Enhancing Quality Control in Industry 4.

0:
2023 First International Conference on Cyber Physical Systems, Power Electronics and Electric Vehicles (ICPEEV) | 979-8-3503-2356-6/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICPEEV58650.2023.10391889

Advanced Image Processing for Automated Defect

Detection
Makund Arora
Department of Computer Science Engineering,
Dayalbagh Educational Institute, Deemed University,
Agra, India.
Email: [email protected]

Abstract—With the advent of Industry 4.0, the integration of

advanced technologies has revolutionized manufacturing
processes. Quality control plays a crucial role in ensuring
product reliability and customer satisfaction. Traditional
manual inspection methods are time-consuming, subjective,
and prone to errors. This research paper proposes a novel
approach to enhance quality control in Industry 4.0 through
advanced image processing techniques for automated defect
detection. The research will involve extensive experimentation
and evaluation of the proposed system using real-world
manufacturing data. The results of this study will contribute
to the advancement of defect detection systems in the
manufacturing industry and provide valuable insights for
improving quality control processes in the era of Industry 4.0.

Keywords—Artificial Intelligence, Image Processing, Visual

Computing, Automated Defect Detection, Industry 4.0.
Fig. 1. Multi-Scale Saliency Defect Detection Algorithm
I. INTRODUCTION
C. Objectives
A. Background The main objectives of this research paper are as follows:
This research study aims to develop a comprehensive a. To develop an advanced image processing system for
methodology for automated defect detection and automated defect detection in the manufacturing industry.
classification using image processing and machine learning b. To explore and implement state-of-the-art image
techniques. The "Multi-Scale Saliency Defect Detection processing techniques and machine learning algorithms for
Algorithm" aims to improve accuracy, efficiency, and defect identification and classification.
reliability of defect identification in the manufacturing c. To evaluate the performance of the proposed system and
industry. The algorithm (Fig1) identifies and localizes compare it with existing manual inspection methods.
defects in images captured from manufacturing lines or d. To assess the system's robustness, efficiency, and
inspection systems with high accuracy and speed. This accuracy in real-time defect detection.
automated quality control system increases productivity e. To provide insights and recommendations for enhancing
and improves product quality, monitors manufacturing quality control processes in the context of Industry 4.0.
processes, assists in object detection and classification, and
aids in data-driven decision-making. II. LITERATURE REVIEW

B. Problem Statement In the field of image processing, various advancements

The manufacturing industry seeks innovative solutions for and applications have been made. Dastres& Soori (2021)
improving quality control and product quality in Industry and Prabha et al. (2021) discuss the benefits and recent
4.0. This research paper proposes an advanced image achievements in image processing systems, including
processing system for automated defect detection, defect detection algorithms for industrial products'
combining Multi-Scale Saliency Defect Detection surfaces. Fang et al. (2020) conduct a comprehensive
Algorithm and machine learning algorithms. The system's survey on surface defect detection technologies for metal
performance evaluation compares accuracy, processing materials. Azamfirei et al. (2023) review current trends in
speed, and efficiency against manual inspection methods, automation for quality inspection, focusing on achieving
and real-time defect detection and robustness assessment Zero Defect Manufacturing (ZDM). Lotheta Ajay (2022)
validate its applicability in challenging scenarios. explores the applications of image processing and pattern
recognition in industrial engineering, as well as the use of
cloud computing for image processing. Bulnes et al.
(2016) propose a methodfor efficient storage and access
of data related to periodical defects in web materials.

Authorized licensed use limited to: Dayananda Sagar University. Downloaded on January 30,2024 at 09:45:54 UTC from IEEE Xplore. Restrictions apply.
 Yang et al. (2017) presents an optical surface inspection and industrial processes. It is characterized by the
system for metal surfaces. Edwards (2015) discusses the integration of digital technologies, automation, data
use of neural networks in advanced services like image analytics, and the Internet of Things (IoT) into traditional
search and speech recognition. manufacturing systems.
In the context of quality improvement and industrial
engineering, Antony et al. (2022, 2023) examine (ii) Quality Control in Industry 4.0:
organizational readiness factors for implementing Quality control is a critical aspect of manufacturing,
Quality 4.0. Singh et al. (2020) focus on AI-based ensuring that products meet specific standards and
optimization techniques for supply chain management. requirements. In Industry 4.0, quality control is greatly
Mani &Patwardhan (2009) propose a hybrid constraint enhanced through the use of digital technologies.
handling technique for adaptive algorithms. Katiyar et al.
(2021) discusses defect detection using image processing (iii) The Connection between Industry 4.0 and Quality
and deep learning models. Maggipinto et al. (2019) Control in Industry 4.0
highlights the importance of anomaly detection in Industry 4.0 and quality control are closely linked. The real-
production systems. Cho et al. (2019) discuss the use of time data and advanced analytics capabilities in Industry
external wall insulation for energy efficiency. Schmitt et 4.0 enable more effective and efficient quality control. By
al. (2015) develop a real- time quality control system for identifying and addressing quality issues early in the
textile samples. Hoshikawa et al. (2019) explore image- production process, manufacturers can reduce defects,
based cell/colony analysis for regenerative medicine. improve product consistency, and enhance overall product
Habart et al. (2016) design an algorithm for segmenting quality. This not only leads to cost savings but also
pancreatic islets in microscopic images. Shen et al. improves customer satisfaction and ensures that products
(2012) describe a machine vision system for bearing meet the highest quality standards.
defect inspection. Manish et al. (2018) investigates image
processing techniques for surface finish and defect C. Image Processing Techniques For Defect Detection
inspection in machined surfaces. Manzano et al. (2020) The selected image processing techniques in this research
develop a low-cost machine vision system for industrial paper are chosen to develop an advanced defect detection
production processes. Jun et al. (2020) proposes a system for the manufacturing industry. Image acquisition,
framework for defect prediction in continuous-flow preprocessing, segmentation, feature extraction,
manufacturing. Psarommatis et al. (2020, 2022) explore classification, and post-processing stages are incorporated
Zero Defect Manufacturing (ZDM) and its adoption in to achieve accurate and automated defect identification.
quality improvement. Powell et al. (2021) discuss the These techniques enable the enhancement of image quality,
application of product-oriented machine learning the identification of regions of interest, the extraction of
algorithms for defect detection. Westphal & Seitz (2021) meaningful features, and the classification of defects using
design convolutional neural networks for defect machine learning algorithms. By employing these
classification in additive manufacturing. V et al. (2019) techniques, the research aims to improve quality control
focus on automated detection and characterization of processes, enhance product quality, and increase
defects using pulse thermography. Shepard (2003) manufacturing efficiency.
presents a model for thermal signal reconstruction and
defect detection. Förstner (2000) discusses pre- D. Machine Learning Approaches For Defect
processing techniques for feature extraction in digital Classification
images. This research study utilizes machine learning algorithms
for defect classification in automated defect detection in the
A. Significance Of The Study manufacturing industry. These algorithms, such as support
Advanced image processing systems for automated defect vector machines, random forests, neural networks, and
detection in the manufacturing industry are crucial for CNNs, can learn patterns and predict defects based on
reducing human error, improving product quality, and segmented regions. The goal is to develop a robust and
reducing human error. These systems use advanced efficient system capable of handling various defect types
techniques and machine learning algorithms to identify and improving the overall accuracy of the automated defect
defects in real-time, enhancing effectiveness over time. detection process. The evaluation and comparison of these
Evaluations against existing manual inspection methods approaches will help identify the most effective algorithm
demonstrate the system's superiority, making it practical in for defect classification in the manufacturing industry.
high-speed manufacturing environments. This work
contributes to Industry 4.0 advancements, enhancing E. Existing Research And Gaps
quality control processes and customer satisfaction. Existing research on defects and manufacturing processes
still lacks generalizability. More comprehensive studies
B. Industry 4.0 And Quality Control should consider a wider range of defects and manufacturing
Industry 4.0 and quality control are two interconnected processes. Machine learning approaches need improvement
concepts that have gained significant importance in the in classification accuracy and robustness. Integrating real-
manufacturing and industrial sectors. time defect detection and analysis is crucial for enhancing
(i) Industry 4.0: efficiency and productivity. Comparative studies on
Industry 4.0, often referred to as the Fourth Industrial automated defect detection systems versus manual
Revolution, represents a paradigm shift in manufacturing inspection methods would provide insights for adoption.

Authorized licensed use limited to: Dayananda Sagar University. Downloaded on January 30,2024 at 09:45:54 UTC from IEEE Xplore. Restrictions apply.
Integrating defect detection systems with existing quality devices, such as cameras or sensors. It is essential to ensure
control processes and Industry 4.0 frameworks is limited, consistent lighting conditions and proper calibration to
but exploring their integration and real-time analytics could minimize variations in image quality. Additionally, data
contribute to overall quality improvement. pre-processing techniques, including image resizing,
normalization, and noise reduction, may be applied to
III. METHODOLOGY enhance the quality and consistency of the dataset.

The research study on automated defect detection and B. Image Processing Pipeline
classification in the manufacturing industry follows a step- The image processing pipeline involves a series of
by-step methodology(Fig.2) . Firstly, a dataset of images or sequential steps to pre-process and analyse the captured
videos capturing the manufacturing process is collected. images. These steps may include noise removal, image
Then, pre-processing techniques are applied to enhance enhancement, edge detection, segmentation, and region of
image quality. Manual annotation of defects in the dataset interest (ROI) extraction. Noise removal techniques, such
is performed to create ground truth labels. Relevant features as median filtering or Gaussian filtering, can be applied to
are extracted from the annotated regions to represent defect eliminate unwanted artifacts. Image enhancement
characteristics. Machine learning models, such as support techniques, like contrast adjustment or histogram
vector machines, random forests, or neural networks, are equalization, can be employed to improve the visibility of
selected and trained on the annotated data. defects. Edge detection algorithms, such as Canny edge
detection, can identify potential defect boundaries.
Segmentation techniques, such as thresholding or
clustering, can be used to separate defects from the
background. Finally, ROI extraction is performed to isolate
the regions containing defects for further analysis.

C. Residual Learning Process

The residual learning process in the proposed system
involves fitting a series of stacked layers to approximate the
underlying mapping represented by H(x), where x denotes
the input to the first layer. Instead of expecting these layers
to directly approximate H(x), a hypothesis is made that they
can asymptotically approximate the residual function,
which is H(x) - x. This assumption holds true when the
input and output dimensions are the same.
To reformulate the original function, it is expressed as F(x)
+ x, where F(x) represents the desired mapping and x is the
input. F(x) and F(x) + x can asymptotically approximate the
desired functions. However, the ease of learning may differ
between the two forms.
The motivation for this reformulation arises from the
degradation problem observed in deep neural networks.
Ideally, if the added layers in a deep model could function
Fig. 2. Methodology of Image Processing System as identity mappings, a deeper model should not have a
higher training error than its shallower counterpart.
The trained models are evaluated using appropriate metrics However, in practice, the degradation problem occurs,
to assess their classification performance. Performance indicating that solvers struggle to approximate identity
optimization is conducted through fine-tuning and mappings using multiple nonlinear layers.
parameter adjustments. Real-time implementation of the To address this issue, shortcut connections or skip
models into a defect detection system is carried out. connections are introduced. These shortcuts are
Comparative analysis is conducted to compare the system's incorporated every few stacked layers, allowing the
performance against manual inspection methods. Finally, iterative application of the residual learning concept. A
the results are validated, interpreted, and discussed to building block is utilized to represent this concept, enabling
provide insights and identify areas for improvement. By the layers to explicitly approximate the residual function.
following this methodology, the study aims to develop an By incorporating these shortcuts and adopting the residual
effective and efficient automated defect detection and learning approach, deep neural networks can effectively
classification system for the manufacturing industry. learn complex mappings and mitigate the degradation
problem.
A. Data Collection And Pre-Processing In the residual learning approach,
To develop and evaluate the proposed automated defect y = F(x, {Wi}) + x.---------(1)
detection system, a comprehensive dataset of high- Here, x and y represent the input and output vectors of the
resolution images of industrial components with known layers, and F(x, {Wi}) represents the residual mapping that
defects is required. The data collection process involves needs to be learned. The introduction of shortcut
capturing images of the components using suitable imaging connections in Eq. (1) does not introduce additional

Authorized licensed use limited to: Dayananda Sagar University. Downloaded on January 30,2024 at 09:45:54 UTC from IEEE Xplore. Restrictions apply.
parameters or computational complexity. This practical
advantage facilitates fair comparisons between plain and F. Model Training And Evaluation
residual networks, as they can have the same number of In the era of Industry 4.0, the integration of a Preactor
parameters, depth, width, and computational cost (except scheduling system with neural network modelling in image
for the negligible element-wise addition). It is crucial to processing offers an innovative solution for optimizing
ensure that the dimensions of x and F are equal in Eq. (1). production processes (Fig. 4). By leveraging the power of
If they differ, a linear projection Ws can be applied through neural networks, this approach enables real-time decision-
the shortcut connections to match the dimensions, making and adaptive scheduling strategies based on visual
y = F(x, {Wi}) + Wsx.--------(2) data analysis. The system accurately identifies bottlenecks,
The use of a square matrix Ws is possible, but experiments monitors machine performance, and predicts potential
have shown that employing the identity mapping is production disruptions, empowering industries to
sufficient for addressing the degradation problem and is proactively address challenges, minimize downtime, and
more efficient. Thus, Ws is only utilized when dimension improve resource allocation. With this integration,
matching is necessary. manufacturers can enhance overall productivity and
It is important to note that although the above notations efficiency, staying agile and responsive in dynamic
simplify the explanation using fully-connected layers, they environments, and fully harnessing the potential of Industry
are also applicable to convolution layers. The function F(x, 4.0 to drive their success.
{Wi}) through eq.2 can represent multiple convolution
layers, and the element-wise addition is performed channel
by channel on two feature maps.

D. Feature Extraction And Selection

Once the ROIs are identified, relevant features need to be
extracted to characterize the defects. These features can
include texture descriptors, shape descriptors, colour
histograms, or statistical measures. The choice of features
depends on the nature of the defects and the specific
requirements of the classification model. Feature selection
techniques, such as principal component analysis (PCA) or
feature ranking algorithms, can be employed to reduce the
dimensionality of the feature space and enhance the
efficiency and effectiveness of defect classification.

E. Machine Learning Model Development

Fig. 4. Model of Image Processing

Fig. 3. Machine Learning Model

Machine learning models are developed to classify the .

extracted features into different defect categories. Various Fig. 5. Neural Network Architecture for Industry 4.0
algorithms, including SVMs, random forests, or CNNs, can General Model
be utilized based on the complexity and size of the dataset.
The choice of the model is influenced by factors such as G. TESTING AND SENSITIVITY ANALYSIS
classification accuracy, computational efficiency, and the The trained network model was evaluated to assess its
ability to handle high-dimensional feature spaces. The capability in predicting the response of experimental data
model architecture and hyperparameters are determined that were not included in the training process. The network
through experimentation and optimization techniques, such was then subjected to tests using training, cross-validation,
as cross-validation or grid search.

Authorized licensed use limited to: Dayananda Sagar University. Downloaded on January 30,2024 at 09:45:54 UTC from IEEE Xplore. Restrictions apply.
and testing data sets. By comparing the desired outputs with
the outputs generated by the network, the results were A. Hardware And Software Setup
analyzed for any variations or disparities.(Arora, 2011) The implementation of the automated defect detection
system requires appropriate hardware and software
configurations. The hardware setup may include high-
resolution cameras or sensors for image capture, a
computing system with sufficient processing power and
memory to handle image processing tasks, and any
necessary peripherals. The software setup may involve
programming languages (Python, C++), libraries
(OpenCV, scikit-learn, TensorFlow), and development
environments for implementing the image processing
algorithms, machine learning models, and user interface
components.

B. Integration Of The Image Processing System

The image processing system needs to be integrated into
the existing manufacturing environment or production line.
This integration involves establishing communication
between the image capture devices, computing system, and
other components of the manufacturing process. The image
processing pipeline should be integrated into the system's
Fig. 6. Statistical Analysis of MSE workflow, ensuring real-time processing and defect
detection. Proper synchronization and data transfer
An artificial neural network (ANN) model based on the mechanisms should be implemented to facilitate seamless
single layer recurrent backpropagation algorithm. During integration with minimal disruption to the production line.
training, the input is propagated forward through the
network to compute the output value of each unit using a C. User Interface Design
forward pass (Fig.5). The output vector is compared with A user interface is essential to interact with the automated
the desired vector, resulting in an error signal for each defect detection system. The user interface can provide
output unit. The weights of the network are adjusted to functionalities such as initiating the defect detection
minimize the error, and this process is repeated for several process, displaying real-time defect detection results, and
iterations until the network can produce the desired output enabling user feedback or manual intervention when
for a given input. The number of layers network structure necessary. The design of the user interface should consider
includes n- hidden neurons in each layer, which describe user-friendliness, ease of navigation, and clear visual
the dynamics of industry output. The performance of the representation of defect information. It should also
network was evaluated using mean square error (MSE) incorporate appropriate controls and feedback mechanisms
criterion. The training, testing, and cross-validation data to ensure efficient operation and user satisfaction.
were divided into eg. 60%, 20%, and 20% respectively. The
network simulation achieved a minimum MSE for training V. RESULTS AND DISCUSSION
and for cross-validation up to eight decimal places. The
results obtained by the Neural Network simulation for This approach leverages neural networks' capabilities to
training, cross-validation, and testing data sets are shown in enable real-time decision-making and adaptive scheduling
Figure 6. strategies based on visual data analysis. By effectively
identifying bottlenecks, monitoring machine performance,
and predicting potential production disruptions, this
integrated system empowers industries to proactively
address challenges, minimize downtime, and improve
resource allocation. This leads to enhanced overall
productivity and efficiency, allowing manufacturers to stay
agile and responsive in dynamic environments and fully
utilize the potential of Industry 4.0 to drive success. The
trained neural network model's evaluation demonstrates its
ability to predict responses of experimental data not used
during the training process. The network undergoes testing
using various data sets, including training, cross-validation,
and testing data. By comparing the model's generated
outputs with the desired outputs, any variations or
Fig. 7. Graphical Representation of System’s Real Time
disparities are analyzed. This process assesses the accuracy
Decision
and generalization capabilities of the model. The use of a
single-layer recurrent backpropagation algorithm for the
IV. SYSTEM IMPLEMENTATION
artificial neural network's structure is outlined. During

Authorized licensed use limited to: Dayananda Sagar University. Downloaded on January 30,2024 at 09:45:54 UTC from IEEE Xplore. Restrictions apply.
training, forward and backward passes adjust the network's Statistical analysis can be employed to determine if the
weights to minimize errors iteratively. The model's automated system outperforms or is comparable to manual
performance is evaluated using the mean square error inspection methods.
(MSE) criterion. The data is divided into training, testing,
and cross-validation sets, and the model achieves minimal C. Robustness And Efficiency Analysis
MSE values for training and cross-validation. The The robustness and efficiency of the automated defect
network's ability to learn and generalize is demonstrated by detection system should be analyzed, Robustness
the results obtained across these data sets. The system's assessment involves testing the system's performance under
ability to make real-time decisions, predict disruptions, and various challenging conditions, such as changes in lighting,
adaptively schedule tasks based on visual data analysis can variations in defect types or sizes, and complex
significantly enhance productivity and efficiency in backgrounds. The system's ability to handle such variations
manufacturing environments (Fig.7). The thorough testing, without compromising accuracy is crucial. Efficiency
evaluation, and implementation strategies outlined analysis focuses on measuring the system's processing
underscore the potential benefits of this innovative time, resource utilization, and scalability. Real-time defect
approach for industries seeking to leverage Industry 4.0 detection capabilities should be evaluated, and the system's
technologies to their advantages. performance should be monitored to ensure it meets the
desired operational requirements.
Table.1: Correct Classification in Percentage The results and discussion section should present the
evaluation outcomes, compare them with expectations and
Classifier No. of Testing Training Cross existing literature, and provide insights into the system's
Approaches Iteration Validation strengths, limitations, and potential areas for improvement.
I R1 92.4 93.4 91.2 Visual representations, such as graphs, charts, or images,
R2 90.6 92.7 93.7 can be utilized to enhance the presentation and
R3 91.2 94.2 94.5 understanding of the results.
II R1 90.4 88.4 86.3
R2 87.3 89.7 85.6 VI. CONCLUSION
R3 90.5 91.4 88.5
III R1 87.5 83.4 79.4 In this research paper, we proposed an advanced image
R2 85.3 83.7 78.5 processing system for automated defect detection in the
R3 86.3 82.9 79.3 context of Industry 4.0. Through the implementation and
IV R1 88.7 83.5 90.2 evaluation of the system, several key findings have
R2 83.4 82.9 88.7 emerged. Firstly, the developed system successfully
R3 86.3 89.8 83.5 automated the defect detection process, eliminating the
I -mi-SVM average, II -MIL-Boost, need for manual inspection and reducing the chances of
III- MILES, IV- MILES human error. The system demonstrated accurate and
The study reveals that each classifier shown in Table1 has efficient identification and classification of defects in real-
distinct capabilities in identifying specific scenarios. The time, leading to improved product quality and operational
mi-SVM average classifier is effective in detecting faulty efficiency. Performance evaluation metrics, such as
screws or sockets but less competent in identifying faulty accuracy, precision, recall, and F1-score, indicated the
labels. Other classifiers also have similar abilities in system's effectiveness in defect detection compared to
detecting production defects. However, labels are not manual inspection methods. The system also showed
particularly good at detection. Primary edge detection robustness and adaptability in handling variations in
involves scanning and dividing an image into sections, lighting conditions, image quality, and complex
using three fundamental steps: smoothing, edge backgrounds. Furthermore, the integration of machine
enhancement, and edge localization. learning algorithms enhanced defect classification
capabilities and increased the system's accuracy.
A. Performance Evaluation Metrics
The performance of the automated defect detection system A. Contributions To Knowledge
needs to be evaluated using appropriate metrics. This research advances Industry 4.0 quality control by
Performance evaluation metrics can include accuracy, automating defect detection using advanced image
precision, recall, F1-score, and confusion matrix. These processing techniques and machine learning algorithms. It
metrics provide insights into the system's ability to improves the overall quality control process, emphasizes
correctly classify defects, minimize false positives and real-time processing, and expands knowledge on defect
false negatives, and provide an overall assessment of its classification using machine learning.
performance.
B. Limitations And Future Work
B. Comparison With Manual Inspection Methods This research has some limitations, including the lack of
To assess the effectiveness of the automated defect labeled defect datasets for machine learning models. To
detection system, a comparison with manual inspection improve efficiency, future work should focus on labeling
methods should be conducted. This can involve evaluating techniques or exploring semi-supervised or unsupervised
the system's performance against human inspectors in learning approaches. Real-world testing and integrating
terms of detection accuracy, consistency, and efficiency. sensor data, such as temperature or vibration, can enhance

Authorized licensed use limited to: Dayananda Sagar University. Downloaded on January 30,2024 at 09:45:54 UTC from IEEE Xplore. Restrictions apply.
defect detection capabilities. Cost-effectiveness analysis Manufacturing, 27(2), 431–445.
https://doi.org/10.1007/s10845-014-0876-9
and optimization of hardware and software components are
crucial for practical implementation and widespread [10] Cao, Q., Zanni-Merk, C., Samet, A., Reich, C., Beuvron, F. de
adoption in the manufacturing industry. B. de, Beckmann, A., &Giannetti, C. (2022). KSPMI: A
In conclusion, this research paper presents a comprehensive Knowledge-based System for Predictive Maintenance in
Industry 4.0. Robotics and Computer-
study on enhancing quality control in Industry 4.0 through
IntegratedManufacturing,74,102281.
advanced image processing for automated defect detection. https://doi.org/10.1016/j.rcim.2021.102281
The proposed system has demonstrated its potential to
revolutionize quality control processes, improve product [11] Chen, B., Wan, J., Shu, L., Li, P., Mukherjee, M., & Yin, B.
(2018). Smart Factory of Industry 4.0: Key Technologies,
quality, and increase operational efficiency. The findings
Application Case, and Challenges. IEEE Access,6,6505–
and insights from this research contribute to the knowledge 6519. https://doi.org/10.1109/ACCESS.2017.2783682
base in the field, providing a foundation for further
advancements in automated defect detection and quality [12] Cho, S.-H., Lee, K.-T., Kim, S.-H., & Kim, J.-H. (2019). Image
Processing for Sustainable Remodeling: Introduction to Real-
control in the era of Industry 4.0.
time Quality Inspection System of External Wall Insulation
Works. Sustainability, 11(4), 1081.
ACKNOWLEDGMENT https://doi.org/10.3390/su11041081

[13] Czimmermann, T., Ciuti, G., Milazzo, M., Chiurazzi, M.,

I express gratitude to Prof. Monika Aggarwal, Prof. P.K.
Roccella, S., Oddo, C. M., & Dario, P. (2020). Visual-Based
Kalra, Prof. C. Patwardhan, and Prof. Ajay Saxena for their Defect Detection and Classification Approaches for Industrial
invaluable support and guidance in my research paper. I also Applications—A SURVEY. Sensors, 20(5), 1459.
thank Director, Prof. P.K. Kalra, for his inspiration and https://doi.org/10.3390/s20051459
guidance.
[14] Dastres, R., & Soori, M. (2021). Advanced Image Processing
Systems Article in. In International Journal of Imaging and
REFERENCES Robotics. www.ceserp.com/cp-jour

[1] Aceto, G., Persico, V., & Pescapé, A. (2020). Industry 4.0 and [15] De Mitri, O., Stahl, J., Jauch, C., &Distante, C. (2019). Image
Health: Internet of Things, Big Data, and Cloud Computing for acquisition, evaluation and segmentation of thermal cutting
Healthcare 4.0. Journal of Industrial Information Integration, edges using a mobile device. In S. Negahdaripour, E. Stella, D.
18, 100129. https://doi.org/10.1016/j.jii.2020.100129 Ceglarek, & C. Möller (Eds.), Multimodal Sensing:
Technologies and Applications (p. 24). SPIE.
[2] Adaway, B. (1983). Industrial applications of image https://doi.org/10.1117/12.2527993
processing. Computer Graphics 83, (Online Publications,
Pinner), 555–568. https://doi.org/10.2478/aucts-2014-0004 [16] Dhanasekar, J., Sharan, A. P., Natarajan, M. A., Nizamudeen,
A., Krishnan, H. M., & Kumar, S. R. S. (2022). Design and
[3] Antony, J., McDermott, O., & Sony, M. (2022). Quality 4.0 Implementation of a Defect Identification Using Image
conceptualisation and theoretical understanding: a global Processing Technique (pp. 734–751).
exploratory qualitative study. The TQM Journal, 34(5), 1169– https://doi.org/10.1007/978-981-19-1677-9_65
1188. https://doi.org/10.1108/TQM-07-2021-0215
[17] Dogo, E. M., Salami, A. F., Aigbavboa, C. O., &Nkonyana, T.
[4] Antony, J., Sony, M., McDermott, O., Jayaraman, R., & Flynn, (2019). Taking Cloud Computing to the Extreme Edge: A
D. (2023). An exploration of organizational readiness factors Review of Mist Computing for Smart Cities and Industry 4.0
for Quality 4.0: an intercontinental study and future research in Africa. In Edge Computing (pp. 107–132).
directions. International Journal of Quality & Reliability https://doi.org/10.1007/978-3-319-99061-3_7
Management, 40(2), 582–606.
https://doi.org/10.1108/IJQRM-10-2021-0357 [18] Edwards, C. (2015). Growing pains for deep learning.
Communications of the ACM, 58(7), 14–16.
[5] J K Arora (2011) “Artificial Neural Network modelling for the https://doi.org/10.1145/2771283
System of blood flow through tapered artery with mild stenosis
,International Journal of Mathematics Trends and Technology- [19] Fang, X., Luo, Q., Zhou, B., Li, C., & Tian, L. (2020).
July to Aug 2011, ISSN: 2231- 5373, Research progress of automated visual surface defect detection
http://www.internationaljournalssrg.org for industrial metal planar materials. Sensors
(Switzerland), 20(18), 1–35.
[6] Awcock, G. J., & Thomas, R. (1995). Applied Image https://doi.org/10.3390/s20185136
Processing. Macmillan Education UK.
https://doi.org/10.1007/978-1-349-13049-8 [20] Förstner, W. (2000). Image Preprocessing for Feature
Extraction in Digital Intensity, Color and Range Images. In
[7] Azamfirei, V., Psarommatis, F., &Lagrosen, Y. (2023). Geomatic Method for the Analysis of Data in the Earth
Application of automation for in-line quality inspection, a Sciences (pp. 165–189). Springer Berlin Heidelberg.
zero-defect manufacturing approach. In Journal of https://doi.org/10.1007/3-540-45597-3_4
Manufacturing Systems (Vol. 67, pp. 1–22). Elsevier B.V.
https://doi.org/10.1016/j.jmsy.2022.12.010 [21] Godina, R., & Matias, J. C. O. (2019). Quality Control in the
Context of Industry 4.0. In Industrial Engineering and
[8] Bhattacharya, A., & Cloutier, S. G. (2022). End-to-end deep Operations Management II (pp. 177–187).
learning framework for printed circuit board manufacturing https://doi.org/10.1007/978-3-030-14973-4_17
defect classification. Scientific Reports, 12(1), 12559.
https://doi.org/10.1038/s41598-022- [22] H. (2019). Automated Visual Inspection of Glass Bottle
16302-3 Bottom With Saliency Detection and Template Matching.
IEEE Transactions on Instrumentation and Measurement,
[9] Bulnes, F. G., Usamentiaga, R., Garcia, D. F., &Molleda, J. 68(11), 4253–4267.
(2016). An efficient method for defect detection during the https://doi.org/10.1109/TIM.2018.2886977
manufacturing of web materials. Journal of Intelligent

Authorized licensed use limited to: Dayananda Sagar University. Downloaded on January 30,2024 at 09:45:54 UTC from IEEE Xplore. Restrictions apply.
 [23] Habart, D., Švihlík, J., Schier, J., Cahová, M., Girman, P.,
Zacharovová, K., Berkov, Z., Kříž, J., Fabryová, E., Kosinová,
L., Papáčková, Z., Kybic, J., &Saudek, F. (2016). Automated
Analysis of Microscopic Images of Isolated Pancreatic Islets.
Cell Transplantation, 25(12), 2145–2156.
https://doi.org/10.3727/096368916X692005

[24] Hans Raj, K., Sharma, R. S., Srivastava, S., &Patvardhan, C.

(2000). Modeling of manufacturing processes with ANNs for
intelligent manufacturing. International Journal of Machine
Tools and Manufacture, 40(6), 851–868.
https://doi.org/10.1016/S0890-6955(99)00094-2

[25] Jalayer, M., Jalayer, R., Kaboli, A., Orsenigo, C., &Vercellis,
C. (2021). Automatic Visual Inspection of Rare Defects: A
Framework based on GP-WGAN and Enhanced Faster R-
CNN. 2021 IEEE International Conference on Industry 4.0,
Artificial Intelligence, and Communications Technology
(IAICT),221–227.
https://doi.org/10.1109/IAICT52856.2021.9532584

[26] Jasperneite, J., Sauter, T., &Wollschlaeger, M. (2020). Why

We Need Automation Models: Handling Complexity in
Industry 4.0 and the Internet of Things. IEEE Industrial
Electronics Magazine, 14(1), 29–40.
https://doi.org/10.1109/MIE.2019.2947119

[27] Javaid, M., Haleem, A., Pratap Singh, R., & Suman, R. (2021).
Significance of Quality 4.0 towards comprehensive
enhancement in manufacturing sector. Sensors International, 2,
100109. https://doi.org/10.1016/j.sintl.2021.100109

[28] ll, D., Eleftheriadis, R., &Myklebust, O. (2021). Digitally

Enhanced Quality Management for Zero Defect
Manufacturing. Procedia CIRP, 104, 1351–1354.
https://doi.org/10.1016/j.procir.2021.11.227

[29] Psarommatis, F., & Bravos, G. (2022). A holistic approach for

achieving Sustainable manufacturing using Zero Defect
Manufacturing: a conceptual Framework. Procedia
CIRP, 107, 107–112.
https://doi.org/10.1016/j.procir.2022.04.018

[30] Saberironaghi, A., Ren, J., & El-Gindy, M. (2023). Defect

Detection Methods for Industrial Products Using Deep
Learning Techniques: A Review. Algorithms, 16(2), 95.
https://doi.org/10.3390/a16020095

[31] Schmitt, R., Fürtjes, T., Abbas, B., Abel, P., Kimmelmann, W.,
Kosse, P., & Buratti, A. (2015). Real-Time Machine Vision
System for an Automated Quality Monitoring in Mass
Production of Multiaxial Non-Crimp Fabrics. IFAC-
PapersOnLine, 48(3), 2393–2398.
https://doi.org/10.1016/j.ifacol.2015.06.446

[32] & Wu, H. (2016). Using GF-2 Imagery and the Conditional
Random Field Model for Urban Forest Cover Mapping.
Remote Sensing Letters, 7(4), 378–387.
https://doi.org/10.1080/2150704X.2016.1142682
[33] Westphal, E., & Seitz, H. (2021). A machine learning method
for defect detection and visualization in selective laser
sintering based on convolutional neural networks. Additive
Manufacturing, 41, 101965.
https://doi.org/10.1016/j.addma.2021.101965

[34] Yang, H., Haist, T., Gronle, M., &Osten, W. (2017).

Simulation of microscopic metal surfaces based on measured
microgeometry. Tm - TechnischesMessen, 84(7–8), 493–501.
https://doi.org/10.1515/teme-2017-
0019

Authorized licensed use limited to: Dayananda Sagar University. Downloaded on January 30,2024 at 09:45:54 UTC from IEEE Xplore. Restrictions apply.