基于改进YOLOv5s的高压线路异物检测  

High Voltage Line Foreign Object Detection Based on Improved YOLOv5s

作　　者：顾铭杰 姚聪颖 时乘 姚军财 GU Ming-jie;YAO Cong-ying;SHI Cheng;YAO Jun-cai(School of Power Engineering,Nanjing Institute of Technology,Nanjing Jiangsu 211167,China;School of Computer Engineering,Nanjing Institute of Technology,Nanjing Jiangsu 211167,China)

机构地区：[1]南京工程学院电力工程学院,江苏南京211167 [2]南京工程学院计算机工程学院,江苏南京211167

出　　处：《计算机仿真》2025年第3期103-110,共8页Computer Simulation

基　　金：国家自然科学基金(61301237);江苏省自然科学基金面上项目(BK20201468);江苏省高校“青蓝工程”中青年学术带头人资助项目(苏教师函[2022]29号)。

摘　　要：为了精准的检测出高压线路上的异物,提出了一种基于YOLOv5s的改进算法。算法首先为了对不同尺寸的目标增加合适的权重,增强网络的多尺度特征融合能力,以有效地捕捉不同大小目标的特征,采用F-Bi特征提取网络;其次,将YOLOv5的目标检测头替换为更高效的解耦头Decouple Head,加快模型的收敛速度;最后,使用W-IoU损失函数对模型原始损失函数C-IoU进行替换,提高锚点框的质量。实验结果表明,改进的YOLOv5s算法,其mAP50为96.3%,较原始YOLOv5s模型提升了2.6%;mAP50-95为68%,较原始YOLOv5s模型提升了2%,符合对输电线路巡检的实时性、准确性要求。In order to accurately detect foreign objects on high-voltage lines,an improved algorithm based on YOLOv5s is proposed.Firstly,in order to add appropriate weights to targets of different sizes and enhance the multiscale feature fusion ability of the network,the algorithm adopts F-Bi feature extraction network to effectively capture the features of targets of different sizes.Secondly,the target detection head of YOLOv5 is replaced with a more efficient decoupling head(Decouple Head),to accelerate the convergence speed of the model.Finally,the W-IoU loss function is used to replace the original loss function C-loU of the model,to improve the quality of the anchor box.The experimental results show that the improved YOLOv5s algorithm has a good performance with mAP50 of 96.3%,which is 2.6% higher than the original YOLOv5s model;and the mAP50-95 is 68%,which is 2% higher than the 0-riginal YOLOv5s model.These results show that it meets the real-time and accuracy requirements for transmission line inspection.

关 键 词：异物检测 深度学习 损失函数 高压线路 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术] TM743[自动化与计算机技术—计算机科学与技术]