机械与电子2025，Vol.43Issue(12)：38-44,51,8.
基于轻量化深度学习驱动的输电线路异物入侵实时监测方法
A Real-time Monitoring Method for Foreign Object Intrusion in Transmission Lines Driven by Lightweight Deep Learning
摘要
Abstract
To address the challenge of reliable online detection and tracking of foreign object intrusions on transmission lines,a lightweight deep-learning-driven method of real-time monitoring is proposed.It overcomes the high computational complexity of existing algorithms and their incompatibility with re-source-limited field environments.Firstly,this paper proposes a YOLOv7-seg foreign object image seg-mentation model based on category aggregation for identifying and cropping foreign objects from complex backgrounds.Secondly,the ConvNeXt model is trained using the triple loss function to enable it to extract features from clipped images,and a standard feature database is generated by using this model and the training data.Then,this paper proposes a novel feature-assisted IoU multi-object tracking algorithm to ensure stronger scene adaptability on hardware with low computing power.Finally,a framework for real-time tracking and detection of foreign objects is proposed and deployed on edge devices for experimental verification.The experimental results show that the method proposed in this paper can be deployed on low-cost edge devices and has good adaptability and robustness.关键词
异物检测/输电线路/YOLOv7-seg/边缘设备/ConvNeXtKey words
foreign object detection/transmission lines/YOLOv7-seg/edge devices/ConvNeXt分类
信息技术与安全科学引用本文复制引用
吴奇伟,薛海,王真,刘征宇,胡妍捷..基于轻量化深度学习驱动的输电线路异物入侵实时监测方法[J].机械与电子,2025,43(12):38-44,51,8.基金项目
国家电网科技项目资助(J2024120) （J2024120）
