Machine vision of textile testing and quality research

Abstract

In today’s globalized economy, various industries are promoting product transformation and upgrading. The textile industry is also facing a harsh international situation and fierce market competition. While constantly promoting the upgrade of automation equipment, the original quality control mode relying on manual testing has been unable to meet the modern production requirements and the market demand for product quality. This paper investigates the product inspection and quality control in the textile industry at home and abroad, and puts forward the application of machine vision technology in textile automated inspection and quality control, so As to strengthen the product quality control system and improve the product’s physical quality. By studying the composition of machine vision detection technology, this paper studies the two core technologies of image acquisition and image processing in machine vision, summarizes and analyzes the common defects of textile Products, and proves that the common defects can be detected and repaired in time by machine vision detection method.

Keywords

Machine vision defect automatic detection image processing textiles

1 Introduction

At present, most domestic textile companies still use traditional manual testing for quality control methods for product appearance defects. The defect of this detection method is that the product quality control ability is unstable, and it is greatly influenced by subjective consciousness [1 –4]. The quality control of products mainly depends on the quality awareness, responsibility, operational proficiency, work ability and physical state of the test personnel. According to relevant research, the time for a tester to continuously maintain the best working condition is less than 30 minutes. According to the difference in the ability of the tester, the detection rate of the defective product is controlled between 45% and 65% [5 –8]. The product defect rate caused by textile defects accounts for about 85% of the overall non-performing rate of the textile industry. Therefore, the detection and quality control of textile appearance defects is the most important link in textile quality control [9, 10]. With the continuous upgrading of automated production equipment, the production efficiency of textiles continues to increase, and the quality control model relying on manual inspection has been unable to meet increasingly stringent product standards and modern production requirements. The purpose of this paper is to make rational use of machine vision applications, timely and accurately identify textile defects in the textile production process and repair them to avoid product quality loss caused by appearance defects [11 –15].

2 The status quo of textile testing research

2.1 Research purposes and significance

As one of the pillar industries of China, the textile industry occupies a considerable proportion in China’s social and economic construction. At the same time, China’s textile industry is also facing a harsh international situation and fierce market competition. The increasing product standards and tariffs of developed countries in the international market restrict the development of enterprises; many domestic enterprises are fiercely competitive, and efficient automated production lines are put into use. The supply and demand imbalance in the textile market. In this market environment, low-price competition between enterprises has seriously affected the healthy and orderly development of China’s textile industry. How to strengthen the quality control system of products, improve the physical quality of products, and rely on excellent quality to occupy the market, “winning by quality” has become the development trend and urgent demand of China’s textile industry. There are many kinds of textile products. The testing standards and quality control points of each product are different according to the production raw materials, production process and product use [16 –18]. However, the appearance quality of the products is definitely the focus of product quality control. At present, most domestic textile enterprises still adopt the most traditional detection methods for product appearance detection methods and quality control methods. Through manual visual inspection, the control of product appearance defects mainly depends on the testers’ own capabilities. Including the quality awareness, responsibility, operational proficiency, work ability and physical state of the inspectors. With the continuous updating of automation equipment, the quality control model relying on manual inspection has been unable to meet the modern production requirements and the market demand for product quality [19 –22].

Defects in the appearance of textiles directly affect the appearance of the product, and may also increase the processing difficulty of the subsequent production process. In serious cases, there will be hidden quality hazards and even scrap. The application of machine vision technology in textile automatic inspection and quality control is a major breakthrough in the quality control system of the textile industry. Compared with the traditional detection method, the combination of electronic information technology and high-speed camera, combined with automated production and transmission equipment, unified detection standards and efficient detection frequency, improve the speed and quality of inspection, and reduce the labor cost of enterprises. In today’s globalized economy, the implementation of machine vision inspection technology can greatly enhance the competitiveness of enterprises. It is of great significance and social value to promote the technological transformation and transformation of the textile industry and promote the positive and healthy development of the entire industry [23 –25].

2.2 Research status at home and abroad

The application principle of machine vision technology in the textile industry is to collect and classify textile surface information through information automatic detection technology to realize defect detection and classification. As the future development trend and research focus of textile standard testing and quality control, the research and application of machine vision technology involves many disciplines including electronic information science, materials, machinery, and electrical information. The detection methods can be summarized into five categories, namely statistical methods, spectral methods, model-based methods, learning methods and structural methods. Statistical methods are often used in the research of many institutions and scholars. Such detection methods can be subdivided into autocorrelation functions, gray level co-occurrence matrices, mathematical morphology and parting geometry methods. The autocorrelation function has the ability to detect various rules and roughness textures, but needs to be supported by a standardized framework. The gray level co-occurrence matrix can complete the characterization of the pixel, and it has invariance under certain circumstances, but the amount of data is large and the fineness is not enough. Mathematical morphology can filter and extract specific regional features in a targeted manner, which is particularly sensitive to mathematical geometry, but lacks visual support. The fractal geometry method can describe the natural texture characteristics, but requires big data support.

With the development of science and technology, machine vision inspection technology continues to deepen, extracting textile surface texture information and applying it to machine vision inspection, utilizing the characteristics and laws of texture in textile processing, and realizing products by identifying textile surface texture comparison standards. Identification detection of defects. China’s textile testing technology started relatively late with foreign developed countries. By learning the advanced experience and scientific research results of foreign developed countries, China’s textile testing technology is catching up while developing and innovating. At present, the domestic textile automatic testing equipment is still mainly imported equipment, which is characterized by the early start of the production company, rich experience, mature and stable product technology, but there are still regional differences caused by language barriers, low after-sales service., the applicability is single, the price is too high and so on. The development of domestic automated testing equipment is still in its infancy. Although some research results have been achieved, due to the high technicality of the testing system of this type of equipment, there is a strong demand for the technical strength reserve of the development enterprise, and it has not matured yet. Promotion of equipment use.

3 The machine vision detection technology composition

The application of machine vision technology involves many disciplines, including electronic information science, materials, machinery, electrical information and so on. As the future development trend and research focus of textile standard testing and quality control, its core composition can be divided into two parts, namely image acquisition part and image processing part. The product automation inspection process is shown in Fig. 1.

Fig.1

Process diagram of textile automated inspection image acquisition.

3.1 Image acquisition

The main components of textile image acquisition in machine vision technology include automated production equipment and information extraction system. As shown in Fig. 2, after the textile raw materials are processed by the loom, they are pulled forward by the winding equipment and the conveying mechanism, and are detected by the testing equipment. Ensure the quality of the appearance of the textile. The process of the testing equipment is positioned between the completion of the textile processing and the winding, ensuring the online final inspection after the final processing of the textile, before winding the package, and timely repairing the defects after the defects are found.

Fig.2

Schematic diagram of image acquisition of textile automated inspection.

The automated production line is evenly distributed with an infrared light source, and a high-definition information acquisition camera is mounted on the other side of the symmetry. As shown in Fig. 3, the textile is leveled and conveyed at a uniform speed through automated production equipment. The automated production line is adjusted according to the size and thickness of the materials used in the textile. The infrared light source and the high-definition information acquisition camera are symmetrically mounted on both sides of the equipment. The selection of the infrared illuminating source needs to take into account the influence of light in the working environment, and minimize the degradation of image acquisition quality caused by the surrounding environment. The pattern of the textile surface is processed according to different textile processes and has its unique specificity. The patterns on the surface of the product each week have their own unique characteristics, which are summarized as four characteristics, namely periodicity, roughness, uniformity and directionality. When traditional textiles pass through an image capture device, the product information collected by the high definition camera is regular and specific. When an abnormal defect occurs on the surface of the product, the information collected by the high-definition camera has a significant deviation, and can be quickly and accurately located and recognized after being processed by the electronic information image.

Fig.3

Schematic diagram of textile image acquisition.

3.2 Image processing

The textile image processing process in machine vision technology consists of 8 steps.As shown in Table 1, captured video image sequences, ROI extraction, image preprocessing, image enhancement and segmentation, threshold comparison, gray image binarization, feature region extraction, defect recognition and processing.The application of machine vision inspection technology in the quality control of textile defects improves the accuracy and speed of textile defect detection.The labor cost of manual testing is reduced, and the quality control ability of textiles and the market competitiveness of products are improved.

Table 1
Flow chart of textile image processing

Serial number Project Task

1 Capture video image sequence Read video images captured by the camera into system memory

2 ROI extraction Rapid identification of defects; if there are no defects, the image data is discarded directly, otherwise subsequent processing is performed

3 Image preprocessing Enhance image features, extract effective regions, median filter denoising

4 Image enhancement and segmentation Enhanced image features, segmentation and extraction of feature regions

5 Threshold comparison Comparing the gray value of each pixel of the grayscale image with the binarization threshold

6 Grayscale image binarization By threshold comparison, if the gray value of the pixel is greater than the threshold, the gray value of the pixel is set to 255, and conversely, the gray value of the pixel is set to 0, thereby obtaining a binary image.

7 Feature area extraction Calculating the area of the feature area based on the contour shape of the binary image extraction feature area

8 Defect recognition and processing Comparing the calculated area of the feature area with the area threshold set by the detection system (determined as the minimum area of the defect), if it is greater than the area threshold set by the system, it is considered a defect, and vice versa, it will be considered ordinary Cloth; if identified as a defect

Serial number	Project	Task
1	Capture video image sequence	Read video images captured by the camera into system memory
2	ROI extraction	Rapid identification of defects; if there are no defects, the image data is discarded directly, otherwise subsequent processing is performed
3	Image preprocessing	Enhance image features, extract effective regions, median filter denoising
4	Image enhancement and segmentation	Enhanced image features, segmentation and extraction of feature regions
5	Threshold comparison	Comparing the gray value of each pixel of the grayscale image with the binarization threshold
6	Grayscale image binarization	By threshold comparison, if the gray value of the pixel is greater than the threshold, the gray value of the pixel is set to 255, and conversely, the gray value of the pixel is set to 0, thereby obtaining a binary image.
7	Feature area extraction	Calculating the area of the feature area based on the contour shape of the binary image extraction feature area
8	Defect recognition and processing	Comparing the calculated area of the feature area with the area threshold set by the detection system (determined as the minimum area of the defect), if it is greater than the area threshold set by the system, it is considered a defect, and vice versa, it will be considered ordinary Cloth; if identified as a defect

3.3 Common defect problems

The processing technology of textiles is divided into two parts: spinning and weaving. The spinning production process includes cotton cleaning, carding, carding, drawing, roving and spinning; the weaving production process includes winding, warping, sizing, weaving, weaving and finishing. Each process in textile processing has corresponding equipment parameters and process standards. Once equipment abnormalities or personnel operations are not in place, defective products with appearance defects may occur. At the same time, each process has a certain rate of non-performing, and the more processes, the lower the pass rate of the product. An overview of common defects in textile processing is given in Table 2. Depending on the shape characteristics of the defects, they can be classified into three types, namely linear defects, planar defects and point defects. Linear defects are characterized by the difference in latitude and longitude length. The main reason for this is caused by irregular wire ends. The shape characteristics of the planar defects are arranged in a plane, and the main factor that occurs is that the contour boundaries of the raw materials are irregular. The generation of point defects often presents a linear or planar array distribution, and the main factors that occur are pinholes and small holes.

Table 2
Common fabric defect categories

Serial number Defect category Overview

1 Hole A hole is a warp or weft yarn that is broken or three or more together.

2 Mildew Mildew on the surface of the fabric in a humid environment

3 Jumping and jumping The phenomenon of flower jumping is that three or more warp or weft yarns are separated from the cloth surface, and the mesh is present on the surface cloth. Elephant. The yarn skipping is skipped by five or more of the warp or weft yarns, and is linear on the face cloth.

4 Double weft Double weft is the phenomenon of weaving two weft yarns together

5 Dense road The thin road is composed of thin weft and dense road. Weak latitude means that the density of the weft yarn is less than the density of the standard yarn. Phenomenon; the opposite of thin latitude is the secret road

6 De-wefting Means the weaving of three or more weft yarns on the surface of the fabric

7 Rare Means that the weft density is less than five or more per centimeter

8 Weft Weft shrinkage means that in the weaving of the cloth, the weft yarn is made in the process of interlacing with the warp yarn due to the uneven weft yarn. Knitting into fabrics, where weft yarns appear on fabric or fabric surfaces

9 Shrink It refers to the wrinkle phenomenon of strips or lumps on the surface of the fabric, because the warp tension is uneven during the weaving of the warp and weft.

10 Double classic Double menstruation means juxtaposition in a fabric that is all warp, on the premise that the original fabric is unchanged More than one warp yarn phenomenon

11 Cloud weaving Cloud weaving refers to the phenomenon that the density of weft yarns on the cloth surface is rare and regular, and they are arranged in a regular manner.

12 Comprehensive wear The comprehensive misalignment refers to the phenomenon that the fabric is not damaged as required, and the heald is misplaced, causing the texture on the fabric to be damaged.

13 Road The roadway indicates that the warp yarns on the surface of the fabric are straight and have a rare dense phenomenon.

14 Cotton ball Cotton ball refers to the phenomenon that the yarn becomes spherical due to friction during weaving.

15 Sundries Miscellaneous refers to the phenomenon of weaving foreign matter into fabrics, such as paper scraps, hair, dander, metal objects, etc.

Serial number	Defect category	Overview
1	Hole	A hole is a warp or weft yarn that is broken or three or more together.
2	Mildew	Mildew on the surface of the fabric in a humid environment
3	Jumping and jumping	The phenomenon of flower jumping is that three or more warp or weft yarns are separated from the cloth surface, and the mesh is present on the surface cloth. Elephant. The yarn skipping is skipped by five or more of the warp or weft yarns, and is linear on the face cloth.
4	Double weft	Double weft is the phenomenon of weaving two weft yarns together
5	Dense road	The thin road is composed of thin weft and dense road. Weak latitude means that the density of the weft yarn is less than the density of the standard yarn. Phenomenon; the opposite of thin latitude is the secret road
6	De-wefting	Means the weaving of three or more weft yarns on the surface of the fabric
7	Rare	Means that the weft density is less than five or more per centimeter
8	Weft	Weft shrinkage means that in the weaving of the cloth, the weft yarn is made in the process of interlacing with the warp yarn due to the uneven weft yarn. Knitting into fabrics, where weft yarns appear on fabric or fabric surfaces
9	Shrink	It refers to the wrinkle phenomenon of strips or lumps on the surface of the fabric, because the warp tension is uneven during the weaving of the warp and weft.
10	Double classic	Double menstruation means juxtaposition in a fabric that is all warp, on the premise that the original fabric is unchanged More than one warp yarn phenomenon
11	Cloud weaving	Cloud weaving refers to the phenomenon that the density of weft yarns on the cloth surface is rare and regular, and they are arranged in a regular manner.
12	Comprehensive wear	The comprehensive misalignment refers to the phenomenon that the fabric is not damaged as required, and the heald is misplaced, causing the texture on the fabric to be damaged.
13	Road	The roadway indicates that the warp yarns on the surface of the fabric are straight and have a rare dense phenomenon.
14	Cotton ball	Cotton ball refers to the phenomenon that the yarn becomes spherical due to friction during weaving.
15	Sundries	Miscellaneous refers to the phenomenon of weaving foreign matter into fabrics, such as paper scraps, hair, dander, metal objects, etc.

3.4 Quality control

The texture of the textile surface has a specific uniformity characteristic. When the surface of the textile has defects, its original law is broken and the continuity is destroyed. Such defects can be classified into statistical feature distortion, directional feature distortion, and unstructured distortion. The statistical characteristic distortion has certain regularity and continuity, which is different from the conventional texture of the textile surface and has continuous repeatability and traceability. The directional characteristic distortion mainly manifests in the anomaly in the latitude and longitude direction. Unstructured distortions appear to be disordered, non-directional, and can contain all defects except statistical feature distortion and directional feature distortion.

Common defects can be detected and repaired in time using machine vision detection methods. As shown in Fig. 4, in the machine vision application, the textile inspection technology operation flow chart ensures that clear image information is collected by focusing the image, and the collected image information is spliced and detected to confirm whether there is a defect and the result display is confirmed. In the whole operation process, image focusing is the primary work of the whole detection system, and it is also the guarantee of subsequent image stitching and flaw detection, ensuring that high-definition images are the guarantee of detection accuracy. The image acquisition work in the first half and the image processing work in the second half are connected to each other. When the picture clarity is abnormal, the system prompts and switches to the initial position of the whole process to re-focus and image capture; Also, switch to the initial position of the process.

Fig.4

Flow chart of textile defect detection.

4 Summary

With the rapid development of science and technology, more and more work that was originally done by manual has realized automated production. It is certain that electronic information technology will be more applied to the textile industry in the future development. Through the application of machine vision technology to achieve textile defect detection and quality control, the quality value and economic value generated by it have been reflected. The implementation of machine vision inspection technology can greatly enhance the market competitiveness of enterprises and promote the upgrading and transformation of textile industry technology. Promoting the positive and healthy development of the entire industry is of great significance and social value. It is certain that the future textile industry and even the entire manufacturing industry will continue to be the development trend of automation and information. In today’s globalized economy, the automation equipment of the textile industry is rapidly updated. The efficient and efficient production mode has higher requirements for product quality control. Accelerating investment in technology research and development, and developing more rapid and accurate machine vision technology is imperative.

Footnotes

Acknowledgments

The research presented in this paper was supported by Guangdong textile industry intelligent detection Engineering Technology Research Center(DGPT), DongGuan,China.The authors acknowledge2019Guangdong University Students’Science and Technology Innovation Fostering Special Fund (“Climbing Program” Special Fund) Project(pdjh 2019a0898).

References

Wei

, Research on on-line detection image acquisition system for woven fabric defect based on machine vision, Donghua University, 2016.

Guangxin

, Research on automatic detection of fabric defects based on machine vision, Tianjin Polytechnic University, 2016.

Yujie

, Research on fabric flaw detection and intelligent recognition based on visual inspection technology, Fuzhou University, 2016.

Zhuo

, Research on real-time online detection system of fabric based on machine vision, Xi’an Engineering University, 2016.

Min

, Shuqin

and Jia

, Application of visually significant small jacquard fabric defect detection, Journal of Textiles 12(10) (2016), 38–42.

Chong

, Research and design of on-line detection system for dyeing color difference based on machine vision, Zhejiang University of Science and Technology, 2015.

Verma

and Dubey

Prof. S.

, Unipath and Multipath Reactive Routing Protocol in MANET with Fixed Pause Time and Varying Speed, International Journal of Innovative Research in Computer and Communication Engineering (IJIRCCE) 6(5), 2018.

Liu

, Research on automatic detection system of fabric, Tianjin Vocational and Technical Normal University, 2015.

Guoying

, Shulan

and Jun

, Study on fabric flaw detection method under different illuminations, Computer Science and Applications (9) (2014), 181–186.

10.

Zhiwen

, Yingzhan

, Yang

and Shengfu

, Research and implementation of textile defect detection system based on machine vision, Industrial Control Computer 27(11) (2014), 20–24.

11.

Zhoufeng

, Quanjun

, Chunlei

, et al., Fabric defect detection algorithm based on local statistics and overall saliency[J], Journal of Textiles 15(11) (2014), 62–67.

12.

Fangling

, Design of automatic detection system for textile warp and weft density based on machine vision, Jiangsu University of Science and Technology, 2013.

13.

Daixin , Research on surface quality inspection system of mesh fabric based on machine vision, Hubei University of Technology, 2012.

14.

Sahu

and Singh

, Review Paper on Multi-Level Discrete Wavelet Transform using Different Technique, International Journal of Innovative Research in Computer and Communication Engineering (IJIRCCE) 6(3), 2018.

15.

Oleena

S.H.

, Discrete Element Simulation of Centrifuging Mode of Granular Particles and Their Heat Transfer in a Rotating Cylinder, International Journal of Innovative Research in Computer and Communication Engineering (IJIRCCE) 6(2), 2018.

16.

Conghua

, Research and design of textile defect detection system based on machine vision, Wuhan University of Technology, 2012.

17.

Ning

, Research on automatic inspection machine system based on machine vision, Xi’an Engineering University, 2012.

18.

Wei

, Research on the detection and quality evaluation of textile appearance based on machine vision, Hubei University of Technology, 2009.

19.

Rong

, Junyu

and Ruichun

, Textile defect detection system based on machine vision, Journal of Chuxiong Normal University 8(6) (2005), 1–3.

20.

Paul

and Rho

, Probabilistic model for M2M in IoT networking and communication, Telecommunication Systems 62(1) (2016), 59–66.

21.

Xiangyang

, Yong

, Jiangyou

, et al., Current status and application prospects of key technologies of machine vision, Journal of Kunming University of Science and Technology 9(4) (2004), 21–24.

22.

Tsai

H.C.

, Chen

B.W.

, Wang

J.F.

and Paul

, Enhanced long-range personal identification based on multimodal information of human features, Multimedia Tools and Applications 73(1) (2014), 291–307.

23.

Slowly

, Research on automatic detection and classification of fabric defects based on machine vision, Xi’an Polytechnic University, 2012.

24.

Hang

, Research on on-site detection and classification of textiles, Xi’an Engineering University, 2012.

25.

Wei

, Research on textile color classification and color difference detection system [D], Xi’an Polytechnic University, 2015.