Challenges in the segmentation and analysis of X-ray Micro-CT image data

Abstract:

Pore scale modeling of fluid flow is becoming increasing popular among scientific disciplines. With increased computational power, and technological advancements it is now possible to create realistic models of fluid flow through highly complex porous media by using a number of fluid dynamic techniques. One such technique that has gained popularity is lattice Boltzmann for its relative ease of programming and ability to capture and represent complex geometries with simple boundary conditions. In this study lattice Boltzmann fluid models are used on macro-porous silt loam soil imagery that was obtained using an industrial CT scanner. The soil imagery was segmented with six separate automated segmentation standards to reduce operator bias and provide distinction between phases. The permeability of the reconstructed samples was calculated, with Darcy’s Law, from lattice Boltzmann simulations of fluid flow in the samples. We attempt to validate simulated permeability from differing segmentation algorithms to experimental findings. Limitations arise with X-ray micro-CT image data. Polychromatic X-ray CT has the potential to produce low image contrast and image artifacts. In this case, we find that the data is unsegmentable and unable to be modeled in a realistic and unbiased fashion.