工业场景下的钢材表面缺陷实时检测网络

doi:10.3778/j.issn.1002-8331.2505-0316

摘要/Abstract

摘要： 针对工业生产中，金属表面的缺陷检测任务面临着缺陷尺度差异大、特征提取困难和推理实时性差等问题，提出了一种新型高效的改进模型MBAC-YOLO（YOLO with multi-head convolution，bio-inspired hybrid attention module and contextual enhancement）。该方法包括了三个创新的模块：多头卷积模块MCM（multi-head convolution module），可显著提高模型感受野及主干网络的局部特征提取能力；仿生启发混合注意力模块BHAM（bio-inspired hybrid attention module），可增强空间与通道的边缘信息理解及颈部网络特征表现力；全局上下文增强模块GCEM（global context enhancement module），用于生成自适应权重并构建全局上下文信息交互。为评估所提方法的表现，分别采用了NEU-DET数据集和GC10-DET数据集进行验证，结果表明，MBAC-YOLO的准确率（mAP50）分别提升了5.8个百分点和5.3个百分点，检测速度（FPS）达到了185.19和182.0，表明了所提出的方法在检测精度和实时性上具有明显优势。

关键词: 缺陷检测, 特征提取, 注意力机制, 上下文信息

Abstract: For defect detection on metal surfaces in industrial production, facing challenges including large variations in defect scales, feature extraction difficulties, and poor real-time inference, this paper proposes an efficient model named MBAC-YOLO (YOLO with multi-head convolution, bio-inspired hybrid attention module and contextual enhancement). It integrates three innovative modules： the multi-head convolution module (MCM), enhancing the model’s receptive field and backbone network’s local feature extraction; the bio-inspired hybrid attention module (BHAM), strengthening understanding of spatial/channel edge information and boosting feature representation in the neck network; and the global context enhancement module (GCEM), generating adaptive weights and constructing global contextual interaction. Performance is validated using NEU-DET and GC10-DET datasets. Results demonstrate MBAC-YOLO achieves mAP50 imp-rovements of 5.8 percentage points and 5.3 percentage points, with FPS reaching 185.19 and 182.0, indicating significant advantages in both accuracy and real-time performance.

Key words: defect detection, feature extraction, attention mechanism, context information

仵大奎, 葛承昆, 周文举, 高艺友. 工业场景下的钢材表面缺陷实时检测网络[J]. 计算机工程与应用, 2026, 62(7): 85-95.

WU Dakui, GE Chengkun, ZHOU Wenju, GAO Yiyou. Real-Time Detection Network for Steel Surface Defects in Industrial Scenarios[J]. Computer Engineering and Applications, 2026, 62(7): 85-95.

参考文献

[1] 魏静，史庆丰，沈飞，等.工业外观检测中的图像扩增方法综述[J].自动化学报，2025，51（7）：1423-1462.
WEI J，SHI Q F，SHEN F，et al.A review of image augmentation methods in industrial cosmetic inspection[J].Acta Automatica Sinica，2025，51（7）：1423-1462.
[2] 何彦良，邢博文，辛蕾，等.基于超声波法的GIS盆式绝缘子缺陷检测[J].绝缘材料，2025，58（4）：126-133.
HE Y L，XING B W，XIN L，et al.Defect detection of GIS basin insulator based on ultrasonic method[J].Insulating Materials，2025，58（4）：126-133.
[3] 孙明健，程星振，王艳，等.基于光声信号的高铁钢轨表面缺陷检测方法[J].物理学报，2016，65（3）：351-360.
SUN M J，CHENG X Z，WANG Y，et al.Method for detecting high-speed rail surface defects by photoacoustic signal[J].Acta Physica Sinica，2016，65（3）：351-360.
[4] 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，2017，39（6）：1137-1149.
[5] HE K M，GKIOXARI G，DOLLáR P，et al.Mask R-CNN[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2020，42（2）：386-397.
[6] CHEN M Q，YU L J，ZHI C，et al.Improved faster R-CNN for fabric defect detection based on Gabor filter with genetic algorithm optimization[J].Computers in Industry，2022，134：103551.
[7] REDMON J，FARHADI A.YOLOv3：an incremental improvement[EB/OL].[2025-05-07].https：//arxiv.org/abs/1804.02767.
[8] ZHAO J Q，ZHANG X H，YAN J W，et al.A wheat spike det-ection method in UAV images based on improved YOLOv5[J].Remote Sensing，2021，13（16）：3095.
[9] YI H，LIU B，ZHAO B，et al.Small object detection algorithm based on improved YOLOv8 for remote sensing[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing，2024，17：1734-1747.
[10] LIU W，ANGUELOV D，ERHAN D，et al.SSD：single shot multibox detector[M]//Lecture notes in computer science.Cham：Springer，2016：21-37.
[11] HUANG X H，ZHU J H，HUO Y.SSA-YOLO：an improved YOLO for hot-rolled strip steel surface defect detection[J].IEEE Transactions on Instrumentation and Measurement，2024，73：5040017.
[12] ZHANG H，FU W，WANG X M，et al.An efficient model for metal surface defect detection based on attention mechanism and multi-scale feature[J].The Journal of Supercomputing，2024，81（1）：40.
[13] ZHAO C，SHU X，YAN X，et al.RDD-YOLO：a modified YOLO for detection of steel surface defects[J].Measurement，2023，214：112776.
[14] 窦智，高浩然，刘国奇，等.轻量化YOLOv8的小样本钢板缺陷检测算法[J].计算机工程与应用，2024，60（9）：90-100.
DOU Z，GAO H R，LIU G Q，et al.Small sample steel plate defect detection algorithm of lightweight YOLOv8[J].Computer Engineering and Applications，2024，60（9）：90-100.
[15] 马磊，李晔，王宇翔.YOLOv8-FD：YOLOv8改进的钢板表面缺陷检测方法[J].计算机工程与应用，2024，60（24）：211-221.
MA L，LI Y，WANG Y X.YOLOv8-FD：YOLOv8 improved method for detecting surface defects on steel plates[J].Computer Engineering and Applications，2024，60（24）：211-221.
[16] LIN W F，WU Z H，CHEN J Y，et al.Scale-aware modulation meet transformer[C]//Proceedings of the 2023 IEEE/CVF International Conference on Computer Vision.Piscataway：IEEE，2024：5992-6003.
[17] WOO S，PARK J，LEE J Y，et al.CBAM：convolutional block attention module[C]//Proceedings of the 15th European Conference on Computer Vision.Cham：Springer，2018：3-19.
[18] QUAN Y，ZHANG D，ZHANG L Y，et al.Centralized feature pyramid for object detection[J].IEEE Transactions on Image Processing，2023，32：4341-4354.
[19] GAUTAMA T，VAN HULLE M M.Function of center-surround antagonism for motion in visual area MT/V5：a modeling study[J].Vision Research，2001，41（28）：3917-3930.
[20] WESTHEIMER G.Center-surround antagonism in spatial vision：retinal or cortical locus?[J].Vision Research，2004，44（21）：2457-2465.
[21] CAO Y，XU J R，LIN S，et al.GCNet：non-local networks meet squeeze-excitation networks and beyond[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision Workshop.Piscataway：IEEE，2019：1971-1980.
[22] ZHANG Y，YE M，ZHU G Y，et al.FFCA-YOLO for small object detection in remote sensing images[J].IEEE Transactions on Geoscience and Remote Sensing，2024，62：5611215.
[23] 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，2020，69（4）：1493-1504.
[24] CHENG X，YU J B.RetinaNet with difference channel attention and adaptively spatial feature fusion for steel surface defect detection[J].IEEE Transactions on Instrumentation and Measurement，2021，70：2503911.
[25] LIN T Y，GOYAL P，GIRSHICK R，et al.Focal loss for dense object detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2020，42（2）：318-327.
[26] LIU R Q，HUANG M，GAO Z M，et al.MSC-DNet：an efficient detector with multi-scale context for defect detection on strip steel surface[J].Measurement，2023，209：112467.
[27] ZHANG Y，ZHANG H F，HUANG Q Q，et al.DsP-YOLO：an anchor-free network with DsPAN for small object detection of multiscale defects[J].Expert Systems with Applications，2024，241：122669.
[28] LIU G H，CHU M X，GONG R F.Semi-SSDDet：an adaptive semi-supervised steel surface defect detection[J].Signal，Image and Video Processing，2025，19（10）：798.
[29] CHEN F Q，DENG M L，GAO H，et al.AP-Net：a metallic surface defect detection approach with lightweight adaptive attention and enhanced feature pyramid[J].Cluster Computing，2024，27（3）：3837-3851.
[30] LI P H，TAO H J，ZHOU H，et al.Enhanced multiview attention network with random interpolation resize for few-shot surface defect detection[J].Multimedia Systems，2024，31（1）：36.
[31] HAN K，WANG Y H，TIAN Q，et al.GhostNet：more features from cheap operations[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway：IEEE，2020：1577-1586.
[32] YANG J N，LIU S L，WU J J，et al.Pinwheel-shaped convolution and scale-based dynamic loss for infrared small target detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2025：9202-9210.
[33] HU J，SHEN L，ALBANIE S，et al.Squeeze-and-excitation networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2020，42（8）：2011-2023.
[34] WANG X L，GIRSHICK R，GUPTA A，et al.Non-local neural networks[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway：IEEE，2018：7794-7803.
[35] LIU Y，LI H F，HU C，et al.Learning to aggregate multi-scale context for instance segmentation in remote sensing images[J].IEEE Transactions on Neural Networks and Lear-ning Systems，2025，36（1）：595-609.