Advanced Segmentation for Industrial Materials Using Machine Learning

Every quantitative analysis of a set of micrographs involves some form of segmentation. A region could be a mineral fragment, a grain in a metal, a pore in a composite, an oil contamination on the surface, a blood cell – any area differentiable from a neighboring area. By analysing these regions or the borders between regions, we get useful information. There are several standards for determining useful microstructural properties by image analysis of segmented images – e.g. for grain size, inclusion content, and porosity. These analyses, however, depend on the accuracy and reliability of the segmentation results.