Startseite Fabric pattern recognition using image processing and AHP method
Artikel Open Access

Fabric pattern recognition using image processing and AHP method

  • Manal Al Saleh EMAIL logo , Thouraya Hamdi und Mohamed Jmali
Veröffentlicht/Copyright: 11. Februar 2025
Artikel downloaden (PDF)
AUTEX Research Journal
Aus der Zeitschrift AUTEX Research Journal Band 25 Heft 1

Abstract

This article aims to provide methods and a system to recognize the woven fabric patterns automatically. Two different approaches were designed to identify the woven patterns in this article: image analysis and the analytic hierarchy process (AHP). In this work, identifying the woven fabric patterns is based on the measurement of some characteristics and parameters that clearly describe and explain the contexture of the fabric. The methods used were essentially pattern recognition and image processing, which aim to numerically calculate four critical parameters which can give a clear idea of the texture and the patterns of fabrics: the continuity of diagonal lines, the section succession in the vertical direction, section succession in the horizontal direction, and isolated blocks. Then, we used the AHP to classify basic weave patterns into three classes representing the fundamental weaves: plain, twill, and satin. Finally, the statistical approach used to evaluate the efficiency of the proposed method concerning classification showed high correlation values between the predicted results and the experimental values. This study has practical implications in the textile industry, especially in the clothing sector. The manufacturers can use the results to automatically identify the woven fabric patterns without using the manual examination method.

Keywords: image processing; pattern recognition; fabric pattern; AHP; textile analysis

1 Introduction

The conventional means of pattern identification require manual operations, which are time-consuming and easily strain the operator’s eyes [1]. Consequently, it is interesting to develop a computerized method for recognizing and identifying the weaves of the fabric. Furthermore, typical methods call for manual operations with simple tools based on visual inspection by the operator to examine and identify the pattern structure [2]. Thus, generating image processing and programed data investigation procedures would be interesting in determining the fabric’s features [3].

Classification and identification are different problems in fabric pattern recognition; classification involves dividing into classes and separating into groups according to some common attributes, while identification involves identifying and determining the name or characteristics of one item. In the case of recognizing fabric weaves, classification makes it possible to group the tested weaves according to the classes of the fundamental weaves. However, pattern identification is much more precise and complicated because we must determine each sample’s name and structural characteristics.

Image processing consists of eliminating parasitic information and storing the relevant information for pattern recognition [4]. Image processing techniques were mainly employed for fabric pattern identification [5]. Many researchers have exploited the advantages of artificial intelligence to solve the problem of recognizing the weaves of fabrics [6]. They combined image processing techniques and artificial intelligence methods. Among the methods used, some researchers combined image processing techniques with a neural network to solve the problem [7]. They worked on developing an efficient system that minimizes the number of inputs in the architecture of neural networks.

In other studies, researchers used the warp and weft floats to determine weave patterns [8] or the fast Fourier transform techniques [9]. Advanced methods like KFCM and IDM have been carefully used to identify the fabric pattern and repetition for yarn-dyed fabrics [10]. Also, we mention the gradient histogram among the tools developed in image processing and used in recognizing the woven fabric pattern [11]. Image processing has proven to be an effective method for analyzing fabric structures and identifying types of patterns. In their study, Zheng et al. presented a new way to recognize yarn-dyed fabrics with distorted repeat patterns using multi-scale density [12].

In the study of Huang et al. [13], a new approach based on the digital processing of the fabric was used. It consists of capturing the images with a maximum resolution. Next, the images were archived to be manipulated and processed with algorithms that provide measurable parameters. Then, the neural networks and image transformation technology were presented to classify the fabric weaves. In another study [14], an autocorrelation function was employed to define the fabric pattern ratio. In this work, the displayed image of the fabrics was obtained and treated automatically to achieve excellent results. The results demonstrated that three types of basic weaves could be classified with precision, and the parameters of the yarn count, such as weft count and blending, could also be obtained. Other researchers have also used optical coherence tomography (OCT) to recognize fabric weave patterns [15].

Furthermore, advanced methods like learning vector quantization have automatically allocated the woven fabric structure [16]. In addition, several researchers have worked on the systematic discovery of construction characteristics of yarn-dyed fabric [17,18]. Some others presented an original mutated genetic and imperialist rival algorithm for composition determination of woven fabric images [19].

In image analysis, the conversion stage from a color image to a grayscale image directly influences the time and speed of data processing. Also, histogram equalization aims to assign the pixels’ intensity values to obtain a more uniform distribution in the gray level and enhance the image’s visual presentation [20]. Kang et al. developed an automatic fabric assessment system to analyze the fabric structure and objectively assess the quality automatically. First, the images were filtered by Gaussian filtering and histogram equalization. Then, several fabric building parameters were measured by programs based on image processing [21].

In recent years, image analysis has significantly enhanced the capabilities of woven fabric pattern recognition processing [22]. Iqbal Hussain et al. [23] also contributed to this cause, showing that deep convolutional neural networks could be employed for both the recognition and classification of woven fabrics.

Sun et al. [24] identified the potential of spectroscopy-based pattern recognition methods for classifying textile fabrics and found them adequate for fabric analysis. Anila et al. [25] examined fabric texture analysis and weave pattern recognition through intelligent processing methods and concluded that advanced technologies might be useful in textile evaluation.

New ideas in structural textile pattern recognition and processing using hypergraphs were presented to improve methods of textile analysis [26]. In parallel, color pattern recognition for yarn-dyed fabrics contributed to the knowledge of textile classification techniques [27]. Moreover, Guo et al. [28] designed an automatic recognition method to determine the repeat size of weave patterns in woven fabric images, which significantly advances fabric analysis techniques.

A notable innovation introduced a video-based automatic fabric pattern recognition system with a Bayesian-optimized convolutional neural network to enhance the efficiency of fabric analysis [29]. In addition, using multi-layer extraction techniques to identify complex patterns in fabric enhanced the pattern recognition capability [30]. Pattern driven color pattern recognition methods that enhance the design process of printed fabric motifs were also explored by Zheng [31] in order to contribute to advancements in textile design methodologies. Meng et al. [32] developed a multi-task, multi-scale convolutional neural network to enhance the accuracy of fabric analysis for the automatic recognition of woven fabric patterns. Xiao et al. [33] also developed a transform invariant low-rank textures technology-based automatic recognition system for woven fabric patterns to increase the precision and efficiency of fabric pattern analysis. Finally, Sabuncu and Özdemir applied OCT to recognize fabric weave patterns. They discussed fundamental techniques of fabric weave identification, as well as automated methods for checked and colored fabrics, with a focus on striped and colored fabrics, to show the versatility of OCT in textile analysis [34,35,36].

In this article, we used digital image analysis to evaluate the specific characteristics of fabric construction. We began with a series of image processing steps and progressed to a stage of information extraction on the fabric texture. After that, we opted to use the analytic hierarchy process (AHP). Generally, the main problems of pattern recognition are the description of the input data and the extraction of characteristics of these patterns that are applied to decrease the difficulty.

Based on the analysis of the current research and the previous works, we can obtain some key differences between our study and others in the literature. First, our study combines two distinct methods – image analysis and AHP – to recognize the woven fabric patterns. Most other studies focus on a single method: deep learning, image processing, or specific algorithms. We introduced four critical parameters – continuity of diagonal lines (CDL), section succession in the vertical/horizontal directions (SSV/SSH), and isolated blocks (IBL) – to quantitatively describe the fabric’s texture and pattern. At the same time, most other studies do not emphasize this specific combination of measured parameters.

Additionally, we used a statistical method to assess how well our method was performing, and we found a perfect correlation between our predictions and experimental results. This study was also extended to practical applications in the textile industry, especially for manufacturers who want to implement an automated solution to identify woven fabric patterns. Although some studies may discuss applications, focusing on the practicality and the direct benefits to the clothing sector makes our approach different.

Our study is distinct in that we attempted to develop an automatic recognition system with analytical and hierarchical methods, which is not usual in the literature. These differences describe the innovative parts of our research and their contribution to the woven fabric pattern recognition field.

2 Materials and methods

2.1 Materials

In this work, we used 72 fabric samples. This database has different types of weaves (plain, twill, and satin). The samples presented have different basic structural parameters, such as the yarn count and the thickness of the fabric. Table 1 gives an idea of the range of values of these criteria for the patterns tested. These parameters are essential in the problem of pattern classification. However, our study is the first that highlights the determination by an image analysis process.

Table 1

Statistical data of structural parameters

Number of warp yarns/cm Number of weft yarns/cm Thickness (mm)
Max Min Average Max Min Average Max Min Average
Plain 40 12 21.12 32 10 18.08 0.71 0.48 0.60
Twill 38 14 24.71 30 10 20.06 0.72 0.45 0.57
Satin 40 17 25.46 36 14 20.20 0.70 0.46 0.56

In Figure 1, we present some samples of the weaves of the fabric studied in our work. First, the fabrics were captured with adequate resolution to be processed effectively later. The method consists of putting the tested sample in a scanner while ensuring the warp and weft yarns are parallel to the device’s edges.

Figure 1 Presentation of the basic pattern. (a) Plain weave, (b) Twill weave, and (c) satin Weave.
Figure 1

Presentation of the basic pattern. (a) Plain weave, (b) Twill weave, and (c) satin Weave.

Then, the digital data were transferred to a computer program developed under MATLAB. The main objective of this part was to analyze the images and calculate the digital characteristics.

In this figure, we have presented the three basic patterns of fabrics. Our primary goal in all the rest of the work was to identify these fabrics automatically. To separate and easily classify them, we tried to find measurable values for each type of fabric. So, we have listed their main characteristics as follows.

  1. An alternating and consecutive crossing between the warp and weft yarns characterizes plain weaves.

  2. Twill weaves are identified and characterized primarily by diagonal lines either in the right or left direction.

  3. The importance of the concept of floating yarns recognizes satin weaves.

2.2 Experimental method for determining the weave of a fabric

The NFG 07-154 standard, which applies to all fabrics, aims to determine the weave ratio, hence determining the crossing of the warp and weft yarns. The weave ratio represented on the weave paper is retained to represent the weaves of the fabric. Care was taken to arrange different yarns of the weave when there is more than one thread of warp or weft. The tools required to apply this standard method were weave paper, a low-power magnifying glass, tweezers, scissors, and an unwinding needle.

2.3 Presentation of the analysis of variance (ANOVA) method

ANOVA is a method and approach based on statistical and mathematical principles. It classifies the obtained variance data into various components for interpretation. A one-way ANOVA is used for three or more data groups to determine the relationship between the dependent and independent variables.

An ANOVA test is a process that determines if examination or experimental results are significant. In other words, they assist in making the best decision to decline the null hypothesis or allow the alternate hypothesis. In addition, it is a system for testing groups and confirming if there is a distinction between them.

The resulting numerical data are processed to validate the ANOVA calculations, and the critical P and F values are verified using a comparative study.

2.4 AHP

AHP is a system for classifying and interpreting complex judgments by practicing math and psychology.

We present below the different calculation steps:

Step 1: Begin the AHP by determining the alternatives that are required to be judged.

Step 2: Establish the problem and criteria.

Step 3: Set priority among criteria applying pairwise comparison.

Step 4: Verify consistency.

Step 5: Take the relevant weights.

3 Results and discussion

3.1 Image processing operation

This study aims to achieve clear images that can be quickly processed for effective results. Starting from an original image containing all digital information, we wanted to obtain binary images which were easy to process and manipulate without losing valuable data.

The process adopted in analyzing the fabric image can be summarized in several steps. First, image acquisition is implemented to get the original image using appropriate imaging technologies to ensure that all the necessary digital information is included. Then, the image is preprocessed to reduce noise, adjust the contrast, and normalize it to prepare it for further analysis. After that, for segmentation, the featured image of interest is divided into different images of the region based on relation thresholding to edge the detection background.

After that, we pass to a binary image in which the value of each pixel is either 0 (which represents the background) or 1 (which represents the foreground). This simplification is helpful for faster processing and analysis. The binary image analysis tries to recognize and extract important features such as shape and size.

In the last part of treatments, algorithms are used to develop judgments established on the extracted features and characteristics to make decisions.

This systematic approach generally improves the quality of processed images and guarantees that helpful information is preserved for analysis and decision-making.

3.2 Determination of measurable variables

We subsequently present some measurable parameters using image processing techniques. These characteristics were used in this study as inputs in optimization and decision-support systems.

3.2.1 Rate of CDL

CDL is defined as the rate of pixels having the same intensity in the same diagonal direction compared to the total number of pixels in this direction (Figure 2). Then, we calculated the average of all the values of the ratios found.

Figure 2 Schematic representation of diagonal lines in an image. (a) Diagonal line slanting to the right and (b) Diagonal line slanting to the left.
Figure 2

Schematic representation of diagonal lines in an image. (a) Diagonal line slanting to the right and (b) Diagonal line slanting to the left.

The rate of CDL was calculated using the following formula:

(1) CDL = 1 K × ∑ i = 1 K Max ( x 1 i , x 2 i , … , x ji ) N i × 100 ,

with K being the total number of adjacent diagonal lines in an image and N i the total number of pixels in the ith diagonal line.

Each ith diagonal line was divided into j sections of pixels; each nth section (n: 1, 2, …, j) is characterized by a set of adjacent pixels of the same value (0 or 1). The number of pixels was symbolized by x ni .

3.2.2 Rate of SSV

As shown in Figure 3, we defined N as the number of vertical lines. Each vertical line was denoted by n, with n = 1. N: we divided each nth vertical line into several sections. Each section was defined as the existence of an adjacent series of black pixels and a series of white pixels. Each section was symbolized by S n,i with i = 1. K: it is the total number of sections in each vertical line. We also defined X B,n,i as the total number of black pixels in the ith section in the nth vertical line and X W,n,i is the total number of white pixels in the ith section in the nth vertical line.

Figure 3 Schematic presentation of SSV.
Figure 3

Schematic presentation of SSV.

The rate of SSV was calculated using the following formula:

(2) SSV = 1 N × ∑ n = 1 N 1 K × ∑ i = 1 K Min ( X B , n , i , X W , n , i ) Max ( X B , n , i , X W , n , i ) × 100 .

3.2.3 Rate of SSH

As shown in Figure 4, the rate of SSH was calculated in the same way as the previous SSH parameter but with a reversal in the directions. It was calculated using the following formula:

(3) SSH = 1 M × ∑ m = 1 M 1 H × ∑ j = 1 H Min ( X B , m , j , X W , m , j ) Max ( X B , m , j , X W , m , j ) × 100 ,

Figure 4 Schematic presentation of SSH.
Figure 4

Schematic presentation of SSH.

with M being the number of horizontal lines, H the number of sections in one horizontal direction, S m,j the reference of the jth section in the mth horizontal direction (m: 1…M; j: 1…H), X B,m,j the total number of black pixels in the jth section in the mth horizontal line, and X W,m,j the total number of white pixels in the jth section in the mth horizontal line.

3.2.4 Rate of IBL

A block of similarly isolated pixels was defined as a set of adjacent pixels with the same value (0 or 1) and surrounded by other pixels with different values in all the possible directions, as shown in Figure 5.

Figure 5 Presentation of all the possible directions of a pixel.
Figure 5

Presentation of all the possible directions of a pixel.

For example, if we designate the blocks using a set of black pixels, we look for the blocks surrounded by white pixels, as shown in Figure 6.

Figure 6 Schematic presentation of IBL.
Figure 6

Schematic presentation of IBL.

We carried out our calculations on an image of dimensions N × M. The rate of IBL was calculated using the following formula:

(4) IBL = Max NB white NP white , NB black NP black × 100 ,

with NBwhite being the number of white pixel blocks, NBblack the number of black pixel blocks, NPwhite the total number of black pixels in the whole image, and NPblack the total number of black pixels in the entire image.

3.3 Presentation of numerical values for processed images

We used the previously mentioned image parameter definitions. Next, we applied our sample database to determine the fabric’s weave properties. Finally, we have provided a statistical analysis in Table 2 to explain the process better.

Table 2

Statistical data of image parameters

Image features Minimum Average Maximum Standard deviation Variance
CDL 5 29.37 60 18.32 335.81
SSV 4 28.30 70 20.37 415.28
SSH 12 34.86 69 16.68 278.23
IBL 10 23.01 60 13.30 176.91

3.4 Using the AHP to pattern recognition

We used the AHP in our research to determine the suitable fabric pattern. For this, we started to specify the AHP for the following classification: fabric image parameters (CDL, SSV, SSH, and IBL) and fabric patterns (plain, twill, and satin).

This system has four inputs representing the image characteristics (CDL, SSV, SSH, and IBL) and three outputs related to fabric patterns (plain, twill, and satin). To explain the calculation process and determine the correct class of fabric pattern by the AHP, a case study of an image of one fabric-defined characteristic is presented in Table 3. Then, through this calculation we tried to determine the corresponding fabric pattern.

Table 3

A study case of fabric characteristics

CDL SSV SSH IBL
14 35 43 15

In Table 4, we have presented the pairwise comparison matrix between the criteria. Then, we have displayed the normalization matrix in Table 5. After that, we calculated the consistency ratio (CR), as shown in Table 6.

Table 4

Pairwise comparison matrix of criteria

CDL SSV SSH IBL
CDL 1 1/3 1/3 1/2
SSV 3 1 1/2 3
SSH 3 2 1 3
IBL 2 1/3 1/3 1
Sum 9.00 3.67 2.17 7.50
Table 5

Normalization matrix

CDL SSV SSH IBL Weight (%) Consistency
CDL 0.11 0.09 0.15 0.07 0.11 4.1
SSV 0.33 0.27 0.23 0.40 0.31 4.2
SSH 0.33 0.55 0.46 0.40 0.44 4.2
IBL 0.22 0.09 0.15 0.13 0.15 4.1
Table 6

Consistency ratio for Input parameters

RI 0.90
n 4
γ max 4.12
CI 0.04
CR 0.05

Following a complete mathematical approach, we proceeded to finalize and validate the AHP results. The most crucial criticism we seek to determine is the CR:

(5) CR = CI RI ,

where RI is the random index (Table 7) and CI is the consistency index, which was calculated through the following equation

(6) CI = γ max − n n − 1 ,

where n is the order of the matrix and γ max is the average of all the consistency values.

Table 7

Random index (RI)

n 1 2 3 4 5 6 7 8 9 10
RI 0 0 0.58 0.90 1.12 1.24 1.32 1.41 1.45 1.49

3.4.1 Obtaining an overall relative score for each alternative

To validate the calculation, we calculated each pairwise comparison matrix’s CR. To explain the procedure adopted, we presented above the pairwise comparison matrix for each class of fabric pattern for the criterion: the CDL (Table 8), the normalization matrix (Table 9), and CR calculation (Table 10).

Table 8

Pairwise comparison matrix for each class of fabric pattern for the CDL criterion

Plain weave Twill weave Satin weave
Plain weave 1 1/5 2
Twill weave 5 1 7
Satin weave 1/2 1/7 1
Sum 6.50 1.34 10.00
Table 9

Normalization matrix

Plain weave Twill weave Satin weave Weight (%) Consistency
Plain weave 0.15 0.15 0.20 0.17 3.01
Twill weave 0.77 0.74 0.70 0.74 3.03
Satin weave 0.08 0.11 0.10 0.09 3.00
Table 10

Consistency ratio for Output parameters

RI 0.58
n 3
γ max 3.01
CI 0.01
Ratio 0.01

We continued determining the pairwise comparison matrix for each fabric pattern class for the other criteria (SSV, SSH, and IBL). Finally, we obtained the optimal result (pattern fabric) through the solution matrix (Table 11).

Table 11

Matrix of solution

CDL SSV SSH IBL Result
Plain weave 0.17 0.59 0.22 0.75 0.54
Twill weave 0.74 0.16 0.10 0.07 0.18
Satin weave 0.09 0.25 0.68 0.18 0.28

As shown in Table 11, it is clear that the higher weight (54%) corresponds to the pattern fabric class: plain weaves. Therefore, this AHP calculation approach was applied to all the samples.

4 Evaluation of the results

4.1 Study of the effectiveness of the method used

By comparing the results found by the AHP and the results determined experimentally, as shown in Figure 7, we see a perfect correlation between the two processes. To analyze the results found numerically, we adopted the following values in the rest of our work: plain (1), satin (2), and twill (3).

Figure 7 Correlation between the AHP and the experimental method.
Figure 7

Correlation between the AHP and the experimental method.

In general, the coefficient of determination (R 2) is a statistical measure that verifies the reliability of a processed method. In our case and as shown in Figure 7, R 2 equals 0.948. Therefore, this value is very close to 1. These results indicate the excellent correlation between the values found experimentally and those determined by the AHP. To better analyze the impact, we proposed adding a step of calculation of some statistical parameters, which can inform us more straightforwardly the correlation between the results. The values are mentioned in Table 12. Among these data, we can cite the following parameters:

Table 12

Statistical analysis of the correlation between the manual method and AHP

R MAPE MSE RMSE
0.973 3.472 0.041 0.204

R: correlation coefficient,

MAPE: mean absolute percent error,

MSE: mean squared error, and

RMSE: root mean squared error.

By analyzing the statistical data in Table 12, we observed that the value of the correlation coefficient (R = 0.973) was very high; it was very close to 1. This result allowed us to conclude a strong correlation between the values determined experimentally and those calculated by the AHP. Also, we found that the value of MAPE was well below 10% (MAPE = 3.47%), making it possible to conclude the excellent correlation. On the other hand, we noticed that the values of MSE and RMSE were very low (MSE = 0.041; RMSE = 0.204). These parameters measure how accurately the model predicts the response.

4.2 Study of the digital characteristics of images by the ANOVA method

Our study gave much importance to finding reliable image characteristics that digitally and perfectly describe the fabric’s weave. Therefore, we planned to use the following ANOVA, which makes it possible to determine whether there were any statistically significant differences between the means of the groups (fabric patterns). This stage is critical to verify the effectiveness of these parameters in classifying and diversifying pattern fabrics (plain, twill, and satin). For this, we chose the parameter “CDL” as a calculated variable.

In ANOVA, null hypothesis means that there is no difference among group means. However, if any group differs significantly from the overall group mean, then the ANOVA will report a statistically significant result. Table 13 provides a detailed report on the statistical data for the three fabric patterns we wanted to define and classify automatically.

Table 13

Detailed report of the statistical analysis based on the CDL criterion

Groups Number of samples Sum Average Variance
Plain 25 376 15.04 49.04
Twill 32 1545 48.28 33.43
Satin 15 194 12.93 71.35

In Table 14, we applied the ANOVA method to determine both the probability (p-value) and the critical value for F. Using the p-value in the ANOVA output allowed us to choose whether the differences between some of the means are statistically significant. Analyzing the critical value for F can also reject or maintain the null hypothesis.

Table 14

ANOVA based on the CDL criterion

Source of variations Sum of squares Degree of freedom Mean of squares F Probability Critical value for F
Between groups 20630.51 2 10315.25 221.56 9.259 × 10−31 3.12
Within groups 3212.36 69 46.55
Total 23842.87 71

From this table, we assumed that the p-value was smaller than the designated threshold value (5%). In our case, p = 9.25 × 10−31. Also, the F value in our test was more significant than the F critical value (221.56 > 3.12). This result allowed us to conclude it was necessary to reject the null hypothesis. This result was very encouraging because we could classify and separate the pattern fabrics reliably and efficiently.

4.3 Study of the digital characteristics of images by a post hoc test

From the previous part, we presumed that the differences between some of the means were statistically significant. For this, we analyzed the data using the post hoc test method. This type of test was only used after finding a statistically significant result and needing to determine the primary source of this difference. Post hoc tests attempt to control the experiment’s error rate (usually alpha = 0.05). In Table 15, we analyzed the data using the t-test method (two-sample assuming equal variances). We analyzed the weave groups of the fabrics in pairs.

Table 15

t-Test: two-sample assuming equal variances based on the CDL criterion

t-Test 1 t-Test 2 t-Test 3
Plain Twill Plain Satin Twill Satin
Mean 15.04 48.28 15.04 12.93 48.28 12.93
Variance 49.04 33.43 49.04 71.35 33.43 71.35
Observations 25 32 25 15 32 15
Pooled variance 40.24 57.26 45.23
Hypothesized mean difference 0 0 0
Degrees of freedom 55 38 45
t Stat −19.63 0.85 16.79
P(T ≤ t) one-tailed 8.16 × 10−27 0.19 2.62 × 10−21
t-Critical one-tailed 1.67 1.68 1.67
P(T ≤ t) two-tailed 1.63 × 10 −26 0.39 5.24 × 10 −21
t-Critical two-tailed 2.00 2.02 2.01

We divided the original alpha level (0.05) by the number of tests (3). Then, the result was a Bonferroni-corrected value (0.0167), which is the current threshold. In the previous table, we analyzed the row corresponding to (P (T ≤ t) two-tailed). As a result, we found the following results:

  1. In the first case (t-Test 1), we found a significant difference between the plain and the twill. This result is observed because 1.63 × 10−26 < 0.0167.

  2. In the second case (t-Test 2), we do not find a significant difference between the plain and the satin. This result is observed because 0.39 > 0.0167.

  3. In the third case (t-Test 3), we found a significant difference between the twill and the satin. This result is observed because 5.24 × 10−21 < 0.0167.

This study has statistically and scientifically shown the impact and significant difference between the different fabric patterns studied based on the “CDL” criterion. To have a more general idea of the effect of various measures on identifying fabric patterns, we proposed in Table 16 a statistical study based on all the parameters. In this table, we calculated the data’s mean and standard deviation related to all the parameters characterizing the image. We studied the classes of fabrics patterns (plain, twill, and satin) in this part.

Table 16

Statistical presentation of data based on all criteria

Plain Twill Satin
CDL Mean 15.04 48.28 12.93
Standard deviation 7.00 5.78 8.44
SSV Mean 53.76 16.18 11.73
Standard deviation 11.24 5.53 5.88
SSH Mean 53.32 22.03 31.46
Standard deviation 9.96 5.82 12.98
IBL Mean 17.64 16.53 45.80
Standard deviation 5.39 4.93 9.45

After that, we graphically present these data to highlight a comparison between the means and the standard deviations.

Figure 8 shows that the characteristics chosen to describe the images of the weaves of the fabrics could classify and distinguish the groups efficiently. Indeed, the parameter “CDL” gave significantly higher values for the twill weaves with some approximate equality for the other two types of weaves. We noticed an apparent difference in the parameter “IBL” for the satin weaves compared to the two other weaves. Finally, to differentiate the plain weaves, we found the two criteria “SSV” and “SSH”. We observed a clear difference between the plain weave and the other two types, with a slight difference between the twill and the satin for the parameter.

Figure 8 Comparative study for statistical data between classes of fabric weaves based on all criteria. (a) Rate of CDL. (b) Rate of SSV. (c) Rate of SSH. (d) Rate of IBL.
Figure 8

Comparative study for statistical data between classes of fabric weaves based on all criteria. (a) Rate of CDL. (b) Rate of SSV. (c) Rate of SSH. (d) Rate of IBL.

4.4 Importance of integrating image processing and AHP

Our approach is better than the fabric’s previous pattern methods in terms of recognition. While conventional methods have been limited to the extraction of features and classification using simple algorithms, our approach of integrating image processing with the AHP has the following benefits.

First, this combination made the decision better because it offered a step-by-step procedure that enabled one to state the characteristics of the fabric in question based on its importance. It thus beats systems that are solely based on numerical values. Second, our approach is more accurate because it could consider the subjective features of fabric patterns, which are often not considered in similar methods.

Our method was also found to be versatile. AHP could be applied in any culture since it is possible to include expert judgments and thus tailor the evaluation to the application or the user’s opinion. In addition, AHP enabled a comprehensive analysis of the characteristics of fabric patterns from qualitative and quantitative points of view, something that other methods have limited to numerical data.

4.5 Practical implications and limitations

4.5.1 Practical implications

The methods presented in this study for automatically recognizing woven fabric patterns have practical implications for everyday life, especially in the textile and clothing industries. The time and labor required for manual inspections can be significantly decreased, leading to increased efficiency on the production lines. This automation also enhances quality control, as common defects or inconsistencies can be rapidly detected, which assures that higher quality standards are maintained. Designers can also use the system to quickly analyze and classify fabric patterns, which will, in turn, stimulate creativity for the creation of new designs.

4.5.2 Limitations

Despite the advantages of the proposed methods, they have some limitations. For example, the efficiency of image analysis depends very much on the quality of the images captured; poor lighting or low camera resolution can lead to failures in pattern recognition. Moreover, the system may not work well with very detailed or completely unusual fabric patterns. Some of the parameters used in the analysis, for example, CDL and other parameters, may be sensitive to minor variations in texture, leading to inconsistent results across different fabric types.

5 Conclusions

In this work, we used image processing techniques to eliminate the noise introduced in the image by the filtering method and homogenize the distribution of pixel intensity to obtain more significant results. The present article used image processing to identify the weaving fabric’s density. The detection method consisted of applying filtering, histogram equalization, and thresholding to obtain a binary image. After that, we applied a calculation program based on recognizing shapes and the extraction of digital information from the images of the weaves of fabrics. Then, we used the AHP to classify the weaves by building a system whose inputs are the previously determined image characteristics. The output is the determination and identification of the nature of the pattern fabric (plain, twill, and satin). Finally, we used statistical tools (ANOVA method) to finalize our work to check the separation between the pattern classes. By evaluating our system, we found a strong correlation between the results obtained by our system and the actual results determined by experimentation.

Acknowledgements

The authors extend their appreciation to the Deputyship for Research & Innovation, Ministry of Education in Saudi Arabia for funding this research work through the project number RI-44-0375.

  1. Funding information: This research was financially supported by the Deputyship for Research & Innovation, Ministry of Education in Saudi Arabia (RI-44-0375).

  2. Author contributions: All authors have assumed complete responsibility for the content of this manuscript, granted their approval for its submission to the journal, reviewed all findings, and endorsed the final version. Manal Al Saleh, Thouraya Hamdi, and Mohamed Jmali were involved in developing the models, conducting experiments, and performing analytical and statistical analyses.

  3. Conflict of interest: The authors declare there is no conflict of interest.

  4. Ethical approval: The conducted research is not related to either human or animal use.

  5. Data availability statement: The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

References

[1] Xin, B., Hu, J., Baciu, G., Yu, X. (2009). Investigation on the classification of weave pattern based on an active grid model. Textile Research Journal, 79(12), 1123–1134.10.1177/0040517508101459Suche in Google Scholar

[2] Kuo, C. F. J., Shih, C. Y., Ho, C. E., Peng, K. C. (2010). Application of computer vision in the automatic identification and classification of woven fabric weave patterns. Textile Research Journal, 80, 2144–2157.10.1177/0040517510373630Suche in Google Scholar

[3] Ajallouian, F., Tavanai, H., Palhang, M., Hosseini, S. A., Sadri, S., Matin, K. (2009). A novel method for the identification of weave repeat through image processing. The Journal of The Textile Institute, 100, 195–206.10.1080/00405000701660244Suche in Google Scholar

[4] Jeng-Jong, L. Applying a co-occurrence matrix to automatic inspection of weaving density for woven fabrics. Textile Research Journal, 72(6), 486–490.10.1177/004051750207200604Suche in Google Scholar

[5] Pan, R., Gao, W., Liu, J., Wang, H. (2011). Automatic recognition of woven fabric pattern based on image processing and BP neural network. The Journal of the Textile Institute, 102, 19–30.10.1080/00405000903430255Suche in Google Scholar

[6] Lim, J., Kim, S. (2011). Analysis of woven fabric structure using image analysis and artificial intelligence. Fibers and Polymers, 12, 1062–1068.10.1007/s12221-011-1062-8Suche in Google Scholar

[7] Boong, S. J., JiHyan, B. (2003). Automatic recognition of woven fabric patterns by an artificial neural network. Textile Research Journal, 73, 645–650.10.1177/004051750307300714Suche in Google Scholar

[8] Kang, T. J., Kim, C. H., Oh, K. W. (1999). Automatic recognition of fabric weave patterns by digital image analysis. Textile Research Journal, 69, 77–83.10.1177/004051759906900201Suche in Google Scholar

[9] Xu, B. (1996). Identification fabric structures with fast Fourier transform techniques. Textile Research Journal, 66, 496–506.10.1177/004051759606600803Suche in Google Scholar

[10] Jing, J., Xu, M., Li, P. (2015). Automatic recognition of weave pattern and repeat for yarn-dyed fabric based on KFCM and IDMF. Optik, 126, 2876–2883.10.1016/j.ijleo.2015.07.025Suche in Google Scholar

[11] Xiao, Z., Nie, X., Zhang, F., Geng, L. (2014). Recognition for woven fabric pattern based on gradient histogram. Journal of the Textile Institute, 105, 744–752.10.1080/00405000.2013.847542Suche in Google Scholar

[12] Zheng, D., Wang, L. (2017). Multi-scale density detection for yarn-dyed fabrics with deformed repeat patterns. Textile Research Journal, 87, 2524–2540.10.1177/0040517516673331Suche in Google Scholar

[13] Huang, C. C., Liu, S. C., Yu, W. H. (2000). Woven fabric analysis by image processing, Part I: Identification of weave patterns. Textile Research Journal, 70, 481–485.10.1177/004051750007000603Suche in Google Scholar

[14] Kuo, C., Shih, C. Y., Lee, J. Y. (2004). Automatic recognition of fabric weave patterns by fuzzy C-means clustering method. Textile Research Journal, 74(2), 107–111.10.1177/004051750407400204Suche in Google Scholar

[15] Sabuncu, M., Özdemir, H. (2016). Recognition of fabric weave patterns using optical coherence tomography. Journal of the Textile Institute, 107, 1406–1411.Suche in Google Scholar

[16] Jing, J., Wang, J., Li, P., Li, Y. (2011). Automatic classification of woven fabric structure by using learning vector quantization. Procedia Engineering, 15, 5005–5009.10.1016/j.proeng.2011.08.930Suche in Google Scholar

[17] Pan, R., Gao, W., Liu, J., Wang, H., Zhang, X. (2010). Automatic detection of structure parameters of yarn-dyed fabric. Textile Research Journal, 80, 1819–1832.10.1177/0040517510369411Suche in Google Scholar

[18] Zheng, D., Han, Y., Hu, J. L. (2014). A new method for classification of woven structure for yarn-dyed fabric. Textile Research Journal, 84, 78–95.10.1177/0040517513483858Suche in Google Scholar

[19] Fasahat, F., Payvandy, P. (2017). A novel hybrid genetic and imperialist competitive algorithm for structure extraction of woven fabric images. Journal of the Textile Institute, 108, 893–905.10.1080/00405000.2016.1197494Suche in Google Scholar

[20] Burel, G. (2001). Introduction au traitement d’images, simulation sous Matlab. Hermes- Science, Lavoisier.Suche in Google Scholar

[21] Kang, T. J., Choi, S. H., Kim, S. M., Oh, K. W. (2001). Automatic structure analysis and objective evaluation of woven fabric using image analysis. Textile Research Journal, 71, 261–270.10.1177/004051750107100312Suche in Google Scholar

[22] Zhang, R., Xin, B. (2016). A review of woven fabric pattern recognition based on image processing technology. Research Journal of Textile and Apparel, 20(1), 37–47.10.1108/RJTA-08-2015-0022Suche in Google Scholar

[23] Iqbal Hussain, M. A., Khan, B., Wang, Z., Ding, S. (2020). Woven fabric pattern recognition and classification based on deep convolutional neural networks. Electronics, 9(6), 1048.10.3390/electronics9061048Suche in Google Scholar

[24] Sun, X., Zhou, M., Sun, Y. (2015). Classification of textile fabrics by use of spectroscopy-based pattern recognition methods. Spectroscopy Letters, 49(2), 96–102.10.1080/00387010.2015.1089446Suche in Google Scholar

[25] Anila, S., Rani, K. S. S., Saranya, B. (2018). Fabric texture analysis and weave pattern recognition by intelligent processing. Journal of Telecommunication, Electronic and Computer Engineering (JTEC), 10(1–13), 121–127.Suche in Google Scholar

[26] Ngo, V. M., Helmer, S., Le-Khac, N. A., Kechadi, M. T. (2021). Structural textile pattern recognition and processing based on hypergraphs. Information Retrieval Journal, 24(2), 137–173.10.1007/s10791-020-09384-ySuche in Google Scholar PubMed PubMed Central

[27] Zheng, D., Wang, P., Zhou, J., Ho, K. C. (2019). Color pattern recognition for yarn? dyed fabrics. Color Research & Application, 44(1), 88–97.10.1002/col.22263Suche in Google Scholar

[28] Guo, Y., Ge, X., Yu, M., Yan, G., Liu, Y. (2019). Automatic recognition method for the repeat size of a weave pattern on a woven fabric image. Textile Research Journal, 89(14), 2754–2775.10.1177/0040517518801197Suche in Google Scholar

[29] Mamun, A. A., Nabi, M. M., Islam, F., Bappy, M. M., Uddin, M. A., Hossain, M. S., et al. (2023). Streamline video-based automatic fabric pattern recognition using Bayesian-optimized convolutional neural network. The Journal of the Textile Institute, 115(10), 1878–1891.10.1080/00405000.2023.2269760Suche in Google Scholar

[30] Kumar, K. S., Bai, M. R. (2020). Deploying multi layer extraction and complex pattern in fabric pattern identification. Multimedia Tools and Applications, 79(15), 10427–10443.10.1007/s11042-019-7421-ySuche in Google Scholar

[31] Zheng, D. (2021). Pattern driven color pattern recognition for printed fabric motif design. Color Research & Application, 46(1), 207–221.10.1002/col.22564Suche in Google Scholar

[32] Meng, S., Pan, R., Gao, W., Zhou, J., Wang, J., He, W. (2021). A multi-task and multi-scale convolutional neural network for automatic recognition of woven fabric pattern. Journal of Intelligent Manufacturing, 32(4), 1147–1161.10.1007/s10845-020-01607-9Suche in Google Scholar

[33] Xiao, Z., Guo, Y., Geng, L., Wu, J., Zhang, F., Wang, W., et al. (2018). Automatic recognition of woven fabric pattern based on TILT. Mathematical Problems in Engineering, 2018(1), 9707104.10.1155/2018/9707104Suche in Google Scholar

[34] Sabuncu, M., Ozdemir, H. (2016). Recognition of fabric weave patterns using optical coherence tomography. The Journal of The Textile Institute, 107(11), 1406–1411.10.1080/00405000.2015.1114791Suche in Google Scholar

[35] Sabuncu, M., Ozdemir, H., Akdogan, M. U. (2017). Automatic identification of weave patterns of checked and colored fabrics using optical coherence tomography. IEEE Photonics Journal, 9(5), 1–8.10.1109/JPHOT.2017.2742148Suche in Google Scholar

[36] Sabuncu, M., Özdemir, H. (2018). Recognition of weave patterns of striped fabrics using optical coherence tomography. Fibres & Textiles in Eastern Europe, 26(3(129)), 98–103.10.5604/01.3001.0011.7311Suche in Google Scholar
Received: 2024-11-25
Revised: 2025-01-14
Accepted: 2025-01-15
Published Online: 2025-02-11

© 2025 the author(s), published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Artikel in diesem Heft

  2. Study and restoration of the costume of the HuoLang (Peddler) in the Ming Dynasty of China
  3. Texture mapping of warp knitted shoe upper based on ARAP parameterization method
  4. Extraction and characterization of natural fibre from Ethiopian Typha latifolia leaf plant
  5. The effect of the difference in female body shapes on clothing fitting
  6. Structure and physical properties of BioPBS melt-blown nonwovens
  7. Optimized model formulation through product mix scheduling for profit maximization in the apparel industry
  8. Fabric pattern recognition using image processing and AHP method
  9. Optimal dimension design of high-temperature superconducting levitation weft insertion guideway
  10. Color analysis and performance optimization of 3D virtual simulation knitted fabrics
  11. Analyzing the effects of Covid-19 pandemic on Turkish women workers in clothing sector
  12. Closed-loop supply chain for recycling of waste clothing: A comparison of two different modes
  13. Personalized design of clothing pattern based on KE and IPSO-BP neural network
  14. 3D modeling of transport properties on the surface of a textronic structure produced using a physical vapor deposition process
  15. Optimization of particle swarm for force uniformity of personalized 3D printed insoles
  16. Development of auxetic shoulder straps for sport backpacks with improved thermal comfort
  17. Image recognition method of cashmere and wool based on SVM-RFE selection with three types of features
  18. Construction and analysis of yarn tension model in the process of electromagnetic weft insertion
  19. Influence of spacer fabric on functionality of laminates
  20. Design and development of a fibrous structure for the potential treatment of spinal cord injury using parametric modelling in Rhinoceros 3D®
  21. The effect of the process conditions and lubricant application on the quality of yarns produced by mechanical recycling of denim-like fabrics
  22. Textile fabrics abrasion resistance – The instrumental method for end point assessment
  23. CFD modeling of heat transfer through composites for protective gloves containing aerogel and Parylene C coatings supported by micro-CT and thermography
  24. Comparative study on the compressive performance of honeycomb structures fabricated by stereo lithography apparatus
  25. Effect of cyclic fastening–unfastening and interruption of current flowing through a snap fastener electrical connector on its resistance
  26. NIRS identification of cashmere and wool fibers based on spare representation and improved AdaBoost algorithm
  27. Biο-based surfactants derived frοm Mesembryanthemum crystallinum and Salsοla vermiculata: Pοtential applicatiοns in textile prοductiοn
  28. Predicted sewing thread consumption using neural network method based on the physical and structural parameters of knitted fabrics
  29. Research on user behavior of traditional Chinese medicine therapeutic smart clothing
  30. Effect of construction parameters on faux fur knitted fabric properties
  31. The use of innovative sewing machines to produce two prototypes of women’s skirts
https://doi.org/10.1515/aut-2025-0028
Schlagwörter für diesen Artikel
image processing; pattern recognition; fabric pattern; AHP; textile analysis
Creative Commons
BY 4.0

Artikel in diesem Heft

  2. Study and restoration of the costume of the HuoLang (Peddler) in the Ming Dynasty of China
  3. Texture mapping of warp knitted shoe upper based on ARAP parameterization method
  4. Extraction and characterization of natural fibre from Ethiopian Typha latifolia leaf plant
  5. The effect of the difference in female body shapes on clothing fitting
  6. Structure and physical properties of BioPBS melt-blown nonwovens
  7. Optimized model formulation through product mix scheduling for profit maximization in the apparel industry
  8. Fabric pattern recognition using image processing and AHP method
  9. Optimal dimension design of high-temperature superconducting levitation weft insertion guideway
  10. Color analysis and performance optimization of 3D virtual simulation knitted fabrics
  11. Analyzing the effects of Covid-19 pandemic on Turkish women workers in clothing sector
  12. Closed-loop supply chain for recycling of waste clothing: A comparison of two different modes
  13. Personalized design of clothing pattern based on KE and IPSO-BP neural network
  14. 3D modeling of transport properties on the surface of a textronic structure produced using a physical vapor deposition process
  15. Optimization of particle swarm for force uniformity of personalized 3D printed insoles
  16. Development of auxetic shoulder straps for sport backpacks with improved thermal comfort
  17. Image recognition method of cashmere and wool based on SVM-RFE selection with three types of features
  18. Construction and analysis of yarn tension model in the process of electromagnetic weft insertion
  19. Influence of spacer fabric on functionality of laminates
  20. Design and development of a fibrous structure for the potential treatment of spinal cord injury using parametric modelling in Rhinoceros 3D®
  21. The effect of the process conditions and lubricant application on the quality of yarns produced by mechanical recycling of denim-like fabrics
  22. Textile fabrics abrasion resistance – The instrumental method for end point assessment
  23. CFD modeling of heat transfer through composites for protective gloves containing aerogel and Parylene C coatings supported by micro-CT and thermography
  24. Comparative study on the compressive performance of honeycomb structures fabricated by stereo lithography apparatus
  25. Effect of cyclic fastening–unfastening and interruption of current flowing through a snap fastener electrical connector on its resistance
  26. NIRS identification of cashmere and wool fibers based on spare representation and improved AdaBoost algorithm
  27. Biο-based surfactants derived frοm Mesembryanthemum crystallinum and Salsοla vermiculata: Pοtential applicatiοns in textile prοductiοn
  28. Predicted sewing thread consumption using neural network method based on the physical and structural parameters of knitted fabrics
  29. Research on user behavior of traditional Chinese medicine therapeutic smart clothing
  30. Effect of construction parameters on faux fur knitted fabric properties
  31. The use of innovative sewing machines to produce two prototypes of women’s skirts
Jetzt anmelden und 20% sparen
Institutioneller Zugang
Wie funktioniert Authentifizierung?
Haben Sie eine Idee, wie wir unsere Website verbessern können?
Bitte schreiben Sie uns.
© 2025 De Gruyter Brill
Heruntergeladen am 9.9.2025 von https://www.degruyterbrill.com/document/doi/10.1515/aut-2025-0028/html
Button zum nach oben scrollen