Analysis of scheelite ore with scanning electron microscopy (2D) and X-ray computer tomography (3D), and correlation of data with Deep Learning methods


Analysis of scheelite ore with scanning electron microscopy (2D) and X-ray computer tomography (3D), and correlation of data with Deep Learning methods

Hellmuth, F.

This thesis demonstrates a combination of three consecutive methods to achieve reliable phase classification in a 3D-image of particulate material obtained by X-ray computed tomography (CT). The method consists of 1.) sample preparation to minimise the effect of image artefacts on the CT-scan and enable the subsequent analysis with Scanning Electron Microscopy (SEM) techniques, 2.) Alignment of 2D SEM-based phase classification to a specific location in the 3D-image and 3.) implementation of the 2D-mineral classification as training data for a Convolutional Neural Network (CNN). The trained neural network enables phase segmentation of the particles in the 3D-image based on the 2D-phase classification.

Keywords: X-Ray Computed Tomography; Scheelite; Deep learning; Mineralogy; Scanning Electron Microscopy

  • Master thesis
    TU Bergakademie Freiberg, 2021
    Mentor: Prof. Bernhard Schulz; Dr. Axel Renno
    71 Seiten

Permalink: https://www.hzdr.de/publications/Publ-33782
Publ.-Id: 33782