摘要
输电线路的分割有助于电力系统的智能化管理与维护。针对现有的输电线路分割算法精度不高的问题,提出了一种基于改进U-Net的输电线图像分割算法。首先模型以U-Net为基础架构,然后在跨层连接处引入改进后的CBAM模块,该模块由通道注意力与多尺度空间注意力组成。通过在输电线分割数据集上进行实验,MIoU达到了77.0%,F1-Score为85.6%。实验结果表明,基于此,提出的算法能够准确分割输电线路。
The segmentation of transmission lines contributes to the intelligent management and maintenance of power systems.A transmission line image segmentation algorithm based on improved U-Net is proposed to address the issue of low accuracy in existing transmission line segmentation algorithms.Firstly,the model is based on U-Net architecture,and then an improved CBAM module is introduced at the cross layer connection,which consists of channel attention and multi-scale spatial attention.Through experiments on the transmission line segmentation dataset,MIoU reached 77.0%and F1-Score is 85.6%.The experimental results show that the algorithm proposed in this paper can accurately segment transmission lines.
作者
焦阳阳
Jiao Yangyang(Shanxi Energy Internet Research Institute,Taiyuan Shanxi 030032,China)
出处
《山西电子技术》
2025年第4期108-109,共2页
Shanxi Electronic Technology
关键词
语义分割
卷积神经网络
注意力机制
输电线路
semantic segmentation
convolutional neural networks
attention mechanism
transmission line
