Sistema de visão computacional para a caracterização da grafita usando microfotografias

Abstract

Caterials Sciences field uses Computational Vision systems to determine size and/or amount of grains, welding control, modeling of alloy elements, among other. The present paper has as main objective to develop and validate the SVRNA system (Microstructure Segmentation for Computational Vision based on Artificial Neural Networks), which, combined with Arti¯cial Neural Network, uses mathematical morphology technics to accomplish the constituent segmentations from white cast iron of semi-automatic form, and graphite automatic classi¯cation from nodular, malleable and gray cast iron. Segmentation and quanti¯cation results of this materials are compared between SVRNA and a commercial program more used in this domain. Comparative analysis between this methods showed that SVRNA present best results. It has concluded, therefore, which the proposed system can be used in applications in Material Sciences field for microstructure segmentation and quantification in metallic materials, reducing the analyze time, and obtained accurate results.