Caracterização de minério de ferro por visão computacional

Abstract

Ore classification is useful for understanding and controlling industrial beneficiation processes. The present work contributes to the developement of a computer vision system for automated classification of iron ores by image analysis of optical micrographs. For this purpose, morphological feature extraction algorithms were implemented and feature based classification algorithms were developed. These methods were successfully applied to the classification of hematite grains into four classes determined by their shape (lamellar or granular) and porosity (porous and compact) as characterized by just four features, namely, circularity, aspect ratio, fiber ratio and porosity.