Change detection

After performing supervised classification on two different S2 scenes what is the best way to assess the change detection between the images? Ideally i would like to identify areas of change and to be able to tell what class the features have changed from and to.

Additionally, is there any way of creating a confusion matrix in SNAP to assess the accuracy of the classification approach?

Please take a look at this thread,

Source of the thread

I think it’s not supported by SNAP, but you could perform it in qgis,

If you have two classified images with less than 10 classes, you can multiply one by 10 and then add both rasters up to one product consisting of pixels with 11, 12, 13…