Macquarie University
01whole.pdf (4 MB)

Generative Adversarial Networks for magnetic potential field image super-resolution

Download (4 MB)
posted on 2022-11-09, 00:29 authored by Anthony Benn

Magnetic data are one of the most common geophysical techniques in mineral and resources exploration. Developed nations have repositories of high-resolution geophysical datasets. However, these resources do not exist for many regions worldwide. Consequently, there is demand for new methodologies of image enhancement, to capitalise on existing low-resolution datasets.

Generative Adversarial Networks (GANs) are a type of deep learning algorithm that have been used fruitfully on image enhancement problems in the computer vision field, but to date have received little attention in geophysics. This research uses GANs to enhance low-resolution magnetics data. Thirty thousand pairs of high- and low-resolution images are constructed from existing magnetic datasets in Australia. The dataset is split into training and validation sets and the training tiles are input into a GAN model for training. The GAN model attempts to predict the high-resolution images from the low-resolution input, effectively “learning” the geophysical and spatial characteristics of the data, and the transformation between resolutions.

This new GAN-driven resolution enhancement model demonstrates the potential of this technique to be used on low-resolution datasets, particularly in areas where investment in high-resolution datasets is limited. GANs have wide-ranging applications to minerals exploration, groundwater studies, and planetary research.


Table of Contents

Chapter 1: Introduction -- Chapter 2: Background -- Chapter 3: Methods -- Chapter 4: Results -- Chapter 5: Discussion and conclusion -- Chapter 6: References


A thesis submitted in partial fulfillment of the requirements for the degree of Master of Research

Awarding Institution

Macquarie University

Degree Type

Thesis MRes


Thesis MRes, Macquarie University, Department of Earth and Environmental Sciences, 2022

Department, Centre or School

Department of Earth and Environmental Sciences

Year of Award


Principal Supervisor

Steven Hansen

Additional Supervisor 1

Craig O'Neill


Copyright: The Author Copyright disclaimer:




61 pages