基于深度卷积神经网络的织物瑕疵识别方法
Fabric Defect Recognition Method Based on Deep Convolution Neural Network
DOI: 10.12677/CSA.2021.112035, PDF,  被引量    国家自然科学基金支持
作者: 张卷卷*：中国移动通信集团浙江有限公司，浙江 杭州；董 骅：中国信息通信研究院，北京；朱 涛：杭州电子科技大学自动化学院，浙江 杭州
关键词: 卷积神经网络；深度学习；滑动窗口；瑕疵识别；Convolution Neural Network； Deep Learning； Sliding Window； Fabric Quality Recognition
摘要: 针对传统基于图像特征提取的瑕疵检测方法过于依赖特征提取效果，且泛化能力较差以及人工质检存在的效率低、易受主观因素影响等问题，本文提出了一种基于深度卷积神经网络的图像瑕疵识别方法，基于ResNet50卷积神经网络，构建了分类模型。并增大了输入网络的图像尺寸；采用多种图像变换增强数据；修改损失函数让模型更加关注困难样本。该模型在测试集上的AUC (Area Under Curve)值可以达到0.905，同时F1分数达到了0.81。此外本文提出了一种基于滑动窗口检测的瑕疵识别方法，提高对图像中细节的关注，大幅提升了原模型的分类性能。
Abstract: Aiming at the problems of traditional defect detection methods based on image feature extraction, such as too much dependence on the effect of feature extraction, poor generalization ability, low efficiency of artificial quality inspection and vulnerability to subjective factors, an image defect recognition method based on deep convolution neural network is proposed. Based on ResNet50 convolution neural network, the classification model is constructed, and the size of the input network image is increased, various image transformations are used to enhance the data, and the loss function is modified to make the model pay more attention to the difficult samples. The AUC (Area Under Curve) value of the model on the test set can reach 0.905, and the F1 score can reach 0.81. In addition, a defect recognition method based on sliding window detection is proposed in this paper, which can improve the attention of image details and greatly improve the classification performance of the original model.
文章引用：张卷卷, 董骅, 朱涛. 基于深度卷积神经网络的织物瑕疵识别方法[J]. 计算机科学与应用, 2021, 11(2): 344-355. https://doi.org/10.12677/CSA.2021.112035

参考文献

[1] Kumar, A. (2008) Computer-Vision-Based Fabric Defect Detection: A Survey. IEEE Transactions on Industrial Elec-tronics, 55, 348-363.
[2] 李精忠, 张津铭. 一种基于傅里叶变换的光滑边界面状要素Morphing方法[J]. 武汉大学学报•信息科学版, 2017, 42(8): 1104-1109.
[3] 吴亚新. 基于灰度共生矩阵的纺织品瑕疵识别算法[J]. 电脑知识与技术, 2017, 13(28): 204-205.
[4] 张靖, 周明全, 张雨禾, 等. 基于马尔科夫随机场的散乱点云全局特征提取[J]. 自动化学报, 2016, 42(7): 1090- 1099.
[5] Hamdi, A.A., Sayed, M.S., Fouad, M.M., et al. (2018) Unsuper-vised Patterned Fabric Defect Detection Using Texture Filtering and K-Means Clustering. IEEE International Conference on Innovative Trends in Computer Engineering (ITCE), Aswan, 2-4 February 20, 130-144.
[6] 李春雷, 高广帅, 刘洲峰, 等. 应用方向梯度直方图和低秩分解的织物疵点检测算法[J]. 纺织学报, 2017, 38(3): 149-154.
[7] 薛婷婷, 刘秀平, 张凯兵, 等. 基于学习的Gabor滤波器多样式布匹瑕疵检测[J]. 西安工程大学学报, 2017(6): 775-780.
[8] Jing, J., Zhang, H., Wang, J., et al. (2013) Fabric Defect Detection Using Gabor Filters and Defect Clas-sification Based on LBP and Tamura Method. Journal of the Textile Institute, 104, 18-27.
[9] Ngan, H.Y.T., Pang, G.K.H. and Yung, N.H.C. (2010) Performance Evaluation for Motif-Based Patterned Texture Defect Detection. IEEE Transactions on Automation Science & Engineering, 7, 58-72. [Google Scholar] [CrossRef] 
[10] 景军锋, 范晓婷, 李鹏飞, 等. 应用深度卷积神经网络的色织物缺陷检测[J]. 纺织学报, 2017, 38(2): 68-74.
[11] 张宏伟, 张凌婕, 李鹏飞. 基于深度卷积神经网络的织物花型分类[J]. 纺织高校基础科学学报, 2017, 30(2): 261-265.
[12] Krzhevaky, A., Sutskever, I. and Hinton, G.E. (2012) ImageNet Classification with Deep Convolutional Neural Networks. International Conference on Neural Information Processing Systems, Doha, 12-15 November 2012, 1097-1105.
[13] Szegedy, C., Liu, W., Jia, Y., et al. (2015) Going Deeper with Convolutions. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Boston, 7-12 June 2015, 1-9.
[14] Masci, J., Meier, U., Ciresan, D., et al. (2012) Steel Defect Classification with Max-Pooling Convolutional Neural Networks. The 2012 IEEE International Joint Conference on Neural Networks (IJCNN), Brisbane, 10-15 June 2012, 1-6.
[15] He, K, Zhang, X., Ren, S., et al. (2001) Deep Residual Learning for Image Recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Kauai, 8-14 December 2001, 770-778.
[16] Dyk, D.A.V. and Meng, X.L. (2001) The Art of Data Augmentation. Journal of Computational & Graphical Statistics, 10, 1-50.
[17] Datar, M., Gionis, A., Indyk, P., et al. (2002) Maintaining Stream Statistics over Sliding Windows. Siam Journal on Computing, 31, 1794-1813.
[18] Shore, J. and Johnson, R. (1980) Axiomatic Deri-vation of the Principle of Maximum Entropy and the Principle of Minimum Cross-Entropy. IEEE Transactions on In-formation Theory, 26, 26-37.
[19] Lin, T.Y., Goyal, P., Girshick, R., et al. (2017) Focal Loss for Dense Object Detec-tion. IEEE Transactions on Pattern Analysis & Machine Intelligence, 39, 2999-3007.
[20] Theodoridis, S. (2015) Sto-chastic Gradient Descent. Machine Learning, 161-231.
[21] Russakovsky, O., Deng, J., Su, H., et al. (2015) ImageNet Large Scale Visual Recognition Challenge. International Journal of Computer Vision, 115, 211-252.