钢材表面缺陷轻量化识别算法研究

doi:10.3901/JME.2025.08.018

摘要/Abstract

摘要： 针对传统钢材表面缺陷检测方法存在检测效率低、检测精度差、实时性差等问题，提出一种基于改进YOLOv5的轻量化钢材表面缺陷识别算法，适用于低成本、低算力、低内存的硬件设备的工程部署。首先，通过K-means++算法对NEU-DET数据集重新聚类生成自适应锚框，优化先验框和真实框之间的匹配度；然后通过改进激活函数和损失函数，HardSwish激活函数降低计算成本的同时提高了稳定性，SIoU损失函数可以有效地加快了网络收敛速度；其次，为了提取目标区域的丰富的细节信息，在原始YOLOv5算法基础上添加CA坐标注意力模块；再借鉴结构重参化引入改进的Repvgg block，同时将模型通道数减半，进一步增加模型的工程部署性；最后，通过消融试验和一系列对比试验，证明本算法性能上的优越性，较YOLOv5原算法，参数量减少约73.6%，浮点计算量减少约72.3%，同时mAP值提升1.5%。研究结果为钢材表面缺陷精细化检测提供了新的方法及思路，对提高钢材产品质量具有实际意义。

关键词: 钢材表面缺陷, 深度学习, 轻量化, YOLOv5, 卷积神经网络

Abstract: Aiming at the problems of low detection efficiency, poor detection accuracy and poor real-time performance of traditional steel surface defect detection methods, An algorithm for surface defect identification of lightweight steel based on improved YOLOv5 is proposed, which is suitable for the engineering deployment of hardware equipment with low cost, low computational power and low memory. Firstly, K-means++ algorithm is used to cluster anchor boxes in the NEU-DET dataset to optimize the matching degree between the prior box and the ground-truth box. Then, by improving the activation function and loss function, HardSwish activation function reduces the computational cost and improves the stability at the same time, while SIoU loss function can effectively accelerate the convergence speed of the network. Secondly, in order to extract rich details of the target region, the CA coordinate attention module is added on the basis of the original YOLOv5 algorithm. Furthermore, the improved Repvgg block is introduced by referring to the structure reparameterization, and the number of model channels is halved to further increase the engineering deployability of the model. Finally, through ablation experiment and a series of comparative experiments, the performance of the proposed algorithm is proved to be superior. Compared with the original YOLOv5 algorithm, the number of parameters is reduced by about 73.6%, the floating point calculation is reduced by about 72.3%, and the mAP value is increased by 1.5%. The results provide a new method and idea for the fine inspection of steel surface defects and have practical significance for improving the quality of steel products.

Key words: steel surface defect, deep learning, light weight, YOLOv5, convolutional neural network

中图分类号:

TP29

王祺, 叶仁传, 马国杰, 马佩珏, 范杰, 杨文龙. 钢材表面缺陷轻量化识别算法研究[J]. 机械工程学报, 2025, 61(8): 18-31.

WANG Qi, YE Renchuan, MA Guojie, MA Peijue, FAN Jie, YANG Wenlong. Lightweight Surface Defect Detection Method of Steel[J]. Journal of Mechanical Engineering, 2025, 61(8): 18-31.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: http://www.cjmenet.com.cn/CN/10.3901/JME.2025.08.018

http://www.cjmenet.com.cn/CN/Y2025/V61/I8/18

参考文献

[1] 胡贵超. 基于计算机视觉的钢材表面缺陷检测系统的研究[D]. 哈尔滨：哈尔滨理工大学，2013. HU Guichao. Research on steel surface defect inspection system based on computer vision[D]. Harbin：Harbin University of Science and Technology，2013.
[2] 彭向前. 产品表面缺陷在线检测方法研究及系统实现[D]. 武汉：华中科技大学，2008. PENG Xiangqian. Study on methogology of product surface defects online detection and system implementation[D]. Wuhan：Huazhong University of Science and Technology，2008.
[3] GIRSHICK R，DONAHUE J，DARRELL T，et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]// 2014 IEEE Conference on Computer Vision and Pattern Recognition，23-28 June 2014，Columbus OH，USA，2014：580-587.
[4] HE K，ZHANG X，REN S，et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence，2015，37(9)：1904-1916.
[5] GIRSHICK R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision，07-13 December，2015，Santiago，Chile，2015：1440-1448.
[6] REN S，HE K，GIRSHICK R，et al. Faster R-CNN：Towards real-time object detection with region proposal networks[J]. Advances in Neural Information Processing Systems，2015，28：91-99.
[7] HE K，GKIOXARI G，DOLLAR P，et al. Mask R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision，22-29 October，2017，Venice，Italy，2017：2961-2969.
[8] REDMON J，DIVVALA S，GIRSHICK R，et al. You only look once：Unified，real-time object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，27-30 June 2016，Las Vegas，NV，USA，2016：779-788.
[9] LIU W，ANGUELOV D，ERHAN D，et al. SSD：Single shot multibox detector[C]//European Conference on Computer Vision，Springer，Cham，2016：21-37.
[10] TIAN Z，SHEN C，CHEN H，et al. Fcos：Fully convolutional one-stage object detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision，27 Feb. 2020，Seoul，Korea，2019：9627-9636.
[11] 陈一潇，阿里甫·库尔班，林文龙，等. 面向拥挤行人检测的CA-YOLOv5[J]. 计算机工程与应用，2022，58(9)：238-245. CHEN Yixiao，ALIFU·Kuerban，LIN Wenlong，et al. CA-YOLOv5 for crowded pedestrian detection[J]. Computer Engineering and Applications，2022，58(9)：238-245.
[12] 孙泽强，陈炳才，崔晓博，等. 融合频域注意力机制和解耦头的YOLOv5带钢表面缺陷检测[J]. 计算机应用，2023，43(1)：242-249. SUN Zeqiang，CHEN Bingcai，CUI Xiaobo，et al. Strip steel surface defect detection by YOLOv5 algorithm fusing frequency domain attention mechanism and decoupled head[J]. Journal of Computer Applications，2023，43(1)：242-249.
[13] 韩强，张喆，续欣莹，等. 基于FF R-CNN钢材表面缺陷检测算法[J]. 太原理工大学学报，2021，52(5)：754-763. HAN Qiang，ZHANG Zhe，XU Xinying，et al. Steel surface defect detection based on FF R-CNN[J]. Journal of Taiyuan University of Technology，2021，52(5)：754-763.
[14] SHEN H，LI S，GU D，et al. Bearing defect inspection based on machine vision[J]. Measurement，2012，45(4)：719-733.
[15] HE D，XU K，ZHOU P. Defect detection of hot rolled steels with a new object detection framework called classification priority network[J]. Computers & Industrial Engineering，2019，128：290-297.
[16] BOIKOV A，PAYOR V，SAVELEV R，et al. Synthetic data generation for steel defect detection and classification using deep learning[J]. Symmetry，2021，13(7)：1176.
[17] LI Shun，WANG Xiaoqiang. YOLOv5-based defect detection model for hot rolled strip steel[J]. Journal of Physics：Conference Series，2022，2171：012040.
[18] HE K，ZHANG X，REN S，et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，27-30 June 2016，Las Vegas，NV，USA，2016：770-778.
[19] HUANG G，LIU Z，van der MAATEN L，et al. Densely connected convolutional networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，9 Nov. 2017，Honolulu，HI，USA，2017：4700-4708.
[20] TAN M，LE Q. Efficientnet：Rethinking model scaling for convolutional neural networks[C]//International Conference on Machine Learning，PMLR，2019：6105-6114.
[21] LIN T Y，DOLLAR P，GIRSHICK R，et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，9 Nov. 2017，Honolulu，HI，USA，2017：2117-2125.
[22] LIU S，QI L，QIN H，et al. Path aggregation network for instance segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，18-23 June 2018，Salt Lake City，UT，USA，2018：8759-8768.
[23] KLAMBAUER G，UNTERTHINER T，MAYR A，et al. Self-normalizing neural networks[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems，8 June，2017，2017：972-981.
[24] HOWARD A，SANDLER M，CHU G，et al. Searching for mobilenetv3[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision，27 Oct. 2019，Seal，Korea，2019：1314-1324.
[25] REZATOFIGHI H，TSOI N，GWAK J Y，et al. Generalized intersection over union：A metric and a loss for bounding box regression[C]//Proceedings of the IEEE/ CVF Conference on Computer Vision and Pattern Recognition，15-20 June，2019，Long Beach，CA，USA，IEEE，2019：658-666.
[26] ZHENG Zhaohu，WANG Ping，LIU Wei，et al. Distance-IoU loss：Faster and better learning for bounding box regression[C/CD]// AAAI Conference on Artificial Intelligence，2019.
[27] Zheng Z，Wang P，Ren D，et al. Enhancing geometric factors in model learning and inference for object detection and instance segmentation[J]. IEEE Transactions on Cybernetics，2022，52(8)：8574-8586.
[28] GEVORGYAN Z. SIoU Loss：More powerful learning for bounding box regression[J]. arXiv Preprint arXiv：2205.12740，2022.
[29] HU J，SHEN L，SUN G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，18-23 June 2018，Salt Lake City，UT，USA，2018：7132-7141.
[30] HOU Q，ZHOU D，FENG J. Coordinate attention for efficient mobile network design[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition，20-25 June，2021，Nashvile，TN，USA，2021：13713-13722.
[31] 熊鹏文，陈志远，廖俊杰，等. 基于改进卷积注意力机制的触觉图像识别[J]. 东南大学学报，2024，54(1)：175-182. XIONG Pengwen，CHEN Zhiyuan，LIAO Junjie，et al. Tactile image recognition based on improved convolutional attention mechanism[J]. Journal of Southeast University，2024，54(1)：175-182.
[32] DING X，ZHANG X，MA N，et al. Repvgg：Making vgg-style convnets great again[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition，20-25 June，2021，Nashvile，TN，USA，2021：13733-13742.
[33] SONG K，YAN Y. A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects[J]. Applied Surface Science，2013，285：858-864.