Fabric wrinkle characterization and classification using modified wavelet coefficients and support-vector-machine classifiers

Abstract

In this paper we present a novel wrinkle evaluation method that uses modified wavelet coefficients and optimized support-vector-machine (SVM) classifications to characterize and classify the wrinkling appearance of fabric. Fabric images were decomposed with the wavelet transform, and five parameters were defined, based on the modified wavelet coefficients, to describe wrinkling features, such as orientation, hardness, density, and contrast. These parameters were also used as the inputs of optimized SVM classifiers to obtain overall wrinkle grading in accordance with the standard American Association of Textile Chemists and Colorists smoothness appearance (SA) replicas. The SVM classifiers, based on a linear kernel and a radial-basis-function kernel, were used in the study. The effectiveness of this evaluation method was tested by 300 images of five selected fabrics that had different fiber contents, weave structures, colors, and laundering cycles. The cross-validation tests on the SA classifications indicated that the SA grades of more than 75% of these diversified samples could be recognized correctly. The extracted wrinkle parameters provided useful information for textile, appliance, and detergent manufactures to inspect wrinkling behaviors of fabrics.

A Projector-Camera System for Ironing Support with Wrinkle Enhancement

“…Let horizontal, vertical, and diagonal wavelet coefficients of k-th wavelet transformation be C Hk , C V k , and C Dk , respectively. Based on the finding of Sun, et al [18] in intensity, scale, and direction among the six levels of AATCC smoothness appearance replicas. So, C H4 , C V 4 , C D4 , C H5 , C V 5 , and C D5 are utilized to calculate four features: energy, hardness, density, and contrast.…”

“…Therefore, research on automatic wrinkle detection based on visual features has been underway. Features representing the change of the fringe pattern observed by irradiation with slit light [14], features representing the unevenness of pixels or the ratio of edges in an image from an RGB camera [21], and features obtained from an infrared camera (IR camera) have been considered to recognize wrinkle [18].…”

“…The first class of features are derived from pyramid-structured wavelet decomposition, which were proposed by Sun, et al [18]. The method relies on the characteristics in which wrinkles normally appear in lower frequency domain than the texture of clothes such as tucks and yarns.…”

“…We calculated an LFD against three contiguous scales by Formula (6), and obtained five LFDs from the following triples: (3, 4, 5), (5, 6, 7), (7,8,9), (9,10,11), and (17,18,19). The notation LF D k represents an LFD calculated from triples of (k-1, k, k+1), where k=4, 6, 8, 10, and 18.…”

Data mining and machine learning in textile industry

“…SVM has been applied to different textile problems, such as predicting fabric type 29 and fabric parameters, 30 predicting yarn properties, [31][32][33] fiber identification, 34 and color management 35 in textile printing and dyeing. Similar to the ANN, SVM has also been applied for quality management 36,37 in the textile sector, especially for defect classification 38 in textile textures.…”