SEM ADVANCED IMAGE ANALYSIS TOOLS APPLIED TO CEMENT AND CONCRETE MICROSTRUCTURE: A REVIEW OF 'SPECTRO-SPATIAL' STRATEGIES AND CASES OF APPLICATION

摘要

We describe a methodology to get phase classification from scanning electron microscopy (SEM) using combination of backscattered image (BSE) and fast X-ray maps of selected chemical elements. We apply this in the field of cement and concrete based materials. This development uses a set of tools, which are well known in satellite image classification, which combine in the same interface a classifier based on support vector machine (SVM), an advanced filtering based on Markow random fields (MRF) and the construction of binary partition tree allowing the user to modify the final classification. We then illustrate 3 different cases of application. The first one is related to quantification of amorphous and mineral phases in selected cementitious materials (more and more used in new binders). The second example is related to the hydration degree measurement in new binders containing ordinary Portland cement (OPC) and these supplementary cementitious materials (SCMs). Last we measure the evolution of phases in function of depth for mortars that have been degraded in sulphate solution.