Deep learning extends 2D microXRF to the volume in XCT data for 3D mineralogy

The term “Deep XFCT” was coined in this work published in MDPI Energies, where Dragonfly’s deep learning was utilized to extend 2D surface-based microXRF to segment 3D volumes of XCT data. MicroXRF is very useful for chemical mapping in 2D. XCT is excellent for volumetric imaging. Combining these two methods is possible using deep learning to get the best of both worlds – resulting in 3D mineral mapping.

According to lead author Patrick Tung: “Dragonfly was instrumental in being able to quickly and easily perform all the analysis in a cohesive software package. Not only was it possible adjust and optimise the registration between the different image modalities, but also to apply deep learning protocols for segmentation and visualisation.”

 

Video Presentation

External Content

This is a Youtube video.
To be able to watch, you need to accept Marketing Cookies.

Manage Consent

I agree with external content being displayed to me. Personal data may be transmitted to third-party platforms. More Information - Data protection

Publication

Tung, P.K.M., Halim, A.Y., Wang, H., Rich, A., Marjo, C. and Regenauer-Lieb, K., 2022. Deep-XFCT: Deep Learning 3D-Mineral Liberation Analysis with Micro-X-ray Fluorescence and Computed Tomography. Energies 2022, 15, 5326 (https://doi.org/10.3390/en15155326).

Research Center

Tyree X-Ray Facility
(https://www.unsw.edu.au/engineering/our-schools/minerals-and-energy-resources-engineering/our-research/facilities/tyree-x-ray)

Images

3D rendering showing the CT data with deep learning segmentation of minerals in color.