I am struggling to develop a rule set for individual tree crown dilineation in eCognition. I have realised that the crown area i get from eCognition is very different from the one i got from the field. I am using Worldview-2. If anyone has experience on how to improve my rulesets, i will appreciate.
I have tried Imagine Objective but since i am working over a large plantation, it crashes and when i use only one band, the results are really awkward. I will appreciate other platforms but I still believe that eCognition can do the job but i am getting frustrated and the clock is ticking! Thank you again.
Hi Abel..
i also struggle in eCognition.i have similar problems too. but i think you extract small area firstly, because worldview 2 has quite large data set.
i use approximately 2000*2000 pixels on quickbird. and i try to develop a rule set in this way.
Hi,
My approach is now very simple and i will explain it to you below. i was thinking that i could somehow share my projects so that you can have a general view of the rulesets that i developed.
1. I removed other bands and remained with 5, NIR, RE, R, G and B. Determined scale parameter with ESP Tool. Did multiresolution segmentation with more band weight on NIR.
2.I used mean NIR values for the objects together with NDV I and Greeness to assign classes.
3. I refined the class shapes using morphology and watershed transformation with settings that i determined from the field.
The results are not really good but its something and like i said for now the clock can not let me improve them, i will have to be content with them for now.
Cheers and best wishes
Abel
http://www.tandfonline.com/doi/abs/10.1080/13658810903174803#preview
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing