HOW TO USE AN EQUAL NUMBER OF REFERENCE POINTS FOR ACCURACY ASSESSMENT

[ivaninnocent]

I have a land cover classification raster of a an island with these classes; water, forest, oil palm, settlement, wetland and arable land. The water(ocean) class sorrounding the island occupies 72% of the classification area despite not being very useful for my analysis,. If i use the default stratified sampling in SCP, most reference points will fall in water which i want to avoid. Is there a way to conduct my accuracy assessment by assigning / allocating equal numbers of reference points per class? so if in total i want to use N=600 points then each class of the 6 classes can be stratified and allocated 100 points. ArcGIS pro has this option but am trying to see how to do it in SCP. Could we expect more sample allocation options for stratification e.g equal allocation or mixed approaches in SCP version 8.

The other inquiry is on the updating of accuracy or sample points. How can i re-use the same reference or accuracy points on the accuracy assessment of classifications of same raster by random forest(RF), support vector machine(SVM) and maximum likelihood.(ML) I can decide to use same file of saved sample/reference /ground truth points to do accuracy assessment for the 3 classification rasters of same image if comparing kappa. In ARCGIS pro, this is possible but am struggling to find it in SCP.

otherwise thank you for this great and wonderful plugin.

[semiautomaticgit]

Hello @ivaninnocent , thank you for using SCP.
At the moment, improvements at the allocation methods are not foreseen for the release 8 of SCP, however I'll consider adding more options in the future.
About your second question, if I understand correctly, you can repeat the accuracy assessment using the same reference raster or vector with different classifications in the tool Accuracy.