Industrial Defect Detection Algorithm with Knowledge Distillation Integrating Multi-Scale Features

doi:10.3778/j.issn.1002-8331.2407-0518

Abstract

Abstract: With the improvement of automation level in industrial production lines, enterprises have increasingly strict requirements for product quality, and defect detection has become an important task in assisting automated production. However, due to the complex structure of some industrial products, various unknown types of defects may occur during the production process, and it is difficult to obtain defect samples, which makes industrial defect detection still challenging. In order to improve the detection efficiency of industrial defects in structural types, a knowledge distillation industry defect detection algorithm MSFFKD_DD is proposed, which integrates multi-scale features. Firstly, the artificial synthesis anomaly module SAM_AB is proposed to generate pseudo defect sample images using normal sample images, simulate unknown types of defects, and introduce pseudo defect features in the knowledge distillation process to enhance the defect detection capability of the algorithm. Then, a feature fusion module MSFFM is designed to enhance the ability of the algorithm to extract product detail features by fusing shallow and deep features. At the same time, SSIM loss is introduced into the loss function to improve the segmentation accuracy of industrial product defects with complex structures. Experiments are conducted on the industrial dataset MVTec AD, and the image level AUROC, pixel level AUROC, and AUCPRO of the MSFFKD_DD algorithm reach 98.4%, 97.2%, and 95.1%, respectively, effectively improving the accuracy and segmentation precision of industrial defect detection for structural types.

Key words: defect detection, industrial detect, multi-scale feature, knowledge distillation

摘要： 随着工业生产线自动化程度的提高，企业对产品质量的要求逐渐严格，缺陷检测成为辅助自动化生产的一项重要任务。然而由于部分工业产品结构复杂，生产过程中可能出现各种未知类型的缺陷，且缺陷样本难以获取，使得工业缺陷检测仍具有一定的挑战性。为了提高结构类型工业缺陷的检测效果，提出融合多尺度特征的知识蒸馏工业缺陷检测算法MSFFKD_DD。提出人工合成异常模块SAM_AB，使用正常样本图像生成伪缺陷样本图像，模拟未知类型的缺陷，在知识蒸馏过程中引入伪缺陷特征，使算法具备更强的缺陷检测能力；设计特征融合模块MSFFM，通过融合浅层特征和深层特征，增强算法对产品细节特征的提取能力，同时在损失函数中引入SSIM损失，提高对具有复杂结构工业产品缺陷的分割精度。在工业数据集MVTec AD上进行实验，MSFFKD_DD算法的图像级AUROC、像素级AUROC和AUCPRO分别达到98.4%、97.2%和95.1%，有效提升了结构类型工业缺陷检测的准确率和分割精度。

关键词: 缺陷检测, 工业缺陷, 多尺度特征, 知识蒸馏

WANG Ling, WANG Minghui, WANG Peng, BAI Yan’e. Industrial Defect Detection Algorithm with Knowledge Distillation Integrating Multi-Scale Features[J]. Computer Engineering and Applications, 2025, 61(21): 276-286.

王玲, 王明慧, 王鹏, 白燕娥. 融合多尺度特征的知识蒸馏工业缺陷检测算法[J]. 计算机工程与应用, 2025, 61(21): 276-286.

References

[1] 罗东亮, 蔡雨萱, 杨子豪, 等. 工业缺陷检测深度学习方法综述[J]. 中国科学: 信息科学, 2022, 52(6): 1002-1039.
LUO D L, CAI Y X, YANG Z H, et al. Survey on industrial defect detection with deep learning[J]. Scientia Sinica (Informationis), 2022, 52(6): 1002-1039.
[2] COLLIN A S, DE VLEESCHOUWER C. Improved anomaly detection by training an autoencoder with skip connections on images corrupted with Stain-shaped noise[C]//Proceedings of the 25th International Conference on Pattern Recognition. Piscataway: IEEE, 2021: 7915-7922.
[3] YAN X D, ZHANG H D, XU X M, et al. Learning semantic context from normal samples for unsupervised anomaly detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2021: 3110-3118.
[4] DEFARD T, SETKOV A, LOESCH A, et al. PaDiM: a patch distribution modeling framework for anomaly detection and localization[C]//Proceedings of International Conference on Pattern Recognition. Cham: Springer, 2021: 475-489.
[5] RUDOLPH M, WANDT B, ROSENHAHN B. Same same but DifferNet: semi-supervised defect detection with normalizing flows[C]//Proceedings of the 2021 IEEE Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2021: 1906-1915.
[6] ZAVRTANIK V, KRISTAN M, SKO?AJ D. DRAEM-a discriminatively trained reconstruction embedding for surface anomaly detection[C]//Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2022: 8310-8319.
[7] WANG W Y, MI C F, WU Z H, et al. A real-time steel surface defect detection approach with high accuracy[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 5005610.
[8] 张涛源, 谢新林, 谢刚, 等. 融合Transformer的带钢缺陷实时检测算法[J]. 计算机工程与应用, 2023, 59(16): 232-239.
ZHANG T Y, XIE X L, XIE G, et al. Real-time strip steel defect detection algorithm fused with transformer[J]. Computer Engineering and Applications, 2023, 59(16): 232-239.
[9] 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.
[10] BERGMANN P, FAUSER M, SATTLEGGER D, et al. Uninformed students: student-teacher anomaly detection with discriminative latent embeddings[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 4182-4191.
[11] CAO Y K, XU X H, LIU Z G, et al. Collaborative discrepancy optimization for reliable image anomaly localization[J]. IEEE Transactions on Industrial Informatics, 2023, 19(11): 10674-10683.
[12] XIAO Q F, WANG J, LIN Y F, et al. Unsupervised anomaly detection with distillated teacher-student network ensemble[J]. Entropy, 2021, 23(2): 201.
[13] SALEHI M, SADJADI N, BASELIZADEH S, et al. Multiresolution knowledge distillation for anomaly detection[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 14897-14907.
[14] 刘涛. 基于深度学习的磁瓦在线检测技术研究[D]. 杭州: 浙江大学, 2021.
LIU T. Research on online inspection technology of magnetic tiles based on deep learning[D]. Hangzhou: Zhejiang University, 2021.
[15] WANG G D, HAN S M, DING E R, et al. Student-teacher feature pyramid matching for anomaly detection[C]//Proceedings of the British Machine Vision Conference, 2021: 349.
[16] 程泽平. 基于正常样本学习的图像缺陷检测算法[D]. 北京: 北京邮电大学, 2022.
CHENG Z P. The image defect detection algorithm based on normal sample learning[D]. Beijing: Beijing University of Posts and Telecommunications, 2022.
[17] DENG H Q, LI X Y. Anomaly detection via reverse distillation from one-class embedding[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 9727-9736.
[18] RUDOLPH M, WEHRBEIN T, ROSENHAHN B, et al. Asymmetric student-teacher networks for industrial anomaly detection[C]//Proceedings of the 2023 IEEE/CVF Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2023: 2591-2601.
[19] 杨珂浩, 于龙, 高仕斌, 等. 基于无监督知识蒸馏的高铁绝缘子缺陷检测算法[J]. 电气化铁道, 2023, 34(1): 9-14.
YANG K H, YU L, GAO S B, et al. An algorithm for detection of insulator defects for high-speed railway based on unsupervised knowledge distillation[J]. Electric Railway, 2023, 34(1): 9-14.
[20] BATZNER K, HECKLER L, K?NIG R. EfficientAD: accurate visual anomaly detection at millisecond-level latencies[C]//Proceedings of the 2024 IEEE/CVF Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2024: 127-137.
[21] ZHANG X, LI S Y, LI X, et al. DeSTSeg: segmentation guided denoising student-teacher for anomaly detection[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 3914-3923.
[22] REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 6517-6525.
[23] 高春艳, 秦燊, 李满宏, 等. 改进YOLOv7算法的钢材表面缺陷检测研究[J]. 计算机工程与应用, 2024, 60(7): 282-291.
GAO C Y, QIN S, LI M H, et al. Research on steel surface defect detection with improved YOLOv7 algorithm[J]. Computer Engineering and Applications, 2024, 60(7): 282-291.
[24] CARION N, MASSA F, SYNNAEVE G, et al. End-to-end object detection with transformers[J]. arXiv:2005.12872, 2020.
[25] ZHENG Q H, TIAN X Y, YU Z G, et al. MobileRaT: a lightweight radio transformer method for automatic modulation classification in drone communication systems[J]. Drones, 2023, 7(10): 596.
[26] ZHENG Q H, WANG R Y, TIAN X Y, et al. A real-time transformer discharge pattern recognition method based on CNN-LSTM driven by few-shot learning[J]. Electric Power Systems Research, 2023, 219: 109241.
[27] 高琦, 付皓宇, 吴晓军, 等. 基于无监督学习的连铸铸坯缺陷检测方法[J]. 重型机械, 2024(3): 32-40.
GAO Q, FU H Y, WU X J, et al. Defect detection method for continuous casting billets based on unsupervised learning[J]. Heavy Machinery, 2024(3): 32-40.
[28] HINTON G, VINYALS O, DEAN J. Distilling the knowledge in a neural network[J]. arXiv:1503.02531, 2015.
[29] WANG Z, JING J F. Pixel-wise fabric defect detection by CNNs without labeled training data[J]. IEEE Access, 2020, 8: 161317-161325.
[30] TIEN T D, NGUYEN A T, TRAN N H, et al. Revisiting reverse distillation for anomaly detection[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 24511-24520.