I was wondering how I could extract "crown area" of single trees in a forest from a high-resolution image with only 3 band (RGB) in ArcMap. ( No NIR to use NDVI as any threshold).

I need a fast easy-to-do way to come up with a shapefile with polygons (made of several pixels depending on size of each individual tree). Here was I was thinking : Focal Statistic - Conditional (Raster calculator)- Extracr Attributes- Convert to polygon

I just need trees (to calculate their diameter at the end) in the forest. No matter the species or whatever.

I am using eCognition but I want to try ArcMap/ ( It's ok if its not 100% accurate!!! )

  • Could you provide a screenshot showing the type of trees and surrounding vegetation? What type of 3-band imagery are you using? What is the spatial resolution?
    – Aaron
    Oct 22 '14 at 18:55
  • Its just an image taken from Google-earth! There is just one type of species but surrounding areas are mainly bare land, bareland+grass and some urban areas which I am not interested in any of them! I just want to throw away any none-tree objects and then find the CPA.
    – G.Z
    Oct 22 '14 at 18:58
  • Have you tried calculating any texture metrics in eCognition?
    – Aaron
    Oct 22 '14 at 19:00
  • Nope! I used brightness and Mean-Layer 2 for classification of shadow, forest, bareland, ...I wanted to extract polygons as trees to ArcMap but then it wasn't giving good results. Now I just want to use ArcMap to find trees(brighter and higher in pixel value) and make polygons for each single one.
    – G.Z
    Oct 22 '14 at 19:02
  • Focal Statistic - Conditional (Raster calculator)- Extract Attributes- Convert to polygon.... This one works somehow but the final polygons are not exactly matching the tree-positions! It needs serious improvements!
    – G.Z
    Oct 22 '14 at 19:05