The idea behind the Generate Random Sample tool is that you either generate a bunch of random points in your image, or from a set of ROIs. It sounds like you probably want to generate them in your classification image, right? So, then you can check all of those point locations in the high resolution images to see if they're properly classified. In that case, you would choose Classification → Post Classification → Generate Random Sample → Using Ground Truth Image. In that case, you don't need to choose any ROIs. You'll just get a point ROI output, which contains the random points you were after.
You do need to choose which of your classes you want points to be generated in, though. For example, you *could* leave out one or more classes that you're not particularly interested in, or the unclassified class. It sounds like you probably want to choose them all, so you'll get some points in all of your classes. But I could be totally wrong, because what do I know?
I hope this helps.
|