Expression of individual yarn for structure analysis of textile fabric based on segmentation of three dimensional CT image

Abstract

In this paper, a method for expressing individual woven yarns in a textile fabric is proposed for structure analysis of the textile fabric from its three-dimensional (3D) X-ray computed tomography (CT) image. In this method, the voxels in the 3D CT image are assigned to each yarn on the basis of the yarn positional information, which is a sequence of the yarn center points. In order to obtain the yarn positional information, the filament orientation is first estimated at all the voxels in the 3D CT image by correlating the voxel values with a filament model. Each filament is then reconstructed by connecting the voxels of the filaments on the basis of the estimated filament orientation. Each yarn is then reconstructed by binding the reconstructed filaments using a newly defined dissimilarity between two filaments. Finally, the yarn positional information is obtained by thinning the reconstructed yarns. The voxels in the 3D CT image are assigned to each yarn using newly defined dissimilarity between the voxel and the extracted positional information of the yarn. Using the assigned voxels, each individual yarn is expressed by the volume-rendering technique. The effectiveness of the proposed yarn expression method is discussed by experimentally applying this method to the 3D image of a plain knitted fabric.