Abstract
Sorting of minerals based on near infrared (NIR) analysis is promising because many minerals have distinct “fingerprints” in the NIR region. An experimental system was customised in order to obtain accurate NIR reflectance spectra. As a test application, the NIR spectra of transects across pre-classified copper ore particles were measured. Matrices containing correlation coefficients of particle pixel–pixel spectra were subjected to principal component analysis. In addition to providing insight into the homogeneity of the particle surface, the results were used to identify key spectral features which could be used to sort product from waste particles. With a selected spectral range, it was found that the classification improved when two standard pixels relating to product and waste were inserted into the transects. While testing an equal number of particles from each rock type, it was found that a correct classification was made in 82% of all rocks. It was found that moisture had little to no effect on the sorting method.
Get full access to this article
View all access options for this article.
