[1]张晓君,陈 平,刘 宾.基于 YOLOv12 的轻量化带钢板表面缺陷检测算法研究[J].机械与电子,2026,44(05):49-55.
　ZHANG Xiaojun,CHEN Ping,LIU Bin.Research on Lightweight Surface Defect Detection Algorithm Based on YOLOv12 for Strip Steel[J].Machinery & Electronics,2026,44(05):49-55.
点击复制

基于 YOLOv12 的轻量化带钢板表面缺陷检测算法研究()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
44
期数:
2026年05期
页码:
49-55
栏目:
智能检测
出版日期:
2026-05-27

文章信息/Info

Title:
Research on Lightweight Surface Defect Detection Algorithm Based on YOLOv12 for Strip Steel
文章编号:
1001-2257 ( 2026 ) 05-0049-07
作者:
张晓君; 陈 平; 刘 宾
中北大学信息探测与处理山西省重点实验室,山西 太原 030051
Author(s):
ZHANG Xiaojun ;  CHEN Ping ;  LIU Bin
( Shanxi Key Laboratory of Signal Capturing and Processing , North China University , Taiyuan 030051 , China )
关键词:
缺陷检测; DCM 模块;  YOLOv12 ; 轻量化卷积
Keywords:
defect detection ;  DCM module ;  YOLOv12 ;  lightweight convolution
分类号:
TP391.4 ;TG142
文献标志码:
A
摘要:
针对带钢板表面缺陷类型多样、缺陷尺度差异显著以及工业生产线环境下检测精度与稳定性不足等问题,提出一种基于 YOLOv12 的轻量化带钢板表面缺陷检测算法。该算法重点刻画缺陷检测过程中模型参数之间的复杂依赖关系,尤其关注特征提取、特征融合与计算复杂度之间的协同作用,以克服传统方法的效率和精度受限的问题。首先引入 DCM 模块,对 YOLOv12 中的 C3k2 与 A2C2f 结构进行重构与增强,有效提升了网络的特征表征能力与上下文信息建模能力,从而改善缺陷目标的判别性能。同时结合 LSFM-Conv 轻量化卷积结构,显著降低了模型的冗余参数与计算开销,提高了推理效率和实际部署的可行性,进而缓解了小样本缺陷检测场景下因样本稀缺带来的性能退化问题。对比实验结果表明,与原始 YOLOv12 模型相比,所提算法在参数量减少至原模型约 92% 的同时,精确率、召回率和平均精度均值( mAP )分别提升了 0.5 百分点、4.0 百分点和 1.3 百分点。实验结果验证了该算法在带钢板表面缺陷检测任务中的有效性与鲁棒性,具有较好的综合性能和实际应用价值。
Abstract:
To address the challenges posed by diverse surface defect types , significant variations in defect scales , and insufficient detection accuracy and stability in industrial production line environments , this paper proposes a lightweight surface defect detection algorithm for steel strips based on YOLOv12.The algorithm focuses on characterizing the complex dependencies among model parameters during defect detection , particularly emphasizing the synergistic effects between feature extraction , feature fusion , and computational complexity to overcome the efficiency and accuracy limitations of traditional methods.Initially , a DCM module is introduced to reconstruct and enhance the C3k2 and A2C2f structures within YOLOv12 , significantly improving the network ’ s feature representation and contextual modeling capabilities , thereby refining the discriminative performance for defect targets.Concurrently , a LSFM-Conv lightweight convolutional structure is integrated to substantially reduce redundant parameters and computational overhead , boosting inference efficiency and practical deployment feasibility.This mitigates performance degradation caused by sample scarcity in small sample defect detection scenarios.Comparative experiments demonstrate that compared to the original YOLOv12 model , the proposed LSFM-DCM-YOLO achieves a 0.5 percentage points increase in precision , a 4.0 percentage points increase in recall , and a 1.3 percentage points increase in mean average precision ( mAP ), while reducing the number of parameters to approximately 92% of the original model.Experimental results validate the effectiveness and robustness of this method in steel plate surface defect detection tasks , demonstrating strong overall performance and practical application value.

参考文献/References:

[ 1 ] Bogdoll D , Nitsche M , Z?llner J M.Anomaly detection in autonomous driving : a survey [ C ] ∥2022 IEEE / CVF Conference on Computer Vision and Pattern Recognition , 2022 : 4488-4499.

[ 2 ] Wang Jinwen , Chen Ting , Xu Xinke , et al.An improved YOLOv8 model for strip steel surface defect detection [ J ] .Applied Sciences , 2024 , 15 ( 1 ): 52.
[ 3 ] Song Kechen , Yan Yunhui.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.
[ 4 ] Ma Hongtao , Zhang Zhisheng , Zhao Junai.A novel ST YOLO network for steel-surface-defect detection [ J ] .Sensors , 2023 , 23 ( 22 ): 9152.
[ 5 ] Brown T B , Mann B , Ryder N , et al.Language models are few-shotlearners [ C ] ∥34th Conference on Neural Information Processing Systems ( NIPS 2020 ), 2020 : 1877-1901.
[ 6 ] Zhu Shuxian , Zhou Yajie.MRP-YOLO : an improved YOLOv8 algorithm for steel surface defects [ J ] .Machines , 2024 , 12 ( 12 ): 917.
[ 7 ] Wang Yang , Wang Hongyuan , Xin Zihao.Efficient detection model of steel strip surface defects based on YOLO V7 [ J ] .IEEE Access , 2022 , 10 : 133936-133944.
[ 8 ] Gao Yang , Lv Gang , Xiao Dong , et al.Research on steel surface defect classification method based on deep learning [ J ] .Scientific Reports , 2024 , 14 : 8254.
[ 9 ] Damacharla P , Rao A M V , Ringenberg J , et al.TLU Net : a deep learning approach for automatic steel surface defect detection [ C ] ∥2021 International Conference on Applied Artificial Intelligence ( ICAPAI ) .New York : IEEE , 2021 : 1-6.
[ 10 ] Liu Mingshan , Huang Yifei , Xie Yuanli.Design of steel surface defect detection system based on deep learning [ C ] ∥2022 IEEE International Conference on Artificial Intelligence and Computer Applications ( ICAICA ) .New York : IEEE , 2022 : 743-749.
[ 11 ] Xu Zihao , Hu Jinyan , Xiao Xingyi , et al.GEB-YOLO : optimized YOLOv7 model for surface defect detection on aluminum profiles [ J ] .Engineering Proceedings , 2024 , 75 ( 1 ): 28.
[ 12 ] Tang Yunchao , Huang Zhaofeng , Chen Zheng , et al. Novel visual crack width measurement based on backbone double-scale features for improved detection automation [ J ] .Engineering Structures , 2023 , 274 : 115158.
[ 13 ] Li Dongming , Wang Erfu , Li Zhiyi , et al.STE-YOLO : A surface defect detection algorithm for steel strips [ J ] .Electronics , 2025 , 14 ( 1 ): 54.
[ 14 ] Ouyang Daliang , He Su , Zhang Guozhong , et al.Efficient multi-scale attention module with cross-spatial learning [ C ] ∥2023 IEEE international conference on acoustics , speech and signal processing ( ICASSP ) .New York : IEEE , 2023 : 1-5.
[ 15 ] Zhang Hu , Zu Keke , Lu Jian , et al.EPSANet : an efficient pyramid squeeze attention block on convolutional neural network [ C ] ∥Computer Vision ACCV2022 , 2022 : 1161-1177.
[ 16 ] Cao Yue , Xu Jiayui , Lin Stephen , et al.Gcnet : Non local networks meet squeeze-excitation networks and beyond [ C ] ∥2019 IEEE / CVF International Conference on Computer Vision Workshop ( ICCVW ) .New York : IEEE , 2019 : 1971-1980.

相似文献/References:

[1]陆 晓,高 超,蒋承伶,等.基于双注意力机制和级联检测框架的金具螺栓缺陷检测研究[J].机械与电子,2024,42(01):63.
　LU Xiao,GAO Chao,JIANG Chengling,et al.Research on Bolt Defect Detection Based on Dual-attention Mechanism and Cascade Detection Framework[J].Machinery & Electronics,2024,42(05):63.
[2]赖肖强,胡 泓.基于机器视觉算法和改进 MobileNetV2 的晶圆缺陷检测研究[J].机械与电子,2025,(01):34.
　LAI Xiaoqiang,HU Hong.Research on Wafer Defect Detection Based on Machine Vision Algorithms and Improved MobileNetV2[J].Machinery & Electronics,2025,(05):34.
[3]黄心玥,王明泉,耿宇杰,等.基于 YOLOv8 的铝合金铸造轮毂缺陷检测算法研究与分析[J].机械与电子,2025,(02):26.
　HUANG Xinyue,WANG Mingquan,GENG Yujie,et al.Research and Analysis of Defect Detection Algorithm for Aluminum Alloy Casting Wheels Based on YOLOv8[J].Machinery & Electronics,2025,(05):26.
[4]曹振锋,王明泉,路宇鹏,等.基于 YOLOv8 的轻量化轮毂缺陷检测算法研究[J].机械与电子,2025,(06):31.
　CAO Zhenfeng,WANG Mingquan,LU Yupeng,et al.Research on the Lightweight Wheel Hub Defect Detection Algorithm Based on YOLOv8[J].Machinery & Electronics,2025,(05):31.
[5]刘 骞,陈茂林.基于深度学习的少样本光伏边框划痕检测[J].机械与电子,2025,(08):47.
　LIU Qian,CHEN Maolin.Few-shot Scratch Detection on Photovoltaic Frame Based on Deep Learning[J].Machinery & Electronics,2025,(05):47.
[6]王立杰,高嘉伟,徐相龙,等. 基于SDA-YOLOv8的PCBA 缺陷检测算法研究[J].机械与电子,2026,44(01):45.
　WANG Lijie,GAO Jiawei,XU Xianglong,et al. Research on PCBA Defect Detection Algorithm Based on SDA -YOLOv8[J].Machinery & Electronics,2026,44(05):45.

备注/Memo

备注/Memo:
收稿日期: 2025-12-26
基金项目:山西省重点研发计划项目( 202302150401012 );山西省自然基金资助项目( 20210302124190 );国家自然科学基金资助项目( 62201520 , 62301508 );太原市关键核心技术攻关“揭榜挂帅”项目( 2025TYJB02 )
作者简介:张晓君 ( 1994- ),男,山西忻州人,硕士研究生,研究方向为图像处理;陈 平( 1983- ),男,安徽池州人,博士,教授,博士研究生导师,研究方向为图像处理、工业检测与识别等。
更新日期/Last Update: 2026-08-25