I’ve been diving into Sentinel-2 imagery for land cover classification and it’s been a game changer… The 10-meter resolution really helps in distinguishing between different vegetation types. I’m curious if anyone has tips on post-processing techniques that enhance classification accuracy further.
, I know the struggle with classification accuracy! I found that applying a simple NDVI threshold can help highlight vegetation types more clearly, especially with Sentinel-2’s 10-meter resolution. It’s saved me some headaches in post-processing.
Have you tried using machine learning classifiers like Random Forest after applying pre-processing techniques like atmospheric correction? It’s helped me boost accuracy significantly when working with Sentinel-2 imagery. Just remember that while accuracy can be improved, it often involves a trade-off with processing time.
And one technique that really helped me is using a stratified sampling approach for training your model. It ensures that you have a balanced representation of all land cover types, which can enhance classification results significantly. I’ve seen great improvements by combining this with cloud masking techniques.