Automatic Ore Texture Analysis for Process Mineralogy

Geologists often classify ore textures on the basis of the origin of the rock in which they form and the grade of one or more minerals within the ore (Sinclair and Blackwell, 2002, pp 32-36).  Metallurgists, on the other hand, need to know how best to process the ore in order to extract the valuable minerals from it (Petruk, 2000). Geological classification has the following disadvantages:it requires the services of skilled economic geologists, of whom the supply has been dwindling for many years;it is subjective, depending on the perceptions of the individual geologists who make the classification; andit may relate only weakly to processing requirements of the ore. We describe methodology for automatic classification of ore textures from images of drill core. The images analysed are æclassifiedÆ images, in which a particular mineral (or mineral group) is assigned to each pixel. The new methodology is intended to:provide a quantitative and reproducible alternative to traditional geological texture classification,relate to process mineralogy of the ore, andimprove the ore-texture-related input to orebody domaining and geostatistical block models.