Evaluating fabric pilling/wrinkling appearance using 3D images
Abstract
Fabric appearance is usually the highest priority consideration for consumers. Pilling and wrinkling are two major factors which cause the fabric to have a worse appearance after a certain service period. In order to prevent more piling and wrinkling, a fabric pilling and wrinkling severity evaluation is very important. Traditional visual examination needs at least three trained experts to judge each sample, which is both subjective and time-consuming. Objective, high efficiency, and automatic pilling and wrinkling evaluation based on computer processing techniques are now being developed quickly. In this study, an integrated fabric pilling and wrinkling measurement system based on stereovision was developed. The hardware part of the system consists of a pair of consumer high resolution cameras and a mounting stage, which is affordable and portable in comparison with other 3D imaging systems. A novel pilling detection algorithm focusing on 3D image local information was proposed to extract three pilling features including pilling density, pilling average height, and pilling average size. The logistic regression classifier was applied for pilling severity classification because it showed a good accuracy with 80% on the 120 3D pilling images. A fast wrinkle detection algorithm with leveled 3D fabric surface was developed to measure wrinkle density, hardness, tip-angle, and roughness. According to these four wrinkling features, 180 3D wrinkling images were tested by the logistic regression classifier with an overall 74.4% accuracy in comparison with visual judging results. Both pilling and wrinkling results obtained from the proposed automatic 3D fabric pilling and wrinkling severity evaluation system were consistent with the subjective visual evaluation results. The system is ready for practical use.