Data type after mosaicking

I am learning about OBIA with eCognition and use Sentinel-2 imagery for that. I already pre-processed the S2L1C into S2L2A with sen2cor with no problem. After that I did S2Resampling and subset it. Then I mosaic them with only the important band (2, 3, 4, 8, 12). The file type I choose for the output is geotiff and it seems work when I tried the RGB in SNAP. I also remove B#_count and valid-pixel expression. But when I tried to do segmentation in eCognition, it showed a weird result unlike when I did with the raw data (S2L1C).

Segmentation with S2L1C

Segmentation with S2L2A

Segmentation with pre-processed S2L2A (S2Resampling and subset)

Segmentation with pre-processed and mosaic-ed S2L2A

I found out that the data type changed after mosaic, from uint16 to float32. Is there any option to change the data type output for Mosaic? Thank you.

I once found out that eCognition cannot handle some types of float rasters, leading to the segmentation pattern of the last example. Can you try to export the raster as 16bit integer, for example with “Convert Datatype” and select GeoTiff as an output?

Thank you for the feedback. As you said, I tried to mosaic the pre-processed S2L2A and then run the geotiff with “Convert Datatype”. I selected the uint16 and excluded the B#_count. Then I import the converted geotiff in eCognition and it seems kinda work. I think the “pixel” is smaller than what I did from mosaic-ing in ArcMap. I need to step up the number in segmentation process (from 10 to 100).

Result from Convert Datatype

Thank you for your feedback. Now I can cut down the process a bit (not using other application to mosaic-ing the imagery).