I created image objects in eCognition using multi-resolution segmentation and it created a lot of tiny objects. I've been having issues with Remove Objects, so instead I exported the image objects to Arc and removed them manually there.

I would like to now train the classification model on the manually-edited image object layer. To do this, I imported the shapefile as a thematic layer, and ran the Vector-Based Segmentation algorithm. 

I thought that this would create image objects that directly matched the polygons in the shapefile, however the result of vector-based segmentation re-created the small image objects I had removed in ArcPro.

The white lines are the vector layer, and the blue (with the tiny image objects) is the resultant image object layer with the tiny image objects again.

Is there a way to prevent these tiny objects from being recreated? 
Thank you!


  • Khrystyna Bezborodova

    Hello Maya,

    Could you please share the ruleset and the data?
    You can send it to our Support if it is more convenient for you!


  • Matthias Staengel

    Hi Maya,

    Somehow it probably still recognizes the "old" vector file. Please open a new project and load the "new" shapefile into that project to see if this is still the case. Maybe you still have those objects stored on a certain level? If that is the case please delete all levels before you run the vector-based segmentation again.

    Btw., you also could improve the initial segmentation or rework the objects in such a manner that you do not need to go the extra mile via a third party software. We have plenty of algorithms there that could help you, like the mentioned "remove objects" (would be interesting to understand why it did not work in your case) or for example the "Spectral Difference Segmentation".

    Hope this issue is already solved.



Recent Discussions

Need more help?

Enter Knowledge Base Contact Support