Detection of Fabric Defective Areas Based on Clustering and Morphological Operators

At various stages of fabric production, there are numerous damages to the surface of the fabric. Regardless of the causes of the failures, precise identification of their types helps to correctly classify the fabric and thus provides a high percentage of the quality control process. Quality control of fabrics is of great importance in order to improve product quality and maintain a competitive market. Identification of faulty areas in automated methods is of great importance. In this paper, a new method is presented in the clustering of faulty zones based on clustering as well as morphological operators. In the proposed method, after preprocessing necessary to improve image quality, the first step is to cluster the image to create similar areas. Then morphological operators are applied to extract the defective area. The defective area is represented by the active contour algorithm. Although many methods such as local binary patterns and other methods have been proposed, the speed of detection of these algorithms is low and has high computational complexity. The proposed method is implemented on the CMUPIE database and evaluated using accuracy assessment criteria and accuracy criteria. The accuracy of identifying defective areas in the proposed method is 93.82%, and the precision of detecting defective areas in the suggested method is 98.33% which are significantly improved compared to similar methods.