Hi,
I work on a diachronic map of vegetation, I use RGB NDVI technique , this technique is based on the use of three ndvi taken in three different dates to quantify the change of the green biomass through the three dates (in my case : 1984, 2001, 2011), In this technique a non-supervised classification was used to quantify the changes. Currently, I find a problem to produce the matrix of confusion, When I read in the literature authors says that the matrix of confusion is done on the basis of ROI collected from ground field , in my case it is impossible to collect points view that the dates are very distant (1984,2001,2011)....thank you.
Thank you very much.