Hello,


I have a question about the exact Large-Scale Mean-Shift segmentation 
workflow. I've set up a workflow to process large amounts of tiles multi 
threaded, which is working nicely. Before processing larger datasets I've 
ran multiple test on smaller sets to come up with the best segmentation 
settings. After running a test on approximately 900 hectares I found that 
about 80% of the segmentation was successful. Some errors are due to cloud 
shadow/aligning artifacts, but other errors I don't fully understand. Below 
an example where the segmentation clearly missed the segments of the 
grass/road on the left side, while on the right side it segmented 
correctly. This missing of segments happens throughout the image, with 
different objects and spectral ranges, for I would say about 15% of the 
total area.  



<https://lh3.googleusercontent.com/-io3WGg61RMM/V4yzeaoR8PI/AAAAAAAA36E/V4AG9doY_98mu3NtrQIr1K6i-4UP3Vl8QCLcB/s1600/segmen.PNG>













My question is, why is it missing this segment, while on similar 
objects/spectral ranges this does not happen?
The image consists of a RGB and NIR band stretched from 0 to 255. My 
segmentation settings are: -ranger 7.5 -spatialr 10 -minsize 0 with tile 
sizes of 1000. The spectral euclidean distance should be more than 7.5 for 
the road and grass segments. What am I missing?

Many thanks!

Kind regards,
Darell

-- 
-- 
Check the OTB FAQ at
http://www.orfeo-toolbox.org/FAQ.html

You received this message because you are subscribed to the Google
Groups "otb-users" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/otb-users?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"otb-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to