Quantitative Mineral Assessment of Apatite, Calcite/Dolomite, and Phlogopite Powder Mixtures by Using VSWIR Reflectance

Quantitative assessment of mineral powders made from mono/two-mineral blocks sampled from the SiilinjÄrvi apatite ore deposit is the target of this paper. The blocks were composed of apatite, calcite/dolomite, and phlogopite. The blocks were powdered and mixed into a set of 21 well-documented mixtures. Their visible to short-wave infrared (VSWIR) reflectance spectra were recorded by using the FieldSpecFR Spectrometer. Linear unconstrained unmixing of the spectra resulted in highly biased estimates in spite of the fact that the mean absolute residual error of the unmixing was very small. The bias, which is actually induced by camouflage between mineral classes, could approximately be removed by a polynomial function. This resulted in unbiased estimates for apatite calcite/dolomite and phlogopite. Best results were obtained when the wavelengths were limited to the SWIR area only. This two-part procedure consisting of the linear unmixing and the camouflage correction gives one more aspect to subpixel classification. Training data are given not only for the classes but also for their interaction. In this paper, the mean absolute residual error of the estimated volumetric fractions of apatite, calcite/dolomite, and phlogopite from the spectra was 1.32%, which is smaller than the result in the previous VSWIR studies concerning the assessment of mineral powders. This reflectance method offers a very quick tool for several mineral-related applications.