基于自监督卷积和无参数注意力机制的工业品表面缺陷检测

doi:10.6040/j.issn.1672-3961.0.2024.183

摘要/Abstract

摘要： 目前,基于深度学习的工业品表面缺陷检测存在着负样本数据集少、样本易受复杂工业环境影响无法提取有效特征、标注数据集需要耗费大量人工成本的问题。为解决上述问题,本研究提出一个基于知识蒸馏的工业品表面缺陷检测模型,并在此模型中加入自监督预测卷积模块和无参数注意力机制。该模型将教师网络所学习到的丰富特征知识传递给学生网络,有效提高模型特征表达能力并对缺陷进行像素级定位。该模型在 MVTec-AD数据集进行试验,并与各类试验方法的结果进行对比,检测指标和定位指标在不同模型曲线下面积A_ROC上有所提升,证明该方法可以有效提高模型的检测和定位能力。

关键词: 工业品表面缺陷检测, 深度学习, 知识蒸馏, 自监督预测, 注意力机制

Abstract: Currently, surface defect detection in industrial products using deep learning faced several challenges. These included a limited number of negative sample datasets, susceptibility to complex industrial environments that hindered effective feature extraction, and high labor cost for labeling datasets. In order to solve the above problems,a knowledge distillation-based surface defect detection method for industrial products was proposed. The model incorporated a self-supervised predictive convolution module and a parameter-free attention mechanism. It transferred the rich feature knowledge learned by the teacher network to the student network. This approach effectively improved the model feature expression and localized defects at the pixel level. The model was experimented on the MVTec-AD dataset, and the comparison results with the state-of-the-art experimental methods showed that its detection and localization metrics were improved on A_ROC. The results proved that the method could improve the model's detection and localization capabilities.

Key words: surface defect detection of industrial products, deep learning, knowledge distillation, self supervised prediction, attention mechanism

中图分类号:

TP391

周群颖,隋家成,张继,王洪元. 基于自监督卷积和无参数注意力机制的工业品表面缺陷检测[J]. 山东大学学报 (工学版), 2025, 55(4): 40-47.

ZHOU Qunying, SUI Jiacheng, ZHANG Ji, WANG Hongyuan. Industrial product surface defect detection based on self supervised convolution and parameter free attention mechanism[J]. Journal of Shandong University(Engineering Science), 2025, 55(4): 40-47.

参考文献

[1] 孙博言, 王洪元, 刘乾, 等. 基于多尺度和注意力机制的混合监督金属表面缺陷检测[J]. 智能系统学报, 2023, 18(4): 886-893. SUN Boyan, WANG Hongyuan, LIU Qian, et al. Hybrid supervised metal surface defect detection based on multi-scale and attention[J]. CAAI Transactions on Intelligent Systems, 2023, 18(4): 886-893.
[2] 李键, 李华, 胡翔坤, 等. 基于深度学习的表面缺陷检测技术研究进展[J]. 计算机集成制造系统, 2024, 30(3): 774-790. LI Jian, LI Hua, HU Xiangkun, et al. Research progress of surface defect detection technology based on deep learning[J].Computer Integrated Manufacturing System, 2024, 30(3):774-790.
[3] CHUA L O. CNN: a paradigm for complexity[M].Santiago, Chile: World Scientific Publishing Co. Pte Ltd., 1999: 529-837.
[4] NG H F. Automatic thresholding for defect detection[J]. Pattern Recognition Letters, 2006, 27(14): 1644-1649.
[5] BROMLEY J, GUYON I, LECUN Y, et al. Signature verification using a "siamese" time delay neural network[J]. Advances in Neural Information Processing Systems, 1993, 7(4):669-688.
[6] WANG Y S, YAO H X, ZHAO S C. Auto-encoder based dimensionality reduction[J]. Neurocomputing, 2016, 184: 232-242.
[7] ZHENG X Q, WANG H C, CHEN J, et al. A generic semi-supervised deep learning-based approach for automated surface inspection[J]. IEEE Access, 2020, 8: 114088-114099.
[8] 成科扬, 丁杨柳, 詹永照,等. 基于回顾蒸馏学习的无监督工业品缺陷检测方法[J]. 南京大学学报(自然科学版), 2022, 58(6): 1030-1040. CHENG Keyang, DING Yangliu, ZHAN Yongzhao, et al. Unsupervised industrial product defect detection method based on retrospective distillation learning [J].Journal of Nanjing University(Natural Science Edition), 2022, 58(6): 1030-1040.
[9] BUCILUĂ C, CARUANA R, NICULESCU-MIZIL A. Model compression[C] //Proceedings of the 12th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Philadelphia, USA. ACM, 2006: 535-541.
[10] HINTON G, VINYALS O, DEAN J. Distilling the knowledge in a neural network[EB/OL].(2015-03-09)[2024-03-12]. https://arxiv.org/abs/1503.02531
[11] GOU J P, YU B S, MAYBANK S J, et al. Knowledge distillation: a survey[J]. International Journal of Computer Vision, 2021, 129(6): 1789-1819.
[12] YU L, YAZICI V O, LIU X L, et al. Learning metrics from teachers: compact networks for image embedding[C] //Knowledge Distillation: A Survey.Long Beach, USA: IEEE, 2019: 2907-2916.
[13] PARK W, KIM D, LU Y, et al. Relational knowledge distillation[C] //2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach, USA:IEEE, 2019: 3962-3971.
[14] CHEN M H, LAINA I, VEDALDI A. Training-free layout control with cross-attention guidance[C] //2024 IEEE/CVF Winter Conference on Applications of Computer Vision(WACV). Waikoloa, Hawaii, USA:IEEE, 2024: 5331-5341.
[15] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C] //2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City, USA:IEEE, 2018: 7132-7141.
[16] WANG Q L, WU B G, ZHU P F, et al. ECA-net: efficient channel attention for deep convolutional neural networks[C] //2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). 2020. Seattle, WA, USA:IEEE, 2020: 13-19.
[17] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C] // Computer Vision-ECCV 2018. Munich,Germany: Springer International Publishing, 2018: 3-19.
[18] LI X, WANG W H, HU X L, et al. Selective kernel networks[C] //2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach,USA: IEEE, 2019: 510-519.
[19] DENG J, DONG W, SOCHER R, et al. ImageNet: a large-scale hierarchical image database[C] //2009 IEEE Conference on Computer Vision and Pattern Recognition. Miami, USA:IEEE, 2009: 248-255.
[20] BERGMANN P, FAUSER M, SATTLEGGER D, et al. MVTec AD: a comprehensive real-world dataset for unsupervised anomaly detection[C] //2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach, USA:IEEE, 2019: 9592-9600.
[21] HENDRYCKS D, GIMPEL K, NOVELLO P, et al. A baseline for detecting misclassified and out-of-distribution examples in neural networks[EB/OL].(2016-10-07)[2024-03-12]. https://arxiv.org/abs/1610.02136v3
[22] SABOKROU M, POURREZA M, FAYYAZ M, et al. AVID: adversarial visual irregularity detection[C] // Computer Vision-ACCV 2018. Perth, Australia: Springer International Publishing, 2019: 488-505.
[23] SCHLEGL T, SEEBÖCK P, WALDSTEIN S M, et al. Unsupervised anomaly detection with generative adversarial networks to guide marker discovery[C] // Information Processing in Medical Imaging. Orlando, USA: Springer International Publishing, 2017: 146-157.
[24] BERGMANN P, LÖWE S, FAUSER M, et al. Improving unsupervised defect segmentation by applying structural similarity to autoencoders[EB/OL].(2018-07-05)[2024-03-12]. https://arxiv.org/abs/1807.02011v3
[25] RUFF L, VANDERMEULEN R, GOERNITZ N, et al. Deep one-class classification[C] // Proceedings of the international conference on machine learning. Stockholm, Sweden: ICML, 2018: 4393-4402.
[26] DEHAENE D, FRIGO O, COMBREXELLE S, et al. Iterative energy-based projection on a normal data manifold for anomaly localization[EB/OL].(2020-02-10)[2024-03-12]. https://arxiv.org/abs/2002.03734v1
[27] SALEHI M, SADJADI N, BASELIZADEH S, et al. Multiresolution knowledge distillation for anomaly detection[C] //2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Nashville, USA: IEEE, 2021: 14902-14912.
[28] NAPOLETANO P, PICCOLI F, SCHETTINI R. Anomaly detection in nanofibrous materials by CNN-based self-similarity[J]. Sensors, 2018, 18(1): 209-224.
[29] DAI Y M, GIESEKE F, OEHMCKE S, et al. Attentional feature fusion[C] //2021 IEEE Winter Conference on Applications of Computer Vision(WACV). Waikoloa, Hawaii, USA:IEEE, 2021: 3560-3569.

相关文章 15

[1]	李丰,文益民. 融合多尺度视觉和文本语义特征的图像描述生成算法[J]. 山东大学学报 (工学版), 2025, 55(3): 80-87.
[2]	董明书,陈俐企,马川义,张珠皓,孙仁娟,管延华,庄培芝. 沥青路面内部裂缝雷达图像智能判识算法研究[J]. 山东大学学报 (工学版), 2025, 55(3): 72-79.
[3]	薛冰冰,王勇,杨维浩,王川,于迪,王旭. 基于ETC收费数据的高速公路交通流数据修复及实时预测[J]. 山东大学学报 (工学版), 2025, 55(3): 58-71.
[4]	王禹鸥,苑迎春,何振学,王克俭. 改进RoBERTa、多实例学习和双重注意力机制的关系抽取方法[J]. 山东大学学报 (工学版), 2025, 55(2): 78-87.
[5]	邹正标,刘毅志,廖祝华,赵肄江. 动态交通流量预测的时空注意力图卷积网络[J]. 山东大学学报 (工学版), 2024, 54(5): 50-61.
[6]	常新功,苏敏惠,周志刚. 基于进化集成的图神经网络解释方法[J]. 山东大学学报 (工学版), 2024, 54(4): 1-12.
[7]	索大翔,李波. 基于Gromov-Wasserstein最优传输的输电线路小目标检测方法[J]. 山东大学学报 (工学版), 2024, 54(3): 22-29.
[8]	宋辉,张轶哲,张功萱,孟元. 基于类权重和最小化预测熵的测试时集成方法[J]. 山东大学学报 (工学版), 2024, 54(3): 36-43.
[9]	聂秀山,巩蕊,董飞,郭杰,马玉玲. 短视频场景分类方法综述[J]. 山东大学学报 (工学版), 2024, 54(3): 1-11.
[10]	刘新,刘冬兰,付婷,王勇,常英贤,姚洪磊,罗昕,王睿,张昊. 基于联邦学习的时间序列预测算法[J]. 山东大学学报 (工学版), 2024, 54(3): 55-63.
[11]	高泽文,王建,魏本征. 基于混合偏移轴向自注意力机制的脑胶质瘤分割算法[J]. 山东大学学报 (工学版), 2024, 54(2): 80-89.
[12]	李璐,张志军,范钰敏,王星,袁卫华. 面向冷启动用户的元学习与图转移学习序列推荐[J]. 山东大学学报 (工学版), 2024, 54(2): 69-79.
[13]	陈成,董永权,贾瑞,刘源. 基于交互序列特征相关性的可解释知识追踪[J]. 山东大学学报 (工学版), 2024, 54(1): 100-108.
[14]	李家春,李博文,常建波. 一种高效且轻量的RGB单帧人脸反欺诈模型[J]. 山东大学学报 (工学版), 2023, 53(6): 1-7.
[15]	王旭晴,魏伟波,杨光宇,宋金涛,吕婷,潘振宽. 基于算法展开的图像盲去模糊深度学习网络[J]. 山东大学学报 (工学版), 2023, 53(6): 35-46.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed