Higher Order Connectivity for Determining Contiguous Neighbours

[Ioannis Dalezios]

Hi everyone,

I am currently working on a microstructural prediction model and I am using DREAM.3D v6.5 to analyze the synthetic microstructure in order to get statistical measurements, such as grain size and morphology distributions.

When using the "Segment Features (Misorientation)" and "Find Feature Sizes" filters to segment the grains, I get an abnormally large amount of tiny grains, like the image below shows.

I believe this might be related to the connectivity algorithm used in the segmentation process. From what I understand, DREAM3D uses 6-connectivity (face-to-face neighbors only) to determine contiguous neighbors during the grain segmentation. I'm wondering if there is a way to increase the order of connectivity to 26, in order to account for edges and corners, as well. If not, how difficult could it be to implement this?
Any insights, suggestions or alternative approaches would be greatly appreciated!

[Jack]

What kind of morphology are you looking for? 

--
You received this message because you are subscribed to the Google Groups "dream3d-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dream3d-user...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dream3d-users/33c62a65-5bee-444b-a725-ac4d452405a8n%40googlegroups.com.

[Michael Jackson]

Dear Ioannis,

      Just a few sanity checks: Your orientations that you are importing are in radians correct? Do you have a ground-truth example microstructure that you could share that shows your segmentation versus what DREAM.3D v6.5 produces? 

Or maybe I am not understanding the question correctly and you are using DREAM.3D v6.5 to generate a synthetic microstructure and you are getting a large amount of tiny (single voxel) grains? If that is the case then in the filter “Initialize Synthetic Volume” you will need to set a smaller “Resolution Value” so that more voxels can represent the grain size that you have asked the synthetic generator to create.

--
Mike Jackson                    mike.j...@bluequartz.net
BlueQuartz Software         www.bluequartz.net

President/Owner               Dayton, Ohio

Principal DREAM.3D Developer

--

You received this message because you are subscribed to the Google Groups "dream3d-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dream3d-user...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dream3d-users/33c62a65-5bee-444b-a725-ac4d452405a8n%40googlegroups.com.

[Ioannis Dalezios]

Yes, the orientations are indeed in radians, but we are using the "Segment Features (Scalar)" filter, as our synthetic microstructures have a single integer value assigned to each orientation.

Our model simulates the LPBF process and uses some geometric assumptions during solidification, so it shouldn't produce many grains smaller than an initial cube size that we seed (usually 10μm size) and we also post-process the results using the "Potts Model" filter on Dream.3D, so any small grains left should be gone. The grains that are produced from the model are very irregularly shaped, so maybe there is a segmentation error in the corner-connected grains. We can share a .dream3d file if needed, but there are multiple GBs, so we'll need to find a way to send it to you.

--

Ioannis Dalezios

Graduate Student in Materials Science and Engineering

Carnegie Mellon University

224-256-0817