锻压技术

基于E-YOLO的冷轧板材表面缺陷检测方法

英文标题：Surface defect detection method on cold rolled sheet metal based on E-YOLO

单位：1. 冀南技师学院装备制造工程系 2.河北工程大学机械与装备学院 3.河北省智能工业装备技术重点实验室 4.河北省冀南新区现代装备制造协同创新中心

分类号：TH113.1；TG335.5

出版年,卷(期):页码：2025,50(2):125-131

摘要：

在实际的制造系统中，冷轧板材表面缺陷会导致材料强度下降造成安全风险，而传统检测方法的准确率与稳定性难以保证。因此，提出一种轻量且高效的缺陷检测模型E-YOLO（Efficient-YOLO）。针对YOLOv8模型在小目标检测、特征提取效率及模型推理速度上的不足，对其结构进行了创新优化，修改了原始模型的低效颈部连接结构，引入多分支特征融合机制，并创新性地采用了特征重提取结构来增强模型对细微缺陷特征的感知能力。最后，实验表明，相较于YOLOv8，E-YOLO的检测准确率提高了7.3%；相较于大型模型Faster RCNN，其检测速度提升约18倍，为冷轧板材表面缺陷的高效、准确检测提供了一种可行的方法。

In practical manufacturing systems, surface defects on the cold rolled sheet metal can lead to a decrease in material strength and pose a safety risk, while the accuracy and stability of traditional detection methods are difficult to guarantee. Therefore, a lightweight and efficient defect detection model E-YOLO (Efficient-YOLO) was proposed, and in response to the shortcomings of the YOLOv8 model in small target detection, feature extraction efficiency and model inference speed, the structural innovations and optimizations about YOLOv8 model were conducted. The inefficient neck connection structure of the original model was modified, the perception ability of the model for subtle defect features was enhanced by introducing a multi-branch feature fusion mechanism and innovatively adopting a feature re-extraction structure. Finally，experiments show that compared with YOLOv8, E-YOLO improves the detection accuracy by 7.3% and increases the detection speed by approximately 18 times compared with the larger model Faster RCNN, providing a feasible way for efficient and accurate detection of surface defects on cold rolled sheet metal.

基金项目：

河北省专精特新小巨人企业科技特派团资助项目（SJ240140123）；河北工程大学创新基金项目（SJ2401002049）

作者简介：陈栋(1991-)，男，硕士，讲师,E-mail：2313030095@qq.com;通信作者：刘欣宜(1993-)，女，博士，讲师,E-mail：liuxy93@126.com

参考文献：

[1]Wang Y C, Gao L, Gao Y P, et al. A new graph-based semi-supervised method for surface defect classification[J]. Robotics and Computer-Integrated Manufacturing, 2021, 68: 102083.

[2]Kim H M, Rho Y W, Yoo H R, et al. A study on the measurement of axial cracks in the magnetic flux leakage NDT system[A]. Proceedings of the 2012 IEEE International Conference on Automation Science and Engineering (CASE 2012)[C]. Seoul, Korea：IEEE, 2012.

[3]Jeon Y J, Choi D C, Lee S J, et al. Defect detection for corner cracks in steel billets using a wavelet reconstruction method[J]. Journal of the Optical Society of America A, 2014, 31(2): 227-237.

[4]Zhang B Y, Yang H, Yin Z P. A region-based normalized cross correlation algorithm for the vision-based positioning of elongated IC chips[J]. IEEE Transactions on Semiconductor Manufacturing, 2015, 28(3): 345-352.

[5]豆远航,崔学习,钟馨平,等.基于深度学习的非接触应变测量网格图像修复技术[J].锻压技术,2024,49(8):195-204.

Dou Y H, Cui X X, Zhong X P, et al. Grid image restoration technology for non-contact strain measurement based on deep learning [J]. Forging & Stamping Technology,2024,49(8):195-204.

[6]Ren S Q, He K M, Girshick R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 39(6): 1137-1149.

[7]Redmon J, Divvala S, Girshick R, et al. You Only Look Once: Unified, real-time object detection[A]. Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR)[C].Las Vegas Nevada，USA：IEEE，2016.

[8]He Y, Song K C, Meng Q G, et al. An end-to-end steel surface defect detection approach via fusing multiple hierarchical features[J]. IEEE Transactions on Instrumentation and Measurement, 2019, 69(4): 1493-1504.

[9]Du W Z, Shen H Y, Fu J Z, et al. Approaches for improvement of the X-ray image defect detection of automobile casting aluminum parts based on deep learning[J]. NDT & E International，2019，107：102144.

[10]Chen W, Gao Y P, Gao L, et al. A new ensemble approach based on deep convolutional neural networks for steel surface defect classification[J]. Procedia CIRP, 2018, 72:1069-1072.

[11]刘艳菊,王秋霁,张惠玉,等.基于改进SSD的工件表面缺陷检测[J].热加工工艺,2024,53(2):134-139.

Liu Y J, Wang Q J, Zhang H Y, et al. Defect detection of workpiece surface based on improved SSD [J]. Hot Working Technology, 2024, 53 (2): 134-139.

[12]Tan M X, Pang R M, Le Q V. EfficientDet：Scalable and efficient object detection[A]. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition[C]. Seattle，WA，USA：IEEE，2020.

[13]Wang P Q, Chen P F, Yuan Y, et al. Understanding convolution for semantic segmentation[A]. Proceedings of the 2018 IEEE Winter Conference on Applications of Computer Vision (WACV)[C]. Lake Tahoe，NV，USA：IEEE，2018.

[14]Hou Q B, Zhou D Q, Feng J S. Coordinate attention for efficient mobile network design[A]. Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition[C]. Nashville, TN, USA：IEEE，2021.

[15]Han K, Wang Y H, Tian Q, et al. GhostNet: More features from cheap operations[A]. Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition[C]. Seattle, WA, USA：IEEE，2020.

[16]Lin T Y, Maire M, Belongie S, et al. Microsoft COCO: Common objects in context[A].Computer Vision-ECCV 2014[C].

Berlin,Germany:Springer International Publishing, 2014.

[17]Sun P, Zhang R, Jiang Y, et al. Sparse R-CNN: End-to-end object detection with learnable proposals[A]. Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition[C]. Nashville, TN, USA：IEEE，2021.

服务与反馈：

【文章下载】【加入收藏】