I have a task to be tackled of which I am not sure, wheter it is even possible. I need to classify single trees/clusters of trees in a forest automatically using sentinel-2 data. Supervised classification should be extremely trick in my opinion. I have some ground data, so that certain pixels are identified as a specific kind of tree in a point layer. Is there any way to get the automatic classification done? Especially when that scheme should not only work on a single image but on several images of the same area?

  • The spatial resolution of Sentinel-2 is 10 meters, see the tech docs. There is a good academic article explaining what is possible with imagery of around 15m spatial resolution. – John Powell Mar 24 '17 at 10:23
  • The article talks about tree classification with a 1m resolution. I can see that that would work. However I am asking because the 10m spatial resolution seems a bit too coarse for the task. I was hoping someone might have experience with this. (Also the article seems a bit old. A lot has happened since 2004 - thanks anyway!). – Homerun_ Mar 24 '17 at 10:48
  • I think it is way too coarse, which is why I mentioned it. You can't extract features finer than pixel res, no matter how many look angles you have. We use such data (individual tree heights and extents) and I know for a fact that the company that does it uses 25cm resolution aerial imagery. – John Powell Mar 24 '17 at 10:54
  • Hm yeah, that sounds plausible. Just what I thought. Thank you :) – Homerun_ Mar 24 '17 at 10:57
  • Out of curiosity, what are you trying to classify the trees for? – Aaron Mar 25 '17 at 16:29

NASA offer an advanced webinar course https://arset.gsfc.nasa.gov/land/webinars/advanced-land-classification that uses the Semi-Automatic classification plugin in QGIS. It will take you through step by step with Landsat imagery. You can use the same steps for Sentinel imagery. The Semi-automatic classification plugin comes with the best user documentation I have seen for a QGIS plugin. It's well worth a look and the time to work out how to use it.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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