Automatic Evaluation of Fabric Pilling Using a 3-D Non-contact Scanning System

To comprehensively understand the fabric pilling phenomena and exactly grade the degree of pilling, the overall fabric surface ruggedness as well as pill characteristics such as pill number, area, and population density should be evaluated with a 3D noncontact scanning system, which obtains 3D surface data with high accuracy. The fractal dimension calculated by a wavelet-fractal method and the standard deviation of mean curvature are used as descriptors of fabric surface ruggedness. Wavelet reconstruction makes it easier the localization and characterization of pills. Karhunen-Loeve (KL) transform is employed to reduce dimension and speed the pilling grading procedures. Bayes classifier, minimum distance classifier, k-nearest neighbors classifier, and neural network are applied to the fabric pilling grading. The experimental results show that the automatic fabric pilling measurement system proposed in this study is effective and feasible