ID-YOLO: A Multimodule Optimized Algorithm for Insulator Defect Detection in Power Transmission Lines

被引:1
作者
Zhang, Qiang [1 ]
Zhang, Jianing [1 ]
Li, Ying [1 ]
Zhu, Changfei [1 ]
Wang, Guifang [1 ]
机构
[1] Liaoning Tech Univ, Sch Elect & Control Engn, Huludao 125105, Peoples R China
基金
中国国家自然科学基金;
关键词
Feature extraction; Insulators; YOLO; Convolution; Power transmission lines; Data mining; Standards; Defect detection; Computer architecture; Location awareness; feature fusion; insulator defect (ID); real-time detection; you only look once (YOLO);
D O I
10.1109/TIM.2025.3527530
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
Insulators play a crucial role in providing electrical isolation in power transmission lines, and timely detection of their defects is vital to avoid severe human life and property losses. In the context of drone inspections of power transmission lines, accurate and timely detection and localization of insulator defects (IDs) are of paramount importance. Considering the inadequacy of the you only look once (YOLO) series of algorithms in extracting features of insulators and their defects in complex backgrounds, we have designed a method called ID-YOLO to address this challenge. First, we develop the global convolution (GConv) module to integrate spatial and channel information, thereby enhancing the effectiveness of feature extraction. Second, we built the C3-global pooling fusion (C3-GPF) module, aimed at strengthening focus on key data during the feature extraction and fusion stages. Third, we develop the multiscale information fusion (MSIF) module to balance the algorithm's detection accuracy and speed, ensuring superior performance in practical applications. Fourth, we built the weighted feature information fusion (WFIF) module to further enhance the fusion capability of key information. Finally, we adopt the SCYLLA-IoU (SIoU) loss function to replace the original CIoU, thereby improving the algorithm's localization precision and accelerating convergence speed. The experimental results indicate that ID-YOLO achieves an average precision (AP) of 90.9%, representing a 3.3% improvement over the baseline YOLOv5s algorithm. In addition, ID-YOLO achieves a detection speed of 90 frames per second (FPS), meeting the requirements for real-time detection. Practical test results demonstrate that the ID-YOLO algorithm significantly improves detection precision while effectively addressing the challenges associated with multiobject and small-object detection, showcasing its potential application in detecting IDs in power transmission lines.
引用
收藏
页数:11
相关论文
共 35 条
  • [11] Huang S., 2022, P 8 INT C COMP ART I, P573
  • [12] Li C, 2022, arXiv, DOI [arXiv:2209.02976, DOI 10.48550/ARXIV.2209.02976]
  • [13] Involution: Inverting the Inherence of Convolution for Visual Recognition
    Li, Duo
    Hu, Jie
    Wang, Changhu
    Li, Xiangtai
    She, Qi
    Zhu, Lei
    Zhang, Tong
    Chen, Qifeng
    [J]. 2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, : 12316 - 12325
  • [14] Feature Pyramid Networks for Object Detection
    Lin, Tsung-Yi
    Dollar, Piotr
    Girshick, Ross
    He, Kaiming
    Hariharan, Bharath
    Belongie, Serge
    [J]. 30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017), 2017, : 936 - 944
  • [15] MTI-YOLO: A Light-Weight and Real-Time Deep Neural Network for Insulator Detection in Complex Aerial Images
    Liu, Chuanyang
    Wu, Yiquan
    Liu, Jingjing
    Han, Jiaming
    [J]. ENERGIES, 2021, 14 (05)
  • [16] SSD: Single Shot MultiBox Detector
    Liu, Wei
    Anguelov, Dragomir
    Erhan, Dumitru
    Szegedy, Christian
    Reed, Scott
    Fu, Cheng-Yang
    Berg, Alexander C.
    [J]. COMPUTER VISION - ECCV 2016, PT I, 2016, 9905 : 21 - 37
  • [17] Infrared Image Target Detection of Substation Electrical Equipment Using an Improved Faster R-CNN
    Ou, Jianhua
    Wang, Jianguo
    Xue, Jian
    Wang, Jianping
    Zhou, Xian
    She, Lingcong
    Fan, Yadong
    [J]. IEEE TRANSACTIONS ON POWER DELIVERY, 2023, 38 (01) : 387 - 396
  • [18] Redmon J., 2018, arXiv, DOI [10.48550/arXiv.1804.02767, DOI 10.48550/ARXIV.1804.02767]
  • [19] You Only Look Once: Unified, Real-Time Object Detection
    Redmon, Joseph
    Divvala, Santosh
    Girshick, Ross
    Farhadi, Ali
    [J]. 2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2016, : 779 - 788
  • [20] Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
    Ren, Shaoqing
    He, Kaiming
    Girshick, Ross
    Sun, Jian
    [J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2017, 39 (06) : 1137 - 1149