Analysis of ore particles based on textural pattern recognition

Changes in mill feed conditions such as particle size distribution and ore type have a significant impact on grinding performance, but are process variables that have remained largely unexploited in control and optimisation strategies. This study documents the development of a machine vision strategy used to characterise particle features in ore feed systems. In contrast to algorithm based methods which explicitly target individual particle recognition, this paper explores a textural approach, viz. variance and range textural operators for ore type characterisation and surface particle size estimation. Promising preliminary results were obtained for mean particle size predictions on an industrial mill feeder.