Core logging is often carried out by many geologists with different levels of skill and experience who collectively do not interpret and record their observations in the same way over the life of a project. This has traditionally led to logged datasets often being viewed as inconsistent and lacking any real auditability. Computer vision techniques address some of these issues by extracting geological information from drill core imagery with significantly improved consistency and detail compared to current manual logging. Veins are one of the most neglected and poorly logged datasets due to their complexity, scale, and volume throughout an ore body. Consequently, logged vein data often comprises subjective estimates or averages and sub-sampled detail that is inconsistently logged. Using computer vision-based techniques, we analyse traditional RGB core photography to generate new types of high-resolution vein data including vein morphology segmentations at the pixel scale as well as vein area and percentage. These novel vein data are then interrogated in 3D to demonstrate how these high-resolution vein characteristics can provide new geological insights and improve ore body knowledge. This work demonstrates the advantages of computer vision logging techniques based on their ability to create new types of logging data with improved consistency. Computer vision logging also benefits geologists by allowing them to move beyond routine and repetitive work and focus more on higher level technical tasks.

Open-Access Online Publication: May 29, 2023