Detection of hydrophobicity grade of insulators based on AHC-YOLO algorithm
Authors:
Open Access
- 20 March 2025
- journal article
- research article
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-12
Abstract
Thanks to the rapid development of image processing technology, the efficiency and accuracy of power inspection have been enhanced through deep learning techniques. However, during on-site inspections, the complexity of the background images of composite insulators often makes it difficult to directly extract key features for accurately assessing hydrophobicity levels. Moreover, considering the real-time requirements for insulator hydrophobicity detection in practical operations, the model must be lightweight to speed up the detection process. To address this issue, this paper proposes a YOLO algorithm for the precise detection of composite insulator hydrophobicity. The algorithm integrates a high-performance GPU network (HGNetv2), a mixed local channel attention mechanism (MLCA), lightweight convolution (CSPPC), and the Inner-WIoU loss function, significantly reducing the network’s burden and improving the accuracy of recognizing composite insulator sheds and classifying their hydrophobicity levels. By adopting a strategy of identifying insulator sheds and then classifying their hydrophobicity levels, precise detection of hydrophobicity is achieved. Experimental results show that the proposed AHC-YOLO algorithm has increased the detection accuracy of sheds by 5.77%, with GFLOPs reduced to 5.8. In the task of classifying hydrophobicity levels, the Top-1 accuracy has been improved by 4.994%, with GFLOPs reduced to 1.9. These achievements not only meet the needs for the detection and classification of composite insulator hydrophobicity but also further demonstrate the effectiveness and superiority of the algorithm through ablation and comparative experiments.Keywords
Funding Information
- Natural Science Foundation of Hebei Province (20181210100359413, 20181210100359413, 20181210100359413, 20181210100359413, 20181210100359413, 20181210100359413, 20181210100359413, 20181210100359413)
- Fundamental Research Funds for the Central Universities (2020MS093, 2020MS093, 2020MS093, 2020MS093, 2020MS093, 2020MS093, 2020MS093, 2020MS093)
This publication has 25 references indexed in Scilit:
- Run, Don't Walk: Chasing Higher FLOPS for Faster Neural Networks2023
- BiFormer: Vision Transformer with Bi-Level Routing Attention2023
- YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors2023
- Mixed local channel attention for object detectionEngineering Applications of Artificial Intelligence, 2023
- Recent Advances in Unmanned Aerial Vehicles: A ReviewArabian Journal for Science and Engineering, 2022
- Coordinate Attention for Efficient Mobile Network Design2021
- Insulator Defect Recognition Based on Global Detection and Local SegmentationIEEE Access, 2020
- YOLO9000: Better, Faster, Stronger2017
- Deep Residual Learning for Image Recognition2016
- You Only Look Once: Unified, Real-Time Object Detection2016