Article Details

Abstract

In the current field of defect detection for tower insulators, there is a problem that the insulator recognition is not high in foggy weather and the defective insulator is difficult to identify correctly. Therefore, an improved YOLOv5 defect detection model is proposed. Based on the YOLOv5 algorithm, the model introduces three different Attention Mechanisms of SEnet, CBAMnet and ECAnet into the three feature layers of different sizes in the backbone network feature extraction to enhance the feature extraction ability. In addition, the Focal Loss function is introduced to strengthen the control of the weight of easy classification samples and difficult classification samples. The experimental results show that the improved YOLO detection model has an AP value of 85.67% for normal insulators, an AP value of 97.33% for defective insulators, and a mAP value of 91.50%. Compared with the traditional YOLOv5 model, it increases by 0.29%, 0.48%, and 0.38% respectively. At the same time, its normal regression rate and defect regression rate are increased by 5.52% and 3.95%, respectively. The improved model can better identify normal and defective insulators, which has certain reference value for subsequent insulator related research.

Keywords

insulator defect; Attention Mechanism; loss function; Object Detection; YOLOv5 algorithm

References

1. [1] Lu Yiming Insulator Defect Detection Based on MobileNetV3-YOLOv5 Network [D]. Shijiazhuang: Shijiazhuang Railway University, 2023
2. [2] Wang Chenglong Research on insulator defect detection algorithm based on improved YOLOv5n [D]. Daqing: Northeast Petroleum University, 2023
3. [3] Yang Longhuan, Dong Xiaojie, Huang Shiyi, etc Research on Defect Detection Algorithm for Lightweight Insulators Based on YOLOV5 [J]. Communication and Information Technology, 2023 (5): 13-18
4. [4] Su Chao, Peng Zerong, Huang Zhiwen Insulator defect detection based on image processing [J]. Industrial Control Computer, 2016, 29 (12): 18-20
5. [5] ZHAO Z B，LIU N，WANG L. Localization of Multiple Insulators by Orientation Angle Detection and Binary Shape Prior Knowledge [J].IEEE Transactions on Dielectrics and Electrical Insulation，2015，22（6）：3421-3428.
6. [6] Hu Haote, Xu Jian, Huang Ying, etc Defect detection of insulators in transmission towers based on improved Faster R-CNN [J]. Information Technology and Informatization, 2023 (7): 63-66
7. [7] Ai Shangmei, Zhou Jianfeng, Zhang Bichao, etc Research on Defect Detection of Photovoltaic Modules Based on Improved SSD Algorithm [J]. Smart Power, 2023, 51 (12): 53-58
8. [8] REDMON J，FARHADI A. YOLOv3: An Incremental Improvement [J/OL].arXiv:1804.02767 [cs.CV].[2024-09-23].https://doi.org/10.48550/arXiv.1804.02767.
9. [9] REDMON J，DIVVALA S，GIRSHICK R，et al. You Only Look Once: Unified, Real-Time Object Detection [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Las Vegas：IEEE，2016：779-788.
10. [10] REDMON J，FARHADI A. YOLO9000: Better, Faster, Stronger [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu：IEEE，2017：6517-6525.
11. [11] Xiong Shaoto, Tong Anke, Yang Chaocheng, etc Insulator defect detection based on improved YOLOv5 [J]. Industrial Control Computer, 2023, 36 (12): 70-72
12. [12] JOCHER G，CHAURASIA A，STOKEN A，et al. ultralytics/yolov5: v6.1-TensorRT, TensorFlow Edge TPU and OpenVINO Export and Inference [EB/OL].(2022-2-22)[2024-09-19].https://zenodo.org/records/6222936.
13. [13] HU J，SHEN L，ALBANIE S. Squeeze-and-Excitation Networks [J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2020，42（8）：2011-20231.
14. [14] WOO S，PARK J，LEE J Y，et al. CBAM: Convolutional Block Attention Module [C]//Computer Vision-ECCV 2018. Munich：Springer，2018：3-19.
15. [15] WANG Q L，WU B G，ZHU P F，et al. ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Seattle：IEEE，2020：11531-11539.
16. [16] LIN T-Y，GOYL P，IRSHICK R，et al. Focal Loss for Dense Object Detection [J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2020，42（2）：318-327.

About the authors

Zeng Weiguo (1999-), male, Han ethnicity, from Jingzhou, Hubei, is currently pursuing a master's degree in image recognition and object detection; Corresponding author: Liu Xiaoqun (1968-), male, Han ethnicity, from Zhangjiakou, Hebei Province, China. He is a master's supervisor, professor, and master's student, with a research focus on computer networks and information security; Hao Juan (1989-), female, Han ethnicity, from Baoding, Hebei Province, associate professor, master's student, research direction: image processing.