Hello!

 

I have been using eCognition for class projects and now my own final project for a couple months now. I am out of formal training at this point and I'm just using eCognition TV tutorials so far to get better with the software. Today I watched the Morphology video and when I try to apply it to my study area, it is so slow I have to cancel it. I've had this problem with other things, too, like trying to using segmentation algorithms on certain classes in my project.

 

I'm trying to figure out if there is some kind of ball park figure as to when my project is too lage, like too many pixels, or if there are setting in my machine I can tweak (I JUST got a brand new, fast, gaming computer recently specifically for this and deep learning models, I'm trying to finish grad school! ), or if there might be something within eCognition i can tweak. I have already lowered my study area three times, and I don't want to make it any smaller. I really want to keep the landscape features that I have.

 

I have a third-foot (so about 10 cm) resolution imagery, and a 1-ft nDSM I created elsewhere. I have a point cloud in the project, too, but I don't think it's getting used when I try morphology. I have an exact 1 square US survery mile area I'm working with. The multriresolution segmentations, even at small scale, don't take that long, usally around 5 minutes lately. So I'm wondering why using the morphology takes  SO crazy long?? 

 

any advice appreciated!

 

Audrey

 

1 comment

  • Michael Wang

    Hi Audrey,

    There could be a number of reasons why eCognition is slowing down, including the possibility that the image is too big. However, one very common reason for it is the use of Haralick Texture features. If you have a bunch of Haralick textures pulled up in the Object information window or have it in your classification tasks, this can cause a massive slow down even on really powerful PC. If this is the case, I would highly recommend you to remove these and see if it improves performance at all.

    best regards,

    Mike

Recent Discussions

Need more help?

Enter Knowledge Base Contact Support