Texture-Based Garments Defect Detection Method Using Machine Learning

Fabric defects identification is significantly important in textile industry. Various activities are involved in detecting fabric quality. The total profits of the industries depend on the defects on the quality of fabrics that produces loss. Existing defect identification methods are conducting in different organizations with the help of human defect inspectors who draw manually defect pattern. However, these detection techniques have few drawbacks like complication, inaccuracy, negligence, tediousness, and exhaustion in addition to time-consuming which causes decrease performance in failure finding. For the sake of overcoming these problems, a framework that depends on processing of image has been introduced to efficiently and effectively identify the defects of fabrics. The given introduced work consists of image segmentation, morphological methods, feature extraction, and classification. We have used edge detection for image segmentation, ICA for feature extraction, and SVM for image classification. The results achieved from the given framework are superior than existing framework. The experiments have been conducted in Python environment using TILDA datasets.