From Material scan to Digital Model: Image Analysis & Image Processing

A key capability of GeoDict is the generation of 3D geometric material models in many forms: statistical, analytical, surface-based, but most importantly, voxelized. A voxel is filled with a single material, be it porous or oriented. Starting with a scan of real media is the natural way to digitize a material. Initially, GeoDict could import synchrotron scans, and later, micro-CT and FIB-SEM. Currently, also 2D scans , such as SEM and thin sections, have become very interesting because a 3D model from 2D scans may be created with the help of AI.

Using image processing capabilities, gray value scans of densities are turned into 3D digital materials. The scan are rotated to align the natural material axes with the image axes, cropped for scan artefacts, or re-scaled to arrive at voxels. Then, filters are applied to aid the further processing and analysis steps. Afterwards, two classifications of the scan lead to structures - representations where the meaning of the content of the individual voxel is an index. This index may either be the number of a material or an object.

Image analysis capabilities are available to understand materials by leading to classifications of the pore space, fibers, grains, or other microstructure of which the material is made of. For a true digital material, constituent properties need to be assigned to the individual objects or, at least, materials in the structure. The GeoDict Innovation Conference showcases challenges and opportunities arising from all the above.