I'd like to apply a classifier developed on one image, image1, to a different image, image2, without training the classifier on image2. That is, apply the classifier (be it MLC, NN, SVM, etc) to image2 as if it were image1. For example, if I believed that there were no spectral differences between the two images, then I should not have to collect training sites and retrain the classifier on image2, I should just be able to directly apply the classifier trained on image1 and expect the same classification accuracy.
I cannot figure out a way to accomplish this in ENVI. In IPW there were classification routines that returned the classifier itself, which could then be applied to any other images, provided the input bands were the same. Does anyone know how to do this in ENVI?
Best,
Josh
|