I have created a land use land cover map using the SCP plugin of QGIS. I want to perform accuracy assessment of this classified image by creating 100 sample points for each class and importing them into Google Earth. This option, however, is unavailable in SCP. There is only an option to provide a reference shapefile (created from my ROI training file) or a raster image for the assessment.

If this is not possible, is there an option for performing stratified random sampling for each class using the raster itself, without having to polygnize it?

  • Could you not just generate a polygon covering the entire extent of the area you want to assess? You could then generate the random points in that and assign them with whatever land cover they fall in? Or am I missing something? Commented Mar 28, 2017 at 10:55
  • What I want is to create 100 random points within each class. Creating a single polygon for my study site will not assure that all classes are each represented by 100 random points. If there is no automatic way of doing this, then I will probably polygonize my classified image and generate random points for each class. Commented Mar 28, 2017 at 11:15
  • Okay. That's why that was a comment not a solution; ). As a fall back, you could generate thousands of random points and attribute with your values...then select a random 100 from the set for each of classes. Commented Mar 28, 2017 at 19:28

1 Answer 1


I have used r.sample.category to create 100 points for each class using the classified image itself without the need to polygonize it. The data type should be integer and not float.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.