I have been trying to roughly polygonize vegetation areas from Imagery (RGB). I tried using Iso Cluster Unsupervised Classification (Spatial Analyst) but it is still too detailed for my purpose. I just need to roughly draw polygons wherever I have dense vegetation as can be seen here:enter image description here

Any idea/ thought to put me on the right track is much appreciated.


Vegetation extraction is a bit more complex than running the spatial analysis tools that you named. For better results I would suggest the following:

  • run analysis on a 4 band image (e.g. R,G,B,NIR)
  • change image to be symbolized as 432 for RGB not 321
  • create training samples that represent vegetation and run a supervised classification

These steps will give you a fair result, however you will still get many false returns (e.g. shadows by buildings, or green water bodies...etc).

For best results you will have to use a remote sensing software (including the three bullets above) that analyzes both spectral and textual objects of the imagery. I asked a question about what free software exists that can do this type of analysis, see link below:

Land Cover Feature Extraction from Satellite Imagery

  • Say you wanted to digitize buildings in a similar way? Any thoughts on how to do that? – GISKid Nov 1 '13 at 19:34
  • @GISKid, yes same idea but different training set. – artwork21 Nov 1 '13 at 19:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.