Hi,

I conducted multiresolution segmentation in current release of eCognition on a WorldView 3 image with eight spectral bands, utilizing the parameters seen below:

The resulting segments were then exported as polygon vector to ArcMap for further analysis:

Upon using the 'intersect' tool in ArcGIS, I discovered several self-overlaps in the vector file. Specifically, out of 66,829 segment polygons, 210 were identified as self-overlaps!

However, when using 'Select by Location' to select the instances where features from the target layer 'are within the source layer feature', the result is 105 segments fulfilling the given criteria. I inspected the results and observed that 105 features are actually Self overlaps. This has been further confirmed by 'Count Overlapping Features' tool available in ArcGIS Pro.

I am seeking to understand why multiresolution segmentation produces these self-overlaps. Additionally, is it possible to identify and rectify such instances within eCognition before exporting the segments to ArcMap, thereby ensuring that no anomalies are present in the exported vector data?

Any assistance or insights you could provide would be greatly appreciated.