基于DualAxis-SEAtt注意力机制输电线路关键目标检测算法
DOI:
CSTR:
作者:
作者单位:

1.国网江苏省电力有限公司南京供电分公司;2.江苏方天电力技术有限公司

作者简介:

通讯作者:

中图分类号:

基金项目:

国网江苏省电力有限公司科技项目(J2025036)


CHENG Yubao1,ZHAO Yanjie1,ZHANG Huan1,ZHANG Yuxi1,
Author:
Affiliation:

State Grid Jiangsu Electric Power Co., Ltd. Nanjing Power Supply Branch

Fund Project:

State Grid Jiangsu Electric Power Science and Technology Project (J2025036)

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    无人机输电线路光电巡检图像常受复杂背景与成像噪声影响,目标纹理和边缘特征易被削弱,导致细长部件、小尺度目标及局部缺陷检测性能下降。现有方法对通道特征与空间结构信息的协同利用不足,难以在抑制背景冗余的同时增强关键目标的特征响应。为此,本文提出了一种基于光电图像特征协同增强的输电线路部件检测方法。该方法以YOLOv8n为基础,在特征提取网络中引入多通道光电特征协同增强模块DualAxis-SEAtt自适应强化与目标纹理、边缘及空间结构相关的特征响应,抑制复杂背景中的冗余信息。基于自建数据集的实验结果表明:与YOLOv8n相比,所提方法检测性能明显提升,召回率在各注意力机制对比方法中最高,有效提高了目标覆盖能力并减少漏检;在多种光照场景和成像退化条件下,表现出更好的检出能力和稳定性;与传统图像处理方法相比,复杂背景和遮挡场景下鲁棒性更强。该方法可为复杂光环境下输电线路关键部件及缺陷的光电检测提供技术支持。

    Abstract:

    Optoelectronic images acquired during unmanned aerial vehicle (UAV)-based inspection of transmission lines are often affected by complex backgrounds and imaging noise, which can weaken target textures and edge features and consequently degrade the detection performance for slender components, small-scale objects, and local defects. Existing methods provide insufficient coordinated utilization of channel features and spatial structural information, making it difficult to suppress background redundancy while simultaneously enhancing the responses of critical target features. To address these issues, this paper proposes a transmission-line component detection method based on the collaborative enhancement of optoelectronic image features. Built upon YOLOv8n, the proposed method introduces a multi-channel optoelectronic feature collaborative enhancement module, termed DualAxis-SEAtt, into the feature extraction network. The module adaptively strengthens feature responses associated with target textures, edges, and spatial structures while suppressing redundant information in complex backgrounds. Experimental results on a self-constructed dataset demonstrate that the proposed method achieves considerable performance improvements over YOLOv8n. In particular, it obtains the highest recall among the compared attention-based methods, thereby improving target coverage and reducing missed detections. It also exhibits better detection capability and stability under various illumination conditions and imaging degradations. Compared with conventional image-processing methods, the proposed method shows stronger robustness in complex-background and occlusion scenarios. The proposed method can provide technical support for the optoelectronic detection of critical transmission-line components and defects under complex lighting conditions.

    参考文献
    相似文献
    引证文献
引用本文
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2026-04-08
  • 最后修改日期:2026-07-13
  • 录用日期:2026-07-27
  • 在线发布日期:
  • 出版日期:
文章二维码