基于Transformer的生成对抗网络的柔印标签在线检测

doi:10.3778/j.issn.1002-8331.2403-0106

摘要/Abstract

摘要： 为了及时把控柔印标签的生产质量，融合Transformer设计了一种生成对抗网络，提出了一种柔印标签在线检测方法。为了解决实际生产中缺陷样品稀少、样本分布不均的问题，提出了一种新颖的噪声添加方案模拟缺陷样本，仅使用合格的柔印标签样本训练模板生成器。结合跳连接和Transformer block设计了一种生成对抗网络，以及相应的损失函数，以提高生成器对模板的表示能力。设计了一种基于自适应阈值的缺陷评估方案实现柔印标签检测。实验结果表明，提出的检测方法可以在每个样本38 ms的合理检测时间下，达到2.23%平均误检率、0%平均漏检率、0.983 F1-score的检测性能，在数据集上优于现有异常检测深度学习方法。

关键词: 图像处理, 缺陷检测, 生成对抗网络, 柔印标签, 随机噪声

Abstract: To control the production quality of flexographic labels in a timely manner, this paper integrates Transformer architecture to design a generative adversarial network and proposes an online detection method for flexographic labels. To address the issue of rare defective samples and uneven sample distribution in actual production, the paper introduces a novel noise addition scheme to simulate defective samples, training the template generator solely with qualified flexographic label samples. A generative adversarial network is designed by combining skip connections and Transformer blocks, along with corresponding loss functions, to enhance the generator’s capability to represent templates. Lastly, a defect evaluation scheme based on adaptive thresholds is devised for the detection of flexographic labels. Experimental results indicate that the detection method proposed in this paper achieves a detection performance with an average false positive rate of 2.23%, an average false negative rate of 0%, and an F1-score of 0.983 within a reasonable detection time of 38 ms per sample, outperforming existing anomaly detection deep learning methods on this dataset.

Key words: image processing, defect detection, generative adversarial network, flexographic printing labels, random noise

龙进良, 蓝学深, 蔡念, 燕舒乐, 肖盼, 许少秋, 周映红. 基于Transformer的生成对抗网络的柔印标签在线检测[J]. 计算机工程与应用, 2025, 61(11): 306-315.

LONG Jinliang, LAN Xueshen, CAI Nian, YAN Shule, XIAO Pan, XU Shaoqiu, ZHOU Yinghong. Online Detection of Flexographic Printing Labels Based on Transformer-Based Generative Adversarial Network[J]. Computer Engineering and Applications, 2025, 61(11): 306-315.

参考文献

[1] COSNAHAN T, WATT A A, ASSENDER H E J M T P. Flexography printing for organic thin film transistors[J]. Materials Today: Proceedings, 2018, 5(8): 16051-16057.
[2] LORENZ A, SENNE A, ROHDE J, et al. Evaluation of flexographic printing technology for multi-busbar solar cells[J]. Energy Procedia, 2015, 67: 126-137.
[3] YU T, ZENG S, LUO Y, DING Y. Application research of online label defect detection based on machine vision[C]//Proceedings of the IEEE 6th Information Technology and Mechatronics Engineering Conference, 2022: 1398-1402.
[4] KHAN M A, AHMED F, KHAN M D, et al. A smart and robust automatic inspection of printed labels using an image hashing technique[J]. Electronics, 2022, 11(6): 955.
[5] LI D, LI J, FAN Y, et al. Printed label defect detection using twice gradient matching based on improved cosine similarity measure[J]. Expert Systems with Applications, 2022, 204: 117372.
[6] JIA L, CHEN C, LIANG J, et al. Fabric defect inspection based on lattice segmentation and Gabor filtering[J]. Neurocomputing, 2017, 238: 84-102.
[7] HAMDI A A, SAYED M S, FOUAD M M, et al. Unsupervised patterned fabric defect detection using texture filtering and K-means clustering[C]//Proceedings of the International Conference on Innovative Trends in Computer Engineering, 2018: 130-144.
[8] LI M, WAN S, DENG Z, et al. Fabric defect detection based on saliency histogram features[J]. Computational Intelligence, 2019, 35(3): 517-534.
[9] WU J, LE J, XIAO Z, et al. Automatic fabric defect detection using a wide-and-light network[J]. Applied Intelligence, 2021, 51(7): 4945-4961.
[10] LIU Q, WANG C, LI Y, et al. A fabric defect detection method based on deep learning[J]. IEEE Access, 2022, (10): 4284-4296.
[11] CHEN M, YU L, 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.
[12] LIU Z, HUO Z, LI C, et al. DLSE-Net: a robust weakly supervised network for fabric defect detection[J]. Displays, 2021, 68: 102008.
[13] COHEN N, HOSHEN Y J A P A. Sub-image anomaly detection with deep pyramid correspondences[J]. arXiv:2005. 02357, 2020.
[14] LEE S, LEE S, SONG B C J I A. CFA: coupled-hypersphere-based feature adaptation for target-oriented anomaly localization[J]. IEEE Access, 2022, 10: 78446-78454.
[15] DEFARD T, SETKOV A, LOESCH A, et al. Padim: a patch distribution modeling framework for anomaly detection and localization[J]. arXiv:2011.08785, 2020.
[16] RUDOLPH M, WEHRBEIN T, ROSENHAHN B, et al. Fully convolutional cross-scale-flows for image-based defect detection[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2022: 1088-1097.
[17] ZHOU Y X, XU X, SONG J K, et al. MSFlow: multiscale flow-based framework for unsupervised anomaly detection[J]. IEEE Transactions on Neural Networks and Learning Systems, 2024.
[18] YU J, ZHENG Y, WANG X, et al. Fastflow: unsupervised anomaly detection and localization via 2D normalizing flows [J]. arXiv:211107677, 2021.
[19] AKCAY S, ATAPOUR-ABARGHOUEI A, BRECKON T P. Ganomaly: semi-supervised anomaly detection via adversarial training[C]//Proceedings of the 14th Asian Conference on Computer Vision, 2019: 622-637.
[20] MOUSAKHAN A, BROX T, TAYYUB J J A. Anomaly detection with conditioned denoising diffusion models[J]. arXiv:2305.15956, 2023.
[21] PENG J, SHAO H, XIAO Y, et al. Industrial surface defect detection and localization using multi-scale information focusing and enhancement GANomaly[J]. Expert Systems with Applications, 2024, 238: 122361.
[22] CHEN K, CAI N, WU Z, et al. Multi-scale GAN with transformer for surface defect inspection of IC metal packages[J]. Expert Systems with Applications, 2023, 212: 118788.
[23] WU Z, CAI N, CHEN K, et al. GAN-based statistical modeling with adaptive schemes for surface defect inspection of IC metal packages[J]. Journal of Intelligent Manufacturing, 2024, 35: 1811-1124.
[24] ZHENG Q H, ZHAO P H, ZHANG D L, et al. MR‐DCAE: manifold. regularization‐based deep convolutional autoencoder for unauthorized broadcasting identification[J]. International Journal of Intelligent Systems, 2021, 36(12): 7204-7238.
[25] XIAO P, YAN S, LONG J, et al. An adaptive coarse-to-fine framework for automatic first article inspection of flexographic printing labels[J]. Expert Systems with Applications, 2023, 227: 120241.
[26] HUBER P J J B I S M, distribution. Robust estimation of a location parameter[J]. Breakthroughs in Statistics: Methodology and Distribution, 1992: 492-518.
[27] WANG Z, BOVIK A C, SHEIKH H R, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004, 13(4): 600-612.
[28] JIANG Y, CHANG S, WANG Z. TransGAN: two pure transformers can make one strong gan, and that can scale up[J]. arXiv:2102.07074, 2021.