Assoc Prof. Steve Micklethwaite - Honours Projects

Unravelling Earthquakes using Deep Learning

Supervisor(s): Steven Micklethwaite (EAE), Andrew Casey (Physics), Tom Drummond (Engineering), Sam Thiele (EAE)
Field of Study: Structural geology, artificial intelligence, image analysis
Support Offered:  High Performance Compute, labelled datasets

Earthquakes lead to enormous damage and loss of life. Because earthquakes occur unexpectedly and nucleate many kilometres deep within the Earth our best understanding of how they operate is limited to laboratory experiments and direct observations of seismic waves. For these reasons high quality geological exposures of ancient earthquakes reveal valuable insights into the earthquake process. Piccaninny Point, Tasmania is one such outstanding exposure, containing a network of faults that moved as earthquakes in the geological past. We have collected digital drone imagery across this outcrop at sub-cm resolution over several hundred metres. In order to extract the full value of the data, deep learning approaches are required to map the geological stratigraphy and the faults. The different datasets available include >400 photos, a 3D point cloud, digital elevation model and a single orthorectified image.  The outcomes of this project will be a preliminary tool that enables us to quantify and better understand fault topology, connectivity and interactions with different Earth materials leading to new insights into earthquake mechanics.

For further information, contact Steven Micklethwaite

What, how many, where - Marine wildlife conservation by machine learning?

Supervisor(s): Steven Micklethwaite (EAE), Rohan Clarke (Biol), Tom Drummond (Eng), Andrew Casey (Physics), Rebecca   McIntosh (Phillip Island Nature Parks)
Field of Study: Biology, artificial intelligence, image analysis
Support Offered: High Performance Computer, labelled datasets
Collaborating Organis: Phillip Island Nature Parks

Seal populations in the Bass Strait appear to be deteriorating, possibly due to marine pollution and climate stresses. Current methods for monitoring the populations are manual and associated with high risks, requiring humans on small boats. Due to the costs and risks, such surveys typically only occur over few years. Recent successful trials using drones resulted in a more accurate census of a seal colony and involved collecting digital images from which wildlife conservationists manually counted the number of seals. We have collected drone digital imagery of seal colonies, and will also obtain thermal imagery. This project aims to fuse both datasets and use a deep learning approach (neural networks with trained data) to be able to quantify the location, size and maturity of seals from Phillip Island Nature Parks. Once the system is obtaining 90-95% pixel accuracy from digital and thermal data, we hope to be able to optimise it to achieve similar results just from digital data. The outcomes of this project will be results that demonstrate the possibility of the approach and a preliminary tool that enables marine wildlife conservationists to automate and monitor the health and behaviour of seal colonies on a regular basis throughout a year.

For further information contact Steven Micklethwaite

Geotrace for Geophysics

Supervisor(s): Steven Micklethwaite (EAE), Lachlan Grose, Samuel Thiele, Robin Armit
Field of Study: Potential field geophysics, GIS, image analysis
Support Offered: Supervision in geophysical interpretation, GIS and Python

Geotrace is a tool developed at Monash University that maps faults, fractures and geological boundaries quickly and efficiently in the open source geographic information software QGIS. This project will explore applications of the tool to potential field geophysical data from the Mount Isa Inlier – one of the world’s premier mineralised districts. Experiments will be derived that assess the efficacy of the tool for geophysical interpretation and its ability to synthesise multiple datasets, quantify feature support and characterise spatial precision. New interpretations of a small portion of the Mount Isa Inlier will then be compared against existing interpretations to assess the power of this approach for prospectivity analyses and interpretation of the regional geology.

For further information, contact Steven Micklethwaite

Ore Deposits

Supervisor(s): Steven Micklethwaite (EAE) + selected others
Field of Study: Economic geology +/- structural geology +/- geophysics +/- geochemistry +/- drones & digital mapping +/- modelling
Support Offered: As appropriate to each project

Each year, opportunities exist for students to engage with honours research into various aspects of ore deposit formation. These are often in collaboration with industry partners, and typically have excellent outcomes in terms of future employment. Students are encouraged to come and discuss what is available.

For further information, contact Steven Micklethwaite

Understanding gold deposit formation in Victoria

Supervisor/s: Andy Tomkins, Steve Micklethwaite, Chris Wilson and Sandy Cruden
Field of study: Economic Geology
Collaborating organisation/s: Fosterville Gold Mine

In this project we will be investigating one of the main gold deposits in Victoria, such as Fosterville or Ballarat. The aim of the study will be to investigate changes in the types of ore minerals that were being deposited during the evolution of the ore forming system. We want to know whether gold was associated with a particular structural stage and with particular minerals within a progressively changing system, and how the chemistry of the ore forming fluids changed over time. Field work will be conducted at one of the Victorian gold mines, and will be supported by the company running the mine – the project is an opportunity for the minerals industry to see your work, and it is not uncommon for the companies supporting honours projects to offer employment after completion of your degree.

For further information, contact AndyTomkins