I have an image where I’m ultimately trying to automate eCognition to map out certain vegetation communities. This is of a wetland restoration where I have significant ground-truthed data.  The wetland is simple – pockets of cattail marsh in an open regenerating meadow, all of which is surrounded by forest.  I used a drone to collect imagery (over 700 photos) and stitched them together in Agrisoft. I imported the basemap (only RGB) and las file into eCogniton. I’ve been attempting to follow numerous videos on the web to determine an approach but most of these videos use true LiDAR, whereas I only have single returns via photogrammetry.  I can clearly see the elevational differences in the dense point cloud and dsm, where each cattail areas are about 1m taller than the surrounding meadows. I've tried classifying a ground layer in las but it's never accurate.  I should add that the overall terrain is on a slight incline, therefore the dsm elevation between each cattail area is different.  Any thoughts or recommended tutorials would be appreciated.   Thanks, and Merry Christmas


1 comment

  • Nils Erik Jørgensen

    Try this:
    Segment your area using dsm.
    classify objects based on neighbour info. It the object has a neighbour that is 1m taller, then it is meadows. and opposite it is cattail. Was this answer to your question?

Recent Discussions

Need more help?

Enter Knowledge Base Contact Support