Raw ore selection by artificial vision

Abstract

Some deposits of inhomogeneous materials, which occur near the surface and are exploitable by open-pit mining, may be characterized in real time prior to mining through ground-surface imagery. If the slice to be stripped is not overly thick, surface texture and color properties may be extrapolated to its entire thickness. Image-analysis techniques for processing ground-surface images acquired in situ or in the laboratory shortly after their acquisition have been developed. These images yield pattern vectors representative of red, green and blue (RGB) color component distributions and hue, saturation and brightness (HSB) texture parameters. The techniques were applied to a sandy ore deposit containing three lithotypes. Geo statistical analysis indicated that the data used to characterize the lithotypes were reliable. The correct recognition of the lithotypes was carried out using a multibarycenter classification algorithm. Such in situ image-analysis procedures provide a means for selecting the ore to be mined, selecting the proper ore-processing method and determining the appropriate blend to be processed. This can be done either before or during mining.