Javeed, Umer1,2, Hill, June1,2 and Thomas, Matilda1,3
1MinEx CRC; 2CSIRO Mineral Resources, Kensington, WA, Australia; 3Geoscience Australia, Canberra, Australia
Rapid and reliable core logging plays a vital role in the discovery of ore deposits. Drilling operations produce large repositories of rock core, which are used to determine the lithology, mineralogy, structure and alteration zones of the core. Geologists have a limited time to analyse the core and the logging process is subjective and may vary depending upon the experience of the geologist. In mineral exploration the identification of veins and characterisation of their form is important for prediction of the presence of vein-hosted mineral deposits and understanding the mineralisation processes. Therefore, the collection of vein information which is accurate and consistent is critical. Red Green Blue (RGB) coloured core images are routinely collected but are mostly used only for visual inspection and record keeping. Automated core analysing systems have the potential to provide rapid and consistent quantitative assessment of core. In this paper, a methodology to detect the veins and fractures by using computer vision techniques is proposed. In this method a simple approach is used to separately define the core from the core tray material. Undesired objects or material are removed from core images by using intensity transformation and morphological image processing operations. If the input image has poor contrast, then contrast stretching techniques are used to enhance the texture of the core. The performance of individual or combined edge detection filters have been analysed to detect veins and fractures in core. The results and effectiveness of these workflows are visually compared with the original core images. The proposed workflows will provide an effective solution to pre-processing of core images prior to the application of sophisticated machine learning techniques to quantify and categorize structural information.
Umer specialises in designing workflows for machine learning and computer vision based core image analysis. June Hill specialises in automating the analysis and interpretation of drill hole data and core images. Matilda is the Project Leader for MinEx CRCs Project 7: Maximising the value of Data and Drilling Through Cover.