Segmentation of 3D image of a rock sample supervised by 2D mineralogical image

Abstract

We present an approach for 2D-to-3D mineralogical information propagation and preliminary results of its implementation. An X-ray microtomography (microCT) scanner is used to obtain a 3D microstructural image of the sample. A scanning electron microscope (SEM) equipped with an energy-dispersive X-ray (EDX) detector provides a supervising 2D mineral map of a rock sample cross section. The 2D and 3D images are then registered using a surface-based algorithm, naturally taking into account specifics of said data. The overlapping area of the images is then used as a training set for supervised mineralogical segmentation of the full 3D sample.