Laser Cladding Cracks Recognition Based on Deep Learning Combined Convolutional Block Attention Module

Lujun Cui; Haiyang Li; Shirui Guo; Xiaolei Li; Yinghao Cui; Bo Zheng; Manying Sun

doi:10.3788/LOP202158.2014001

Laser & Optoelectronics Progress, Volume. 58, Issue 20, 2014001(2021)

Laser Cladding Cracks Recognition Based on Deep Learning Combined Convolutional Block Attention Module

Lujun Cui, Haiyang Li, Shirui Guo^*, Xiaolei Li, Yinghao Cui, Bo Zheng, and Manying Sun

Author Affiliations

School of Mechanical and Electrical Engineering, Zhongyuan University of Technology, Zhengzhou, Henan 450007, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(8)

Fig. 1. Improved U-net structure

Download full size

Fig. 2. CBAM layer for increasing feature map training weight information

Download full size

Fig. 3. Channel attention model structure

Download full size

Fig. 4. Spatial attention model structure

Download full size

Fig. 5. Crack morphologies at different magnifications. (a) 50 times; (b) 100 times; (c) 200 times; (d) 500 times

Download full size

Fig. 6. Test effect. (a)(b) Image to be tested; (c)(d) manual labeling; (e)(f) neural network labeling; (g)(h) adaptive threshold method labeling

Download full size

Table 1. Confusion matrix
View table
Table 1. Confusion matrix
Expert mark Prediction
True False
True True Positive (TP) False Negative (FN)
False False Positive (FP) True Negative (TN)

Table 2. Influence of the position of CBAM on the training parameter quantity and accuracy
View table
Table 2. Influence of the position of CBAM on the training parameter quantity and accuracy
Position None 1 9 1,9 8,9 1,8,9 7,8,9 6,7,8,9
Parameters 1941105 1941368 1941368 1941631 1942051 1942314 1944342 1952921
Accuracy /% 77.1 76.9 78.7 79.2 79.7 79.1 79.8 79.8

Tools

Get Citation

Copy Citation Text

Lujun Cui, Haiyang Li, Shirui Guo, Xiaolei Li, Yinghao Cui, Bo Zheng, Manying Sun. Laser Cladding Cracks Recognition Based on Deep Learning Combined Convolutional Block Attention Module[J]. Laser & Optoelectronics Progress, 2021, 58(20): 2014001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites