Classification of hematite types in iron ores through circularly polarized light microscopy and image analysis

摘要

Brazilian iron ores are predominantly hematitic and may have different textures. In the mining industry, their microstructural characterization is manually performed, by analyzing samples under an optical microscope to identify the hematite textures and estimate'their fractions and crystal size. This procedure is subjective and consequently susceptible to random and systematic errors. The present paper proposes an automatic method for the identification, measurement and classification of hematite crystals in iron ore according to their textural types. The method exploits the use of circularly polarized light to amplify brightness and color differences among hematite crystals, allowing their individualization, and the subsequent morphological analysis and classification into granular, lamellar or lobular classes. The classifier was tested with more than 5400 crystals, reaching a global success rate close to 98%, and success rates per class above 96%.