A Lightweight Real-time Detection System for Power Line Defects Based on DartNet-YOLO

Zitong Xu

doi:10.47297/taposatWSP2633-456930.20250609

PDF
导出
分享
收藏
专辑

A Lightweight Real-time Detection System for Power Line Defects Based on DartNet-YOLO

6卷第9期

浏览量: 0 Downloads: 2

A Lightweight Real-time Detection System for Power Line Defects Based on DartNet-YOLO
A Lightweight Real-time Detection System for Power Line Defects Based on DartNet-YOLO
Jinan Xinhang Experimental Foreign Language School, Jinan, Shandong, 250000, China
2025年6卷第9期页码：150-156
纸质出版：2025-09-30
DOI：10.47297/taposatWSP2633-456930.20250609
Accepted：

Scan QR Code

Zitong Xu. (2025). A Lightweight Real-time Detection System for Power Line Defects Based on DartNet-YOLO. 科技理论与实践（英文版）, 6(9), 150-156.

Xu Zitong.A Lightweight Real-time Detection System for Power Line Defects Based on DartNet-YOLO[J].Theory and Practice of Science and Technology,2025,06(09):150-156.
Zitong Xu. (2025). A Lightweight Real-time Detection System for Power Line Defects Based on DartNet-YOLO. 科技理论与实践（英文版）, 6(9), 150-156. DOI： 10.47297/taposatWSP2633-456930.20250609.

Xu Zitong.A Lightweight Real-time Detection System for Power Line Defects Based on DartNet-YOLO[J].Theory and Practice of Science and Technology,2025,06(09):150-156. DOI： 10.47297/taposatWSP2633-456930.20250609.

摘要

Abstract

The integrity of power transmission lines is essential to the power grid. Addressing the inefficiency and high risk of traditional manual inspection

as well as the issue of existing deep learning algorithms being heavy and difficult to deploy in real-time

this paper proposes an intelligent UAV-based detection system using the lightweight deep learning model DartNet-YOLO.This research first constructs a large-scale

multi-scenario

and finely annotated dataset of power line defects was constructed covering visible light and infrared modalities

incorporating defect severity levels and temporal labels. This study conducted an in-depth optimization of the YOLOv7 baseline on this dataset. The outcome is DartNet-YOLO-S

a lightweight model that leverages the SE attention mechanism and BiFPN for improved feature extraction. The model employs depthwise separable convolutions for a significant reduction in parameters and utilizes knowledge distillation to enhance small object detection accuracy. Experiments showed that the DartNet-YOLO-S model achieved an mAP@0.5 of 87.5% on the self-built test set

with the detection accuracy (AP) for small objects such as insulator cracks increased to 76.3%. The model was successfully deployed on the UAV's onboard platform

Jetson Xavier NX

achieving a real-time detection rate of 28 FPS with power consumption controlled within 8W. Field tests in complex environments demonstrated the robustness and practicality of the system

providing an efficient end-to-end solution for the intelligent operation and maintenance of power transmission lines.

关键词

Keywords

references

Liu, R., Chen, C., Liu, F., Li, Y., Zhang, Z., & Kou, L. (2022). Power system resilience enhancement strategy considering cyber-physical interdependence under disasters: development and prospects:Technology analysis and research prospect. Electric Machines and Control, 26(1), 9–23.

Wu, L., Bi, J., Chang, W., Yang, Y., & Gong, Y. (2018). Comprehensive inspection technology of unmanned aerial vehicle for distribution overhead transmission lines. China Electric Power, *51*(1), 97–101+138.

Zhang, H., Zhang, G., & Qin, Z. (2022). Application and analysis of autonomous power inspection technology for UAV. Rural Electrification, (7), 51–53.

Hu, C., Pei, S., Liu, Y., Fu, M., Geng, Z., & Li, Z. (2024). A real-time defect detection method for transmission lines at edge based on LEE-YOLOv7. High Voltage Engineering, *50*(11), 5047–5057.

Huang, J., Dong, J., Yao, W., Li, Z., & Li, J. (2025). Research on UAV intelligent inspection and fault identification method based on multimodal large image model. Electric System Equipment, (3), 106–108,111.

Smith, J. (2023). Spot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation. *Journal of Computer Vision*, 15(3), 234-256.

Wang, C.-Y., Bochkovskiy, A., & Liao, H.-Y. M. (2022). YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. arXiv preprint arXiv: 2207.02696 .

Tan, M., Pang, R., & Le, Q. V. (2020). Efficientdet: Scalable and efficient object detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp. 10781-10790).

Jocher, G., et al. (2023). Ultralytics YOLOv8. Documentation.

Yu, G., et al. (2021). PP-PicoDet: A better real-time object detector on mobile devices. arXiv preprint arXiv: 2111.00902 .

文章被引用时，请邮件提醒。

Submit

Advancing License Plate Recognition in Complex Environments: A Hybrid Approach Integrating Deep Learning and Image Pre-Processing

Research on Obstacle Avoidance for UAV Patrol Based on Vision

Add Sharpening to the Super-Resolution Image Processing

Intelligent Malware Adversarial Technology Based on Deep Learning

A Review of AI-Based Network Anomaly Traffic Detection

Related Author

Siyuan Zhang School of Engineering, Royal Melbourne Institute of Technology;College of Civil Aviation NUAA, Nanjing University of Aeronautics and Astronautics

Jiamin Zhao Jiangning Campus, School of International Education, Jinling Institute of Technology

Zhixuan Wang School of Automation, Wuhan University of Technology

Kazem Chamran City University Malaysia, SintokKedah Darul Aman

Yuefu Qiu City University Malaysia, SintokKedah Darul Aman

Kehan Li School of Information Engineering, Sichuan Agricultural University

Qinglong Liang International Exchange College, Guangzhou Maritime University

Ziyang Yan School of Mechanical Engineering, Hefei University of Technology,Mechanical Design, Manufacturing, and Automation

Related Institution

College of Civil Aviation NUAA, Nanjing University of Aeronautics and Astronautics

School of Engineering, Royal Melbourne Institute of Technology

Jiangning Campus, School of International Education, Jinling Institute of Technology

School of Automation, Wuhan University of Technology

City University Malaysia, SintokKedah Darul Aman

⁰