Classificação automática do tipo de ferro fundido utilizando reconhecimento de padrões em imagens de microscopia

• Main Author: Rodrigues, Douglas de Araújo
• Publication Date: 2017
• Other Authors: Santos, Gesilane Pereira dos, Fernandes, Marcelo Correia, Santos, José Ciro dos, Freitas, Francisco Nélio Costa, Rebouças Filho, Pedro Pedrosa
• Format: Article
• Language: por
• Source: Repositório Institucional da Universidade Federal do Ceará (UFC)
• dARK ID: ark:/83112/0013000003cht
• Download full: http://www.repositorio.ufc.br/handle/riufc/59520
• topic: Ferro fundido; Processamento digital de imagens; Reconhecimento de padrões; Inteligência computacional aplicada

The cast iron has a microstructure in which the shape of the contained graphite has direct influence in the classification between the several types of this material. The classification is usually made through a visual analysis performed by a specialist using an optical microscope. This work proposes the use of Applied Computational Intelligence in conjunction with extractors of information in metallographic images. The purpose is to assist professionals in the field of Materials Science in the classification of cast iron automatically and reduce the time for classification, restricting as much as possible the faults presented during classification. Two steps were performed for the analysis. In the first one we considered the graphites separately. In a second step, we analyzed the complete image considering all the graphite extracted from it, in which the type of object with greater incidence in the analysis would be the recognition adopted for the whole sample. In both steps, the classifier Support Vector Machine obtained the best results in the recognition of the type of cast iron, with results close to 100%, with a mean reduction of the classification time by 92%. Both the results and the time of the classifications are compared with the specialist's analysis, as well as the results obtained in cast iron classifications that use a neural network approach and a supervised classification using only the shape descriptors. From the results presented, we concluded that the approach is promising and can incorpo-rate commercial software to assist specialists in the field