ADVANCES IN OPTICAL IMAGE ANALYSIS AND TEXTURAL CLASSIFICATION OF IRON ORE FINES

摘要

Optical image analysis (OIA) of micro-porous minerals can result in significant misidentification. Interference between porosity and brighter minerals results in the appearance of darker minerals that are not actually present and partial disappearance of porosity. Misidentifications can also occur due to optical interference/diffraction from certain textural structures. Such misidentifications can significantly distort mineral abundance, liberation and association data, and affect the results of automated textural classification. The novel CSIRO OIA package Mineral3, originally developed for OIA of iron ores and successfully used for other minerals, contains a set of procedures that allow full elimination or significant reduction of such misidentifications in automated mode without any intervention from the operator. These procedures and a newly developed method for more reliable particle separation can be tailored to specific applications depending on factors such as the mineral and textural peculiarities of specific ores, magnification, and particle size. This enables reliable automated textural classification using CSIRO Recognition3 software. Comparisons of OIA results of mineral identification with and without application of these procedures at different magnifications have been performed and are discussed in the paper. The new capability to develop a classification scheme within the Recognition3 package is introduced. This capability significantly simplifies the process of classification scheme development and makes the classification process very flexible. The Mineral3/Recognition3 software suite can automatically process large sets of images obtained automatically within Mineral3, or any other images; for example from QEMSCAN or MLA, to perform textural classification and comprehensive characterisation of studied ore samples.