Análise metalográfica de uma liga de ferro fundido nodular através do processamento de imagens

Abstract

During many years mankind keeps improving the many existent iron alloys according to it’s utility, availability and properties, making it necessary to know as much information as possible regarding the analyzed alloy, as well as obtaining information with quality. One method of acquiring this information is through the metallographic study, which is to perform measurements related to the microstructure of the iron alloy. However, in some industries these measurements are performed manually and visually by a technician, totally or partially, reducing their reliability. In order to offer technological support to the technique that is currently employed, a method is proposed using a segmentation algorithm based on thresholding, capable of automatically performing the segmentation in nodular cast iron alloy images on a micro scale, as well as the calculation of several measures needed for the characterization of the alloy in study. In order to achieve this goal, pre-processing and segmentation algorithms were studied. Then a algorithm was developed that could be capable to deal with non-uniformities in lighting, noise and perform segmentation of regions of interest using local thresholding, and posteriorly performing morphological analizys on the segmented regions in order to extract data about them, so they could be utilized in next stages in the metallographic analizys in the foundry industry.