A novel method based on near-infrared imaging spectroscopy and graph-learning to evaluate the dyeing uniformity of polyester yarn
Zheng Liu,Shiluo Huang,Wei Jin,Ying Mu
DOI: https://doi.org/10.1016/j.engappai.2024.107912
IF: 8
2024-05-01
Engineering Applications of Artificial Intelligence
Abstract:The quality of polyester yarn is mainly affected by its dyeing uniformity. As a result, textile manufacturers need to inspect the dyeing uniformity of polyester yarn. The existing inspection methods rely on the dyeing process which can be viewed as a way to extract the discriminative information of the polyester yarns with different dyeing uniformities. However, the dyeing process is extremely time-consuming and laborious. It is very expensive to obtain the dyeing uniformities of polyester yarns, due to the inefficient dyeing process. To address this problem, an efficient and effective inspection method based on near-infrared imaging spectroscopy and semi-supervised learning is introduced in this paper. This introduced method can obtain the dyeing uniformity of polyester yarn without the dyeing process, which greatly boosts the efficiency of inspection. In the developed method, to acquire the representative information without the dyeing process, near-infrared spectra are applied to represent the molecular structure differences of different polyester yarns. Considering that the dyeing uniformities are expensive to obtain, a novel semi-supervised learning algorithm, termed consensus local graph based on multiple kernel learning (CLGMKL), is proposed in the introduced method to evaluate the dyeing uniformity of polyester yarn. Since redundant wavebands and noise exist in the collected spectra, the relationship among spectra in CLGMKL is learned based on the extracted low-rank matrices which contain the discriminative information of base kernels. Besides, CLGMKL incorporates the label information of spectra. Finally, extensive experiments on the collected spectra of polyester yarns demonstrate the superiority of the proposed method.
automation & control systems,computer science, artificial intelligence,engineering, electrical & electronic, multidisciplinary