目的 旨在提升印刷品质量检测的精度与效率,提出一种基于改进卷积神经网络（CNN）的小样本学习方法。通过改进CNN模型的自动特征提取能力和小样本泛化能力,缓解传统人工检测和机器视觉方法中人工特征设计复杂、缺陷形态适应性不足等问题。方法 本文采用改进的CNN模型,针对印刷品图像的细粒度特征提取和小物体缺陷检测进行优化。通过对网络结构进行调整,并结合多种图像处理技术,提高模型在复杂印刷环境下的检测能力。实验中使用基于小样本学习的技术,进一步增强模型的泛化能力和准确率。结果 实验结果表明,改进后的CNN模型在印刷品缺陷检测任务中的准确率、精确率、召回率和F1分数分别达到98.8%、98.5%、97.1%和98.8%,其中准确率较ResNet-18模型的92.5%提高了6.3个百分点,表明该模型在小尺度缺陷和颜色异常识别方面具有较好的检测效果。结论 本文通过优化CNN模型,显著提升了印刷品质量检测系统的性能,为深度学习在工业质量检测中的应用提供了技术支持,并展现了在实际应用中的巨大潜力。
Abstract
The work aims to propose a print-defect recognition method based on an improved convolutional neural network (CNN) for limited-sample scenarios to improve the precision and efficiency of print quality inspection, and alleviate the problems of complex manual feature engineering and insufficient adaptability to diverse defect morphologies in manual inspection and conventional machine-vision methods by improving the automatic feature extraction ability and limited-sample generalization ability of the CNN model. An improved CNN model was designed for fine-grained feature extraction and small-scale defect recognition in printed-product images. By optimizing the network structure and incorporating multiple image processing techniques, the model was enhanced to improve its robustness and generalization capability under complex printing conditions. Experimental results showed that the improved CNN model achieved an accuracy of 98.8%, a precision of 98.5%, a recall of 97.1%, and an F1-score of 97.8% in the print-defect recognition task. Compared with the ResNet-18 model, which achieved an accuracy of 92.5%, the proposed model improved the accuracy by 6.3 percentage points, demonstrating better recognition performance for small-scale defects and color anomalies. In conclusion, the proposed improved CNN model significantly enhances the performance of print quality inspection, provides an effective technical approach for applying deep learning to industrial quality inspection, and demonstrates great potential in practical applications.
关键词
深度学习 /
卷积神经网络 /
印品缺陷识别 /
小样本
Key words
deep learning /
convolutional neural networks /
print defect recognition /
small sample
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 王军豪. 基于图像处理的印刷品颜色和形状缺陷检测方法研究[D]. 天津: 天津理工大学, 2025.
WANG J H.Research on Detection Method of Color and Shape Defects of Printed Matter Based on Image Processing[D]. Tianjin: Tianjin University of Technology, 2025.
[2] 薛意达. 基于深度学习的印刷纸箱缺陷检测方法研究[D]. 西安: 西安理工大学, 2023: 1-69.
XUE Y D.Research on Defect Detection Method for Printed Cartons Based on Deep Learning[D]. Xi’an: Xi’an University of Technology, 2023: 1-69.
[3] LECUN Y, BOTTOU L, BENGIO Y, et al.Gradient-Based Learning Applied to Document Recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324.
[4] ZHANG E, LI B, LI P, et al.A deep learning based printing defect classification method with imbalanced samples[J]. Symmetry, 2019, 11(12): 1440.
[5] CHEN F Y, FU L, ZHANG Y Q, et al.A Review of Deep Learning-Based Steel Surface Defect Detection[J]. Academic Journal of Science and Technology, 2025, 15(1): 198-202.
[6] MA Y X, YIN J X, HUANG F, et al.Surface Defect Inspection of Industrial Products with Object Detection Deep Networks: A Systematic Review[J]. Artificial Intelligence Review, 2024, 57(12): 333.
[7] KIM Y, IM Y J, VU T T H, et al. Real-Time Detection of Printing Defects with YOLOv5 Models[J]. ICIC Express Letters, Part B: Applications, 2024, 15(3): 271-277.
[8] 王文成, 郑诗翰, 余智科. 基于改进Canny算法的印刷品表面缺陷检测[J]. 现代电子技术, 2025, 48(24): 159-163.
WANG W C, ZHENG S H, YU Z K.Printed Surface Defect Detection Based on Improved Canny Algorithm[J]. Modern Electronic Technique, 2025, 48(24): 159-163.
[9] ZHAO J, KONG L L, LV J C.An Overview of Deep Neural Networks for Few-Shot Learning[J]. Big Data Mining and Analytics, 2025, 8(1): 145-188.
[10] NAIR V, HINTON G E.Rectified linear units improve restricted Boltzmann machines[C]//Proceedings of the 27th International Conference on Machine Learning. Haifa: Omnipress, 2010: 807-814.
[11] HE K, ZHANG X, REN S, 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.
[12] DUBEY S R, SINGH S K, CHAUDHURI B B.Activation Functions in Deep Learning: A Comprehensive Survey and Benchmark[J]. arXiv preprint arXiv:2109. 14545, 2021.
[13] GOODFELLOW I, BENGIO Y, COURVILLE A.Deep Learning[M]. Cambridge: MIT Press, 2016: 75-78.
[14] LECUN Y, BENGIO Y, HINTON G.Deep Learning[J]. Nature, 2015, 521(7553): 436-444.
[15] SMITH L N. Cyclical Learning Rates for Training Neural Networks[PP/OL]. V6. arXiv (2017-04-04). https://doi.org/10.48550/arXiv.1506.01186.
[16] 张昕玥, 陈勇明, 郭俊, 等. 基于卷积神经网络的水稻虫害图像识别[J]. 人工智能与机器人研究, 2025, 14(1): 30-42.
ZHANG X Y, CHEN Y M, GUO J, et al.Pest Image Recognition in Rice Using Convolutional Neural Networks[J]. Artificial Intelligence and Robotics Research, 2025, 14(1): 30-42.
[17] POWERS D M W. Evaluation: From Precision, Recall and F-Measure to ROC, Informedness, Markedness and Correlation[J]. Journal of Machine Learning Technologies, 2011, 2(1): 37-63.
[18] SIMONYAN K, ZISSERMAN A.Very Deep Convolutional Networks for Large-Scale Image Recognition[J]. International Conference on Machine Learning, 2014, 32: 1-14.
[19] HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely Connected Convolutional Networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI. IEEE, 2017: 2261-2269.
基金
北京印刷学院学科建设和研究生教育专项研究生课程项目（21090225001）;文化科技融合背景下的机械工程学科建设（21090126013）
PDF(3695 KB)
渝公网安备 50010702501716号 渝ICP备15012534号-2
