Infrared Image Recognition of Substation Equipment Based on Lightweight Backbone Network and Attention Mechanism

被引：0
作者：
Wang Y. [1 ]
Li Y. [1 ]
Duan Y. [1 ]
Wu H. [1 ]
机构：
[1] College of Electrical and Control Engineering, Xi’an University of Science and Technology, Shaanxi Province, Xi’an
来源：
Dianwang Jishu/Power System Technology | 2023年 / 47卷 / 10期
基金：
中国国家自然科学基金;
关键词：
attention mechanism; infrared image recognition; lightweight; substation equipment; YOLOv5;
D O I：
10.13335/j.1000-3673.pst.2022.2113
中图分类号：
学科分类号：
摘要：
The real-time monitoring of the substation equipment plays an important role in the safe and stable operation of the power grid. To realize the fast and accurate recognition of the substation equipment under complex background environments, a method of infrared image recognition based on the lightweight YOLOv5 is proposed. The introduction of the Ghost convolution in the back-bone network makes the network lighter and the detection speed significantly improved. An attention module based on the information interaction strategy between the channels is added to eliminate the interference of the irrelevant information and improve the saliency of the target. In the feature fusion stage, the improved C3 module with self-attention is utilized to enhance the feature capture ability and improve the network accuracy. In addition, the Cluster NMS (non-maximum suppression) and the EIOU(efficient intersection over union) loss are introduced to accelerate the network convergence. Tests are performed on the data sets containing three types of the substation equipment. The recognition accuracy reaches 93.80% with the detection speed as 0.0011s per picture ideally. Compared with the four classical networks, the experimental results show that the proposed algorithm reduces the average time consumption by 5.42% while improving the accuracy effectively and the storage size of the model is decreased by 26.38 %, which meets the requirements for the accuracy and real-time recognition of the substation equipment, providing conditions for the subsequent equipment fault diagnosis. © 2023 Power System Technology Press. All rights reserved.
引用
收藏
页码：4358 / 4366
页数：8
相关论文
共 25 条
[1]  
WANG Jing, YAO Zoujing, A fault diagnosis method for power equipment based on spatiotemporal features of infrared images[J], Control Engineering of China, 28, 8, pp. 1683-1690, (2021)
[2]  
Sheng HAN, Fan YANG, YANG Gang, Electrical equipment identification in infrared images based on ROI-selected CNN method [J], Electric Power Systems Research, 188, (2020)
[3]  
Wei LIU, SSD：single shot multiBox detector[C], Proceedings of the 14th European Conference on Computer Vision, pp. 21-37, (2016)
[4]  
REDMON J, DIVVALA S, GIRSHICK R, You only look once：unified，real-time object detection[C], Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition, pp. 779-788, (2016)
[5]  
REDMON J，, FARHADI A．YOLO9000：better，faster，stronger[C], Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition, pp. 6517-6525, (2017)
[6]  
REDMON J，FARHADI A．YOLOv3：an incremental improvement [EB/OL]
[7]  
GIRSHICK R, DONAHUE J, DARRELL T, Rich feature hierarchies for accurate object detection and semantic segmentation [C], Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition, pp. 580-587, (2014)
[8]  
GIRSHICK R, Fast R-CNN[C], Proceedings of 2015 IEEE International Conference on Computer Vision, pp. 1440-1448, (2015)
[9]  
REN Shaoqing, HE Kaiming, GIRSHICK R, Faster R-CNN：towards real-time object detection with region proposal networks[C], Proceedings of the 28th International Conference on Neural Information Processing Systems, pp. 91-99, (2015)
[10]  
Bin LI, LI Yalin, ZHU Xinshan, Multi-target detection in substation Scence based on attention mechanism and feature balance [J], Power System Technology, 46, 6, pp. 2122-2132, (2022)
123