Detection of Crimping Defects in X-DR Images of Strain Clamps Based on Multi Attention Mechanism Fusion Network

As an important component of transmission lines, the crimping performance of strain clamps directly affects the stable operation of power systems. In order to quickly and accurately detect crimping defects in strain clamps according to their X-ray digital radiography (X-DR) images, a defect detectio...

Full description

Saved in:
Bibliographic Details
Published inIEEE transactions on power delivery Vol. 39; no. 4; pp. 2125 - 2137
Main Authors Qiu, Zhibin, Li, Junxuan, Zhu, Xuan, Zhou, Zhibiao
Format Journal Article
LanguageEnglish
Published New York IEEE 01.08.2024
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects
Aluminum
attention mechanism
Clamps
Crimping
Data augmentation
defect detection
Defects
Digital imaging
Folding
Frames per second
Image enhancement
Image recognition
Image transmission
Object recognition
Power transmission lines
Quality control
Steel
Strain
strain clamp
Transmission line
Transmission lines
X-ray image
Online AccessGet full text

Cover

More Information
Summary:As an important component of transmission lines, the crimping performance of strain clamps directly affects the stable operation of power systems. In order to quickly and accurately detect crimping defects in strain clamps according to their X-ray digital radiography (X-DR) images, a defect detection method is presented in this article based on multi-attention mechanism fusion network (MAMFNet). Firstly, a strain clamp X-DR image dataset containing six categories of crimping defects is constructed according to relevant defect assessment criteria, and data augmentation and image enhancement are performed. Secondly, the MAMFNet model is built using ShuffleNetV2 and DSC-PANet, incorporating an attention mechanism module to enhance the model's focus on target defects, and utilizing dual non-maximum suppression (Dual NMS) to fuse detection results from different attention mechanisms. Finally, the proposed method is tested using strain clamp X-DR images collected from transmission line inspection. The results show that MAMFNet achieves a mAP of 94.83% and a detection speed of 20 frames per second (FPS), demonstrating superior overall performance compared to other object detection models. This method realizes accurate detection of crimping defects in strain clamps, providing assistance for quality control before commissioning of transmission lines and maintenance during operation.
ISSN:0885-8977
1937-4208
DOI:10.1109/TPWRD.2024.3388022