Identification of microscopic structures in CFRP from X-ray CT based on topological data analysis

X-ray computed tomography (X-ray CT) is widely used to examine the internal structure of carbon fiber-reinforced plastics (CFRPs), yet challenges remain due to the trade-off between field-of-view and spatial resolution. This study presents a framework based on topological data analysis (TDA) using persistent homology to identify carbon fiber positions and cracks in CFRPs from X-ray CT images. Degree-1 persistence diagrams (PD1), obtained via both superlevel-set and sublevel-set filtrations of grayscale X-ray CT images, enable the identification of fiber positions while avoiding artifacts in the resin matrix. The framework is also applied to the detection of particle positions in a scanning electron microscopy (SEM) image. Comparative evaluation with watershed and local thickness algorithms demonstrates the advantages of our method. Segmentation of crack regions is also achieved through a refined application of PD1 using superlevel-set filtration. This framework is applicable to a variety of grayscale images beyond X-ray CT.