Estimation of uncertainty of concentration estimates obtained by image analysis

Images from the scanning electron microscope (SEM) and the optical microscope are widely used in mineral processing to estimate the concentrations of mineral species. The aim of this study was to optimize the procedure used for the determination of concentration estimates from SEM images of ore and rock samples. The goal was especially to develop a method that would allow the estimation of the concentration uncertainty when the image analysis is still in progress. For this study, samples consisting of chromite (FeCr2O4, density 4.7 g cm−3) and quartz (SiO2, density 2.7 g cm−3) were prepared. The chromite concentration in the sample materials varied from 0.1 to 90 wt%. Four replicate microscopic samples for each chromite concentration were prepared and analysed. The total number of samples was 36. Two hundred images and 20 000–30 000 particles per microscopic sample were analysed. Estimation of the sampling error was based on Pierre Gy's sampling theory. Several error components were studied and the error estimates were found to be relatively reliable when they were based on the fundamental sampling error. Copyright © 2002 John Wiley & Sons, Ltd.