Digitising Mining Features
Credit: Australian Resources & Investment, 2021
Client/Partner
Partners: BHP, Trailblazer
Funded by: Resources Technology and Critical Minerals Trailblazer
Timeline
August 2024 – Ongoing
Overview
Validating drill patterns is critical for safety, operational efficiency and cost management in large-scale mining operations. In open-pit environments, drill patterns are carefully designed to maximise safety and minimise costs, relying on precise knowledge of where previous holes have been drilled. Confirming the accuracy of the drilled pattern is therefore essential.
There are two main methods used to verify the location of drill sites: a surveyor enters the field with a GPS altimeter, or aerial drone images and lidar are captured. Both approaches present challenges. The first requires survey teams to operate in hazardous environments near open holes, exposing them to extreme heat and rough terrain. The second demands significant personnel time to process and analyse the data collected. Because the process remains largely manual, only around 2% of holes are actually verified, leaving potential GPS drift or equipment calibration issues undetected.
To address these limitations, the Curtin Institute for Data Science, in partnership with BHP and funded by the Resources Technology and Critical Minerals Trailblazer, designed a phased automation strategy to increase safety, accuracy and verification rates. The first phase focused on developing an expert system to validate drill actuals using post-drill aerial photography, lidar-derived digital elevation models and drill tables. This approach reframed the problem, moving away from scanning entire mages for all possible holes and instead using reported drill locations as reference points to narrow the search space. By extracting features from cropped imagery and DEM data around each reported location, the system could identify candidate drill holes, flag shaded or flattened collars and filter out unreliable data.
The expert system proved capable of automating detection of drill collars from drone or light aircraft imagery, differentiating holes to assign them to specific rigs and identifying calibration issues such as GPS drift. Validation against manually tagged data showed accuracy within approximately one pixel. In most tested cases, the workflow was able to use over 75% of drill holes to assess rig calibration, with verification rates frequently exceeding 70% and sometimes surpassing 95% which is a dramatic improvement over the previous 2% manual verification rate. This freed personnel from high risk field verification tasks, improved operational efficiency and provided BHP with a data driven means of monitoring equipment performance.
“This automated reporting represents a step function improvement for BHP operations!”
This internal feedback from BHP highlights the transformative impact of the project, which has now advanced to its second phase, with a third phase anticipated to further expand capability and scalability across the mining sector.