I realise that point cloud classification of “high vegetation” works best with GSD’s of less than about 8cm/pixel. I do appreciate its efficacy.
Yet, I think I’m not the only user who is plagued by “pits” or depressions in the DTM where trees have been removed. I believe the cause of these pits is that in some cases, some parts of tree canopies are plotted below ground level and are automatically assigned to “ground”.
This is a big ask: Can the point cloud classifier be trained to reject tree canopy points that end up lower than the immediate surrounding terrain? In woodland projects it is tedious to do this manually.
Images to illustrate my problem: