广东电力2025，Vol.38Issue(3)：18-26,9.DOI:10.3969/j.issn.1007-290X.2025.03.003
基于多分支特征分解的输电线路红外与可见光图像融合
Infrared and Visible Image Fusion of Power Transmission Lines Based on Multi-branch Feature Decomposition
摘要
Abstract
In the power system,relying solely on infrared or visible light images is insufficient for comprehensive assessment of transmission line conditions.Image fusion combines the strengths of both,aiding in target detection and 3D modeling.In order to enhance image fusion effect of the transmission lines,this paper proposes a multi-branch feature decomposition network for infrared and visible light image fusion.The network improves feature map decomposition by preserving the original feature map while decomposing base and detail features.In the multi-branch fusion module,weighted fusion is applied based on image characteristics,and a dual-dimensional attention module integrates the features.Finally,the decoder generates the fused image.Compared with existing methods on the transmission line and TNO datasets,the approach outperforms in eight objective evaluation metrics,with entropy(EN),standard deviation(SD),sum of the correlations of differences(SCD),and structural similarity index measure(SSIM)values of 7.26,62.04,1.53,and 0.93,respectively.The fusion results show that our method effectively enhances infrared key information and visible light detail information in transmission line images.关键词
图像融合/输电线路/特征分解/注意力增强Key words
image fusion/power transmission line/feature decomposition/attention enhancement分类
动力与电气工程引用本文复制引用
梁铸威,黄智明,郑日平,谢子轶,许玉格..基于多分支特征分解的输电线路红外与可见光图像融合[J].广东电力,2025,38(3):18-26,9.基金项目
国家自然科学基金项目(62473159) （62473159）
中国南方电网有限责任公司科技项目(030700KC23070011) （030700KC23070011）
