طبقه‌بندی عیوب سطح فولاد با مدلی بهبودیافته از معماری VGG16

نوع مقاله : مقاله پژوهشی

نویسندگان

1 دانشجوی دکتری مهندسی کامپیوتر، دانشگاه آزاد اسلامی، یزد، ایران

2 استادیار دانشکده مهندسی کامپیوتر، دانشگاه آزاد اسلامی، یزد، ایران

10.22034/abmir.2025.23563.1162

چکیده

ورق‌های فولادی به‌عنوان یکی از اجزای کلیدی در صنایع خودروسازی مورد استفاده قرار می‌گیرند. طبقه‌بندی دقیق عیوب سطحی این ورق‌ها، نقشی اساسی در تضمین کیفیت آن‌ها دارد. چالش اصلی در این زمینه، شیوه بازرسی ورق‌های فولادی است زیرا، روش‌های سنتی بازرسی با محدودیت‌های قابل‌توجه ناشی از خطای انسانی مواجه‌اند. هرچند در سال‌های اخیر، الگوریتم‌های یادگیری عمیق به‌عنوان رویکردی مؤثر و خودکار به‌منظور شناسایی عیوب مطرح شده‌اند اما طبقه‌بندی عیوب کوچک و پیچیده، همچنان دشوار باقی‌مانده و نیازمند مدل‌سازی بهینه اطلاعات مکانی و کانالی است. در این پژوهش، مدلی نوین مبتنی بر معماری VGG16 و یادگیری انتقالی ارائه می‌شود که با بهره‌گیری از ماژول توجه فضایی دوبُعدی و مکانیزم بزرگ‌نمایی مبتنی بر توجه بر کاهش نویز پس‌زمینه و تمرکز بر نواحی کلیدی تصویر تمرکز دارد. پس از ارزیابی مدل بر روی دو مجموعه داده NEU-CLS و NEU-DET به‌ترتیب دقت 98/99 و 92/99 درصد به‌دست آمد که نسبت به مدل پایه VGG16 (حداقل 4 درصد) و نسبت به پژوهش‌های پیشین، بهبود (حداقل 018/0 درصد) داشته است. این نتایج، کارایی بالای مدل را در طبقه‌بندی دقیق عیوب سطح فولاد نشان می‌دهد. در پایان، جهت بهره‌برداری عملی، یک وب اپلیکیشن توسعه‌یافته و در اختیار متخصصان قرارگرفته است.

کلیدواژه‌ها

موضوعات

عنوان مقاله [English]

Steel Surface Defect Classification Using an Improved VGG16-Based Model

نویسندگان [English]

  • Touba Torabipour 1
  • Abolfazl Gandomi 2
1 PhD Candidate, Department of Computer Engineering, Islamic Azad University, Yazd, Iran
2 Assistant Professor, Department of Computer Engineering, Islamic Azad University, Yazd, Iran
چکیده [English]

Steel sheets are among the key components used in the automotive industry. Accurate classification of surface defects in these sheets plays a vital role in ensuring product quality. The main challenge in this field lies in the inspection process, as traditional manual inspection methods suffer from significant limitations caused by human error. Although deep learning algorithms have recently emerged as effective and automated approaches for defect detection, the classification of small and complex defects remains challenging and requires optimized modeling of spatial and channel information.In this study, a novel model based on the VGG16 architecture and transfer learning is proposed. The model integrates a two-dimensional spatial attention module and an attention-based magnification mechanism to suppress background noise and focus on the key regions of the image. After evaluation on two benchmark datasets, NEU-CLS and NEU-DET, the proposed model achieved classification accuracies of 99.98% and 99.92%, respectively—showing an improvement of at least 4% over the baseline VGG16 model and approximately 0.018% compared with previous studies. These results demonstrate the superior performance of the proposed approach in accurately classifying steel surface defects. Finally, for practical deployment, a web application was developed and made available to industry experts.

کلیدواژه‌ها [English]

  • Deep Learning
  • Steel Surface Defect Classification
  • Improved VGG16 Architecture
  • Attention Mechanism

مراجع

[1]     K. Gong, D. Xu, F. Guo, Z. Wang, F. Zhang, and C. He, "Few-shot steel strip surface defect classification via self-correlation enhancement and feature refinement," IEEE Transactions on Instrumentation and Measurement, 2025.
[2]     J. Huang, X. Zhang, L. Jia, and Y. Zhou, "An improved you only look once model for the multi-scale steel surface defect detection with multi-level alignment and cross-layer redistribution features," Engineering Applications of Artificial Intelligence, vol. 145, p. 110214, 2025.
[3]     H. Chen, et al., "DCAM-Net: A rapid detection network for strip steel surface defects based on deformable convolution and attention mechanism," IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1-12, 2023.
[4]     S. Ren, K. He, R. Girshick, and J. Sun, "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137-1149, Jun. 2017.
[5]     Y. He, K. Song, Q. Meng, and Y. Yan, "An End-to-End Steel Surface Defect Detection Approach via Fusing Multiple Hierarchical Features," IEEE Transactions on Instrumentation and Measurement, vol. 69, no. 4, pp. 1493-1504, Apr. 2020.
[6]     X. Cheng and J. Yu, "RetinaNet with Difference Channel Attention and Adaptively Spatial Feature Fusion for Steel Surface Defect Detection," IEEE Transactions on Instrumentation and Measurement, vol. 70, pp. 1-11, 2021.
[7]     R. Wei, Y. Song, and Y. Zhang, "Enhanced Faster Region Convolutional Neural Networks for Steel Surface Defect Detection," ISIJ International, vol. 60, no. 3, pp. 539-545, 2020.
[8]     M. Tang, Y. He, J. Liu, K. Song, and Y. Yan, "A strip steel surface defect detection method based on attention mechanism and multi-scale maxpooling," Measurement Science and Technology, vol. 32, no. 11, p. 115401, 2021.
[9]     Z. Guo, C. Wang, Y. Yang, Z. Wang, and F. Li, "Msft-yolo: Improved yolov5 based on transformer for detecting defects of steel surface," Sensors, vol. 22, no. 9, p. 3467, 2022.
[10] Z. Li, X. Wei, M. Hassaballah, Y. Li, and X. Jiang, "A deep learning model for steel surface defect detection," Complex & Intelligent Systems, vol. 10, no. 1, pp. 885-897, 2024.
[11] Y. Gao, G. Lv, D. Xiao, X. Han, T. Sun, and Z. Li, "Research on steel surface defect classification method based on deep learning," Scientific Reports, vol. 14, no. 1, p. 8254, 2024.
[12] X. Zheng, W. Liu, and Y. Huang, "A novel feature extraction method based on Legendre multi-wavelet transform and auto-encoder for steel surface defect classification," IEEE Access, vol. 12, pp. 5092-5102, 2024.
[13] F. Wang, X. Jiang, Y. Han, and L. Wu, "YOLO-LSDI: An Enhanced Algorithm for Steel Surface Defect Detection Using a YOLOv11 Network," Electronics, vol. 14, no. 13, p. 2576, 2025.
[14] X. Li, C. Xu, J. Li, X. Zhou, and Y. Li, "Multi-scale sensing and multi-dimensional feature enhancement for surface defect detection of hot-rolled steel strip," Nondestructive Testing and Evaluation, vol. 40, no. 8, pp. 3669–3692, 2025.
[15] X. Zheng, W. Liu, and Y. Huang, "Legendre multiwavelet-based feature attention guidance lightweight network for accurate steel surface defect classification," Engineering Applications of Artificial Intelligence, vol. 161, p. 112179, 2025.
[16] Courtois, J.-M. Morel, and P. Arias, "Investigating Neural Architectures by Synthetic Dataset Design," in Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 4886-4895, 2022.
[17] K. Dikshit, "NEU Surface Defect Database," Kaggle, available: https://www.kaggle.com/datasets/kaustubhdikshit/neu-surface-defect-database/data, accessed Sep. 28, 2025.
[18] K. C. Song and Y. Yan, "A noise-robust method based on completed local binary patterns for hot-rolled steel strip surface defects," Applied Surface Science, vol. 285, pp. 858-864, 2013.
[19] T. TorabiPour, "Foulad Project," GitHub, available: https://github.com/torabi225/foulad/tree/main, accessed Sep. 28, 2025.
[20] T. Torabipour, A. Gandomi, and M. Ghanimi, "A Two-Stage Method for Diagnosing COVID-19, Leveraging CNN, and Transfer Learning on CT Scan Images," International Journal of Web Research, vol. 6, no. 2, pp. 133-142, 2023.
[21] T. TorabiPour, “Steel Surface Defect Detection App,” Streamlit, available: https://foulad77mappmawvcndm87gpzwhca.streamlit.app/, accessed Sep. 28, 2025.
پژوهش های نظری و کاربردی هوش ماشینی
دوره 3، شماره 2
پاییز و زمستان 1404
آذر 1404
صفحه 73-88

فایل ها

هم رسانی

ارجاع به این مقاله

آمار