Automatic recognition of hematite grains under polarized reflected light microscopy through image analysis

The recognition of hematite grains is an intermediate task that aids the texture characterization of iron ores. Hematite is a strongly anisotropic mineral. Thus, the combined use of a polarizer and an analyzer in reflected light microscopy (RLM) can be used to obtain images that present sufficient contrast to differentiate grains. The present work proposes a methodology for recognizing hematite grains in images obtained with RLM. Three images per field are acquired in different conditions: without polarization in common bright field arrangement; and with polarization under two symmetrical polarizer/analyzer angles. These images are automatically registered. Then, the hematite grains are recognized through a modified region growing segmentation method based on reflectance and textural information. An optimal value for the polarization angle was determined. The results are promising. The vast majority of grains was correctly recognized. The automatically segmented images were compared to edited versions in which all crystals were correctly discriminated. A statistical comparison of crystal size and shape showed no statistical differences, to within 99% confidence, between automatic and edited segmentation results.