Imaging and pattern recognition techniques applied to particulate solids material characterization in mineral processing

The characterization of particulate solids can be carried out by chemical and mineralogical analysis, or, in some cases, following a new approach based on the combined use of i) imaging techniques to detect the surface features of the particles and ii) pattern recognition procedures, to identify and classify the mineralogical composition on the bases of the previously detected “pictorial” features. The aim of this methodology is to establish a correlation between image parameters (texture and color) and physical-chemical parameters characterizing the set of particles to be evaluated. The technique was applied to characterize the raw-ore coming from a deposit of mineral sands of three different lithotypes. An appropriate number of samples for each lithotype has been collected. A vector of attributes (pattern vector), by either texture and color parameters, has been associated to each sample. Image analysis demonstrated as the selected parameters are quite sensitive to the conditions of image acquisition: in fact optical properties may be strongly influenced by physical condition, in terms of moisture content and optics set-up and lighting conditions. Standard conditions for acquisition have been selected according to the in situ conditions during sampling. To verify the reliability of the proposed methodology, images have been acquired under different conditions of humidity, focusing and illumination. In order to evaluate the influence of these parameters on image pictorial properties, textural analysis procedures have been applied to the image acquired from different samples. Data resulting from the processing have been used for remote controlling of the material fed to the mineral processing plant.