DiUS, AWS Advanced Consulting Partner, will be speaking at AWS Summit Sydney 2019 about how we’ve been using artificial intelligence to gain deep insights from geological imagery to reduce costs and increase efficiency during drilling.

Gerd Wittchen, lead ML engineer at DiUS and Brenton Crawford, director and geoscientist, Solve Geosolutions will be talking about how artificial intelligence is extracting new value in mining on Wednesday May 1 from 2.45 to 3.15pm.

The talk will cover our journey from idea to development of a machine learning model trained to extract geological information from drill core imagery, as well as provide learnings for anyone starting their artificial intelligence journey.

We also talk about how we are leveraging AWS Machine Learning services in completely serverless approach to enable the operationalisation of image analysis as a service. This will consolidate important geological analysis to deliver consistent and radically efficient insights from drill core samples.

Using machine learning to make better decisions during mining and exploration

Solve Geosolutions and DiUS started applying image-specific machine learning to core drilling imagery in 2018. The idea started when Solve Geosolutions’ mathematical geologists, who provide specialist computational analysis skills to mineral exploration companies, began exploring the application of image processing algorithms in an effort to automate the analysis of core samples to deliver more consistent results.

Mining and exploration companies gain information about the geology of a deposit from a geologist’s inspection of high-resolution photographs of drill core samples. Studying the veins and texture of these images provide important geological insights on how economically important metals can be extracted from a mineral deposit. This task is tedious and error-prone, and the conclusions can be highly subjective.

Brenton Crawford, director SolveGeosolutions, explained how in an industry such as mining - where improving efficiency and productivity is crucial to profitability, even small improvements to the mineral discovery process can have a significant impact.

“We knew there was a significant opportunity to create better outcomes from core image analysis, but we needed assistance from a partner that had specialist expertise with deep neural networks to help with the complex image processing we required — so we reached out to DiUS.”

Using the power of image segmentation technologies

Starting small with a proof of concept to discover detect mineral veins, the challenge for DiUS was to address the difficulty of the low contrast of veins against the host rock, and their general variability in colour, orientation, length, and width.

Gerd Wittchen, lead ML engineer at DiUS emphasises the importance of leveraging existing research and advances to springboard an approach, as well as having access to the subject matter experts — in this case the geoscientists from Solve Geosolutions.

“We used a deep neural network architecture recently developed by researchers at Facebook AI specialized for object detection and segmentation problems, to build a model that detected veins more completely and faster than a geologist.”

In addition, the team was able to further fast track its approach by using a small dataset and heavily relying on augmenting images to train the model.

The success of the vein detection work lead to a family of equally-successful neural network models built to tackle other geological use cases, including rock texture search-by-example, rock tray fragmentation, rock quality designation, and mineral grain detection.

To consolidate all the geological analysis, DiUS has leveraged AWS Machine Learning Services in a completely serverless approach to enable the operationalisation of image analysis of rock core samples as a service.

Come and see us at AWS Summit Sydney

DiUS is a Gold Sponsor of the AWS Summit Sydney 2019. Swing by our stand at the expo and chat to us about how to help your team get started with artificial intelligence. Our capability spans greenfield machine learning solutions right through to embedding machine learning into IoT and digital products.

The AWS Summit Sydney is running from April 30 to May 2, 2019 at the International Convention Centre in Sydney. The summit is a free event designed to bring the cloud computing community together to learn about the latest AWS services and how they are being applied to solve business problems.

Register for the event.