Volume 9, Issue 6, June – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24JUN794

Automated Product Defect Detection Using Image

Processing Techniques for Effective Sorting and
Quality Assurance : A Survey
Dr. Kavitha K S Mamatha C G
Computer Science and Engineering Computer Science and Engineering
Dayananda Sagar College of Engineering Dayananda Sagar College of Engineering
Bengaluru, India Bengaluru, India

Abstract:- Ensuring product quality and minimizing By integrating this automated defect detection system,
defects is crucial in today's manufacturing industry. manufacturers can significantly enhance efficiency, ensure
Traditional manual inspections are labor-intensive and consistency, and improve overall product quality.
error prone.This paper describes a system designed to Experimental results validate the robustness and
identify defects automatically the YOLOv5 algorithm, effectiveness of the proposed system, showcasing the
known for its accuracy and speed. High-resolution images superior performance of the YOLOv5 model in both
of products are processed with YOLOv5 to identify detection accuracy and processing speed. This research
defects like scratches, dents, and deformations. This underscores the transformative potential of integrating image
system enhances sorting and quality assurance, processing with machine learning to revolutionize quality
improving efficiency and consistency. Experimental control processes across various manufacturing sectors.
results show YOLOv5 superior performance in detection
accuracy and speed compared to traditional methods, II. RELATED WORKS
exploring the feasibility of combining machine learning
and image processing within manufacturing. M. M. Elsayed, et al. [1] Presented a technique
leveraging convolutional neural networks (CNNs) for real-
Keywords:- Automated Defect Detection, Image Processing, time defect identification in manufacturing. This research
Yolov5, Quality Assurance, Manufacturing, Object Detection, concentrated on utilizing CNNs to detect surface flaws on
Machine Learning. metal components, achieving notable accuracy and real-time
capability.
I. INTRODUCTION
J. Zhang, et al. [2] Created a hybrid method that merges
In today's competitive manufacturing landscape, machine-learning with traditional image processing. The
ensuring the highest product quality while minimizing study applied feature extraction techniques, including edge
defects is a critical priority. Traditional manual inspection detection and texture analysis, alongside a support vector
methods, which rely on human effort, are often labor- machine (SVM) to identify defects in textiles.
intensive, time-consuming, and susceptible to errors. This
creates a need for automated systems that can provide K. Lee and H. Park [3] Developed a deep learning
consistent, accurate, and efficient quality assurance. This framework to automate the examination of electronic
paper introduces an advanced system for automated product components. This system made use of deep learning methods
defect detection, utilizing sophisticated image processing to identify and classify defects in solder joints, significantly
techniques. Central to this system is the YOLOv5 (You Only cutting down inspection time compared to manual processes.
Look Once version 5) algorithm, a cutting-edge object
detection model recognized for its exceptional speed and Kumar and G. Kaur [4] Proposed a novel image
precision. The versatility and reliability of the YOLOv5 processing algorithm for ceramic tile defect detection. The
algorithm make it particularly well-suited for real-time algorithm included noise reduction and contrast enhancement
applications in quality assurance. Our approach involves in preprocessing, followed by morphological operations to
capturing high-resolution images of products such as PVC detect cracks and surface anomalies.
pipes and fabrics as they progress along the production line.
These images are then analyzed by the YOLOv5 algorithm to S. Patel, et al. [5] Employed image segmentation to
detect and classify defects with high accuracy. The types of automate defect detection in fabrics. The research utilized
defects identified include scratches, dents, and deformities, watershed segmentation and contour detection to locate and
among others, enabling efficient sorting and quality classify defects such as holes and stains in fabric.
assurance.

IJISRT24JUN794 www.ijisrt.com 1302

Volume 9, Issue 6, June – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24JUN794

Y. Liu, et al. [6] Suggested a multi-scale image analysis K. Tsai, et al. [16] Developed an image-based
technique for automotive part defect detection. This approach inspection system for ceramic products. This system used
used multi-scale image representation and wavelet transform image preprocessing techniques such as histogram
to detect small flaws on the surface that are challenging to equalization and edge detection to identify defects like cracks
identify with standard methods. and surface pits.
N. Gupta, et al. [7] Applied deep convolutional neural
networks (DCNNs) for inspecting food items. The study P. Verma, et al. [17] Employed a combination of
focused on identifying defects in fruits and vegetables using machine vision and deep learning for defect identification in
DCNNs to classify imperfections like bruises and blemishes. pharmaceutical tablets. The study used CNNs to detect
surface defects and irregularities in pharmaceutical tablets,
P. Wong, et al. [8] Developed an automated system for ensuring high accuracy and reliability.
PCB defect identifying using image processing. The system
employed shape recognition and template matching to detect W. Wang, et al. [18] Proposed a defect detection
missing components and misalignments in printed circuit method for electronic components using image processing
boards. and machine learning. This approach combines feature
extraction techniques like texture analysis with SVM
R. Sharma, et al. [9] Introduced a machine vision classifiers to identify defects in electronic components.
system for real-time defect detection in glass manufacturing.
The system used edge detection algorithms and the Hough Y. Huang, et al. [19] Suggested an automated defect
transform to identify cracks and inclusions in glass products. identification system for packaging materials using image
processing. The system employed contour detection and
L. Chen and M. Chen [10] Used a integration of image morphological operations to detect defects such as tears and
processing and neural networks for defect identification in holes in packaging materials.
rubber products. The study integrated texture analysis and
neural network classifiers to detect surface defects such as F. Lin, et al. [20] Developed a deep learning-based
blisters and grooves in rubber sheets. approach for detecting defects in steel products. The approach
used a deep residual network to identify surface defects like
T. Tanaka, et al. [11] Suggested a approach utilizing scratches and dents in steel sheets with high precision.
machine learning and image processing for semiconductor
wafer defect detection. The Principal component analysis III. OUTCOME OF THE SURVEY
(PCA) was used in research for feature reduction and SVM
for classifying defects like scratches and pits on wafer This study acknowledges certain limitations, notably the
surfaces. absence of labeled defect datasets for training machine
learning models. Future research should prioritize techniques
H. Kim and J. Choi [12] Developed an automated for dataset labeling and explore semi-supervised or
system for inspecting automotive paint quality. The system unsupervised learning approaches to enhance efficiency.
used color flaws such as runs, sags, and foreign particles can Additionally, integrating sensor data such as temperature or
be identified using gradient-based approaches and colour vibration could augment defect detection capabilities. Cost-
image processing. effectiveness analysis and optimizing hardware and software
elements are necessary for actual use in practice and
M. Singh, et al. [13] Proposed a new approach for widespread adoption in manufacturing. In conclusion, this
detecting defects in composite materials using image research presents a comprehensive examination of quality
processing. The method utilized thermal imaging and edge control enhancement in Industry through advanced image
detection to identify delamination and voids in composite processing for automated flaws detection. The proposed
structures. system shows promise in revolutionizing quality control
processes, enhancing product quality, and boosting
C. Zhang, et al. [14] Applied a deep learning approach operational efficiency. These findings contribute to the
for detecting defects in solar panels. This method used a knowledge base, laying the groundwork for further
CNN model trained on infrared images to detect micro- advancements in automated defect detection and quality
cracks and hot spots in solar panels. control within the Industry framework. To guarantee the
efficiency of the automated defect detection system,
J. Liu and Y. Wang [15] Created a live flaw robustness and efficiency assessments are imperative.
identification system for textile manufacturing using image Robustness analysis involves testing the system's
processing. The system applied Gabor filters and neural performance under various challenging conditions, including
networks to detect and classify defects like yarn breakages changes in lighting, variations within categories of defects or
and weave irregularities. sizes, and complex backgrounds. Evaluating the system's
real-time defect detection capabilities and monitoring its
performance against operational requirements are essential
aspects of efficiency analysis.

IJISRT24JUN794 www.ijisrt.com 1303

Volume 9, Issue 6, June – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24JUN794

IV. CONCLUSION [9]. Cao, Q., Zanni-Merk, C., Samet, A., Reich, C.,
Beuvron, F. de B. de, Beckmann, A., &Giannetti, C.
In summary, the deployment of automated defect (2022). KSPMI: A Knowledge-based System for
detection in product inspection, utilizing image processing Predictive Maintenance in Industry 4.0. Robotics and
methodologies like the YOLOv5 algorithm, offers a robust ComputerIntegratedManufacturing,74,102281.https://do
strategy for effective sorting and quality assurance across i.org/10.1016/j.rcim.2021.102281
diverse product categories including PVC, PCB, and fabric. [10]. Chen, B., Wan, J., Shu, L., Li, P., Mukherjee, M., &
Through leveraging the capabilities inherent in YOLOv5, this Yin, B. (2018). Smart Factory of Industry 4.0: Key
system facilitates swift and precise defect identification, Technologies, Application Case, and Challenges. IEEE
thereby enhancing overall product quality and reliability. The Access,6,6505–
comprehensive and rapid image analysis enabled by 6519.https://doi.org/10.1109/ACCESS.2017.2783682
YOLOv5 empowers industries to optimize their quality [11]. Cho, S.-H., Lee, K.-T., Kim, S.-H., & Kim, J.-H.
control procedures, mitigate production delays, and (2019). Image Processing for Sustainable Remodeling:
ultimately deliver high-caliber products to end-users. This Introduction to Realtime Quality Inspection System of
innovative approach marks a significant advancement in External Wall Insulation Works. Sustainability, 11(4),
automated defect detection, fostering improvements in 1081. https://doi.org/10.3390/su11041081
manufacturing efficiency and elevating standards of product [12]. Czimmermann, T., Ciuti, G., Milazzo, M., Chiurazzi,
excellence. M., Roccella, S., Oddo, C. M., & Dario, P. (2020).
Visual-Based Defect Detection and Classification
REFERENCES Approaches for Industrial Applications—A SURVEY.
Sensors,20(5),1459.https://doi.org/10.3390/s20051459
[1]. Aceto, G., Persico, V., & Pescape, A. (2020). Industry [13]. Dastres, R., & Soori, M. (2021). Advanced Image
4.0 and Health: Internet of Things, Big Data, and Cloud Processing Systems Article in. In International Journal
Computing for Healthcare 4.0. Journal of Industrial of Imaging and Robotics. www.ceserp.com/cp-jour
Information Integration, 18, 100129. [14]. De Mitri, O., Stahl, J., Jauch, C., & Distante, C. (2019).
https://doi.org/10.1016/j.jii.2020.100129 Image acquisition, evaluation and segmentation of
[2]. Adaway, B. (1983). Industrial applications of image thermal cutting edges using a mobile device. In S.
processing. Computer Graphics 83, (Online Negahdaripour, E. Stella, D. Ceglarek, & C. Möller
Publications, Pinner), 555–568. (Eds.), Multimodal Sensing: Technologies and
https://doi.org/10.2478/aucts-2014-0004 Applications (p. 24). SPIE.
[3]. Antony, J., McDermott, O., & Sony, M. (2022). Quality https://doi.org/10.1117/12.2527993
4.0 conceptualisation and theoretical understanding: a [15]. Edwards, C. (2015). Challenges Faced by Deep
global exploratory qualitative study. The TQM Journal, Learning as It Advances. Communications of the ACM,
34(5), 1169– 1188. https://doi.org/10.1108/TQM-07- 58(7),14–16.
2021-0215 [16]. Fang, X., Luo, Q., Zhou, B., Li, C., & Tian, L. (2020).
[4]. J K Arora (2011) “Artificial Neural Network modelling Advancements in Automated Visual Detection of
for the System of blood flow through tapered artery Surface Imperfections for Industrial Metal Planar
with mild stenosis ,International Journal of Mathematics Materials. Sensors (Switzerland), 20(18), 1–35.
Trends and Technology July to Aug 2011, ISSN: [17]. Forstner, W. (2000). Image Processing Techniques for
22315373,http://www.internationaljournalssrg.org Extracting Features from Digital Intensity, Color, and
[5]. Awcock, G. J., & Thomas, R. (1995). Applied Image Range Images. In Geomatic Method for the Analysis of
Processing. Macmillan Education UK. Data in the Earth Sciences (pp. 165–189). Springer
https://doi.org/10.1007/978-1-349-13049-8 Berlin Heidelberg.
[6]. Azamfirei, V., Psarommatis & Lagrosen, Y. (2023). [18]. Godina, R., & Matias, J. C. O. (2019). Ensuring Quality
Application of automation for in-line quality inspection, in the Era of Industry 4.0. In Industrial Engineering and
a zero-defect manufacturing approach. In Journal of Operations Management II (pp. 177–187).
Manufacturing Systems (Vol. 67, pp. 1–22). [19]. Dhanasekar,J., Sharan, A. P., Natarajan, M. A.,
https://doi.org/10.1016/j.jmsy.2022.12.010 Nizamudeen, A., Krishnan, H. M., & Kumar, S. R. S.
[7]. Bhattacharya, A., & Cloutier, S. G. (2022). End-to-end (2022). Creation and Application of a System for
deep learning framework for printed circuit board Detecting Imperfections Using Techniques in Image
manufacturing defect classification. Scientific Reports, Processing (pp. 734–751).
12(1), 12559. https://doi.org/10.1038/s41598-022- [20]. Dogo, E. M., Salami, A. F., Aigbavboa, C. O., &
16302-3 Nkonyana, T. (2019). Expanding Cloud Computing to
[8]. Bulnes, F. G., Usamentiaga, R., Garcia, D. F., the Extreme Edge: An Evaluation of Mist Computing
&Molleda, J. (2016). An efficient approach for defect for Advancements in Smart Cities and Industry Transl.
identification during the manufacturing of web J. Magn. Japan, vol. 2, pp. 740-741, August 1987
materials. Journal of Intelligent [Digests 9th Annual Conf. Magnetics Japan, p. 301,
Manufacturing,27(2),431–445. 1982].
https://doi.org/10.1007/s10845-014-0876

IJISRT24JUN794 www.ijisrt.com 1304