An intelligent detection and assessment method based on textile fabric image feature. Issue 3 (3rd June 2019)

• Record Type: Journal Article
• Authors: Zhao, Xueqing; Shi, Xin; Liu, Kaixuan; Deng, Yongmei
• Is Part Of: International journal of clothing science and technology. Volume 31:Issue 3(2019)
• Page Start: 390
• Page End: 402
• Publication Date: 2019-06-03
• Subjects: Intelligent detection -- Textile fibers; Clothing and dress -- Periodicals; Textile fabrics -- Periodicals; 677
• DOI: 10.1108/IJCST-01-2018-0005
• Languages: English
• ISSNs: 0955-6222

Abstract

Abstract : Purpose: The quality of produced textile fibers plays a very important role in the textile industry, and detection and assessment schemes are the key problems. Therefore, the purpose of this paper is to propose a relatively simple and effective technique to detect and assess the quality of produced textile fibers. Design/methodology/approach: In order to achieve automatic visual inspection of fabric defects, first, images of the textile fabric are pre-processed by using Block-Matching and 3-D (BM3D) filtering. And then, features of textile fibers image are respectively extracted, including color, texture and frequency spectrum features. The color features are extracted by using hue–saturation–intensity model, which is more consistent with the human vision perception model; texture features are extracted by using scale-invariant feature transform scheme, which is a quite good method to detect and describe the local image features, and the obtained features are robust to local geometric distortion; frequency spectrum features of textiles are less sensitive to noise and intensity variations than spatial features. Finally, for evaluating the quality of the fabric in real time, two quantitatively metric parameters, peak signal-to-noise ratio and structural similarity, are used to objectively assess the quality of textile fabric image. Findings: Compared to the quality between production and pre-processing of textile fiber images, the BM3D filtering method is a very efficient technology to improve the quality of textile fiber images. Compared to the different features of textile fibers, like color, texture and frequency spectrum, the proposed detection and assessment method based on textile fabric image feature can easily detect and assess the quality of textiles. Moreover, the objective metrics can further improve the intelligence and performance of detection and assessment schemes, and it is very simple to detect and assess the quality of textiles in the textile industry. Originality/value: An intelligent detection and assessment method based on textile fabric image feature is proposed, which can efficiently detect and assess the quality of textiles, thereby improving the efficiency of textile production lines.