Digital multispectral techniques and automated image analysis procedures for industrial ore modelling

Abstract The choice of suitable beneficiation strategies is strictly linked to the precise mineral-petrographic, textural and structural characterisation of the ore. Careful analysis and precise modelling of spatial relationships between different mineralogical phases constituting the ore is the basis of any procedure which aims at forecasting the separation results. The textural and structural characterisation of the ore on a macro/microscopic scale is carried out through optical microscopy and in some special cases by SEM analysis. The main purpose of these procedures is that of obtaining information which can be used in numeric form as data for the models. The traditional procedure consists of analysing sections under the microscope to obtain distribution maps for the different mineral phases and making a synthesis of these data so that they can be easily managed inside a numeric procedure. In the last few years the development of procedures based on techniques of optical image processing has greatly reduced the analysis time, allowing a better characterisation of the ore in textural and structural terms. The growing development of research in electronics and computer science and the subsequent availability of hardware and software products, allow handling and processing of full colour digital images, at lower and lower costs. In this paper the problems arising from the adoption of such a digital approach both in terms of quality of results (mineralogical species automatic identification) and in terms of further processing of the data (morphological and morphometrical characteristics and assessment of the mineral species constituting the ore), are described and discussed.