Fiber-level structure recognition of woven textile

In this paper, we present a novel automatic method for woven textile structure recognition in fiber-level. This method is based on digital image analysis techniques. It allows automatic weft yarn and warp yarn crossed-area segmentation through a spatial domain integral projection approach. Secondly, by applying unsupervised fuzzy c-means clustering on multi-scale direction invariant texture features based on gray level co-occurrence matrix, we can classify detected segments into two clusters. Finally, using a fuzzy rule based analysis on texture orientation features, the yarn crossed-area states are automatically determined. To verify the validity of this method, a number of textile images are used. The samples we choose have different weave types: plain or twill, different fiber types and yarn counts. The recognition results match the actual structure of tested samples. A possible 3D representation of woven structure identified is also shown in this paper.