基于DCE-YOLO算法的复杂背景下钢材表面缺陷检测研究*

doi:10.16731/j.cnki.1671-3133.2026.04.012

现代制造工程 ›› 2026, Vol. 547 ›› Issue (4): 91-102.doi: 10.16731/j.cnki.1671-3133.2026.04.012

• 仪器仪表/检测/监控 • 上一篇下一篇

基于DCE-YOLO算法的复杂背景下钢材表面缺陷检测研究^*

刘宗, 胡伟

河南理工大学电气工程与自动化学院,焦作 454150

收稿日期:2025-08-20 发布日期:2026-05-07
通讯作者: 胡伟,博士,副教授,主要研究方向为模式识别与图像处理。E-mail:453795166@qq.com
作者简介:刘宗,硕士研究生,主要研究方向为模式识别与图像处理。E-mail:3115085964@qq.com
基金资助:
^*国家自然科学基金项目(U1804147)

Research on steel surface defect detection in complex backgrounds based on DCE-YOLO algorithm

LIU Zong, HU Wei

School of Electrical Engineering and Automation,Henan Polytechnic University,Jiaozuo 454150,China

Received:2025-08-20 Published:2026-05-07

摘要/Abstract

摘要： 在钢材表面缺陷检测领域,因钢材表面缺陷存在尺度多变且背景复杂的情况,在检测时常出现漏检、误检以及检测精度不佳等问题。为解决上述问题,创新性地提出了一种钢材表面缺陷检测算法DCE-YOLO。首先,提出一种双路径高效下采样,细节保留分支与特征聚焦分支共同作用,提升模型对目标位置的判断能力;其次,设计一种多尺度通道空间自注意力机制,使主干网络能够更高效地提取关键信息;最后,构建了一种基于Sobel算子的边缘信息增强模块,并将其嵌入到C2f模块中,使模型更好地捕捉边缘和纹理。在NEU-DET数据集上实验结果表明,相较于基准模型,所提的DCE-YOLO模型参数量为2.91×10⁶,mAP@0.5和mAP@0.5:0.95分别提升了5.1 %和3.1 %。此外,在GC10-DET数据集上的实验结果证实了该方法具备良好的鲁棒性。

关键词: 钢材表面缺陷, 双路径高效下采样, 自注意力机制, 边缘信息增强

Abstract: In the field of surface defect detection of steel materials,due to the variable scale and complex background of surface defects,problems such as missed detections,false detections and poor detection accuracy are often occur during detection. To address these issues,an innovative steel surface defect detection algorithm,named DCE-YOLO,is proposed. Firstly,a dual-path efficient downsampling is introduced,where the detail retention branch and the feature focusing branch work together to enhance the model′s ability to determine the target location. Secondly,a multi-scale channel spatial self-attention mechanism is designed to enable the backbone network to extract key information more efficiently. Finally,an edge information enhancement module based on Sobel operator is constructed and embedded into C2f block to improve the ability of the model for capturing edges and textures. Experimental results on the NEU-DET dataset show that compared with the baseline model,the proposed DCE-YOLO model,with 2.91×10⁶ parameters,improves mAP@0.5 and mAP@0.5:0.95 by 5.1 % and 3.1 % respectively. Additionally,the experimental results on the GC10-DET dataset confirm the robustness of this method.

Key words: steel surface defects, dual-path efficient downsampling, self-attention mechanism, edge information enhancement

中图分类号:

TP391.41

刘宗, 胡伟. 基于DCE-YOLO算法的复杂背景下钢材表面缺陷检测研究^*[J]. 现代制造工程, 2026, 547(4): 91-102.

LIU Zong, HU Wei. Research on steel surface defect detection in complex backgrounds based on DCE-YOLO algorithm[J]. Modern Manufacturing Engineering, 2026, 547(4): 91-102.

参考文献

[1] 张伟,李强. 高强度冷轧钢带在汽车轻量化中的应用进展[J]. 材料工程,2021,49(3):1-10.
[2] DU Y,CHEN H,FU Y,et al. AFF-Net:a strip steel surface defect detection network via adaptive focusing features[J]. IEEE Transactions on Instrumentation and Measurement,2024,73:2518514.
[3] LV S,OUYANG B,DENG Z,et al. A dataset for deep learning based detection of printed circuit board surface defect[J]. Scientific Data,2024,11(1):811.
[4] SONG X,CAO S,ZHANG J,et al. Steel surface defect detection algorithm based on YOLOv8[J]. Electronics,2024,13(5):988.
[5] TIAN R,JIA M. DCC-CenterNet:a rapid detection method for steel surface defects[J]. Measurement,2022,187:110211.
[6] GIRSHICK R,DONAHUE J,DARRELL T,et al. Region-based convolutional networks for accurate object detection and segmentation[J]. IEEE Transactions on Pattern Analy-sis and Machine Intelligence,2015,38(1):142-158.
[7] REN S,HE K,GIRSHICK R,et al. Faster R-CNN:towards real-time object detection with region proposal networks[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems (NIPS 2015). Montreal:Curran Associates, 2015:91-99.
[8] HE K M,GKIOXARI G,DOLLÁR P,et al. Mask R-CNN[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).[S.l.]:IEEE,2017:2961-2969.
[9] WANG C Y,YEH I H,LIAO H Y M. YOLOv9:learning what you want to learn using programmable gradient information[J]. arXiv preprint arXiv,2402.13616,2024.
[10] WANG A,CHEN H,LIU L,et al. YOLOv10:real-time end-to-end object detection[J]. arXiv preprint arXiv,2405.14458,2024.
[11] LIU W,ANGUELOV D,ERHAN D,et al. SSD:single shot multibox detector[C]//Computer Vision-ECCV 2016. 14th European Conference. Cham:Springer,2016:21-37.
[12] LIN T Y,GOYAL P,GIRSHICK R,et al. Focal loss for dense object detection[C]//2017 IEEE Internation Conference on Computer Vision (ICCV).[S.l.]:IEEE,2017:2980-2988.
[13] 周孟然,王昊男,高立鹏,等. 基于YOLOv5s-FCS的钢材表面缺陷检测[J]. 科学技术与工程,2024,24(14):5901-5910.
[14] ZHOU S,ZENG Y,LI S,et al. Surface defect detection of rolled steel based on lightweight model[J]. Applied Sciences,2022,12(17):8905.
[15] 刘毅,蒋三新. 基于改进YOLOX的钢材表面缺陷检测研究[J]. 现代电子技术,2024,47(9):131-138.
[16] 李相垚,侯红玲,杨澳,等. 面向钢材表面缺陷检测的DCS-YOLOv8算法研究[J/OL]. 机械科学与技术:1-10.[2025-07-09].https://doi.org/10.13433/j.cnki.1003-8728.20240128.
[17] 徐薪羽,沈通,吕佳. 基于改进YOLOv8算法的钢材表面缺陷检测[J]. 自动化应用,2024(15):6-10.
[18] REIS D,KUPECJ,HONG J,et al. Real-time flying object detection with YOLOv8[J]. arXiv preprint arXiv:2301.09972,2023.
[19] 孔思曼,周晨阳,王家华. 基于优化Faster R-CNN算法的金属板材表面缺陷检测[J]. 制造技术与机床,2024(1):171-178.
[20] CAI Z W,VASCONCELOS N. Cascade R-CNN:high quality object detection and instance segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2019,43(5):1483-1498.
[21] ZHANG H,LI F,LIU S,et al. DINO:DETR with improved denoising anchor boxes for end-to-end object detection[J]. arXiv preprint arXiv:2203.03605,2022.
[22] ZHU X,SU W,LU L,et al. Deformable DETR:deformable transformers for end-to-end object detection[J]. arXiv preprint arXiv:2010.04159,2020.
[23] FENG C,ZHONG Y,GAO Y,et al. TOOD:task-aligned one-stage object detection[C]//2021 IEEE International Conference on Computer Vision (ICCV).[S.l.]:IEEE,2021:3490-3499.
[24] LI X,WANG W H,CHEN S,et al. Generalized focal loss:learning qualified and distributed bounding boxes for dense object detection[J]. arXiv preprint arXiv:2006. 04388,2020.
[25] ZHANG S,WANG X J,WANG J Q,et al. Dense distinct query for end-to-end object detection[J]. arXiv preprint arXiv:2303.12776,2023.
[26] 梁礼明,龙鹏威,冯耀. 改进轻量化VTG-YOLOv7-tiny的钢材表面缺陷检测[J]. 光学精密工程,2024,32(8):1227-1240.
[27] 周彦,孟江南,王冬丽,等. 基于多尺度轻量化注意力的钢材缺陷检测[J]. 控制与决策,2024,39(3):901-909.

基于DCE-YOLO算法的复杂背景下钢材表面缺陷检测研究^*

Research on steel surface defect detection in complex backgrounds based on DCE-YOLO algorithm

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 3

编辑推荐

Metrics

[1]	徐慧玲, 尚冠宇. 改进YOLOv11与多模态融合机制的钢材表面缺陷检测方法^*[J]. 现代制造工程, 2026, 547(4): 103-116.
[2]	李希明, 刘业通, 彭世康, 吴湘柠, 李恒强, 蒙艳玫. 基于改进TD3的山地无人作业底盘姿态控制方法^*[J]. 现代制造工程, 2025, 536(5): 1-11.
[3]	阎馨，杨月川，屠乃威. 基于改进SSD的钢材表面缺陷检测[J]. 现代制造工程, 2023, 512(5): 112-120.