TY - JOUR
T1 - An Explainable Laser Welding Defect Recognition Method Based on Multi-Scale Class Activation Mapping
AU - Liu, Tianyuan
AU - Zheng, Hangbin
AU - Bao, Jinsong
AU - Zheng, Pai
AU - Wang, Junliang
AU - Yang, Changqi
AU - Gu, Jun
N1 - Funding Information:
This work was supported by the Fundamental Research Funds for the Central Universities and Graduate Student Innovation Fund of Donghua University under Grant CUSF-DH-D-2020053.
Publisher Copyright:
© 1963-2012 IEEE.
PY - 2022/2/2
Y1 - 2022/2/2
N2 - Vision-based online defect recognition can provide insights for laser welding quality control systems. Although the visual signal contains richer quality information than the one-dimensional signal, the quality features contained in the visual signal are more abstract. To improve the explainability of current convolutional neural networks (CNNs) for laser welding defect recognition (LWDR), a class activation mapping method based on multi-scale fusion features (CAM-MSFF) is proposed. In addition, a multi-scale features adaptive fusion method is proposed with three steps of feature squeeze, feature mapping, and feature recalibrating. In order to facilitate the learning and utilization of multi-scale features by the proposed method, supervisory information is applied to multiple scales. The experimental results show that the proposed CAM-MSFF method has higher accuracy and convergence speed than the conventional model. The results of the explainability tests show that the proposed method can provide a more accurate and human-comprehensible explanation of the model's decision basis.
AB - Vision-based online defect recognition can provide insights for laser welding quality control systems. Although the visual signal contains richer quality information than the one-dimensional signal, the quality features contained in the visual signal are more abstract. To improve the explainability of current convolutional neural networks (CNNs) for laser welding defect recognition (LWDR), a class activation mapping method based on multi-scale fusion features (CAM-MSFF) is proposed. In addition, a multi-scale features adaptive fusion method is proposed with three steps of feature squeeze, feature mapping, and feature recalibrating. In order to facilitate the learning and utilization of multi-scale features by the proposed method, supervisory information is applied to multiple scales. The experimental results show that the proposed CAM-MSFF method has higher accuracy and convergence speed than the conventional model. The results of the explainability tests show that the proposed method can provide a more accurate and human-comprehensible explanation of the model's decision basis.
KW - Class activation mapping (CAM)
KW - convolutional neural network (CNN)
KW - defect recognition
KW - explainable deep learning (XDL)
KW - laser welding
KW - multi-scale
UR - http://www.scopus.com/inward/record.url?scp=85124241805&partnerID=8YFLogxK
U2 - 10.1109/TIM.2022.3148739
DO - 10.1109/TIM.2022.3148739
M3 - Journal article
AN - SCOPUS:85124241805
SN - 0018-9456
VL - 71
JO - IEEE Transactions on Instrumentation and Measurement
JF - IEEE Transactions on Instrumentation and Measurement
ER -