Pilling analysis for textile grading of synthetic fibre material using image processing and machine learning techniques

ABSTRACT

Textile pilling is a persistent issue that results in an unattractive surface on garments, impacting their aesthetic and commercial value. This study employs image processing and machine learning techniques to grade fleece textiles based on pilling evaluation. Two approaches were explored: the first utilized a discrete Fourier transform combined with Gaussian filtering, while the second employed Daubechies wavelets. Binarization was used to isolate textile pilling from the surrounding fabric area. In this study, morphological and topological image processing techniques were applied to extract key features from the image data, creating a comprehensive database for fabric analysis. Machine learning techniques, specifically Support Vector Machine (SVM) and Artificial Neural Networks (ANN), were then used to objectively address the textile grading problem. The Fourier-Gaussian approach achieved classification accuracies of 95.67% with ANN and 92.34% with SVM, while the Daubechies wavelet approach yielded accuracies of 94.23% and 90.67%, respectively. In terms of pilling detection, the Fourier-Gaussian method identified 67 instances of pilling with a pilling area of 1674 units, whereas the Daubechies wavelets method detected 56 instances with a pilling area of 654 units. This automated system enhances textile quality assessment and production efficiency by effectively detecting and quantifying pilling.

Keywords: ANN; SVM; Machine Learning; Pilling; Grade

1. INTRODUCTION

In recent years, the textile industry has increasingly relied on advanced technologies such as image processing and machine learning to improve quality assessment processes [¹]. Pilling, a common issue in fleece textiles, significantly impacts their aesthetic appeal and durability [²]. Traditional methods for pilling assessment are often subjective and labor-intensive, prompting the need for automated, objective techniques [³, ⁴]. This paper focuses on applying image processing and machine learning algorithms for pilling analysis in fleece textiles [⁵]. By harnessing digital image analysis, intricate details of pilling can be captured and quantified with precision, offering manufacturers a robust tool for quality grading [⁶].

A textile grading system refers to a method used to evaluate and categorize textiles based on specific criteria, such as quality, texture, appearance, and performance. This system often involves assessing various characteristics, such as pilling, strength, colourfastness, and overall fabric integrity, to determine the suitability of textiles for different applications. Grading helps in standardizing the quality of textiles and ensures that products meet the desired specifications and industry standards.

Such automated systems enhance accuracy and streamline production processes, reducing human error and operational costs [⁷]. The integration of machine learning further enhances the capability to classify and characterize different pilling levels, providing valuable insights into fabric performance and durability [⁸]. This approach promises to revolutionize textile grading practices, ensuring consistency and reliability in product quality assessment [⁹, ¹⁰]. Through a comprehensive review of existing methodologies and advancements in image processing and machine learning, this study aims to underscore the possible assistances and contests of accepting these machineries in the context of fleece textile [¹¹, ¹²].

Bridging the gap between traditional assessment methods and modern technological innovations, this research contributes to the ongoing evolution of quality control practices in the textile industry [¹³, ¹⁴]. Machine learning techniques in pilling analysis for textile grading of fleece utilize algorithms to automate and enhance the assessment of pilling severity [¹⁵, ¹⁶]. Analyzing digital images of fleece textiles, these techniques extract features and patterns that signify pilling, enabling objective evaluation [¹⁷, ¹⁸]. This integration of machine learning improves efficiency in quality control and supports manufacturers in maintaining consistent product standards and enhancing customer satisfaction through reliable textile grading processes [¹⁹]. Pilling gets worse the more textile grade degradation occurs [²⁰].

The research gap for this study lies in the limited application of advanced image processing and machine learning techniques for automated textile grading, particularly for pilling analysis in synthetic fibers. Traditional methods for assessing pilling are often subjective, labor-intensive, and lack precision. There is a need for an automated system that can consistently and accurately evaluate pilling, which affects the quality and durability of textiles. This study addresses this gap by integrating discrete Fourier transform, Gaussian filtering, and Daubechies wavelets with machine learning algorithms (ANN and SVM) to provide a more objective, efficient, and scalable solution for textile grading.

The study on pilling analysis using image processing and machine learning addresses significant issues in textile grading, such as the subjectivity and inconsistency of traditional methods. By providing an automated and objective approach, the study aims to enhance quality control and production efficiency in the textile industry. The impacts include improved accuracy in detecting pilling, reduced labor costs, and increased reliability in grading synthetic fibres. These advancements can lead to better-quality textiles, longer-lasting garments, and streamlined manufacturing processes. Additionally, the study contributes to the broader adoption of advanced technologies in textile quality assessment.

The critical elements of the pilling were extracted using morphological and topological image processing techniques to create an objective grading system [²¹]. This investigation employed two image processing techniques, and the outcomes were contrasted. In one approach, Daubechies wavelet filtering was used instead of Gaussian filtering when combined with the discrete Fourier transform (DFT). Five characteristics were retrieved for every textile picture in the database [²²]. The data was trained, and the textile grade was objectively classified using SVM and ANN models [²³].

2. METHODOLOGY

2.1. Sample collection and experimental process

Data collection, image processing, feature extraction, model construction, and performance assessment were the five processes in the suggested approach see Figure 1. Process 1 involved gathering 280 representative samples categorized as grade 2, 3, 4, or 5. Eighty samples are made up of each grade. A CCD camera was used to capture images of the fabric, resulting in a grayscale picture dataset for the 280 samples [²⁴]. In order to get the picture, the camera was positioned over the specimen. Process 2, which includes grayscale image smoothing, was achieved by combining the DFT approach with Gaussian filtering. The grayscale images were compressed using the Daubechies wavelet method. The filtered photos were binarized into black and white in order to preserve the information on the textile pilling [²⁵]. Process 3 includes critical elements of the pilling that were extracted using morphological and topological image processing techniques. This study considered five parameters of pilling: the no. of spots, the area, the average area, the ratio of the area, and the density [²⁶]. The database was 280 multiples of five in size.

Figure 1
The methodology flow chart for this study.

Process 4 shows the pilling was categorized using ANN and SVM models. Lastly, a comparison was made between the classification rates of various classifiers. The study’s textiles met the Martindale wear standards of ISO 12945-2:2000 textile grading. This standard contains five grades [²⁷]. The lowest quality textiles are those classified as grade 1, while the finest quality textiles are those classified as grade 5. In this investigation, no grade 1 textiles were used. As a result, 80 samples from grades 2 through 5 were analyzed. Figure 2 displays the look of grades 2 through 5 [²⁸]. Figure 3 mechanism for capturing.

Figure 2
Sample collected for different textile grade.

Figure 3
Mechanism for capturing.

Table 1 contains a list of the experimental equipment specs. This study looked at fleece as the fabric. Cheerful lighting was obtained to photograph textiles using the low-angle oblique illumination method as recommended. The standard method for identifying and analyzing elevated and low faults on a flat shallow is low-angle slanting lighting [²⁹]. Images with uniform illumination are produced, and shadow production is minimized using low-angle oblique lighting. The grayscale images of textile classes 2 through 5 are shown in Figure 4. CCD stands for Charge-Coupled Device. It is a technology used in imaging sensors to capture and convert light into electronic signals, which are then used to produce digital images. CCD sensors are known for their high image quality and sensitivity, making them suitable for various imaging applications.

Table 1
Experimentation equipment specifications.

MODEL	PIXELS	FREQUENCY
Basler A601f-2	656 * 491	58FPS
MODEL	FOCAL LENGTH	MAXIMUM APERTURE RATIO
M2514-MP2	25 mm	1:1.4

Figure 4
The grayscale images of textile classes two through five grades.

2.2. Image processing

2.2.1. Gaussian filtering, along with DFT

High-pass filters are used to emphasize the edges of images, whereas low-pass filters are typically employed to blur or smooth images. The smoothing filter that is most frequently employed is the Gaussian filter. The following is the expression for the Gaussian filter formula Eq. 1:

(1) $$H(u,v)=\frac{1}{2\pi {\vartheta }^2}{e}^{-\left(u^2+v^2\right)/2{\sigma }^2}$$

where s represents the standard deviation of the Gaussian distribution, and uuu and vvv denote the flat and perpendicular detachments from the source, respectively. Discrete and periodic signals are the intended use case for the DFT. Eq. 2 allows us to get an inverse conversion picture [F(x, y)] for the following signals: y = 0, 1, 2,…., N_1 and x = 0, 1, 2,......, M_1. Eq. 3 gives the formula of the inverse DFT.

(2) $$F(u,v)={\displaystyle {\sum }_{u=0}^{M-1}{\displaystyle {\sum }_{Y=0}^{N-1}f(x,y)e^{-j2\pi (\frac{ux}{M}+\frac{vy}{N})}}}$$

(3) $$F(x,y)=1/MN{\displaystyle {\sum }_{u=0}^{M-1}{\displaystyle {\sum }_{Y=0}^{N-1}f(x,y)e^{j2\pi (\frac{ux}{M}+\frac{vy}{N})}}}$$

In this work, image processing was accomplished by combining the DFT and Gaussian filters. Eq. 4 is utilized to filter a textile picture of size M*N.

(4) $$g(x,y)=F^{-1}[H(u,v\left)F\right(u,v)]$$

Figure 5 shows the image filtering procedure. The filtering process was carried out using Matlab software. Sigma, a parameter in the Gaussian filtering conversion, has to have a value assigned to it. For a 3 × 3-pixel picture, Matlab’s default sigma value is 0.5. The study utilized textile pictures at a resolution of 280 × 240 pixels. Figure 6 illustrates how the filtering performance of the original and filtered textile photos was compared over a range of sigma values between 10 and 50. The filtered image gets crisper, and the sigma value is higher. In this investigation, the filter’s mask size was also considered. The degree of picture distortion increases with increasing sigma value. As a result, in this investigation, the filter mask size’s sigma value was set at 40. Figure 7 displays the filtered photos and the matching mask size sigma values.

Figure 5
Image filtering process for this study.

Figure 6
Images that have been filtered and given matching sigma values.

Figure 7
Image processing filter masked size variation.

The Sigma 30 appearing completely black is due to the specific imaging conditions or settings used during data collection. This may result from factors such as exposure time, sensor sensitivity, or contrast settings. The black appearance indicates that the Sigma 30 had no discernible features or contrast at the given conditions, which could be a result of its material properties or the lighting setup.

2.2.2. Daubechies wavelet process

This process is composed of the scaling and wavelet functions, as shown in the following equations (Eqs. 5 and 6).

(5) $$\mathrm{\varnothing }(x)={\displaystyle {\sum }_{K=0}^{N-1}{a}_k}\mathrm{\varnothing }(2x-k)$$

(6) $$\Psi (x)={\displaystyle {\sum }_{K=0}^{M-1}{b}_k\mathrm{\varnothing }(2x-k)}$$

(7) $$\mathrm{\varnothing }(x)=h_0\mathrm{\varnothing }(2x)+h_1\mathrm{\varnothing }(2x-1)+h_2\mathrm{\varnothing }(2x-2)+h_1\mathrm{\varnothing }(2x-2)$$

(8) $$\Psi (x)=h_0\mathrm{\varnothing }(2x-1)+h_1\mathrm{\varnothing }(2x)+h_2\mathrm{\varnothing }(2x+1)+h_3\mathrm{\varnothing }(2x+2)$$

Higher SNR and PSNR values and smaller MSE values indicate better image quality. The image quality was assessed by simulating different scale values in the Daubechies wavelet step and computing the MSE, SNR, and PSNR evaluation indices. The image evaluation results for different scales are shown in Table 2. Six adaptation stages were created for the fabric photographs for this work based on the previously described discussion and Umbaugh’s conversion processes. The wavelet transform’s output is displayed in Figure 8.

Table 2
Results of image-processing evaluation.

IMAGE	MSE	SNR	PSNR
Original	0	Infinite	Infinite
Scale 1	0.123	24.764	35.965
Scale 2	2.934	22.436	33.214
Scale 3	4.564	20.694	28.547
Scale 4	6.543	18.459	25.347
Scale 5	8.543	16.975	22.574
Scale 6	10.395	14.235	19.546
Scale 7	12.438	12.342	17.986

Figure 8
Finding out of wavelets transform.

The wavelet conversion process begins with the convolution of the image in the horizontal direction using a low-pass filter. The process continues with the convolution of the image in the perpendicular direction using the lowpass filter. Finally, the highpass filter is applied again to calculate the intricacy in the vertical track.

2.3. Binarization

The grayscale image of the textile was binarized to black and white using the Matlab DIP toolbox’s im2bw function. As a result, the converted image only contained pixels with values of 0 and 255 that were pure black and white, respectively. 0.5 was the threshold value by default. Consequently, the final value was 255 if the image’s pixel value was more significant than 127.5 (255 ~ 0.5). If the pixel worth was less than 127, the ending rate was set to 0. The outcomes for the threshold values of 0.4, 0.5, and 0.6 are shown in Figure 9.

Figure 9
Outcomes using various thresholds.

2.4. Morphological image processing

The essential features of the filtered textile images were extracted through morphological image processing. The two fundamental processes in morphological image processing are erosion and dilation. An erosion and a dilatation are represented by the given A and the structuring element B (represented by A ⊕ B).

(9) $$AB=(A\mathrm{\Theta }B)\oplus B$$

Erosion reduces the image size throughout the opening process to eliminate unnecessary details. In order to accentuate the critical aspects of the textile image’s pilling, the regions that remain after shrinking are dilated to their original size.

2.5. Image topology

In order to significantly simplify the algorithmic design, image topology entails evaluating discrete items in a 2D digital-image and building a scientific archetypal. Let us say that pixel A is the current one. A’s left neighbor is e, while its top neighbor is o. Figure 10 shows a four-adjacent relationship formed for pixel A by neighboring places.

Figure 10
Four neighbors and adjacent positions for this study.

2.6. Creating data base

The following describes the attributes that are frequently utilized for textile grading. Graded textiles based on the quantity, area, and number of pilling spots. In this study, we collected and analyzed five variables from textile images as part of the grading technique.

(10) $$\text{Average}\text{Pilling Area}\frac{\text{Pilling}\text{Area}}{\text{No}\text{. Of. Pilling Points}}$$

(11) $$\text{Pilling Area Raito=}\frac{\text{Pilling}\text{Area}}{\text{Image Size}}$$

(12) $$\text{Pilling Density=}\frac{\text{No}\text{. Of}\text{. Pilling}\text{Points}}{\text{Image Size}}$$

3. MODEL BUILDING FOR TEXTILE GRADING

3.1. ANN – artificial neural network

An artificial intelligence machine learning technology called an ANN simulates human cognitive processes. It may be used in many different contexts, including nonlinear models. The input, hidden, and output layers are the three layers that typically make up an ANN. The ANN’s input elements included the number of pilling spots, area, average area ratio, and density extracted from the textile image. The dataset was divided into 80% training and 20% test sets.

3.2. SVM – support vector machine

SVMs differentiate between two or more information categories using a hyperplane in n-dimensional space. Commonly utilized fundamental functions in a SVM are the linear, polynomial, and radial functions. When dealing with high-dimensional data, the radial function performs better than the other functions. Thus, the study’s method of textile grading was the radial function. SVM execution was done using the Weka program. The SVM’s input elements were identical to those of the ANN. Here is a description of the SVM’s parameter settings. The punishment function (C value) had a default value of 1, and the sigma value was set at 0.05. The SVM close form was established at 1.0 * 10\6. Eighty percent of the database was the exercise set, and the remaining twenty percent was the test set. The SVM used ten-fold cross-validation.

The three filters used in this study serve different purposes in image processing:

Fourier Ideal Filter: This filter operates in the frequency domain, emphasizing or suppressing certain frequency components. It provides a sharp, idealized separation of frequencies, which can enhance feature detection but may introduce artifacts. Gaussian Filter: Applied in the spatial domain, this filter smooths images by averaging pixel values within a defined radius, reducing noise while preserving edges. Its effect is less sharp compared to the Fourier ideal filter. Daubechies Wavelet: A type of wavelet filter, it captures both frequency and spatial information, offering a multi-resolution analysis with smooth transitions between scales. The Fourier Ideal Filter often gives higher accuracy in pilling detection due to its precise frequency separation, enhancing detail and contrast more effectively than the other filters.

4. RESULTS AND DISCUSSION

4.1. Image processing results

This section presents the image processing findings. The grade 2 textile image processing results for each stage are shown in Figure 11. The textile picture with a diluted background texture was obtained by combining Gaussian filtering with the DFT approach. The original textile picture was filtered and compressed using the Daubechies wavelet approach to reduce noise and improve pilling properties. The picture background was distinguished from the piling via binarization. Still, some white areas were in the picture. After that, morphological techniques were applied to save important information and eliminate unnecessary information. The outcomes of the morphological and binarization approaches in the fourth and fifth rows, respectively. In order to provide dataset needed for ML to finish the last stage of fabric scaling, the pilling characteristics were identified and retrieved.

Figure 11
Results of image processing for the textile image.

4.2. Image classification based on machine learning

ANN and SVM models were evaluated using the dataset. The models’ grading abilities were contrasted. The arrangement correctness of the ANN and SVM when the Fourier-Gaussian approach was applied were 95.67% and 92.34%, respectively, as shown in Tables 3 and 4. ML-based textile grading produces better classification accuracy and takes less processing time than human inspection and grading.

Table 3
Outcomes of the ANN-based Fourier Gaussian technique.

PREDICT ACTUAL	GRADE 2	GRADE 3	GRADE 4	GRADE 5	CORRECT NUMBER	380
Grade 2	77	6	2	0
Grade 3	4	76	1	0	Wrong Number	15
Grade 4	2	1	82	2
Grade 5	1	2	0	86	Accuracy	95.67%

Table 4
Outcomes of the SVM-based Fourier Gaussian technique.

PREDICT ACTUAL	GRADE 2	GRADE 3	GRADE 4	GRADE 5	CORRECT NUMBER	376
Grade 2	79	2	0	1
Grade 3	6	89	0	2	Wrong Number	18
Grade 4	0	1	55	0
Grade 5	0	0	0	61	Accuracy	92.34%

Table 3 presents outcomes from an ANN using the Fourier Gaussian technique to predict grades ranging from 2 to 5. It shows a detailed breakdown of correct and incorrect predictions across different grades. The model achieved an impressive accuracy of 95.67%, correctly predicting 380 instances out of the total 395 samples. Specifically, it accurately identified 77 out of 85 Grade 2 cases, 76 out of 81 Grade 3 cases, 82 out of 87 Grade 4 cases, and 86 out of 89 Grade 5 cases. The small number of incorrect predictions highlights the effectiveness of the ANN-based approach in this classification task. Table 4 presents results from an SVM-based Fourier Gaussian technique for predicting grades 2 to 5. The model achieved an accuracy of 92.34%, correctly predicting 376 out of 407 instances. It accurately identified 79 out of 82 Grade 2 cases, 89 out of 97 Grade 3 cases, 55 out of 56 Grade 4 cases, and all 61 Grade 5 cases. Sigma (σ), used in the context of Gaussian filters or statistical measures, is typically expressed in units of the data it relates to. For image processing, Sigma is often measured in pixels, which reflects the standard deviation of the Gaussian distribution applied to the image. For statistical analyses, Sigma is measured in the units of the data set being analyzed (e.g., meters, grams, etc.).

However, there were 18 incorrect predictions across various grades. The SVM-based approach performs well, particularly in correctly classifying Grade 3 and Grade 5 instances. Table 5 illustrates outcomes from an ANN utilizing the Daubechies wavelet method to predict grades 2 to 5. Achieving an accuracy of 94.23%, it correctly identified 378 out of 401 instances. Notably, the model accurately predicted 73 out of 78 Grade 2 cases, 72 out of 78 Grade 3 cases, 84 out of 85 Grade 4 cases, and 81 out of 86 Grade 5 cases. With 15 incorrect predictions, the ANN with Daubechies wavelets demonstrates reliable performance across multiple grade classifications. Table 6 displays results from an SVM employing the Daubechies wavelet method to forecast grades 2 through 5. It attained an accuracy of 90.67%, correctly predicting 372 out of 410 instances. Notably, the SVM accurately identified 72 out of 74 Grade 2 cases, 76 out of 78 Grade 3 cases, 77 out of 83 Grade 4 cases, and 88 out of 95 Grade 5 cases. Despite 15 incorrect predictions, the SVM with Daubechies wavelets demonstrates robust performance across various grade classifications.

Table 5
Outcomes of the ANN-based Daubechies wavelet method.

PREDICT ACTUAL	GRADE 2	GRADE 3	GRADE 4	GRADE 5	CORRECT NUMBER	378
Grade 2	73	4	0	1
Grade 3	2	72	2	2	Wrong Number	15
Grade 4	0	1	84	0
Grade 5	1	4	0	81	Accuracy	94.23%

Table 6
Outcomes of the SVM-based Daubechies wavelet method.

PREDICT ACTUAL	GRADE 2	GRADE 3	GRADE 4	GRADE 5	CORRECT NUMBER	372
Grade 2	72	0	1	1
Grade 3	0	76	2	0	Wrong Number	15
Grade 4	1	0	77	5
Grade 5	2	3	2	88	Accuracy	90.67%

Table 7 compares different textile filtering methods for Grade 2 based on the number of pilings and pilling area. The Fourier-Gaussian method recorded 67 pilings with a total pilling area of 1674, while the Daubechies Wavelets method showed 56 pilings with a smaller pilling area 654. This comparison suggests that the Daubechies Wavelets method may offer more efficient filtering for Grade 2 textile assessments than Fourier-Gaussian techniques. The textile photographs included in this investigation were processed using the technique Figure 12 displays the image processing outcomes from applying an ideal filter to the Fourier transform. Following the screening process, we followed the steps outlined in the methodology section.

Table 7
Comparing different textile filtering methods for Grade 2.

FILTERING METHOD	THE NUMBER OF PILLING	PILLING AREA
Fourier – Gaussian	67	1674
Daubechies Wavelets	56	654

Figure 12
Outcomes of the fourier transformation with optimal filtration.

Table 8 compares the performance of different filters applied in conjunction with ANN and SVM. The Fourier-Ideal filter achieved the highest accuracy with ANN (96.56%) and SVM (94.34%). The Fourier-Gaussian filter followed, showing slightly lower performance for both ANN (95.67%) and SVM (92.34%). The Daubechies wavelet filter had the lowest accuracies, with 94.23% for ANN and 90.67% for SVM. Overall, ANN outperformed SVM across all filters, indicating its superior capability in handling the data processed through these filters.

Table 8
Comparing different textile grading systems.

SI.NO.	FILTER	ANN	SVM
1	Fourier – Ideal	96.56%	94.34%
2	Fourier - Gaussian	95.67%	92.34%
3	Daubechies wavelet	94.23%	90.67%

The study aimed to grade fleece textiles based on pilling evaluation using image processing and machine learning techniques. The outcomes showed that both the Fourier-Gaussian and Daubechies wavelet approaches were effective in detecting pilling, with specific advantages in accuracy and detail extraction. The Fourier-Gaussian filter helps in frequency-based analysis, smoothing noise, while the Daubechies wavelet provides multi-resolution analysis, capturing finer details. These methods were chosen to compare their effectiveness in improving classification accuracy and pilling detection.

5. CONCLUSION

Many ML and filtering approaches have been proposed for assessing textile pilling, yet this research distinguishes itself by comparing novel combinations: the Fourier-Gaussian technique with ANN and the Fourier-Gaussian technique with SVM. These combinations, previously unexplored for scaling fibre materials, demonstrate superior performance. The study employed textile photographs and methods from prior research to validate the efficacy of the proposed techniques. The results indicate that the Fourier-Gaussian approach achieved an accuracy of 95.67% with ANN and 92.34% with SVM, while the Daubechies wavelet approach reached accuracies of 94.23% with ANN and 90.67% with SVM. The Fourier-Gaussian method also detected 67 instances of pilling with a total pilling area of 1674 units, compared to 56 instances with a pilling area of 654 units detected by the Daubechies method. This higher detection rate and accuracy highlight the superior performance of the Fourier-Gaussian approach, especially when using the SVM machine learning method.

This work introduces an alternative to traditional visual textile grading by utilizing a machine learning-based classification and image processing approach. By applying both Daubechies wavelet filtering and Discrete Fourier Transform (DFT) with Gaussian filtering, the study effectively separated pilling from the backdrop using binarization and extracted key characteristics through morphological and topological image processing techniques. The data was then objectively categorized using SVM and ANN. Future research could explore additional image filtering techniques and machine learning algorithms to enhance textile grading further. While the proposed methods show sufficient accuracy, the effectiveness of grading may vary with different colors and materials. A universal scaling system might not be applicable due to the diverse properties of fabric textures and densities; hence, developing customized grading techniques tailored to specific materials is recommended.

6. ACKNOWLEDGMENTS

Princess Nourah bint Abdulrahman University Researchers Supporting Project number (PNURSP2024R733), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia. Research Supporting Project number (RSPD2024R787), King Saud University, Riyadh, Saudi Arabia.

7. BIBLIOGRAPHY

• ^[1] MOHAMED, A.-A, .H, ., EL, -SAYED, I.M., “The impact of large volume atmospheric pressure air glow discharge plasma on enhancing denim fabric (textile) sewing process”, Journal of Taibah University for Science: JTUSCI, v. 18, n. 1, pp. 2303814, 2024. doi: http://doi.org/10.1080/16583655.2024.2303814.
  » https://doi.org/10.1080/16583655.2024.2303814
• ^[2] BISSI, L, ., BARUFFA, G, ., PLACIDI, P, ., et al., “Automated defect detection in uniform and structured fabrics using Gabor filters and PCA”, Journal of Visual Communication and Image Representation, v. 24, n. 7, pp. 838–845, 2013. doi: http://doi.org/10.1016/j.jvcir.2013.05.011.
  » https://doi.org/10.1016/j.jvcir.2013.05.011
• ^[3] CARDAMONE, J, .M, ., DAMERT, W, .C, ., PHILLIPS, J, .G, ., et al., “Digital image analysis for fabric assessment”, Textile Research Journal, v. 72, n. 10, pp. 906–916, 2002. doi: http://doi.org/10.1177/004051750207201009.
  » https://doi.org/10.1177/004051750207201009
• ^[4] GEORGE, J, .S, ., GEORGE, A, ., SEBASTIAN, M, ., et al., A holistic and integrated approach to lifestyle diseases, Palm Bay, Apple Academic Press, 2022.
• ^[5] SHANKER, U., RANI, M., Occurrence, distribution and toxic effects of emerging contaminantsx, Boca Raton, CRC Press, 2024. doi: http://doi.org/10.1201/9781003335757.
  » https://doi.org/10.1201/9781003335757
• ^[6] KADHAR, S, .A, ., GOPAL, E, ., SIVAKUMAR, V, ., et al., “Optimizing flow, strength, and durability in high-strength self-compacting and self-curing concrete utilizing lightweight aggregates”, Matéria (Rio de Janeiro), v. 29, n. 1, pp. e20230336, 2024. doi: http://doi.org/10.1590/1517-7076-rmat-2023-0336.
  » https://doi.org/10.1590/1517-7076-rmat-2023-0336
• ^[7] PARTHASAARATHI, R, ., BALASUNDARAM, N, ., ARASU, N, ., “Analysing the impact and investigating Coconut Shell Fiber Reinforced Concrete (CSFRC) under varied loading conditions”, Journal of Advanced Research in Applied Sciences and Engineering Technology, v. 35, n. 1, pp. 106–120, 2024.
• ^[8] PARTHASAARATHI, R, ., BALASUNDARAM, N, ., NAVEEN ARASU, A, ., “A stiffness analysis of treated and non-treated meshed coir layer fibre reinforced cement concrete”, In: AIP Conference Proceedings, v. 2861, n. 1, AIP Publishing, 2023. doi: http://doi.org/10.1063/5.0158672.
  » https://doi.org/10.1063/5.0158672
• ^[9] SRINIVASAN, S, .S, ., MUTHUSAMY, N, ., ANBARASU, N, .A, ., “The structural performance of fiber-reinforced concrete beams with nanosilica”, Matéria (Rio de Janeiro), v. 29, n. 3, pp. e20240194, 2024. doi: http://doi.org/10.1590/1517-7076-rmat-2024-0194.
  » https://doi.org/10.1590/1517-7076-rmat-2024-0194
• ^[10] KUMAR, S, .N, ., NATARAJAN, M, ., ARASU, A, .N, ., “A comprehensive microstructural analysis for enhancing concrete’s longevity and environmental sustainability”, Journal of Environmental Nanotechnology, v. 13, n. 2, pp. 368–376, 2024. doi: http://doi.org/10.13074/jent.2024.06.242584.
  » https://doi.org/10.13074/jent.2024.06.242584
• ^[11] JAIN, R, ., GOUR, B, ., DUBEY, S, ., “A hybrid approach for credit card fraud detection using rough set and decision tree technique”, International Journal of Computers and Applications, v. 139, n. 10, pp. 1–6, 2016. doi: http://doi.org/10.5120/ijca2016909325.
  » https://doi.org/10.5120/ijca2016909325
• ^[12] KUO, J, ., CHUNG, -FENG, C.-Y.S., HUANG, C.-C, ., et al., “Image inspection of knitted fabric defects using wavelet packets”, Textile Research Journal, v. 86, n. 5, pp. 553–560, 2016. doi: http://doi.org/10.1177/0040517514553872.
  » https://doi.org/10.1177/0040517514553872
• ^[13] JING, J, ., ZHANG, Z, ., KANG, X, ., et al., “Objective evaluation of fabric pilling based on wavelet transform and the local binary pattern”, Textile Research Journal, v. 82, n. 18, pp. 1880–1887, 2012. doi: http://doi.org/10.1177/0040517512449049.
  » https://doi.org/10.1177/0040517512449049
• ^[14] NAVEEN ARASU, A, ., NATARAJAN, M, ., BALASUNDARAM, N, ., et al., “Optimization of high performance concrete by using nano materials”, Research on Engineering Structures Materials, v. 3, n. 9, pp. 843–859, 2023.
• ^[15] KAYSERI, G, .Ö, ., KIRTAY, E, ., “Part 1. Predicting the pilling tendency of the cotton interlock knitted fabrics by regression analysis”, Journal of Engineered Fibers and Fabrics, v. 10, n. 3, pp. 155892501501000305, 2015. doi: http://doi.org/10.1177/155892501501000305.
  » https://doi.org/10.1177/155892501501000305
• ^[16] ROSOCHACKI, W, ., “The problem of evaluating the expiration of the service life of the supporting structure of secondhand cranes”, Archives of Transport, v. 49, n. 1, pp. 37–45, 2019. doi: http://doi.org/10.5604/01.3001.0013.2773.
  » https://doi.org/10.5604/01.3001.0013.2773
• ^[17] LIU, Z, ., LI, C, ., ZHAO, Q, ., et al., “A fabric defect detection algorithm via context-based local texture saliency analysis”, International Journal of Clothing Science and Technology, v. 27, n. 5, pp. 738–750, 2015. doi: http://doi.org/10.1108/IJCST-02-2014-0028.
  » https://doi.org/10.1108/IJCST-02-2014-0028
• ^[18] HUANG, M.-L, ., FU, C, ., “Applying image processing to the textile grading of fleece based on pilling assessment”, Fibers (Basel, Switzerland), v. 6, n. 4, pp. 73, 2018. doi: http://doi.org/10.3390/fib6040073.
  » https://doi.org/10.3390/fib6040073
• ^[19] ALI, M, .M, ., HASHIM, N, ., AZIZ, S, .A, ., et al., “Pineapple (Ananas comosus): a comprehensive review of nutritional values, volatile compounds, health benefits, and potential food products”, Food Research International, v. 137, pp. 109675, 2020. doi: http://doi.org/10.1016/j.foodres.2020.109675. PubMed PMID: 33233252.
  » https://doi.org/10.1016/j.foodres.2020.109675
• ^[20] SAHARKHIZ, S, ., ABDORAZAGHI, M, ., “The performance of different clustering methods in the objective assessment of fabric pilling”, Journal of Engineered Fibers and Fabrics, v. 7, n. 4, pp. 155892501200700413, 2012. doi: http://doi.org/10.1177/155892501200700413.
  » https://doi.org/10.1177/155892501200700413
• ^[21] BHATT, M, .R, ., “Systematic studies on Orchidaceae of Gujarat”, D.Sc. Thesis, Maharaja Sayajirao University of Baroda, India, 2019.
• ^[22] UMBAUGH, S, .E, ., Computer vision and image processing: a practical approach using cviptools with cdrom, Hoboken, Prentice Hall PTR, 1997.
• ^[23] NABI, F, ., MAHMUD, R, ., ISLAM, M, .M, ., “Improving sewing section efficiency through utilization of worker capacity by time study technique”, International Journal of Textile Science, v. 4, n. 1, pp. 1–8, 2015.
• ^[24] YUN, S, .Y, ., KIM, S, ., PARK, C, .K, ., “Development of an objective fabric pilling evaluation method. I. Characterization of pilling using image analysis”, Fibers and Polymers, v. 14, n. 5, pp. 832–837, 2013. doi: http://doi.org/10.1007/s12221-013-0832-x.
  » https://doi.org/10.1007/s12221-013-0832-x
• ^[25] SINGH, V, .K, ., BANSAL, G, ., NEGI, P, ., et al., Characterization of flexural and impact strength of jute/almond hybrid biocomposite, West Conshohocken, ASTM International, 2016.
• ^[26] OJA, M, ., KASKI, S, ., KOHONEN, T, ., “Bibliography of self-organizing map (SOM) papers: 1998–2001 addendum”, Neural Computing Surveys, v. 3, n. 1, pp. 1–156, 2003.
• ^[27] YUN, S, .Y, ., KIM, S, ., PARK, C, .K, ., “Development of an objective fabric pilling evaluation method. II. Fabric pilling grading using artificial neural network”, Fibers and Polymers, v. 14, n. 12, pp. 2157–2162, 2013. doi: http://doi.org/10.1007/s12221-013-2157-1.
  » https://doi.org/10.1007/s12221-013-2157-1
• ^[28] ZHANG, J, ., WANG, X, ., PALMER, S, ., “Objective pilling evaluation of wool fabrics”, Textile Research Journal, v. 77, n. 12, pp. 929–936, 2007. doi: http://doi.org/10.1177/0040517507083522.
  » https://doi.org/10.1177/0040517507083522
• ^[29] ZHENG, D, ., HAN, Y, ., JIN, L, .H, ., “A new method for classification of woven structure for yarn-dyed fabric”, Textile Research Journal, v. 84, n. 1, pp. 78–95, 2014. doi: http://doi.org/10.1177/0040517513483858.
  » https://doi.org/10.1177/0040517513483858

Publication Dates

• Publication in this collection
  08 Nov 2024
• Date of issue
  2024

History

• Received
  22 July 2024
• Accepted
  05 Sept 2024