Machine Learning for Rapid Material Characterisation

Mr Alex Pitt1, Mr Paul McPhee1, Dr Chad Hargrave1

1CSIRO, Pullenvale, Australia

 

Accurate characterisation of the microscopic structure of mined resources such as coal is fundamental to understanding its utility and environmental impact. Such information is also critical to understanding the provenance and potentially hazardous nature of dust, sediments and other environmental samples. Such microscopic analysis typically requires an expert petrographer or environmental scientist to characterise the samples manually, a time consuming and expensive process.

CSIRO has developed an automated Component Grain Analysis (CGA) system to reduce the time required to segment and characterise coal and dust samples. The system manipulates 300 GB sample images with sub-micron pixel resolution, using automated processing to resolve this data intensive task. CGA provides reliable statistics on the distribution of maceral types and impurities in coal samples, and component materials in dust samples.

A recent breakthrough development is the incorporation of machine learning (ML) algorithms to the complex task of particle segmentation. Specifically, convolutional neural networks (CNNs) have been employed due to their demonstrated successes in the semantic segmentation of natural images, their capacity for learning texture, their computational efficiency, and their suitability for distributed computation. State-of-the-art CNN models were trained on microscopy images and ground-truth labels (provided by CSIRO petrographers) and consistently converged to segmentation accuracies on validation data of more than 95% – exceeding the estimated noise in the expert labelling.

Further steps for the development of the ML system include hyper-parameter refinement, label noise reduction, and model augmentation for the automatic characterisation of component materials and intra-particle segmentation for complex particles.


Biography:

Alex Pitt is a software engineer in the Energy division at CSIRO. He graduated from the University of Queensland in 2011 with a B.E. (Software), and worked on data-engineering products for Microsoft in Redmond, Washington. After returning to Australia, he joined the CSIRO where he has worked for the last 6 years on sensor integration, signal processing and computer vision.

 

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.
© 2019 Conference Design Pty Ltd