Advanced AI-Based Solutions for Visual Inspection: A Systematic

Literature Review

Angelo Corallo

, Vito Del Vecchio

and Alberto Di Prizio

Department of Innovation Engineering, University of Salento, Lecce, Italy

Keywords: Visual Inspection, Artificial Intelligence, Machine Learning, Deep Learning, Manufacturing, Industry.

Abstract: Artificial Intelligence (AI)-based solutions, including Machine Learning (ML) and Deep Learning (DL), are

ever more implemented in industry for assisting advanced Visual Inspection (VI) systems. They support

companies in a more effective identification of product defects, enhancing the performance of humans and

avoiding the risks of product incompliance. However, companies often struggle in considering the most

appropriate AI-based solutions for VI and for a specific manufacturing domain. Also, an extensive literature

study focused on this topic seems to lack. On the basis of a Systematic Literature Review, this paper aims to

map the main advanced AI-based VI system solutions (including methods, technologies, techniques,

algorithms) thus helping companies in considering the most appropriate solutions for their needs.

1 INTRODUCTION

In recent years, the adoption of Artificial Intelligence

(AI), including Machine Learning (ML) and Deep

Learning (DL), has significantly improved the

efficiency of industrial solutions in Computer Vision

(CV) and Visual Inspection (VI) activities (Kitaguchi

et al., 2022). This is particularly relevant for

manufacturing companies. These technologies have

transformed the roles and relationships between

machines, humans, automated processes. Traditional

manual VI processes have been replaced by advanced

AI solutions, aiming to enhance the accuracy and

speed through the integrated use of enabling

technologies. Advanced technologies not only

improve the accuracy and speed of analysis, but also

address issues such as defect control. CV emerges as

a discipline for supporting production processes and

high product quality. While traditional techniques

still find integration challenges in data processing,

monitoring, time, quality, and input data (Paneru &

Jeelani, 2021), modern and complex systems need the

implementation of advanced AI solutions for more

efficient and reactive production processes. Despite

several studies are availbale on CV and AI in the

scientific literature (Zhou et al., 2019), a

https://orcid.org/0000-0001-5216-5366

https://orcid.org/0000-0001-9040-783X

https://orcid.org/0009-0003-4818-4718

comprehensive map of AI solutions for VI seems to

lack. Through a Systematic Literature Review (SLR),

the paper maps all the AI-based technologies for VI,

including methods, techniques, algorithms and

enabling technologies. An in-depth analysis reveals

several sectors where advanced AI-based solutions

are crucial. The study contributes to the literature by

providing a comprehensive analysis of AI for VI. It

also serves as a practical guide for industrial

companies for a clear view of AI-VI applications.

2 THEORETICAL

BACKGROUND

2.1 Advanced Visual Inspection

Solutions

VI represents a methodological and technological

approach to detect component defects and control

their quality (Cottrell, 2019). To date, the human eye

is the main tool to ensure product compliance, but in

this scenario, inspection activities can only be carried

out if humans are close to the products. Advanced VI

solutions include not only the use of technologies to

656

Corallo, A., Del Vecchio, V. and Di Prizio, A.

Advanced AI-Based Solutions for Visual Inspection: A Systematic Literature Review.

DOI: 10.5220/0012618000003690

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 26th International Conference on Enterprise Information Systems (ICEIS 2024) - Volume 1, pages 656-664

ISBN: 978-989-758-692-7; ISSN: 2184-4992

support and improve inspection results, but also the

ability to remote sensing (American Bureau of

Shipping, 2022). Traditional inspections include

repetitive tasks that can reduce workers' attention,

increasing the risk of misidentification. Such issues

are critical within manufacturing industries,

especially in sectors such as aerospace and

automotive where product quality must be high

(Brandoli et al., 2021).

Technological advances, associated with the

digitization of industrial processes, assets and

information, and the adoption of enabling

technologies such as the Internet of Things (IoT),

have pushed companies to rely on data and collect

images and videos from their production assets to

perform VI. The availability of digital data is a

relevant factor for performing real-time inspections

(Yang et al., 2021). In this context, automated VI,

also known as Smart VI (El Zant et al., 2021),

represents the most innovative inspection activity,

which involves the use of integrated hardware and

software systems based on AI. In the field of AI, CV

aims to study how computers can interpret and

understand the visual world through the use of

specific learning methods and algorithms (Brownlee,

2022). By training algorithms, companies can

automate the inspection process, saving time and

improving the accuracy and quality of results.

Regardless of the type of algorithms, defect

identification activities and the related systems

solution require knowledge of some critical issues,

such as signal acquisition, signal pre-processing,

feature selection and extraction, fusion and

classification of the data (Piuri et al., 2005). In

manufacturing scenarios, advanced VI solutions are

usually implemented to identify defects such as

corrosion (Brandoli et al., 2021) and scratches (Zhou

et al., 2019) on the surface of components. In the

aerospace sector, identifying defects on turbine

blades is relevant to understand and ensure their

functional compliance (Aust et al., 2021).

By adopting advanced AI-based technology

solutions, organizations can conduct inspections

more quickly and accurately across a wider range of

environments, while also saving costs and keeping

people out of critical areas while improving safety

(IBM, 2023). VI has become a crucial methodological

and technological approach to detect defects from

components and ensure quality control (Cottrell,

2019). It is the technological progress that has

occurred in the digitalization of services and the

adoption of the IoT that drives the collection of data

for VI, facilitating real-time assessments (Yang et al.,

2021). Training algorithms to read images improves

inspection automation, saving time and improving

accuracy. Regardless of the type of algorithm, defect

identification involves critical aspects such as signal

acquisition, preprocessing, feature selection, data

fusion and classification (Piuri et al., 2005).

2.2 Artificial Intelligence for Visual

Inspection

In the context of advanced AI-based VI, the

integration of ML and DL models is critical for

software and hardware systems (Mueller &

Massaron, 2021; Voulodimos et al., 2018). Emulating

human intelligence to analyze unknown data, AI

shows adaptive improvement as the learning sample

size increases (Mueller & Massaron, 2021). DL,

distinct from ML, usually operates on unstructured

data with or without preprocessing, enabling

automatic extraction of components and reducing

dependence on human experts (Voulodimos et al.,

2018).

Moreover, among supervised, unsupervised and

reinforcing learning models, supervised ML is

commonly employed in defect detection algorithms,

which use structured input-output pairs labelled with

expected values. They are mainly composed of

(Mujeeb et al., 2019): i) algorithm, as a set of rules,

often rooted in statistical methods, that extract

recurring patterns from data; ii) model, a

representation of the real context with appropriate

parameters and constraints; iii) training dataset, that

includes data for feeding the learning algorithm with

the association of known output values assigned by

human; iv) test datasets, that after the training phase

allows for evaluating the quality of the model in terms

of precision and accuracy. In particular, supervised

learning algorithms, such as classification and

regression models are commonly used (Yang et al.,

2021; Zaidi et al., 2021). While classification models

return analytical labels associated with binary or

nominal qualitative variables, regression models

identify outputs with continuous numerical values.

The integration of ML and DL represents a

challenge for the development of robust AI solutions

for VI, contributing to advance defect identification

(Mueller & Massaron, 2021; Voulodimos et al.,

2018).

3 RESEARCH METHODOLOGY

This article relies on a robust review methodology,

the systematic literature review (SLR), to explore the

scientific literature and consider diverse research

Advanced AI-Based Solutions for Visual Inspection: A Systematic Literature Review

657

contributions (Xiao & Watson, 2019). Therefore,

based on the SLR procedure suggested by (Corallo et

al., 2023) this article explores the scientific body

regarding advanced AI solutions, including ML, to

perform automated VI in manufacturing. The adopted

SLR protocol consist of four steps (Figure 1).

3.1 Step 1 - Review Planning

This article aims to map AI-based solutions for VI

and answer the Research Question (RQ): “What are

the main AI-based solutions to support VI in

manufacturing?" Scopus (https://www.scopus.com)

has been selected as source of reference (Mishra et

al., 2016). The relevant search keywords

(“Automated Visual Inspection” OR “Visual

Inspection”) AND (“Artificial Intelligence” OR

“Machine Learning”) AND (“Manufacturing” OR

“Industry 4.0”) have been combined. The inclusion

and exclusion criteria were set to refine the review,

selecting documents in line with the SLR objective,

which are in English and belong to the engineering or

computer science fields.

3.2 Step 2 - Review Execution

This step consists of implementing the search query

on Scopus and filtering the results. Articles were

retrieved if the planned keywords appeared in the title,

keywords, or abstract. The query returned 98 papers

which were downloaded to allow content analysis.

However, 13 articles were unavailable, so the final

sample was reduced to 85 papers.

3.3 Step 3 - Analysis

The first analysis of papers allowed a preliminary

screening of the sample based on their title and

abstract. The sample was reduced to 73. The second

phase was based on a qualitative full content analysis

to systematically analyse and organize all key

information in terms of AI technologies, algorithms

and methods.

3.4 Step 4 - Reporting

This step focused on presenting the findings. The

most relevant information has been mapped to

provide an answer to the aforementioned RQ. Several

contributions have been identified in terms of AI-ML

algorithms, software and hardware systems, typical

types of defect inspection, flexible VI solutions, VI

architectures. Furthermore, the reporting phase was

structured by organizing the results based on different

industrial fields of application.

Figure 1: SLR protocol.

4 BODY OF LITERATURE:

AI-BASED SOLUTIONS FOR VI

4.1 Electronic Industry

AI-based solutions for VI play a crucial role in

controlling and testing advanced materials within the

electronic industry, with a primary focus on

semiconductors and Printed Circuit Boards (PCBs).

Recent studies highlight various AI algorithms for

defect detection and classification. (Buckermann et

al., 2021) propose a ResNet50-based Convolutional

Neural Network (CNN) for semiconductor defect

classification, utilizing digital image segmentation.

(Beuth et al., 2020; Chu et al., 2022) introduce an

embedded algorithm for clustering high-dimensional

features of Wafer Container Map (WBM) models,

employing multi-objective optimization. (Hou et al.,

2019) use pre-trained DL and CNN models for wafer

quality analysis, employing the GoogLeNet tool.

(Schlosser et al., 2019, 2022) develop a Stacked Deep

Neural Network (SH-DNN) for fault detection,

evaluating performance based on the F1-score. (Chan

et al., 2021; Saqlain et al., 2020) propose a framework

combining ML and human judgment for weld

inspections, recommending real-time implementation

with CNN-Hough and CNN-Wafer Defect

Identification (CNN-WDI) algorithms. (Weiss, 2020)

explores Deep VI for welding tasks, achieving a

classification accuracy of over 97% in real-time.

(Schwebig & Tutsch, 2020) combine DL with an

optical inspection system for enhanced recognition

accuracy of production errors in power packs. (Dorf

et al., 2018) adopt VI algorithm in mechatronics and

micro electro mechanical systems (MEMS).

(Raveendran & Chandrasekhar, 2022) use light image

processing and DL models for defect detection in

semiconductor and polymeric MEMS substrates.

(Koppe & Schatz, 2021) introduce a ML-based VI

process consisting in image acquisition, labeling and

model development, by leveraging ML as a Service

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

658

(MLaaS) platforms. Chen and Shiu (2022) combine

YOLO technology versions with an automatic optical

inspection platform for quality detection, showcasing

YOLOv3 and YOLOv5 with an accuracy rate

exceeding 70%, while other versions like YOLOv2

and YOLOv4 exhibit lower performance.

4.2 Automotive Industry

The automotive industry faces a critical challenge in

ensuring vehicle manufacturing and quality

compliance. Recent literature introduces various CV

systems addressing this issue. (el Wahabi et al., 2020)

develop a three-layer CNN specifically for vehicle

analysis from images. Their system demonstrates a

99% accuracy during the testing phase of product

development and 84% accuracy during algorithm

training, considering multiple parameters. (Zhou et al.,

2019) propose an automatic inspection system for

defect identification, particularly focusing on scratch

classification. The integrated system includes

hardware and software components such as image

acquisition subsystems, processing subsystems, and

LED light sources. (Chouchene et al., 2020) explore

automated CV systems for analyzing and identifying

non-compliant vehicles. The authors emphasize the

importance of considering five influencing factors

(activity, individual, environmental, organizational,

and social) for effective classification techniques in

inspection performance.

4.3 Ceramic Industry

In (Karangwa et al., 2020), a Fast Region

Convolutional Neural Network (R-CNN) is proposed

to automate the detection of ceramic surface defects

on specular materials. The model adopts the VGG16

architecture for the feature extraction and applies the

selective search algorithm to extract the proposed

regions. By aggregating the features into a matrix, a

92.7% of accuracy was achieved on ImageNet during

the testing activities. After that, the model

demonstrated an average accuracy of more than 94%

in detecting, recognizing and identifying six different

types of defects: breaks (physical damages), cracks

(thin and long signs), dirt (small particles of materials

on the surface), spots (discontinuity of the surface),

pits (reliefs of the glazed surface), pinholes (small

holes on the surface). In (Andrei-Alexandru et al.,

2021), a new iteration of YOLOv4 is deployed to

improve the accuracy and speed of detection within a

ceramic manufacturing environment. YOLOv4

modifies the structure of standard YOLO by adjusting

the depth, width, resolution, and network structure to

accommodate scaling. It also uses a CSPDarknet53

backbone and implements a cross-stage pooling

principles for reducing computational costs. Finally,

(Kadar & Onita, 2019) introduces a CNN solution to

enhance a machine vision system, providing

additional information on the types of product, the

production stage and the detected defects. The

system, equipped with a 10 Mpixel black-and-white

camera, uses image pre-processing techniques within

Vision Builder for Automated Inspection software.

4.4 Aerospace Industry

The aerospace industry, characterised by strong

quality standards and considerations for safety and

security (Brandoli et al., 2021), is a critical sector for

VI. (Brandoli et al., 2021) present a Deep Neural

Network (DNN)-based methodology for automatic

detection of aircraft fuselage corrosion using the D-

Sight aircraft inspection system. Leveraging a pre-

trained CNN model, the system achieves 90.2%

accuracy with InceptionV3 and 92.2% with

DenseNet, supporting aircraft maintenance activities.

(Aust et al., 2021) develop a system for automated

defect detection on engine blades, achieving high

accuracy and recall to aid maintenance decisions.

(Meister et al., 2021) propose a novel classification

approach, integrating CNN and support vector

machine (SVM) models to visualize regions of

interest and compare feature maps with geometric

attributes, enhancing VI. (Beltrán-González et al.,

2020) study a ML-based VI system for aerospace

component defect detection, combining CNN with

long/short-term memory networks (LSTMs) for

improved performance, achieving an average

accuracy of 90.7%. (Aust & Pons, 2022) compare

human operators, image processing algorithms, AI

software, and 3D scanning for various inspections

and defects in the aerospace industry. In approaching

Additive Manufacturing (AM) technology, aerospace

industry supports process engineers in data modelling

for quality assurance of AM welding operations,

proposing the use of Random Forest algorithms with

polar transformation and local binary model for

defect classification (Dasari et al., 2020). (Finney et

al., 2019) collaborate with NASA to develop a system

for inspecting metal components additively produced

with multi-material technologies.

4.5 Additive Manufacturing Industry

Due to increasing product complexity, the demand for

mass customization and technological advancements

in production, AM emerges as a pivotal technology

Advanced AI-Based Solutions for Visual Inspection: A Systematic Literature Review

659

for ensuring high-quality products with intricate

geometries. It not only replaces traditional

manufacturing methods but also demonstrates

improved performance in terms of time and costs

(Dilberoglua & Gharehpapagh, 2017). In various

manufacturing domains like biomedical and

aerospace, where VI systems are essential for

ensuring product quality compliance, AM finds

several applications in the literature.

(Sivabalakrishnan et al., 2020) introduce an AM

system based on IoT to facilitate the customization of

customer orders through mobile or web applications.

The system includes a cloud-based platform that

tracks order and product status, integrating a VI

application to ensure correctness and quality through

image detection and processing. (Gobert et al., 2018)

present a defect detection strategy for melting AM

powder beds, employing supervised ML. This

approach enables real-time defect correction during

production processes through the use of a multi-level

VI system. Visual features are extracted using a

Digital Single Lens Reflex (DSLR) camera and

evaluated with a linear support vector machine

(LSVM) algorithm.

4.6 Textile Industry

In the textile industry, (Sandhya et al., 2021) propose

an AI-based system for automatically detecting tissue

defects. Employing various levels of pre-processing,

tissue images are enhanced using CNN processing

techniques. The Deep Convolutional Neural Network

(DCNN) and a pre-trained network (AlexNet) are

used for training and classifying defects such as color,

cut, hole, thread, and metal contamination. The

system demonstrates a 92.6% accuracy in defect

detection. (Voronin et al., 2021) introduce a two-step

approach that combines novel and traditional

algorithms to enhance image and defect detection.

Their CNN-based defect detection method improves

defect localization compared to traditional methods,

with the analysis showing high method efficiency.

Utilizing the carrier machine as a post-processing

technique significantly it reduces the likelihood of

false alarms. (Tayeh et al., 2020) tackle the challenge

of training CNNs for texture analysis to detect

anomalies and defects based on distance analysis.

They adopt the Triplet Network Model (TNM),

allowing the inclusion of three different images

(anchor, positive, negative) in the CNN algorithm,

which is based on the Euclidean distances.

4.7 Metallurgy Industry

Metallurgy, as an applied science focused on

understanding metal structures and properties,

demands advanced technological systems for

performance and quality monitoring (Cottrell, 2019).

(Thalagala & Walgampaya, 2021) propose a VI

system based on AlexNet CNN architecture and

transfer learning to recognize casting surface defects.

The proposed three-step approach involves dataset

definition, image augmentation and nonparametric

classification, using the k-nearest neighbor algorithm.

The system utilizes a pre-trained model feature

extractor and fine-tuning of hyper-parameters.

(Damacharla et al., 2021) introduce the Transfer

Learning-based U-Net (TLU-Net) framework for

steel surface defect detection, exploring ResNet and

DenseNet encoders. Transfer learning outperforms

random initialization in segmentation and defect

classification, with an 81% improvement for ResNet

and 63% for DenseNet. (Fang et al., 2020) conduct a

comprehensive investigation into 2D and 3D surface

defect detection for flat metal products, emphasizing

the critical link between defect classification

accuracy and detection accuracy in automated VI

systems. Similarly, (Luo et al., 2020) focus on defect

detection in flat steel products by adopting ML

approaches with wavelets and contourlets.

4.8 Smart Manufacturing Industry

This section highlights diverse applications of VI

systems within the broad manufacturing landscape.

(Babic et al., 2021) delve into 3D image-based VI

systems in Smart Manufacturing Systems (SMSs) for

product quality assessment. They acknowledge the

benefits of AI but underscore challenges related to

reflective surfaces, high investment costs for image

analysis software and achieving accuracy with

flexible materials like rubber. (Robotyshyn et al.,

2021) use ML and advocate for binary segmentation

models over multiclass models in product quality

inspection. DL techniques are also considered in (Li

et al., 2020) who develop an Optical Inspection

(OPICA) system using a combination of CNN and

rules-based classification models for edge defect

detection on hard disk recording heads. (Tabernik et

al., 2020) present a segmentation-based DL

architecture for surface anomaly detection,

outperforming DeepLab v3+ and UNet. (El Zant et

al., 2021) integrate intelligent VI into Manufacturing

Execution Systems (MES) for real-time monitoring

and anomaly detection. (Jeong et al., 2021) leverage

ML for controlling chipping defects in display

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

660

production, introducing an automatic VI system with

a DNN. They use TensorFlow v1.13 and Keras v2.2,

evaluating accuracy against U-Net, YOLOv2, and

commercial tools. A conditionally coupled generative

network provides synthetic defect images under

various lighting conditions.

5 DISCUSSION & CONCLUSION

The results of the literature analysis show a

comprehensive overview of AI-based VI solutions

based on ML and DL algorithms. Indeed, the study

allowed for providing an answer to the

aforementioned research question, thus identifying

the most important methodological and technological

solutions for supporting industries in VI activities.

The study showed how such AI-VI solutions are

useful independently from the specific industrial

sector. Indeed, the need to ensure high quality

standards of products or machines and to identify

defects is pivotal in any scenario. Also, the literature

contributions highlighted the importance and

effectiveness of integrating ML and DL techniques

for achieving a more accurate classification of

defects, such as of images. The use of supervised

learning algorithm is fundamental for training

systems and obtaining reliable results.

In addition, among the several algorithm

solutions for VI, Convolutional Neural Networks

seem to be the most used technique, often associated

with the use of other models such as Random Forest,

and Support Vector Machine. Also, the use of

different evaluation metrics such as F-score,

accuracy, precision and recall is suggested.

Furthermore, the study highlighted some issues in

the field of available AI-VI solutions. No ready-to-

use solutions emerge and effective solutions need to

be ad-hoc developed or configured. Also, challenges

emerge in terms of data availability and quality, as

well as training datasets, and the need of important

computational resources for CNN and DNN models.

Despite the paper contribute in extending the

literature and proving suggestions to industrial

companies on AI-based solutions for VI, it suffers

from some limitations in terms of qualitative

approach and industrial validation, which leave future

directions for improvement.

ACKNOWLEDGEMENTS

This study comes form AVIS (Advanced Visual

InSpection) project, RIPARTI (assegni di Ricerca per

riPARTire con le imprese) framework funded by the

Puglia Region, Italy - FESRT-FSE 2014/2020.

REFERENCES

American Bureau of Shipping. (2022, dicembre). The use

of remote inspection technologies. https://ww2.eagle.

org/content/dam/eagle/rules-and-guides/current/other/

242-gn-remote-inspection-tech-dec-2022/rit-gn-dec22

.pdf

Andrei-Alexandru, T., Cosmin, C., Bogdan, P., & Adrian-

Alexandru, T. (2021). Automated ceramic plate defect

detection using ScaledYOLOv4-large. 2021 13th

International Conference on Electronics, Computers

and Artificial Intelligence (ECAI), 1–6. https://doi.org/

10.1109/ECAI52376.2021.9515185

Aust, J., & Pons, D. (2022). Comparative Analysis of

Human Operators and Advanced Technologies in the

Visual Inspection of Aero Engine Blades. Applied

Sciences, 12(4), Articolo 4. https://doi.org/10.3390/

app12042250

Aust, J., Shankland, S., Pons, D., Mukundan, R., &

Mitrovic, A. (2021). Automated Defect Detection and

Decision-Support in Gas Turbine Blade Inspection.

Aerospace, 8(2), Articolo 2. https://doi.org/10.3390/

aerospace8020030

Babic, M., Farahani, M. A., & Wuest, T. (2021). Image

Based Quality Inspection in Smart Manufacturing

Systems: A Literature Review. Procedia CIRP, 103,

262–267. https://doi.org/10.1016/j.procir.2021.10.042

Beltrán-González, C., Bustreo, M., & Del Bue, A. (2020).

External and internal quality inspection of aerospace

components. 2020 IEEE 7th International Workshop on

Metrology for AeroSpace (MetroAeroSpace), 351–355.

https://doi.org/10.1109/MetroAeroSpace48742.2020.9

160103

Beuth, F., Schlosser, T., Friedrich, M., & Kowerko, D.

(2020). Improving Automated Visual Fault Detection

by Combining a Biologically Plausible Model of Visual

Attention with Deep Learning. IECON 2020 The 46th

Annual Conference of the IEEE Industrial Electronics

Society, 5323–5330. https://doi.org/10.1109/IECON

43393.2020.9255234

Borangiu, T. (2006). Interactive learning of scene-robot

models based on AI techniques. In IFAC Proceedings

Volumes (Vol. 39, p. 210). https://doi.org/10.3182/

20060517-3-FR-2903.00120

Brandoli, B., de Geus, A. R., Souza, J. R., Spadon, G.,

Soares, A., Rodrigues, J. F., Komorowski, J., &

Matwin, S. (2021). Aircraft Fuselage Corrosion

Detection Using Artificial Intelligence. Sensors,

21(12), Articolo 12. https://doi.org/10.3390/s21124026

Brownlee, J. (2022, aprile 4). Deep Learning for Computer

Vision: Image Classification, Object Detection, and

Face Recognition in Python. https://b.eruditor.link/

file/3706194/

Advanced AI-Based Solutions for Visual Inspection: A Systematic Literature Review

661

Buckermann, F., Klement, N., Beyer, O., Hütten, A., &

Hammer, B. (2021). Automating the optical

identification of abrasive wear on electrical contact

pins. At - Automatisierungstechnik, 69(10), 903–914.

https://doi.org/10.1515/auto-2021-0021

Chan, K. Y., Yiu, K. F. C., Lam, H.-K., & Wong, B. W.

(2021). Ball bonding inspections using a conjoint

framework with machine learning and human

judgement. Applied Soft Computing, 102, 107115.

https://doi.org/10.1016/j.asoc.2021.107115

Chouchene, A., Carvalho, A., Lima, T. M., Charrua-Santos,

F., Osório, G. J., & Barhoumi, W. (2020). Artificial

Intelligence for Product Quality Inspection toward

Smart Industries: Quality Control of Vehicle Non-

Conformities. 2020 9th International Conference on

Industrial Technology and Management (ICITM), 127–

131. https://doi.org/10.1109/ICITM48982.2020.90803

Chu, M., Park, S., Jeong, J., Joo, K., Lee, Y., & Kang, J.

(2022). Recognition of unknown wafer defect via

optimal bin embedding technique. The International

Journal of Advanced Manufacturing Technology, 121,

1–13. https://doi.org/10.1007/s00170-022-09447-y

Corallo, A., Crespino, A. M., Vecchio, V. D., Lazoi, M., &

Marra, M. (2023). Understanding and Defining Dark

Data for the Manufacturing Industry. IEEE

Transactions on Engineering Management, 70(2), 700–

712. https://doi.org/10.1109/TEM.2021.3051981

Cottrell, S. A. (2019). An Introduction to Metallurgy,

Second Edition (2

ed.). CRC Press. https://doi.org/

10.1201/9780429293917

Damacharla, P., M. V., A. R., Ringenberg, J., & Javaid, A.

Y. (2021). TLU-Net: A Deep Learning Approach for

Automatic Steel Surface Defect Detection. 2021

International Conference on Applied Artificial

Intelligence (ICAPAI), 1–6. https://doi.org/10.1109/

ICAPAI49758.2021.9462060

Dasari, S. K., Cheddad, A., & Palmquist, J. (2020). Melt-

Pool Defects Classification for Additive Manufactured

Components in Aerospace Use-Case. 2020 7th

International Conference on Soft Computing &

Machine Intelligence (ISCMI), 249–254.

https://doi.org/10.1109/ISCMI51676.2020.9311555

Dilberoglua, U. M., & Gharehpapagh, B. (2017). The Role

of Additive Manufacturing in the Era of Industry 4.0 |

Elsevier Enhanced Reader. https://doi.org/10.1016/

j.promfg.2017.07.148

Dorf, R. C., Murakami, A., Lyshevski, S. E., & Grigsby, L.

L. (2018). Electromechanical Systems, Electric

Machines, and Applied Mechatronics. CRC Press.

https://doi.org/10.1201/9780203758687

el Wahabi, A., Hadj Baraka, I., Hamdoune, S., & El

Mokhtari, K. (2020). Detection and Control System for

Automotive Products Applications by Artificial Vision

Using Deep Learning (pp. 224–241).

https://doi.org/10.1007/978-3-030-36671-1_20

El Zant, C., Charrier, Q., Benfriha, K., & Le Men, P. (2021).

Enhanced Manufacturing Execution System “MES”

Through a Smart Vision System. In L. Roucoules, M.

Paredes, B. Eynard, P. Morer Camo, & C. Rizzi (A c.

Di), Advances on Mechanics, Design Engineering and

Manufacturing III (pp. 329–334). Springer

International Publishing. https://doi.org/10.1007/978-

3-030-70566-4_52

Fang, X., Luo, Q., Zhou, B., Li, C., & Tian, L. (2020).

Research Progress of Automated Visual Surface Defect

Detection for Industrial Metal Planar Materials. Sensors

(Basel, Switzerland), 20. https://doi.org/10.3390/s2018

5136

Finney, G. A., Persons, C. M., Whitten, J. R., Centamore,

A. A., & Trevithick, C. P. M. (2019). Evaluation of

technologies for autonomous visual inspection of

additive manufacturing (AM). Dimensional Optical

Metrology and Inspection for Practical Applications

VIII, 10991, 172–183. https://doi.org/10.1117/12.252

0279

Gobert, C., Reutzel, E. W., Petrich, J., Nassar, A. R., &

Phoha, S. (2018). Application of supervised machine

learning for defect detection during metallic powder

bed fusion additive manufacturing using high

resolution imaging. Additive Manufacturing, 21, 517–

528. https://doi.org/10.1016/j.addma.2018.04.005

Hou, D., Liu, T., Pan, Y.-T., & Hou, J. (2019). AI on edge

device for laser chip defect detection. 2019 IEEE 9th

Annual Computing and Communication Workshop and

Conference (CCWC), 0247–0251. https://doi.org/10.11

09/CCWC.2019.8666503

IBM. (2023, luglio 6). AI vs. Machine Learning vs. Deep

Learning vs. Neural Networks: What’s the difference?

IBM Blog. https://www.ibm.com/blog/ai-vs-machine-

learning-vs-deep-learning-vs-neural-

networks/www.ibm.com/blog/ai-vs-machine-learning-

vs-deep-learning-vs-neural-networks

Jeong, E.-Y., Kim, J., Jang, W.-H., Lim, H.-C., Noh, H., &

Choi, J.-M. (2021). A more reliable defect detection and

performance improvement method for panel inspection

based on artificial intelligence. Journal of Information

Display, 22(3), 127–136. https://doi.org/10.1080/1598

0316.2021.1876174

Kadar, M., & Onita, D. (2019). A deep CNN for Image

Analytics in Automated Manufacturing Process

Control. 2019 11th International Conference on

Electronics, Computers and Artificial Intelligence

(ECAI), 1–5. https://doi.org/10.1109/ECAI46879.20

19.9042159

Karangwa, J., Kong, L., You, T., & Zheng, J. (2020).

Automated Surface Defects Detection on Mirrorlike

Materials by using Faster R-CNN. 2020 7th

International Conference on Information Science and

Control Engineering (ICISCE), 2288–2294.

https://doi.org/10.1109/ICISCE50968.2020.00341

Kitaguchi, D., Takeshita, N., Hasegawa, H., & Ito, M.

(2022). Artificial intelligence-based computer vision in

surgery: Recent advances and future perspectives.

Annals of Gastroenterological Surgery, 6(1), 29–36.

https://doi.org/10.1002/ags3.12513

Koppe, T., & Schatz, J. (2021). Cloud-based ML

Technologies for Visual Inspection: A Case Study in

Manufacturing. http://hdl.handle.net/10125/70736

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

662

Li, G., Feng, C., Woldesenbet, A., King, B., Hadavi, H.,

Moku, V., Loken, K., & Kunkel, G. (2020). Deep

Learning based Optical Inspection with Centralized

Analysis for High Volume Smart Manufacturing.

Annual Conference of the PHM Society, 12(1), Articolo

1. https://doi.org/10.36001/phmconf.2020.v12i1.1282

Luo, Q., Fang, X., Liu, L., Yang, C., & Sun, Y. (2020).

Automated Visual Defect Detection for Flat Steel

Surface: A Survey. IEEE Transactions on

Instrumentation and Measurement, 69(3), 626–644.

https://doi.org/10.1109/TIM.2019.2963555

Meister, S., Wermes, M., Stüve, J., & Groves, R. M. (2021).

Cross-evaluation of a parallel operating SVM – CNN

classifier for reliable internal decision-making

processes in composite inspection. Journal of

Manufacturing Systems, 60, 620–639. https://doi.org/

10.1016/j.jmsy.2021.07.022

Mishra, D., Gunasekaran, A., Childe, S. J., Papadopoulos,

T., Dubey, R., & Wamba, S. (2016). Vision,

applications and future challenges of Internet of Things:

A bibliometric study of the recent literature. Industrial

Management & Data Systems, 116(7), 1331–1355.

https://doi.org/10.1108/IMDS-11-2015-0478

Mueller, J. P., & Massaron, L. (2021). Machine learning for

dummies (2nd edition). John Wiley & Sons, Inc.

Mujeeb, A., Dai, W., Erdt, M., & Sourin, A. (2019). One

class based feature learning approach for defect

detection using deep autoencoders. Advanced

Engineering Informatics, 42, 100933. https://doi.org/

10.1016/j.aei.2019.100933

Paneru, S., & Jeelani, I. (2021). Computer vision applica-

tions in construction: Current state, opportunities &

challenges. Automation in Construction, 132, 103940.

https://doi.org/10.1016/j.autcon.2021.103940

Piuri, V., Scotti, F., & Roveri, M. (2005). Computational

intelligence in industrial quality control. IEEE

International Workshop on Intelligent Signal

Processing, 2005., 4–9. https://doi.org/10.1109/WISP.

2005.1531623

Raveendran, S., & Chandrasekhar, A. (2022). Inspecting

and classifying physical failures in MEMS substrates

during fabrication using computer vision.

Microelectronic Engineering, 254, 111696.

https://doi.org/10.1016/j.mee.2021.111696

Robotyshyn, M., Sharkadi, M., & Malyar, M. (2021).

Surface Defect Detection Based on Deep Learning

Approach. 13.

Sandhya, N., Sashikumar, N. M., Priyanka, M., Wenisch, S.

M., & Kumarasamy, K. (2021). Automated Fabric

Defect Detection and Classification: A Deep Learning

Approach. Textile & Leather Review, 4, 315–335.

https://doi.org/10.31881/TLR.2021.24

Saqlain, M., Abbas, Q., & Lee, J. Y. (2020). A Deep

Convolutional Neural Network for Wafer Defect

Identification on an Imbalanced Dataset in

Semiconductor Manufacturing Processes. IEEE

Transactions on Semiconductor Manufacturing, 33(3),

436–444. https://doi.org/10.1109/TSM.2020.2994357

Schlosser, T., Beuth, F., Friedrich, M., & Kowerko, D.

(2019). A Novel Visual Fault Detection and

Classification System for Semiconductor

Manufacturing Using Stacked Hybrid Convolutional

Neural Networks. 2019 24th IEEE International

Conference on Emerging Technologies and Factory

Automation (ETFA), 1511–1514. https://doi.org/10.11

09/ETFA.2019.8869311

Schlosser, T., Friedrich, M., Beuth, F., & Kowerko, D.

(2022). Improving automated visual fault inspection for

semiconductor manufacturing using a hybrid multistage

system of deep neural networks. Journal of Intelligent

Manufacturing, 33(4), 1099–1123. https://doi.org/

10.1007/s10845-021-01906-9

Schwebig, A. I. M., & Tutsch, R. (2020). Intelligent fault

detection of electrical assemblies using hierarchical

convolutional networks for supporting automatic

optical inspection systems. Journal of Sensors and

Sensor Systems, 9(2), 363–374. https://doi.org/10.51

94/jsss-9-363-2020

Sivabalakrishnan, R., Kalaiarasan, A., Ajithvishva, M. S.,

Hemsri, M., Oorappan, G. M., & Yasodharan, R.

(2020). IoT visualization of Smart Factory for Additive

Manufacturing System (ISFAMS) with visual

inspection and material handling processes. IOP

Conference Series: Materials Science and Engineering,

995(1), 012027. https://doi.org/10.1088/1757-

899X/995/1/012027

Tabernik, D., Šela, S., Skvarč, J., & Skočaj, D. (2020).

Segmentation-based deep-learning approach for

surface-defect detection. Journal of Intelligent

Manufacturing, 31. https://doi.org/10.1007/s10845-

019-01476-x

Tayeh, T., Aburakhia, S., Myers, R., & Shami, A. (2020).

Distance-Based Anomaly Detection for Industrial

Surfaces Using Triplet Networks. 2020 11th IEEE

Annual Information Technology, Electronics and

Mobile Communication Conference (IEMCON), 0372–

0377. https://doi.org/10.1109/IEMCON51383.2020.9

284921

Thalagala, S., & Walgampaya, C. (2021). Application of

AlexNet convolutional neural network architecture-

based transfer learning for automated recognition of

casting surface defects. 2021 International Research

Conference on Smart Computing and Systems

Engineering (SCSE), 4, 129–136. https://doi.org/

10.1109/SCSE53661.2021.9568315

Voronin, V. V., Sizyakin, R., Zhdanova, M.,

Semenishchev, E. A., Bezuglov, D., & Zelemskii, A. A.

(2021). Automated visual inspection of fabric image

using deep learning approach for defect detection.

Automated Visual Inspection and Machine Vision IV,

11787. https://doi.org/10.1117/12.2592872

Voulodimos, A., Doulamis, N., Doulamis, A., &

Protopapadakis, E. (2018). Deep Learning for

Computer Vision: A Brief Review. Computational

Intelligence and Neuroscience, 2018, e7068349.

https://doi.org/10.1155/2018/7068349

Weiss, E. (2020). Electronic component solderability

assessment algorithm by deep external visual

inspection. 2020 IEEE Physical Assurance and

Advanced AI-Based Solutions for Visual Inspection: A Systematic Literature Review

663

Inspection of Electronics (PAINE), 1–6.

https://doi.org/10.1109/PAINE49178.2020.9337565

Xiao, Y., & Watson, M. (2019). Guidance on conducting a

systematic literature review. Journal of planning

education and research, 39(1), 93–112.

Yang, J., Xu, R., Qi, Z., & Shi, Y. (2021). Visual Anomaly

Detection for Images: A Survey (arXiv:2109.13157).

arXiv. https://doi.org/10.48550/arXiv.2109.13157

Zaidi, S. S. A., Ansari, M. S., Aslam, A., Kanwal, N.,

Asghar, M., & Lee, B. (2021). A Survey of Modern

Deep Learning based Object Detection Models

(arXiv:2104.11892). arXiv. https://doi.org/10.48550/

arXiv.2104.11892

Zhou, Q., Chen, R., Huang, B., Liu, C., Yu, J., & Yu, X.

(2019). An Automatic Surface Defect Inspection

System for Automobiles Using Machine Vision

Methods. Sensors, 19(3), Articolo 3. https://doi.org/

10.3390/s19030644

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

664