E3S Web of Conferences 538, 04015 (2024) https://doi.org/10.

1051/e3sconf/202453804015
IPFA 2024

Methods of fabric defect detection using expert

systems - a systematic literature review
Ahror Kayumov1,*, Muslimjon Sobirov2, and Khurshid Musayev1
1 Ferghana Branch of the Tashkent University of Information Technologies named after Muhammad
al-Khorazmi, 150107 Ferghana, Uzbekistan
2 Fergana Polytechnic Institute, 150107 Fergana, Uzbekistan

Abstract: This study offers an extensive literature review on fabric defect

detection techniques, commencing with a concise elucidation of
fundamental components within the image acquisition system, including
cameras and lenses. The defect detection methods are systematically
classified into seven categories: structural, statistical, spectral, model-based,
learning, hybrid, and comparative studies. Evaluation of these methods is
conducted based on criteria encompassing accuracy, computational cost,
reliability, rotational/scaling invariance, online/offline operational
capabilities, and sensitivity to noise. The paper aims to provide a nuanced
understanding of the efficacy of various fabric defect detection
methodologies, offering insights into their strengths and limitations across
diverse criteria.Fabric defect detection is a critical aspect of quality control
in textile manufacturing, as it directly impacts the final product's quality.
Expert systems, leveraging advanced computational techniques and domain-
specific knowledge, have emerged as promising tools for automating the
detection of fabric defects. This systematic literature review aims to provide
a comprehensive overview of the various methods employed in fabric defect
detection using expert systems.

1 Introduction
Within the textile industry, fabric manufacturing predominantly occurs through weaving and
knitting machines, utilizing textile fibers as raw materials. These fibers, often derived from
natural elements like cotton, undergo processing to create the fabric. Fabric defects manifest
as irregularities on the surface of the manufactured fabric and are attributable to various
factors such as machine malfunctions, defective yarns, machine spoilage, and excessive
stretching during production. The textile industry classifies over 70 distinct types of fabric
defects, each identified and characterized based on the specific issues encountered in the
manufacturing processes. These defects compromise the overall quality and integrity of the
fabric, necessitating meticulous quality control measures to identify, rectify, and prevent their
occurrence. Understanding the multifaceted nature of fabric defects is crucial for
implementing effective quality assurance practices and ensuring the delivery of high-quality
textiles to consumers[1]. The majority of defects arise either parallel or perpendicular to the

* Corresponding author: [email protected]

© The Authors, published by EDP Sciences. This is an open access article distributed under the terms of the Creative Commons
Attribution License 4.0 (https://creativecommons.org/licenses/by/4.0/).
E3S Web of Conferences 538, 04015 (2024) https://doi.org/10.1051/e3sconf/202453804015
IPFA 2024

direction of motion. Quality standards in fabric assessment classify surface imperfections

into two main categories: alterations in surface color and local irregularities in texture.
Enhancing product quality, automated fabric defect detection systems are witnessing
growing demand in the textile industry. This automated approach involves utilizing image
and video processing techniques to identify faults on the fabric surface. By accurately
pinpointing defects, these systems contribute to the production of high-quality textile
products, meeting the standards and expectations of the industry. The adoption of such
automated fabric defect detection systems reflects a commitment to precision and efficiency
in manufacturing processes, ensuring a superior and defect-free end product.
Fabric defect detection involves the identification of the location, type, and size of defects
present on the fabric surface. Traditionally, fabric defect detection relies on human
inspection, enabling prompt correction of minor flaws. While this method is effective for
immediate intervention, it possesses inherent limitations. Human inspection may fail to detect
defects arising from factors such as carelessness, optical illusions, and minute imperfections
that escape the human eye. As a result, the reliance on manual inspection alone may
compromise the overall accuracy and comprehensiveness of defect identification. To
overcome these limitations and enhance the precision of defect detection, there is a growing
emphasis on incorporating automated systems utilizing advanced technologies like image
processing, ensuring a more thorough and reliable evaluation of fabric quality. This shift
towards automation reflects the industry's commitment to achieving higher levels of accuracy
and efficiency in fabric defect detection processes.
Human inspection faces challenges in detecting defects accurately, consistently, and
efficiently due to factors such as worker boredom, leading to imprecise and uncertain
inspection outcomes. Consequently, automated fabric inspection emerges as a more effective
approach to enhance fabric quality. By leveraging advanced technologies, automated systems
mitigate the shortcomings associated with human inspection, ensuring a higher degree of
accuracy, consistency, and efficiency in defect detection processes. This shift towards
automation not only addresses the limitations of manual inspection but also signifies a
progressive step forward in optimizing fabric quality control within the textile industry.

2 Methods
Automated inspection involves the real-time detection of defects during the production
process. These systems can promptly identify defects as they occur, halting production for
immediate intervention. The automated systems not only detect defects in real-time but also
furnish the operator with comprehensive details about the identified defects. This real-time
feedback loop ensures swift and precise corrective actions, contributing to a more efficient
and high-quality production process.
he forthcoming section provides a breakdown of the components comprising automated
defect detection systems. A recent review conducted by Ngan et al. extensively examined
139 papers specifically focused on fabric defect detection.
Ngan et al. conducted an exhaustive classification of approaches, categorizing them into
seven fundamental groups and further distinguishing between motif-based and non-motif-
based methods. However, a notable observation is that the majority of the reviewed papers
primarily address woven fabric defects, with limited attention given to circular knitting fabric
defects in their comprehensive analysis. Additionally, the paper lacks an informative
explanation of the components of the image acquisition system, leaving a gap in
understanding the intricacies of this crucial aspect in fabric defect detection.
A similar review paper about fabric inspection was previously published by Mahajan et
al. [2]. The current defect detection methods were divided into three categories: statistical,
spectral and model-based. The main problem of this paper was that it was focused on the

2
E3S Web of Conferences 538, 04015 (2024) https://doi.org/10.1051/e3sconf/202453804015
IPFA 2024

uniform fabric textures, but some kinds of fabric have a non-uniform textures. The other
problem of [2] was similar to previous review approach [3] that no information about the
image acquisition system was given. In this paper, the state-of-the-art fabric defect detection
methods in structural, statistical, spectral, model-based, learning, hybrid and comparison
approaches, which have satisfactory results are given. The main contributions of our paper
are as follows: It presents a more comprehensive categorization of approaches of seven
classes (i.e., structural, statistical, spectral, model-based, learning, hybrid and comparison).
It also presents a qualitative analysis for each chosen method. Classification accuracy,
strengths and weaknesses, utilizable in weaving and knitting fabrics are given for each
method. In order to select the components of image acquisition system, it provides the
comparative analysis.

3 Results and Discussions

Acquiring high-quality images during on-loom fabric image acquisition poses several
challenges, with one of the primary difficulties being the selection of the appropriate camera.
In fabric defect detection, two prevalent types of cameras are commonly employed: area scan
cameras and line scan cameras. Line scan cameras stand out for their ability to capture images
from the fabric surface area at high speeds, presenting data in the form of lines. However, to
ensure precision, line scan cameras necessitate synchronization with the moving fabric, a
process facilitated by an encoder. The camera-encoder interface application plays a crucial
role in capturing the true movement of the fabric.
Area scan cameras, on the other hand, capture images of the entire fabric surface
simultaneously. Despite their widespread use, these cameras may encounter challenges when
it comes to high-speed fabric inspection due to their inherent capture mechanism.

Frame grabber camera Line scan cameras Circular knitting machina

Fig 1: Ensuring accurate and real-time acquisition of fabric images

The choice between area scan and line scan cameras is influenced by factors such as the
type of fabric being inspected, the speed of the production line, and the desired level of detail
in defect detection. Line scan cameras are particularly advantageous in scenarios where rapid
image acquisition is crucial, offering the ability to capture a continuous stream of data by
scanning the fabric in a line-by-line fashion [4-10].
For line scan cameras to function optimally, synchronization with the fabric's movement
is imperative. This synchronization is achieved through the use of an encoder, a device that
precisely tracks the fabric's motion. The camera-encoder interface application acts as the
intermediary, ensuring seamless communication between the camera and the encoder,
thereby enabling accurate image capture aligned with the fabric's movement.
In summary, the selection of the appropriate camera, whether area scan or line scan, plays
a pivotal role in addressing the challenges associated with on-loom fabric image acquisition.
The utilization of line scan cameras, accompanied by precise synchronization through

3
E3S Web of Conferences 538, 04015 (2024) https://doi.org/10.1051/e3sconf/202453804015
IPFA 2024

encoders and specialized interface applications, contributes to efficient and high-speed fabric
defect detection during the production process[8].
Fabric defect detection approaches
In this paper, fabric defect detection methods are categorized into seven classes:
structural, statistical, spectral, model-based, learning, hybrid and comparison.
Structural Approaches: Methods that focus on the inherent structure of the fabric to
identify defects, often analyzing patterns and geometrical features.
Statistical Approaches: Techniques that rely on statistical models and analysis to detect
anomalies or deviations in fabric properties.
Spectral Approaches: Methods utilizing spectral information, such as color or
wavelength, to identify defects based on variations in these properties.
Model-Based Approaches: Approaches that employ predefined models or templates to
match against the fabric surface for defect identification.
Learning Approaches: Techniques that involve the use of machine learning algorithms,
allowing systems to learn and adapt based on patterns observed in fabric defects.
Hybrid Approaches: Combined methods that integrate multiple techniques from the
aforementioned classes, leveraging the strengths of different approaches for enhanced defect
detection.
Comparison Studies: Analyses that compare and evaluate the effectiveness of various
defect detection methods, providing insights into their relative strengths and weaknesses [9].

Fig 2: Example defects namely (a) needle breaking, (b) weft curling, (c) slub, (d) hole, (e) stitching (f)
rust stains. (Arrows point to defective regions.)
Six common fabric defects are shown in Fig. 2. Float (Fig. 2(a)) is caused by breaking of
needles, weft curling (Fig. 2(b)) is caused by inserting a highly twisted weft thread, and a
slub (Fig. 2(c)) can be caused by thick places in the yarn or by fly waste being spun in yarn
during the spinning process. Hole (Fig. 2(d)) is a mechanical fault caused by a broken
machine part. Stitching (Fig. 2(e)) is a common fabric defect. This defect is a result of any
undesired motion of the main or auxiliary loom mechanisms. Rust stains (Fig. 2(f)) are caused
by lubricants and rust. Not only do such serious defects make the sale of the fabric impossible,
they also lead to the loss of revenues [3]. A fabric defect detection system improves the
product quality. As a result, automated fabric defect detection systems to manufacture the
high quality of textile products are in increasing demand. This automated system is done by
identifying the faults in fabric surface using the image and video processing techniques.
Fabric defect detection is the determination process ofthe location,type and size ofthe
defects found on the fabric surface. Generally, human inspection is used for fabric defect
detection. It provides instant correction of small defects, but human inspection cannot detect
errors due to carelessness, optical illusion and small defects [3]. However, human inspection
fails on detection defects in terms of accuracy, consistency and efficiency, as workers are
subject to boredom and thus inaccurate, uncertain inspection results are often occurred. Thus,
automated fabric inspection becomes an efficient method forward to improve fabric quality
[11-12].
The main problem of this paper was that it was focused on the uniform fabric textures,
but some kinds of fabric have a non-uniform textures. The other problem of was similar to

4
E3S Web of Conferences 538, 04015 (2024) https://doi.org/10.1051/e3sconf/202453804015
IPFA 2024

previous review approach that no information about the image acquisition system was given.
In this paper, the stateof-the-art fabric defect detection methods in structural, statistical,
spectral, model-based, learning, hybrid and comparison approaches, which have satisfactory
results are given. The main contributions of our paper are as follows: It presents a more
comprehensive categorization of approaches of seven classes (i.e., structural, statistical,
spectral, model-based, learning, hybrid and comparison). It also presents a qualitative
analysis for each chosen method. Classification accuracy, strengths and weaknesses,
utilizable in weaving and knitting fabrics are given for each method. In order to select the
components of image acquisition system, it provides the comparative analysis.

4 Conclusions
This paper systematically explores fabric defect detection methodologies through an
extensive survey of approximately 99 references. The diverse range of approaches identified
in the literature is thoughtfully organized into seven comprehensive categories: Structural,
statistical, spectral, model-based, learning, hybrid, and comparison. In essence, the paper not
only categorizes these approaches but also meticulously examines their fundamental
concepts, shedding light on the intricacies of each category. Furthermore, a detailed
discussion on the strengths and weaknesses of these approaches is presented, providing a
nuanced analysis that aids in understanding the practical implications and trade-offs
associated with each method. By distilling insights from a substantial body of literature, this
survey serves as a valuable resource for researchers, practitioners, and industry professionals
seeking a holistic understanding of fabric defect detection. The comprehensive exploration
of various methodologies, coupled with an insightful analysis of their merits and limitations,
empowers readers with the knowledge needed to make informed decisions when navigating
the complexities of fabric quality control in diverse manufacturing contexts.
This paper presents a survey of fabric defect detection approaches examined in about 99
references. These approaches have been classified into seven categories: Structural,
statistical, spectral, model-based, learning, hybrid and comparison. The main ideas of these
approaches along with their strengths/weaknesses have been discussed. When the developed
methods are examined, each of the vast majority of all the studies is seen to create its own
database. Once the image database was being built, images were either obtained from factory
environment, or broughtto the laboratory and database was created with the proper lighting
setting. Therefore, the reliability and validity of the methods is far from objectivity. Some
studies have used TILDA [4] fabric database. However, this database is difficult to be
obtained by all participants as it is to be paid. Also, Hanbay et al. [5] constructed a novel
fabric database by using a conveyor system which has line scan camera and linear light. This
database contains 3242 defected and 5923 defect-free fabric images. For the development of
objective and reliable methods, anonymously accessed free fabric databases are needed. In
the literature search, studies on yarn and fibers, which are the basic building blocks of woven
and knitted fabrics, are found to be vanishingly small in number. On the other hand, fabric
defects that may occur could be avoided thanks to the evaluation of yarn and fibers before
the production of fabric. In some previous studies, smart yarn modeling and rating systems
have been developed by examining yarn surfaces [6]. In a review study, some studies
conducted on yarns and fibers were examined and a current study is not found. It is thought
that serious studies similar as this one are needed for fabric defect detection, and therefore
the studies will contribute to the textile industry.

5
E3S Web of Conferences 538, 04015 (2024) https://doi.org/10.1051/e3sconf/202453804015
IPFA 2024

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. 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.
4. TILDA Textile Texture-Database, http://lmb.informatik.uni-
freiburg.de/resources/datasets/tilda.en.html (2013)
5. K. Hanbay, M.F. Talu, Ö.F. Özgüven, D. Öztürk, Fabric defect detection methods for
circular knitting machines, in: 23nd Signal Processing and Communications
Applications Conference (SIU), Malatya, 2015, pp. 735–738.
6. Z. Liang, B. Xu, Z. Chi, D. Feng, Exp. Syst. Appl. 39 (2012) 4201–4212.
7. Fazilov, S. K., Mirzaev, N. N., Radjabov, S. S., Dadakhanov, M. K., Asraev, M. A.,
Shamsiev, F. M. (2019). Compusoft, 8(12), 3514-3524.
8. Salomov, U., Abduraxmonov, S., Urishev, O., Juraev, N. (2024). BIO Web of
Conferences 84, 05028
9. Sharibaev, N., Tursunov, A., Djuraev, S. (2022). Journal of Physics: Conference Series
2373, 7, 072043
10. Sharibaev, N., Ergashov, A. (2022). Journal of Physics: Conference Series 2373, 7,
072044
11. Fazliddinov, S., Kuchkarov, B., Sharibaev, N., Abdulkhaev, A., Tulkinov, M. A.
(2022). Journal of Physics: Conference Series 2388, 1, 012173
12. G.K.H.P.C.H. Chan, Fabric defect detection by Fourier analysis, IEEE Trans. Ind.
Appl. 36 (5) (2016) 1267–1276