaggregate(). Packages 'spatial.tools' and 'rgeos' might be needed.
More information about overlaying different spatial geometries:
https://cran.r-project.org/web/packages/sp/vignettes/over.pdf
More information about using R for geoinformatics/GIS:
https://cran.r-project.org/web/views/Spatial.html
R and its packages are free, and if you code your workflow into a script, it'll be easy to reuse, modify and document it.
HTH,
Ákos
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running