Surface Defect Detection of Aerospace Sealing Materials Based on Few-shot Deep Learning

ZHANG Yongqian; ZHU Bin; ZHENG Wen; LIU Zhao; KONG Weiping; SHEN Jiahe

doi:10.19554/j.cnki.1001-3563.2026.01.033

PDF(7796 KB)

Packaging Engineering ›› 2026, Vol. 47 ›› Issue (1) : 283-291. DOI: 10.19554/j.cnki.1001-3563.2026.01.033

Defense Equipment

Surface Defect Detection of Aerospace Sealing Materials Based on Few-shot Deep Learning

ZHANG Yongqian¹, ZHU Bin¹, ZHENG Wen², LIU Zhao^1,*, KONG Weiping¹, SHEN Jiahe¹

Author information +

History +

Abstract

The work aims to propose a surface defect detection model based on few-shot deep learning to address the problems of small sample size and difficulty in identification of aerospace sealing materials. By collecting and preprocessing the sample images of aerospace sealing rings, a dataset suitable for unsupervised algorithms was constructed. On the basis of the RegAD algorithm, a surface defect detection model for sealing materials, named RegAD-RS, was proposed by combining the Residual Network (ResNet) with the Spatial Transformer Network (STN). The model was trained under the conditions that the sample support set contained 2, 4 and 8 normal samples respectively. The results showed that the model could accurately identify the surface defects of materials, with the performance index AUROC (Area Under the Receiver Operating Characteristic Curve) all above 90% and the highest reaching 97.29%, demonstrating strong detection advantages. It improves the detection accuracy of surface defects of special aerospace sealing materials and can effectively ensure the quality and safety of aerospace precision industrial products.

Key words

sealing materials / few-shot / unsupervised learning / defect detection

Cite this article

EndNote

Ris (Procite)

Bibtex

Download Citations

ZHANG Yongqian, ZHU Bin, ZHENG Wen, LIU Zhao, KONG Weiping, SHEN Jiahe. Surface Defect Detection of Aerospace Sealing Materials Based on Few-shot Deep Learning[J]. Packaging Engineering. 2026, 47(1): 283-291 https://doi.org/10.19554/j.cnki.1001-3563.2026.01.033

References

[1] 张薇, 苑改红, 周天一, 等. 基于机器学习的航天装备金属表面缺陷检测方法综述[J]. 航天电子对抗, 2025, 41(3): 43-50.
[2] ZHANG W, YUAN G H, ZHOU T Y, et al.Review of Metal Surface Defect Detection Methods for Aerospace Equipment Based on Machine Learning[J]. Aerospace Electronic Warfare, 2025, 41(3): 43-50.
[3] 侯春佳, 何博侠, 胡金松, 等. 微小型航天密封圈表面缺陷检测[J]. 光子学报, 2024, 53(3): 151-159.
HOU C J, HE B X, HU J S, et al.Surface Defect Detection of Miniature Aerospace Sealing Ring[J]. Acta Photonica Sinica, 2024, 53(3): 151-159.
[4] 葛爱明, 闫峥, 李兴隆, 等. 高光谱遥感图像高分辨率弱小目标检测技术[J]. 飞控与探测, 2025(3): 82-88.
GE A M, YAN Z, LI X L, et al.High Resolution Small Target Detection Technology in Hyperspectral Remote Sensing Images[J]. Flight Control & Detection, 2025(3): 82-88.
[5] TABERNIK D, ŠELA S, SKVARČ J, et al.Segmentation-Based Deep-Learning Approach for Surface-Defect Detection[J]. Journal of Intelligent Manufacturing, 2020, 31(3): 759-776.
[6] HUANG Y B, QIU C Y, YUAN K.Surface Defect Saliency of Magnetic Tile[J]. The Visual Computer, 2020, 36(1): 85-96.
[7] RUDOLPH M, WANDT B, ROSENHAHN B.Same Same but DifferNet: Semi-Supervised Defect Detection with Normalizing Flows[C]// 2021 IEEE Winter Conference on Applications of Computer Vision (WACV). Waikoloa, HI, USA. IEEE, 2021: 1906-1915.
[8] TAO X, GONG X Y, ZHANG X, et al.Deep Learning for Unsupervised Anomaly Localization in Industrial Images: A Survey[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 5018021.
[9] YOUKACHEN S, RUCHANURUCKS M, PHATRAPOMNANT T, et al.Defect Segmentation of Hot-Rolled Steel Strip Surface by Using Convolutional Auto-Encoder and Conventional Image Processing[C]// 2019 10th International Conference of Information and Communication Technology for Embedded Systems (IC-ICTES). Bangkok, Thailand. IEEE, 2019: 1-5.
[10] WU J C, CHEN D J, FUH C S, et al.Learning Unsupervised Metaformer for Anomaly Detection[C]// 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, QC, Canada. IEEE, 2022: 4349-4358.
[11] HUANG C Q, GUAN H Y, JIANG A F, et al.Registration Based Few-Shot Anomaly Detection[C]// Computer Vision - ECCV 2022. Cham: Springer, 2022: 303-319.
[12] HU B, ZHOU N Y, ZHOU Q, et al.DiffNet: A Learning to Compare Deep Network for Product Recognition[J]. IEEE Access, 2020, 819336-19344.
[13] BAI G J, LING C, ZHAO L. Temporal Domain Generalization with Drift-Aware Dynamic Neural Networks[EB/OL].2022: arXiv: 2205.10664. https://arxiv.org/abs/2205.10664
[14] BERTINETTO L, VALMADRE J, HENRIQUES F J, et al.Fully-Convolutional Siamese Networks for Object Tracking[J]. CoRR, 2016, 3: 850-865.
[15] HE K M, ZHANG X Y, REN S Q, et al.Deep Residual Learning for Image Recognition[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA. IEEE, 2016: 770-778.
[16] JADERBERG M, SIMONYAN K, ZISSERMAN A, et al.Spatial Transformer Networks[J]. MIT Press, 2015. DOI: 10.48550/arXiv.1506.02025.
[17] QUAN Y H, CHEN M Q, PANG T Y, et al.Self2Self with Dropout: Learning Self-Supervised Denoising from Single Image[C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA, USA. IEEE, 2020: 1887-1895.
[18] SHAHAM T R, DEKEL T, MICHAELI T.SinGAN: Learning a Generative Model from a Single Natural Image[C]// 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, Korea. IEEE, 2019: 4569-4579.
[19] ZHU D X, WU X D, YANG T B. Benchmarking Deep AUROC Optimization: Loss Functions and Algorithmic Choices[EB/OL].2022: arXiv: 2203.14177. https://arxiv.org/abs/2203.14177
[20] 王津申, 李阳, 王清峰, 等. 高光谱图像序列中的运动弱小目标检测[J]. 飞控与探测, 2019, 2(3): 85-89.
WANG J S, LI Y, WANG Q F, et al.Detection of Moving Dim Targets in Hyperspectral Image Sequences[J]. Flight Control & Detection, 2019, 2(3): 85-89.