Universidade de Fortaleza (UNIFOR), Centro de Ciências Tecnológicas (CCT), Núcleo de Pesquisas Tecnológicas (NPT), Av. Washington Soares, 1321, Sala NPT/CCT, CEP 60.811-905, Edson Queiroz, Fortaleza, CE, Brazil.
Microsc Res Tech. 2011 Jan;74(1):36-46. doi: 10.1002/jemt.20870.
Quantitative metallography is a technique to determine and correlate the microstructures of materials with their properties and behavior. Generic commercial image processing and analysis software packages have been used to quantify material phases from metallographic images. However, these all-purpose solutions also have some drawbacks, particularly when applied to segmentation of material phases. To overcome such limitations, this work presents a new solution to automatically segment and quantify material phases from SEM metallographic images. The solution is based on a neuronal network and in this work was used to identify the secondary phase precipitated in the gamma matrix of a Nickel base alloy. The results obtained by the new solution were validated by visual inspection and compared with the ones obtained by a commonly used commercial software. The conclusion is that the new solution is precise, reliable and more accurate and faster than the commercial software.
定量金相学是一种确定和关联材料微观结构与其性能和行为的技术。通用的商业图像处理和分析软件包已被用于从金相图像中定量材料相。然而，这些通用解决方案也有一些缺点，特别是在应用于材料相的分割时。为了克服这些限制，这项工作提出了一种从扫描电子显微镜金相图像中自动分割和定量材料相的新解决方案。该解决方案基于神经元网络，并且在这项工作中被用于识别镍基合金中γ 基体中析出的次生相。新解决方案获得的结果通过目视检查进行了验证，并与常用商业软件获得的结果进行了比较。结论是，新解决方案比商业软件更精确、可靠、更准确和更快。
