Abstract
Defect detection is of great significance for assessing and controlling the quality of fabrics. However, most traditional detection processes rely on manual visual inspection, resulting in low detection efficiency, ambiguous detection results, and high monitoring costs. In this work, a centroid warp-weft graph-based (C2WG) statistical analysis method is proposed for the detection and evaluation of fabric defects. To reflect the fabric texture variation, the C2WG method is first proposed to find abnormal texture centers. Subsequently, by dual monitoring of local slope and curvature, the location of the abnormal centroid can be accurately determined as texture defects and displayed. Finally, the defect evaluation results under different detection accuracy are obtained by changing the monitoring threshold. Consequently, the defects are classified into different classes. A case study on an industrial design fabric product validates the good performance of the proposed method.