摘要
输电线路绝缘子是电力系统的重要部件,近年来,随着人工智能技术的日益发展,基于无人机智能巡检技术的绝缘子缺陷检测已成为电力设备检测中的研究热点.针对无人机在线巡检的实时性要求高、计算资源有限等问题,提出了两种YOLOv5的轻量化改进方案,将经典YOLOv5分别结合轻量级卷积神经网络MobileNetV3和GhostNet的优点.实验结果表明,改进后的模型在保证检测精度的基础上,有效缩减了模型计算量,降低了算法的复杂度.YOLOv5-MobileNetV3模型计算量降低了 85.4%,检测精度略有下降;YOLOv5-GhostNet模型计算量降低了49.4%,且保持了高检测精度.因此,所提模型更有利于在无人机平台上的部署,实现了对输电线路绝缘子缺陷的实时检测.
Abstract
Insulators of transmission lines are important components of power systems.In recent years,with the increasing development of artificial intelligence technology,insulator defect detection based on UAV intelligent inspection technology has become a research hotspot in power equipment inspection.For the problems of high real-time requirements and limited computational resources of UAV online inspection.Two lightweight im-provements of YOLOv5 that combine the advantages of the classical YOLOv5 with the lightweight convolution-al neural networks MobileNetV3 and GhostNet are proposed,respectively.The experimental results show that the improved model effectively reduces the model computation on the basis of ensuring the detection accuracy.The YOLOv5-MobileNetV3 model reduces 85.4%of computation and slightly decreases the detection accura-cy.The YOLOv5-GhostNet model reduces 49.4%of computation and maintains high detection accuracy.Therefore,the model proposed is more favorable for deployment on UAV platforms to achieve real-time detec-tion of transmission line insulator defects.
基金项目
山西省自然科学基金(202103021224056)
国家科技期刊平台
NSTL
万方数据