I need to perform a segmentation and classification in eCognition but only within the areas (polygons) covered by my thematic layer.
I have successfully performed a chessboard segmentation using my thematic layer and a very large object size to derive image objects from the polygons in my thematic layer. However, I cannot figure out how to perform further segmentations without reverting to segmenting the entire scene. I have tried assigning my image objects to a class and then using a class filter to restrict the next segmentation algorithm to that class only, but it still segments the whole scene.
My desired processing steps are:
1. Perform an initial chessboard segmentation using my thematic layer to derive image objects from the polygons in my thematic layer (I have achieved this).
2. Perform another chessboard segmentation so that every pixel becomes an image object - but only within the areas covered by my thematic layer (I have not been able to achieve this as it segments the whole scene).
3. Perform multi-resolution segmentation in a bottom-up approach - but only within the areas covered by my thematic layer (I have not been able to achieve this due to difficulty with step 2).
I have attached a crude sketch of the three processing steps above in the hope that it clarifies what I am trying to achieve.
I am very grateful for any advice you might have.
Thank you very much.