To address the shortcomings of current surface defect detection algorithm with many parameters, slow detection rate and low accuracy, a defect detection algorithm based on DenseNet network is proposed to mitigate the effects of gradient disappearance and gradient explosion with its more aggressive dense connection mechanism, which also reduces the number of parameters to some extent. Meanwhile, the enhancement effect of SENet network on the effective features is utilized to optimize the network model and enhance the accuracy. Using the strip steel surface defect dataset released by Northeastern University as the base defect sample, the enhancement operation is performed on it. The accuracy of the present algorithm tested on this dataset is as high as 99.44%, which is higher than that of the unimproved DenseNet network, and in terms of computational speed, the present algorithm is only 1.9ms/frame, which obtains a significant improvement compared to the DenseNet network.

Penghui Zhu, Junjie Dai, Haoyuan Chang, Yao Xu, Zhimao Zhang. DenseNet network-based surface defect detection algorithm for strip steel. International Journal of Frontiers in Engineering Technology (2022), Vol. 4, Issue 10: 34-40. https://doi.org/10.25236/IJFET.2022.041006.

[1] Yuan L, Zhang Z, Tao X. The Development and Prospect of Surface Defect Detection Based on Vision Measurement Method[C]// 2016 12th World Congress on Intelligent Control and Automation (WCICA). IEEE, 2016

[2] Dalal N, Triggs B. Histograms of oriented gradients for human detection[C] / /Proc of IEEE Conference on Computer Vision and Pattern Recognition.2005: 886-893.

[3] LOWE D G. Distinctive image features from scale-invariant key points [J]. International Journal of Computer Vision,2004,60 ( 2):91-110. 

[4] Ferreira A J, Figueiredo M A T. Boosting algorithms: a review of methods, theory, and applications [M]/ /Ensemble Machine Learning. New York: Springer, 2012: 35-85.

[5] Xing Jianfu. Convolutional neural network-based surface defect identification and system development for hot-rolled strip steel [D]. Shenyang: Northeastern University, 2017. 

[6] Zhou S, Chen Y, Zhang D. Classifification of surface defects on steel sheet using convolutional neural networks. Materiali in Tehnologije 2017; 51(1):123–31.

[7] Yu F, Koltun V. Multi-scale context aggregation by dilated convolutions [J]. arXiv:1511.07122, 2015.

[8] K. Song, Y. Yan, “A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects,”Applied Surface Science, vol. 285, 2013: 858-864