Tuijin Jishu/Journal of Propulsion Technology

ISSN: 1001-4055
Vol. 44 No. 3 (2023)
__________________________________________________________________________________________

Fabric Defect Analysis In Textile Manufacturing:

Evaluating Methods For Generic And Jacquard
Fabrics
[1]
Ms Shital K Dhamal, [2]Dr Chandani Joshi, [3]Dr D.S Chouhan
[1]
Research Scholar, Department of CSE, SPSU, Udaipur
[2]
Assistant Professor, Department of CSE, SPSU, Udaipur
[3]
Associate Professor, Department of Mathematics, SPSU, Udaipur

Email: [1]shital.dhamal@spsu.ac.in, [2]chandani.joshi@spsu.ac.in, [3]ds.chouhan@spsu.ac.in

Abstract: Regarding maintaining product quality and satisfying customer expectations, fabric defect
analysis is vital in textile manufacturing. This research provides a comprehensive assessment of many
techniques used for analyzing and pinpointing flaws in standard and Jacquard textiles. This study considers
machine vision systems and AI-based algorithms as two examples of cutting-edge automation alongside
more conventional human inspection procedures. Manual inspection techniques are extensively reviewed in
the first section, with particular emphasis on these techniques' inherent subjectivity and resource
inefficiency. More objective and effective defect detection techniques are highlighted as a solution to the
problems caused by human bias. The research then looks into automated methods, examining how recent
developments in image processing, computer vision, and pattern recognition have the potential to greatly
improve the accuracy and speed with which defects may be detected. The major emphasis of this
investigation is the use of AI, namely machine learning and deep learning models, for fabric defect
identification. This demonstrates how AI may revolutionize textile manufacturing by automating flaw
identification and categorization processes. Accuracy, efficiency, scalability, cost-effectiveness, and
adaptation to varied fabric compositions, such as basic and complicated Jacquard textiles, are only some
aspects considered throughout the evaluation. In addition, the research addresses the problems and
opportunities in the field of fabric defect analysis right now. The paper presents prospective improvements,
such as hybrid methods and real-time monitoring systems, to solve current constraints and pave the way for
a more robust defect analysis framework. These innovations aim to contribute to sustainable practices and
customer happiness in the textile manufacturing industry by fostering effective quality monitoring and
production optimization. In conclusion, this study provides a thorough comprehension of fabric defect
analysis procedures, which is helpful for professionals and academics in the field. The results fuel the never-
ending development of quality assurance techniques, resulting in improvements that raise standards, shorten
production times, and give the textile business a fighting chance in the market.
Keywords: Fabric defect analysis, textile manufacturing, defect detection, Jacquard fabrics, generic fabrics.

1. Introduction
Weaving and knitting machines are the workhorses of the textile industry. Textile fibres are the raw
material for making fabric. Cotton is a common example of a natural ingredient used in producing textile fibres.
A fault in the produced fabric surface is a fabric defect. In particular, issues with the machinery, flawed yarns,
machine spoils, and excessive stretching all contribute to flaws in the fabric. The textile industry has classified
more than 70 fabric flaws [1]. Defects often lie along or perpendicular to the direction of motion. Textiles ' two
most common surface flaws are surface colour change and local texture irregularity [2]. Figure 1 depicts six
typical flaws in textiles. “Slubs (Fig. 1(c)) may be formed by thick spots in the yarn or by fly waste being spun
in yarn during the spinning process, whereas the breakage of needles generates float (Fig. 1(a)).”
An example of a mechanical problem induced by a damaged machine component is shown as a hole in
Fig. 1(d). A typical textile flaw is stitching, as seen in Fig. 1(e). This flaw occurs if the primary or secondary
loom mechanisms are inadvertently moved. Lubricants and rust generate rust spots (Fig. 1(f)). Serious flaws
prevent the cloth from being sold, meaning money is lost [3]. A device that can identify flaws in the fabric helps
make safer, higher-quality goods. Therefore, there is a growing need for automated fabric flaw-detecting

3534
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 44 No. 3 (2023)
__________________________________________________________________________________________

systems to ensure the highest quality textile production. This automated device can detect flaws in the fabric's
surface using image and video processing methods.

Fig 1: Needle breaks, curled weft, slubs, holes, sloppy stitching, rust stains, and broken needles are some
common textile flaws. (The arrows show the broken lines.).

Fabric defect detection is the procedure through which the position, kind, and magnitude of fabric flaws
are identified. Fabric flaw identification often relies on human examination. Errors caused by negligence, optical
illusion, and tiny faults are instantly corrected, although human examination cannot discover them [3-5].
Workers are prone to boredom, leading to incorrect, uncertain inspection findings during human inspection,
which fails to detect faults in terms of accuracy, consistency, and efficiency. As a result, automated fabric
inspection is a promising approach to enhancing fabric quality [6, 7]. Automated inspection is a method of
finding flaws in a product as it is being manufactured. These systems are capable of stopping production at the
precise moment a flaw is found, allowing for real-time inspection. When a problem is detected, automated
systems may provide the operator specifics [8-10]. In the following paragraphs, I will describe the many parts of
an automated defect detection system. Ngan et al. [7] recently surveyed 139 works on detecting textile flaws.
They did a more thorough categorization, dividing methods into seven categories. They were also separated into
two groups: motifs and non-motifs. However, most of the publications evaluated deal with issues with woven
fabrics. As a result, the article needed to thoroughly analyze the issues that might arise with circular knitting
fabrics.
However, there needed to be more detail on the picture capture system's constituent parts. Mahajan et
al. [2] earlier published a review study on the topic of fabric inspection. Now, three defect detection techniques
are used: statistical, spectral, and model-based. The fundamental issue with this work was that it only considered
uniform fabric textures, but many types of cloth have irregular patterns. The second issue with [2] is that, like
the prior technique reviewed in [7], details regarding the picture capture equipment need to be provided. This
study provides the current best practices for detecting fabric defects using various methodologies, including
structural, statistical, spectral, model-based, learning, hybrid, and comparison approaches. “Our paper's primary
contributions are as follows: It provides a more thorough breakdown of methods into seven distinct groups
(structural, statistical, spectral, model-based, learning, hybrid, and comparative).” It also includes a qualitative
evaluation of each approach. Each technique's advantages and disadvantages and whether or not it may be used
to create materials suitable for weaving and knitting are outlined. It offers a comparison of options for the image
capture system's components.

2. Fabric Defect Detection Methods

In this study, we divide the methods for detecting flaws in fabrics into two groups: the first, known as
classical algorithms, and the second, known as learning-based algorithms (Figure 1). Most traditional algorithms
are founded on "feature engineering with prior knowledge," which includes statistical, structural, spectral, and
model-based methods. “The learning-based algorithms may be further broken down into two categories:
classical machine learning algorithms and deep learning algorithms.” Machine learning has been more popular
recently and has shown outstanding outcomes across various industries. Mathematical algorithms learn from and
analyze data to generate predictions and judgements in machine learning.

3535
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 44 No. 3 (2023)
__________________________________________________________________________________________

2.1 Traditional Algorithms

A. Statistical Algorithms.
The geographical distribution of grey values in images is used in various statistical approaches, such as
grey-level co-occurrence matrices (G.L.C.M.), autocorrelation analysis, and fractal dimension features [8].
The GLCM-based automatic fabric flaw identification system presented by Raheja et al. A SIGNAL
GRAPH IS BUILT using G.L.M.C. statistics and the distance between pixels. The test picture is then compared
to the non-defective image. In addition, a Gabor filter-based technique is used for fault detection. Higher
detection accuracies and reduced computing complexity are drawn from using GLCM-based algorithms [9, 10].
By deriving the eigenvector of the defect, Anandan et al. [11] combine the G.L.C.M. and curvelet transform
(C.T.), highlighting the fabric defect characteristics. Experiments demonstrate the efficacy of the suggested
algorithm when compared to G.L.C.M. and wavelet-based approaches.
An eigenvalue-based statistical method for flaw detection in fabric pictures was developed by Kumar et
al. [12]. Images of damaged areas of cloth are analyzed using the coefficient of variation. This procedure is
straightforward to implement based on the studies shown.
The membership degree of each fabric area is calculated by Song et al. [13] to identify fabric problems
effectively. The image's extreme point density map is used with the membership function region's characteristics
to determine the prominence of defect areas. In addition, a threshold strategy and morphological processing are
used across the whole scheme. The author claims his system can reliably and quickly identify flaws in fabric
despite interference from noise and background textures.
Gharsallah et al. [14] describe a fabric flaw-detecting method using an enhanced anisotropic diffusion
filter and saliency image characteristics. The latter combines the local gradient magnitude with a saliency map
to distinguish the faulty edge from the background texture, which is beyond the capabilities of standard
anisotropic diffusion algorithms. This method can eliminate the textured backdrop while keeping the faulty edge
in the picture.
Several statistical techniques for detecting flaws in textile fabrics are summarized in Table 1.

Table 1: Algorithms for statistically detecting flaws in textiles

Author Proposed method Dataset Evaluation
Sayed [15] Minimum-error T.I.L.D.A. dataset Detection success rate
thresholding and entropy
filtering
Kumari [16] Similarity calculation KTH-TIPS-I and KTH- False positives and false
based on Sylvester TIPS-II negatives
matrices
Chetverikov and Based on two Brodatz images and Detection success rate
Hanbury [17] fundamental structural T.I.L.D.A. dataset
properties, regularity and
local orientation
(anisotropy)

B. Spectral Approach.
Many other spectrum approaches exist, but some of the most well-known are the Fourier transform,
Gabor transform, wavelet transform, and discrete cosine transform [18–20]. Table 2 details the algorithms
discussed in this survey. “Fabric defect detection applications have been extensively researched and validated
using Fourier transform, wavelet transform, and Gabor transform-based approaches.”
Li et al. [21] proposed automated fabric flaw identification using a multiscale wavelet transform and a
Gaussian mixture model. The "Pyramid" wavelet transform was used to break down a fabric picture, and the
thresholding technique was then used to rebuild the original. The rebuilt picture was then segmented using the
Gaussian mixture model. “The results of the trials show that the suggested method is capable of accurately
identifying and segmenting pictures with defects.”

3536
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 44 No. 3 (2023)
__________________________________________________________________________________________

Using local homogeneity information and the discrete cosine transform (D.C.T.), Rebhi et al. [22]
describe a fabric flaw identification method. After recalculating the homogeneity picture, D.C.T. was applied,
and various energy characteristics were retrieved from each D.C.T. block. Then, the feedforward neural network
classifier is used to make sense of the data.

Table 2: Fabric flaws may be detected using spectral techniques.

Author Proposed method Dataset Evaluation
Sulochan [23] Fuzzy C-means Real and computer- fabric image Detection
clustering with simulated error rate
multiscale wavelet
features
Vermaak et al. [24] Complex wavelet (D.T.C.W.T.) T.I.L.D.A. Detection success rate
transform with two trees dataset
Liu and Zheng [25] information entropy and Industrial Automation A.C.C., true positive
frequency domain Research Laboratory rate(T.P.R.), false
saliency are the Research Associate's positive rate(F.P.R.),
foundations of this Data Set positive predictive value
technique. (PPV), negative
predictive value
(N.P.V.), time, F-
measure
Di et al. [26] In order to get the Hong Kong University's True positive (T.P.), false
quaternion picture's Automation Laboratory positive (F.P.), true
saliency map, we use the Fabric Database Dataset negative (T.N.), and
L0 gradient false negative (F.N.)
minimization method
and the 2D-FRFT.
Jing [27] Image removal using University of Hong Detection success rate
Gabor filters and a Kong's Industrial
golden frame Automation Lab and the
T.I.L.D. database
Mohammed and Gabor feature-enhanced Collected dataset Detection success rate
Alhamdani [28] fuzzy back propagation
neural network

Yapi et al. [29] In the contourlet space, T.I.L.D.A. database (T.P., F.P., TN, and F.N.)
using local textural local precision (P.L.),
distributions learned by local recall (R.L.), and
reinforcement learning local accuracy
(A.C.C.L.)

2.2 Learning-Based Algorithms

A. Classical Machine Learning Algorithms
(1) Algorithms That Use A Dictionary To Learn New Words. Dictionary learning-based algorithms are
effective in several studies in detecting flaws in textile fabrics. “In general, these algorithms begin by learning a
vocabulary from the training or test picture, then rebuild a fabric image free of defects using the dictionary, and
then execute detection by subtracting the rebuilt image from the test image.” In fabric defect inspection, low-
rank representation-based algorithms have recently emerged. Many techniques convert the low-rank
decomposition issue to the nuclear minimization (N.N.M.) problem to optimize the objective function.

3537
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 44 No. 3 (2023)
__________________________________________________________________________________________

An algorithm is proposed by Li et al. that is inspired by models of biological vision. Using a low-rank
representation (L.R.R.) model of biological visual saliency, we can separate the fabric picture into more
noticeable defect parts and background regions that are less noticeable [30].
By contrast, Li et al.'s [31] work modelled the defect area as a sparse structure. Because of this, we may
think of a fabric picture as the combination of a low-rank matrix and a sparse matrix. The described technique
employs eigenvalue decomposition on the blocked image matrix to reduce dimensionality instead of singular
value decomposition (S.V.D.) on the original image matrix. Therefore, this technique is simple and effective,
provided the cloth picture has enough contrast.
Two problems with the low-rank decomposition are highlighted by Shi et al. [32]. One issue is that
current low-rank decomposition models must better pick out faulty zones with strong gradients. Another
drawback is that accurate background knowledge leads to proper segmentation of complicated or tiny fault
areas. Shi et al. propose a low-rank decomposition method that combines gradient information with a structured
network technique to overcome these constraints. The proposed method achieves better results than existing
methods on the point, box, and star databases.
Common examples of traditional machine learning approaches used for fabric fault diagnosis are K-
Nearest Neighbor (K.N.N.) [33] and neural networks [34]. Feature engineering is a crucial part of building any
machine learning system. Mak et al. [35] used support vector data description (S.V.D.D.), a support vector
machine learning technique for one-class classification, to locate textile flaws.
To solve this problem, Zhang et al. suggest using a network of radial basis functions (RBFs). Using a
Gaussian mixture model (G.M.M.) improves the precision with which Gaussian RBF parameters are estimated.
The proposed method is effective on several classification datasets. [36]
Fabric flaws may be identified using an autoencoder-based technique, as proposed by Tian and Li [37].
Using the recurring texture pattern, we identified nearby nondefective patches comparable to each possible
defect patch. We then weighted and aggregated the related latent variables to alter the original latent variable.
The experimental findings show the efficacy of the suggested algorithm.
This issue is treated as a binary problem by Yapi et al. [38]. “To distinguish between the defect and
nondefect classes, a Bayesian classifier (B.C.) was utilized to extract a compact and accurate feature set by
statistical modelling of multiscale contourlet decomposition. This method achieved very accurate detection in
real-time.”

B. Deep Learning Algorithms.

The quality of textile products and the efficiency with which they are produced [41] have benefited
greatly from the recent application of deep learning approaches to the issue of fabric defect identification [39,
40]. However, there are still challenges in using deep learning algorithms within certain sectors despite their
efficacy when dealing with segmentation and classification difficulties [42]. In the first place, the algorithm has
to execute quickly and reliably in real-time since this is a need of the textile manufacturing line itself.
Furthermore, faulty picture data is more difficult to collect than typical defect-free samples, which presents
difficulties in the training process of deep learning [43].
One-stage and two-stage detectors are now available for the deep learning-based object detector [44].
“Table 5 lists some of the traditional deep-learning techniques used for object identification.” In most cases, the
detection accuracy of one-stage detectors falls short of the necessary standards, but their high detection speed
makes them suitable for online detection. While the detection accuracy of two-stage algorithms is better, this
speed makes it challenging to match the algorithm's real-time needs in production situations. Like other areas,
the pros and cons of one-stage and two-stage detection methods hold in fabric defect identification. The two-
stage technique is more precise than the one-stage but is more time-consuming to implement. If the detection
accuracy can be met, the quicker the detection speed will be in its practical use in the textile sector. The
algorithm must be chosen in light of specific use cases to strike a good balance between speed and precision.
• Single-Pass Detection Methods. There is no proposal creation step in a one-stage detection
technique. These algorithms typically consider every pixel in the picture a candidate for an object
and successfully label each interest zone as either a target item or a background.

3538
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 44 No. 3 (2023)
__________________________________________________________________________________________

One one-stage detector proven effective in object identification is the newly suggested single-shot
multi-box detector (S.S.D.). This algorithm takes its cues from the structure of a CNN. Liu et al. [45] have made
certain enhancements to the textile defect scenario, and the experimental results demonstrate the rationale and
efficacy of these changes.
A CNN-based algorithm for on-loom fabric flaw inspection is presented by Ouyang et al. [46]. The
proposed technique augments a convolutional neural network (CNN) with a dynamic activation layer that uses
defect probability data and a paired potential function. This technique performs well on the challenge of
classifying data that is not evenly distributed.
Algorithms based on deep convolutional neural networks (D.C.N.N.s) have seen widespread industry
use due to their successful performance on visual tasks. D.C.N.N. is used by Liu et al. [47] to identify flaws in
fabrics with intricate patterns. This method is tailored for a practical, low-budget textile manufacturing setting.
The detection has been made more reliable by several tweaks. Zhou et al. propose Efficient Defect Detectors
(E.D.D.s), a D.C.N.N. architecture tailored specifically to the issue of fabric defect detection [48]. E.D.D.s use a
scaling approach to modify the input's resolution, depth, and breadth to extract additional low-level data. When
compared to standard fabric fault-detecting methods, the enhancement was superior.

3. Conclusion
An extensive literature evaluation on smart manufacturing techniques for automatically detecting fabric
defects is presented in this research. “Traditional and learning-based algorithms are the two broad groups into
which all the approaches discussed here fall.” No hard lines can be drawn between the various classifications.
Scientists typically use multiple algorithms to get a more accurate detection. In addition to suggesting avenues
for further study, the survey's findings confirm that synergistic approaches provide superior outcomes. “For
completely automated web detection systems, precise, efficient, and reliable fabric defect detection algorithms
are required.”
The use of computer vision for autonomous textile fabric flaw identification has garnered much interest
from scientists. Advances in object identification algorithms, processing power, and sensor technology and
industry promise rapid progress in computer-vision-based textile flaw detection methods.

References
[1] Textile Handbook 2000, Hong Kong Productivity Council, The Hong Kong Cotton Spinners
Association, 2000.
[2] P.M. Mahajan, S.R. Kolhe, P.M. Pati, A review of automatic fabric defect detection techniques, Adv.
Comput. Res. 1 (2009) 18–29.
[3] A. Kumar, Computer vision-based fabric defect detection: a survey, IEEE Trans. Ind. Electron. 55
(2008) 348–363
[4] C. Anagnostopoulos, I. Anagnostopoulos, D. Vergados, G. Kouzas, E. Kayafas, V. Loumos, G.
Stassinopoulos, High-performance computing algorithms for textile quality control, Math. Comput.
Simul. 60 (2002) 389–400.
[5] C. Anagnostopoulos, D. Vergados, E. Kayafas, V. Loumos, G. Stassinopoulos, A computer vision
approach for textile quality control, J. Visualization Comput. Anim. 12 (2001) 31–44.
[6] G.K.H.P.C.H. Chan, Fabric defect detection by Fourier analysis, IEEE Trans. Ind. Appl. 36 (5) (2016)
1267–1276.
[7] H.Y.T. Ngan, G.K.H. Pang, N.H.C. Yung, Automated fabric defect detection—a review, Image Vision
Comput. 29 (2011) 442–458.
[8] M. T. Habib, S. B. Shuvo, M. S. Uddin et al., “Automated textile defect classification by Bayesian
classifier based on statistical features," in Proceedings of the 2016 International Workshop on
Computational Intelligence (I.W.C.I.), Dhaka, Bangladesh, December 2016.
[9] J. L. Raheja, S. Kumar, and A. Chaudhary, "Fabric defect detection based on G.L.C.M. and Gabor
filter: a comparison," Optik, vol. 124, no. 23, pp. 6469–6474, 2013. Security and Communication
Networks 9

3539
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 44 No. 3 (2023)
__________________________________________________________________________________________

[10] J. L. Raheja, B. Ajay, and A. Chaudhary, "Real-time fabric defect detection system on an embedded
D.S.P. platform," Optik, vol. 124, no. 21, pp. 5280–5284, 2013
[11] R. S. Sabeenian, “Fabric defect detection using discrete curvelet transform,” Procedia Computer
Science, vol. 133, pp. 1056–1065, 2018.
[12] P. Senthil Kumar and H. Hafedh, "Detection of defects in knitted fabric images using Eigenvalues,"
International Journal Computer Science Engineering-IJASCSE, vol. 2, no. 3, pp. 7–10, 2013.
[13] L. Song, R. Li, and S. Chen, “Fabric defect detection based on membership degree of regions,” IEEE
Access, vol. 99, p. 1, 2020.
[14] M. B. Gharsallah and E. B. Braiek, “A visual attention system based anisotropic diffusion method for
an effective textile defect detection,” Journal of the Textile Institute, vol. 8, pp. 1–15, 2020.
[15] M. S. Sayed, “Robust fabric defect detection algorithm using entropy filtering and minimum error
thresholding," in Proceedings of the 2016 IEEE 59th International Midwest Symposium on Circuits
and Systems (M.W.S.C.A.S.), Abu Dhabi, U.A.E., October 2016.
[16] R. M. L. N. Kumari, “Sylvester matrix based similarity estimation method for automation of defect
detection in textile fabrics,” Journal of Sensors, vol. 2021, Article ID 6625421, 11 pages, 2021.
[17] D. Chetverikov and A. Hanbury, “Finding defects in texture using regularity and local orientation,”
Pattern Recognition, vol. 35, no. 10, pp. 2165–2180, 2002.
[18] L. Chen, S. Zeng, Q. Gao et al., “Adaptive Gabor filtering for fabric defect inspection,” Journal of
Computers, vol. 31, no. 2, pp. 45–55, 2020.
[19] J. Zhang, Y. Li, and H. Luo, "Defect detection in textile fabrics with optimal Gabor filter and
B.R.D.P.S.O. algorithm," Journal of Physics: Conference Series, vol. 1651, p. 012073, 2020.
[20] B. A. TanveerSajid, “Fabric defect detection in textile images using Gabor filter," I.O.S.R. Journal of
Electrical and Electronics Engineering, vol. 3, no. 2, 2012
[21] P. Li, H. Zhang, J. Jing, R. Li, and J. Zhao, "Fabric defect detection based on multiscale wavelet
transform and Gaussian mixture model method," Journal of the Textile Institute, vol. 106, pp 587–592,
2015
[22] A. Rebhi, I. Benmhammed, S. Abid, and F. Fnaiech, “Fabric defect detection using local homogeneity
analysis and neural network,” Journal of Photonics, vol. 2015, Article ID 376163, 9 pages, 2015.
[23] H. C. Sulochan, "Fabric weave pattern detection based on fuzzy clustering and texture orientation
features in the wavelet domain," Journal of Textile Science & Engineering, vol. 8, no. 6, 2018
[24] N. Vermaak, P. Nsengiyumva, and N. Luwes, “Using the dual-tree complex wavelet transform for
improved fabric defect detection,” Journal of Sensors, vol. 2016, Article ID 9794723, 8 pages, 2016.
[25] G. Liu and X. Zheng, “Fabric defect detection based on information entropy and frequency domain
saliency,” Visual Computer, vol. 24, 2020.
[26] L. Di, H. Long, and J. Liang, “Fabric defect detection based on illumination correction and visual
salient features,” Sensors, vol. 20, no. 18, 2020.
[27] J.-F. Jing, S. Chen, and P.-F. Li, “Fabric defect detection based on golden image subtraction,”
Coloration Technology, vol. 133, 2017
[28] I. S. Mohammed and I. M. Alhamdani, "A fuzzy system for detection and classification of textile
defects to ensure fabric production quality," International Journal of Electrical and Computer
Engineering, vol. 9, no. 5, 2019.
[29] D. Yapi, M. S. Allili, and N. Baaziz, “Automatic fabric defect detection using learning-based local
textural distributions in the contourlet domain,” IEEE Transactions on Automation Science and
Engineering, vol. 15, no. 3, pp. 1014–1026, 2018
[30] C. Li, G. Gao, Z. Liu, M. Yu, and D. Huang, “Fabric defect detection based on biological vision
modelling,” IEEE Access, vol. 6, pp. 27659–27670, 2018
[31] P. Li, J. Liang, X. Shen, M. Zhao, and L. Sui, “Textile fabric defect detection based on low-rank
representation,” Multimedia Tools and Applications, vol. 78, no. 1, pp. 99–124, 2019
[32] B. Shi, J. Liang, L. Di, C. Chen, and Z. Hou, “Fabric defect detection via low-rank decomposition with
gradient information and structured graph algorithm,” Information Sciences, vol. 546, pp. 608–626,
2021.

3540
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 44 No. 3 (2023)
__________________________________________________________________________________________

[33] Y. Wang, N. Deng, and B. Xin, “Investigation of 3D surface profile reconstruction technology for
automatic evaluation of fabric smoothness appearance,” Measurement, vol. 166, p. 108264, 2020
[34] S. Kulkarni, K. Jojare, V. Bhosale, and P. Arude, "Textile fabric defect detection," I.J.A.R.C.C.E., vol.
5, no. 12, pp. 476–478, 2016
[35] K. L. Mak, P. Peng, and K. F. C. Yiu, “Fabric defect detection using morphological filters,” Image and
Vision Computing, vol. 27, no. 10, pp. 1585–1592, 2009.
[36] Y. Zhang, Z. Lu, and J. Li, “Fabric defect classification using radial basis function network,” Pattern
Recognition Letters, vol. 31, no. 13, pp. 2033–2042, 2010.
[37] H. Tian and F. Li, “Autoencoder-based fabric defect detection with cross-patch similarity," in
Proceedings of the 2019 16th International Conference on Machine Vision Applications (M.V.A.),
Tokyo, Japan, May 2019.
[38] D. Yapi, M. Mejri, M. S. Allili, and N. Baaziz, “A learning-based approach for automatic defect
detection in textile images,” IFAC-PapersOnLine, vol. 48, no. 3, pp. 2423–2428, 2015.
[39] D. Siegmund, B. Fu, A. Jose-Garcia et al., “Study and research on detection of fibre defects using key
points and deep learning,” International Journal of Pattern Recognition & Artificial Intelligence, 2020.
[40] S. Das, A. Wahi, S. Keerthika, and N. (ashram, “Defect analysis of textiles using artificial neural
network,” Current Trends in Fashion Technology & Textile Engineering, vol. 6, no. 1, 2020
[41] S. Barua, H. Patil, P. D. Desai et al., “Deep learning-based smart coloured fabric defect detection
system,” Applied Computer Vision and Image Processing, pp. 212–219, 2020.
[42] X. Zhou, Y. Li, and W. Liang, “CNN-RNN based intelligent recommendation for online medical pre-
diagnosis support,” IEEE/ACM Transactions on Computational Biology and Bioinformatics, p. 1, 2020
[43] Y. Li, W. Zhao, and J. Pan, “Deformable patterned fabric defect detection with Fisher criterion-based
deep learning,” IEEE Transactions on Automation Science and Engineering, vol. 14, no. 2, pp. 1256–
1264, 2017.
[44] X. Wu, D. Sahoo, and S. C. H. Hoi, “Recent advances in deep learning for object detection,”
Neurocomputing, vol. 396, pp. 39–64, 2020.
[45] Z. Liu, S. Liu, and C. Li, "Fabric defects detection based on S.S.D.," in Proceedings of the 2nd
International Conference on Graphics and Signal Processing: ICGSP'18, et al., Sydney, Australia,
October 2018.
[46] W. Ouyang, B. Xu, J. Hou, and X. Yuan, “Fabric defect detection using activation layer embedded
convolutional neural network,” IEEE Access, vol. 7, pp. 70130–70140, 2019.
[47] Z. Liu, C. Zhang, C. Li et al., “Fabric defect recognition using optimized neural networks,” Journal of
Engineered Fibers and Fabrics, vol. 14, 2019.
[48] T. Zhou, J. Zhang, H. Su, W. Zou, and B. Zhang, "E.D.D.s: a series of Efficient Defect Detectors for
fabric quality inspection," Measurement, vol. 172, p. 108885, 2021.

3541