X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 20 Nov 2013 02:00 PM by  anon
Rule Image Classifier
 0 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:
New Member


--
20 Nov 2013 02:00 PM
    Background: I'm working in ENVI Classic with Landsat 8 data corrected to reflectance with FLAASH. I created a stack including 7 Landsat OLI bands plus an NDVI image derived from the same scene. I subsetted it and then ran Maximum Likelihood classification and outputted the rule image too. I then tried to run Post Classification > Rule Image Classifier. I have negative values in the Rule Images for each of the 6 classes I created. Questions: 1) Are negative values okay to have in rule images? Help files seem to say that Rule images from Max Likelihood classifications shouldn't have negative values, rather they should have probabilities? 2) Can you enter these same negative values in the 'Threshold' values in the Rule Image Classifier? For example, I left three of the threshold values blank, but the other three I entered values of -40, -40 and -50. These values were chosen with trial and error, rather than looking at the histograms. The resulting reclassification was really good, but if it means my underlying data was bad and the process is not repeatable, then I of course won't use these results. 3) In general, what is ENVI doing with the values you enter in the threshold inputs? Is it the same process with rule images from Max likelihood vs. other supervised classification outputs? Thanks!
    You are not authorized to post a reply.