基于改进YOLOv5的钢板表面缺陷检测

doi:10.16180/j.cnki.issn1007-7820.2025.06.012

摘要/Abstract

摘要：

为满足多数工业场景下钢板表面缺陷检测的需求,针对钢板表面缺陷检测准确率低及小目标缺陷检测率低等问题,文中提出了一种基于改进YOLOv5(You Only Look Once version 5)的钢板表面缺陷检测算法。在YOLOv5的基础上将CBAM(Convolution Block Attention Module)注意力模块嵌入到主干网络中,提高网络检测精度。加入上下文增强模块,提高了算法对小目标的检测性能。使用NWD(Normalized Wasserstein Distance)度量标准代替原YOLOv5中的IoU(Intersection over Union)度量,提高了网络对裂纹缺陷的识别精确度。实验结果表明,钢板表面缺陷检测算法对裂纹、夹杂、斑块、麻点、压入氧化铁皮、划痕6类缺陷的平均检测精度达到了88.9%,每秒帧数达到110.4 frame·s^-1,其中小目标裂纹准确率达到75%。

关键词: 钢板表面缺陷检测, YOLOv5, 注意力模块, 上下文增强模块, 小目标, 位置偏差, NWD度量, IoU度量

Abstract:

To meet the requirements of steel plate surface defect detection in most industrial scenarios, a steel plate surface defect detection algorithm based on improved YOLOv5(You Only Look Once version 5) is proposed to solve the problems such as low detection accuracy of steel plate surface defects and failure to detect small target defects. On the basis of YOLOv5, the CBAM(Convolution Block Attention Module) is embedded into the backbone network to improve network detection accuracy. The context enhancement module is added to improve the detection performance of small targets. The NWD(Normalized Wasserstein Distance) metric is used to replace the original IoU(Intersection over Union) metric in YOLOv5, making the network more accurate in identifying crack defects. Experimental results show that the average detection accuracy of the proposed steel plate surface defect detection algorithm for six types of defects, including crack, inclusion, plaque, pitting, pressed iron oxide, scratch, reaches 88.9%, the frame rate reaches 110.4 frame·s^-1, and the accuracy of small target crack reaches 75%.

Key words: steel plate surface defect detection, YOLOv5, attention module, context enhancement module, small goals, position deviation, NWD measurement, IoU measurement

中图分类号:

TP391

沈庭铅, 鲁玉军, 辛昊, 吴涵超, 汪仕男. 基于改进YOLOv5的钢板表面缺陷检测[J]. 电子科技, 2025, 38(6): 82-88.

SHEN Tingqian, LU Yujun, XIN Hao, WU Hanchao, WANG Shinan. Steel Plate Surface Defect Detection Based on Improved YOLOv5[J]. Electronic Science and Technology, 2025, 38(6): 82-88.

图/表 13

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

表1

表2

图11

参考文献 26

[1]	甘胜丰, 孙林, 曹阳, 等. 多级图像分类系统在硅钢冷轧表面缺陷检测中的应用[J]. 冶金自动化, 2009, 33(2):63-65.
	Gan Shengfeng, Sun Lin, Cao Yang, et al. Application of multi-level image classification system in surface defect detection of cold-rolled silicon strip[J]. Metallurgical Industry Automation, 2009, 33(2):63-65.
[2]	丛家慧, 颜云辉. 视觉注意机制在带钢表面缺陷检测中的应用[J]. 中国机械工程, 2011, 22(10):1189-1192,1221.
	Cong Jiahui, Yan Yunhui. Application of human visual attention mechanism in surface defect inspection steel strip[J]. China Mechanical Engineering, 2011, 22(10):1189-1192,1221.
[3]	向阳, 陈黎, 张晓龙. 基于支持向量机的带钢表面缺陷识别研究[J]. 工业控制计算机, 2012, 25(8):99-101.
	Xiang Yang, Chen Li, Zhang Xiaolong. Research on strip surface defect recognition based on support vector machine[J]. Industrial Control Computer, 2012, 25(8):99-101.
[4]	Girshick R, Donahue J, Trevor D, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]. Columbus: IEEE Conference on Computer Vision and Pattern Recognition,2014:580-587.
[5]	Ren S Q, He K M, Girshick R, et al. FasterRCNN:Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1137-1149.
[6]	He K M, Gkioxari G, Dollár P, et al. Mask R-CNN[C]. Venice: IEEE International Conference on Computer Vision,2017:386-397.
[7]	Redmon J, Divvala S, Girshick R, et al. You only look once:Unified,real-time object detection[C]. Seattle: IEEE Conference on Computer Vision and Pattern Recognition,2016:779-788.
[8]	Redmon J, Farhadi A. YOLO9000:Better,faster, stronger[C]. Honolulu: IEEE Conference on Computer Vision and Pattern Recognition,2017:6517-6525.
[9]	Redmon J, Farhadi A. YOLOv3:An incremental improvement.[EB/OL].(2018-04-08)[2023-10-15]. https://arxiv.org/abs/1804.02767.
[10]	Bochkovskiy A, Wang C Y, Liao H. YOLOv4:Optimal speed and accuracy of object detection[EB/OL].(2020-4-23)[2023-10-15]. https://arxiv.org/pdf/2004.10934.pdf.
[11]	Liu W, Anguelov D, Erhan D, et al. SDD:Single shot MultiBox detector[C]. Amsterdam: The Fourteenth European Conference on Computer Vision,2016:21-37.
[12]	王立中, 管声启. 基于深度学习算法的带钢表面缺陷识别[J]. 西安工程大学学报, 2017, 31(5):669-674.
	Wang Lizhong, Guan Shengqi. Strip steel surface defect recognition based on deep learning[J]. Journal of Xi'an Polytechnic University, 2017, 31(5):669-674.
[13]	李维刚, 叶欣, 赵云涛, 等. 基于改进YOLOv3算法的带钢表面缺陷检测[J]. 电子学报, 2020, 48(7):1284-1292. doi: 10.3969/j.issn.0372-2112.2020.07.006
	Li Weigang, Ye Xin, Zhao Yuntao, et al. Strip steel surface defect detection based on improved YOLOv3 algorithm[J]. Acta Electronica, 2020, 48(7):1284-1292.
[14]	郝用兴, 庞永辉. 基于DenseNet的带钢表面缺陷智能识别系统研究[J]. 河南科技, 2021, 40(3):11-14.
	Hao Yongxing, Pang Yonghui. Research on intelligent recognition system of strip steel surface defects based on DenseNet[J]. Henan Science and Technology, 2021, 40(3):11-14.
[15]	翁玉尚, 肖金球, 夏禹. 改进Mask R-CNN算法的带钢表面缺陷检测[J]. 计算机工程与应用, 2021, 57(19):235-242. doi: 10.3778/j.issn.1002-8331.2010-0446
	Weng Yushang, Xiao Jinqiu, Xia Yu. Strip surface defect detection based on improved Mask R-CNN algorithm[J]. Computer Engineering and Applications, 2021, 57(19):235-242. doi: 10.3778/j.issn.1002-8331.2010-0446
[16]	Lin T, Dollár P, Ross B, et al. Feature pyramid networks for object detection[C]. Honolulu: IEEE Conference on Computer Vision and Pattern Recognition,2017:936-944.
[17]	Liu S, Qi L, Qin H F, et al. Path aggregation network for instance segmentation[C]. Salt Lake City: IEEE/CVF Conference on Computer Vision and Pattern Recognition,2018:8759-8768.
[18]	Zhao Y, Zhao L J, Li C Y, et al. Pyramid attention dilated network for aircraft detection in SAR images[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 18(4):662-666. doi: 10.1109/LGRS.2020.2981255
[19]	Zhao W X, Syafrudin M, Fitriyani N L. CRAS YOLO:A novel multi-category vessel detection and classification model based on YOLOv5s algorithm[J]. IEEE Access, 2023, 11(9):11463-11478.
[20]	殷远齐, 徐源, 邢远新. 基于YOLOv4的车辆目标检测算法[J]. 计算机与现代化, 2022, 38(7):8-14.
	Yin Yuanqi, Xu Yuan, Xing Yuanxin. Vehicle target detection algorithm based on YOLOv4[J]. Computer and Modernization, 2022, 38(7):8-14.
[21]	宿磊, 王立建, 祁阳, 等. 基于IADSA深度迁移网络的金属表面缺陷检测[J]. 机械工程学报, 2023, 59(24):46-55.
	Su Lei, Wang Lijian, Qi Yang, et al. Metal surface defect detection based on IADSA deep transfer network[J]. Journal of Mechanical Engineering, 2023, 59(24):46-55.
[22]	徐忠锴, 刘艳玲, 盛晓娟, 等. 基于改进YOLOv5的变电站典型缺陷自动检测算法[J]. 系统仿真学报, 2024, 36(11):2604-2613. doi: 10.16182/j.issn1004731x.joss.23-0914
	Xu Zhongkai, Liu Yanling, Sheng Xiaojuan, et al. Automatic detection algorithm for substation typical defects based on improved YOLOv5[J]. Journal of System Simulation, 2024, 36(11):2604-2613. doi: 10.16182/j.issn1004731x.joss.23-0914
[23]	何幸, 黄永明, 朱勇. 基于改进YOLOv5的路面坑洼检测方法[J]. 电子科技, 2024, 37(7):53-59.
	He Xing, Huang Yongming, Zhu Yong. Pavement pothole detection method based on improved YOLOv5[J]. Electronic Science and Technology, 2024, 37(7):53-59.
[24]	田志新, 茅健, 林彬彬, 等. 基于多尺度特征融合的钢材表面缺陷分类方法[J]. 电子科技, 2024, 37(2):87-94.
	Tian Zhixin, Mao Jian, Lin Binbin, et al. Classification method of steel surface defects based on multi-scale feature fusion[J]. Electronic Science and Technology, 2024, 37(2):87-94.
[25]	Xiao J S, Zhao T, Yao Y T, et al. Context augmentation and feature refinement network for tiny object detection[EB/OL].(2023-02-14)[2023-11-26]. https://openreview.net/pdf?id=q2ZaVU6bEsT.
[26]	Wang J W, Xu C, Yang W, et al. A normalized gaussian wasserstein distance for tiny object detection[EB/OL]. (2021-10-26)[2023-11-26]. https://arxiv.org/abs/2110.13389.

对比模型	YOLOv5	YOLOv5- CAM	YOLOv5- CAM-NWD	YOLOv5-CAM- NWD-CBAM
裂纹AP/%	69.1	71.3	73.9	75.0
夹杂AP/%	89.9	89.9	90.4	90.3
斑块AP/%	96.0	96.3	96.6	96.9
麻点AP/%	86.3	87.4	87.6	88.0
压入氧化铁皮AP/%	84.2	85.3	86.5	86.7
划痕AP/%	95.0	96.1	96.2	96.4
mAP/%	86.8	87.7	88.5	88.9
FPS/frame·s^-1	114.0	108.7	111.1	110.4

算法	mAP/%	FPS/frame·s^-1
Faster-RCNN	69.9	35.5
SSD	73.4	63.8
YOLOX	78.0	51.3
YOLOv3	90.8	47.0
YOLOv5	86.8	114.0
本文算法	88.9	110.4