Application of Texture Filtering in Clustering of X-ray Computed Tomography Data of Products Made from Polymer Composite Materials

Abstract

X-ray computed tomography (XCT) is one of the most informative methods of nondestructive testing of polymer composite materials (PCMs) and products made of them. One of the important stages of the XCT of PCM products is segmentation, the automation of which is of research interest. In the segmentation process, it is important to identify isotexture zones containing local X-ray density variations. In this paper we investigate the possibilities of three-dimensional texture filtering (Gaussian filter, Gabor filters) in clustering of X-ray computed tomography data by simple linear iterative clustering (SLIC) algorithm and evaluated their efficiency in terms of parameters: the share of mismatch between the boundaries of clusters and the boundaries of segmented areas and sphericity of clusters, as well as the performance in terms of the time to partition the dataset into the required number of clusters. The results of the study show that the application of three-dimensional texture filters improves the clustering accuracy and sphericity of isotexture clusters of PCM product XCT data without any considerable increase in clustering time compared to the raw data. The maximum increase in clustering accuracy was observed when using a combination of Gaussian and Gabor filters, while clustering time increased.