Machine learning for landscape pattern recognition

Dr Gregory Smith1, Dr Thomas Albrecht2, Dr Ignacio González-Álvarez2, Dr Jens Klump2

1CSIRO Data61, Sandy Bay, Australia,

2CSIRO Mineral Resources, Kensington, Australia

 

Mineral exploration, one of Australia’s grand challenges, is complex, costly and traditionally involves significant amounts of field data collection. Automating and quantifying parts of the process can potentially improve both efficiency and accuracy. For example, satellite-derived data such as Digital Elevation Models can image diverse geomorphological surface patterns at scales that would be impracticable for detailed field observations.

We use machine-learning (ML) algorithms and a knowledge-based approach to first describe different landscape types in the field, then classify and automatically map at regional and, potentially, continental scales. The outcome is not just a map, it also enables us to quantify geomorphological elements which can have a substantial impact on the understanding of landscape geochemistry: it allows the evaluation of the best sample media for capturing vertical and lateral geochemical dispersion within the cover associated with basement (fresh) rocks and regolith/critical zone geochemical footprints associated to the presence of mineral deposits at depth.

The talk will focus on the computational and applied ML aspects of the project. We present the classification methodology and evaluate the performance of a range of ML algorithms, including support vector machines, decision trees, and self-organising maps. Derived from the digital elevation model, 48 features were used, and robustness of classification with respect to resolution and spatial extent were tested. A support vector machine yields the best accuracy of 98%.


Biography:

Thomas Albrecht joined CSIRO Mineral Resources in 2018 as Resources Data Scientist. His current research interests include machine learning, modelling, simulation and high-performance computing.

Thomas obtained his PhD from Dresden University of Technology, Germany, in the field of Computational Fluid Dynamics, followed by Post-doc positions at Helmholtz-Centre Dresden-Rossendorf and Monash University.

He has published over 20 articles on a wide range of fundamental problems in fluid mechanics, including rotating flows, separated flows, transition to turbulence and flow control, with applications in geophysics, meteorology, bushfire modelling and turbomachinery. He enjoys windsurfing, cycling, RC airplanes and developing open source code for flight simulation.

ABOUT AeRO

AeRO is the industry association focused on eResearch in Australasia. We play a critical coordination role for our members, who are actively transforming research via Information Technology. Organisations join AeRO to advance their own capabilities and services, to collaborate and to network with peers. AeRO believes researchers and the sector significantly benefit from greater communication, coordination and sharing among the increasingly different and evolving service providers.

Conference Managers

Please contact the team at Conference Design with any questions regarding the conference.
© 2017 Conference Design Pty Ltd