Using Landsat images how do I create my own indices to extract particular landuse? Are there any tool/ software for creating indices?


I would start by manually labeling the pixels in an entire scene that are vs are not part of your desired landuse. Then you could try various predictive model formalizations to build models that predict the marked pixels, keeping the model that gives the best results. (There is an entire field of techniques called "model selection" that helps do this efficiently.)

Logistic regression using various combinations of bands, or a decision tree classifier trained over all bands, are examples of models that would provide an appropriate level of expressibility for an index.

| improve this answer | |
  • Thanks Jeff, I am new to this remote sensing field. What is "model selection"? Where can I find it? – Dinushika Nov 20 '14 at 16:08
  • @Dinushika model selection is merely trying different models (in your case, different candidate indexes) and keeping the best one. In your case "best" is defined according to predicting results of your labeled landuse data. As you are a beginner the easiest approach to model/index selection is probably simple trial and error. – Jeff G Nov 22 '14 at 3:24
  • Oh, now I got it.Thanks for your help to find an answer. – Dinushika Nov 23 '14 at 10:15

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.