Fabric Texture Analysis and Weave Pattern Recognition by Intelligent Processing

Abstract

Coimbatore is a major city in the Indian state of Tamil Nadu located on the banks of the Noyyal River surrounded by the Western Ghats. It is one of the biggest centers of textile manufacturing in India. A fast-growing metropolitan area city, it is home to over 25,000 textile and manufacturing companies and has spawned many new centers of textiles around it. Textile fabric automation and manufacturing has been of great concern over the past decade. This is a remarkable task because of the accidental changes of fabric material properties. Due to the increasing demand of consumers for high-quality textile products, an automatic and objective evaluation of the fabric texture appearance is necessary with respect to geometric structure characteristics, surface, and mechanical properties. The precise measurement of the fabric texture parameters, such as weave structure and yarn counts find wide applications in the textile industry, virtual environments, e-commerce, and robotic telemanipulation. The weave pattern and the yarn count are analyzed and determined for computer simulated sample images and also for the scanned real fabric images. 2-D integral projections are used to identify the accurate structure of the woven fabric and to determine the yarn count. They are used for segmenting the crossed areas of yarns and also to detect the defects like crossed area due to the random distribution of yarns. Fuzzy C-Means Clustering (FCM) is applied to multiscale texture features based on the Grey Level Co-Occurrence Matrix (GLCM) to classify the different crossed-area states. Linear Discriminant Analysis (LDA) is used to improve the classifier performance.