The paper presents an original approach to the localization and analysis of the cracking patterns of cement composites. The lime cement matrix modified with microsilica was evaluated under a two-phase thermal load. For quantitative detection and analysis of thermal cracks, an image-processing method was applied. For this purpose, an original image double-segmentation method was developed using machine-learning algorithms. Among other things, the fractal analysis was used to describe the morphology and the thermal evolution of the cracking patterns. The basic mechanical characteristics were examined and the results indicated a very high correlation between tensile strength and all cracking patterns' parameters. This allows high-quality estimation of the mechanical properties of the lime cement matrix to be carried out on the basis of measurement and evaluation of morphology of the thermal cracking patterns. Knowledge in this field contributes to the development of non-destructive testing methods in cement composites technology, in terms of localization of and tracking the cracking patterns.
Keywords: ImageJ; cement matrix; cracking pattern; cracks detection; fractal dimension; image analysis.