Automatic recognition of fabric structures based on digital image decomposition

Abstract

A method to recognize fabric structures automatically based on digital image decomposition has been introduced. The method includes establishing a Wiener filter adapted to the fabric texture. A woven fabric image can be decomposed into horizontal and vertical subimages by using this Wiener filter. These two subimages contain the weft and warp texture information respectively. After thresholding, the gray-level subimages are transformed into binary images, in which the weft or warp floats range periodically. Then the weaving density can be figured out. Based on the preceding work, the positional information of yarns in every single subimage can directly help to enclose every interlacing point. Considering the variety of gray value in each point unit, warp point and weft point can be distinguished. The basic structures for woven fabric (plain, twill and satin) have been evaluated and it is found that the density for woven fabric can be calculated exactly and the structures can be identified clearly.