| 薛含笑*,万文菲**,徐丹*,王宇***.基于多光谱图像的输变电设备综合状态评估方法研究[J].高技术通讯(中文),2025,35(11):1213~1224 |
| 基于多光谱图像的输变电设备综合状态评估方法研究 |
| Comprehensive state evaluation method of power equipment based on multispectral image |
| |
| DOI：10. 3772 / j. issn. 1002-0470. 2025. 11. 006 |
| 中文关键词: 多光谱图像； 输变电设备； 状态评估； 故障诊断； YOLOv5模型 |
| 英文关键词: multispectral image, power equipment, state assessment, fault diagnosis, YOLOv5 |
| 基金项目: |
| 作者 | 单位 | | 薛含笑* | （*广州商学院信息技术与工程学院广州 511363）
（**工业和信息化部电子第五研究所广州 511370）
（***广州大学人工智能学院 广州 510006） | | 万文菲** | | | 徐丹* | | | 王宇*** | |
|
| 摘要点击次数: 87 |
| 全文下载次数: 62 |
| 中文摘要: |
| 鉴于输变电设备的差异性、故障多样性、环境复杂性和光谱成像的特殊性，单独使用一种光谱图像难以准确评估输变电设备的实际状态，针对这一问题，本文提出了一种基于多光谱图像的输变电设备综合状态评估框架。该方法考虑不同光谱图像的成像原理和结构特征，通过在特征检测层增加锚点设置和优化损失函数设置，来改进现有的目标识别算法YOLOv5(you only look once version 5.0)模型，提高针对输变电设备的目标检测与识别效率。同时，该方法综合考虑了不同光谱图像下不同故障类型诊断的优势，融合多光谱图像的设备状态信息，能够更加全面、准确地综合评估输变电设备的状态。实验结果表明，该方法在所构建的多光谱数据库上对不同故障的诊断准确率均能达到90.00%以上，可以准确评估输变电设备的发热、表面缺陷和放电故障，对输变电设备的安全运行具有重要意义。 |
| 英文摘要: |
| Given the variability of power equipment, the diversity of faults, the complexity of the environment, and the specificity of spectral imaging, it is difficult to accurately assess the actual state of power equipment using the single spectral imaging. This paper proposes a comprehensive state assessment framework for power equipment based on multispectral images. This method considers the imaging principles and structural features of different spectral images, and improves the object recognition algorithm YOLOv5 (you only look once version 5.0) model by adding the anchor setting in the feature detection layer and optimizing the loss function to enhance the efficiency of target detection and recognition for power equipment. Meanwhile, this method comprehensively considers the advantages of diagnosing different types of faults under different spectral images, integrates equipment status information from multispectral images, and can accurately evaluate the fault status of power equipment. Experimental results show that the diagnostic accuracy of the proposed method on the constructed multispectral database is over 90.00%, and can accurately evaluate the heating, surface defects, and discharge faults of power equipment, which is significant for improving the operational safety of power equipment. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
