基于多源数据的架空线绝缘子缺陷无人机智能检测研究

郑怿; 刘敏; 王红旭; 周国亮; 郭晨晨

郑怿,刘敏,王红旭,周国亮,郭晨晨.基于多源数据的架空线绝缘子缺陷无人机智能检测研究[J].电测与仪表,2026,63(8):192-200.
ZHENG Yi,LIU Min,WANG Hongxu,ZHOU Guoliang,guochwnchen.Research on Unmanned Aerial Vehicle Intelligent Detection of Insulator Defects in Overhead Lines Based on Multi-Source Data[J].Electrical Measurement & Instrumentation,2026,63(8):192-200.

基于多源数据的架空线绝缘子缺陷无人机智能检测研究

Research on Unmanned Aerial Vehicle Intelligent Detection of Insulator Defects in Overhead Lines Based on Multi-Source Data

DOI：10.19753 / j.issn1001-1390.2026.08.020

中文关键词: 压缩感知小波变换 YOLOv5网络 MSRN网络坐标注意力机制 CIoU-Loss函数

英文关键词:Compression perception Wavelet transform YOLOv5 network MSRN network Coordinate attention mechanism CIoU Loss function

基金项目:国网冀北电力有限公司科技项目(No.520184220001)

作者中文名	作者英文名	单位
郑怿	ZHENG Yi	国网冀北电力有限公司技能培训中心
刘敏	LIU Min	国网冀北电力有限公司
王红旭	WANG Hongxu	国网冀北电力有限公司技能培训中心
周国亮	ZHOU Guoliang	国网冀北电力有限公司技能培训中心
郭晨晨	guochwnchen	国网冀北电力有限公司技能培训中心

摘要点击次数: 9

中文摘要:

在实际的架空线绝缘子巡检过程中,无人机需要实时处理大量不同类型的数据,这增加了绝缘子缺陷检测的难度。为此,提出基于多源数据的架空线绝缘子缺陷无人机智能检测研究。结合压缩感知算法与小波变换算法,分别针对待融合图像的高频区域与低频区域像素融合规则展开设计,完成多源数据融合,通过融合不同类型的图像,实时准确获取全面的绝缘子信息。在传统YOLOv5网络的输入端增加MSRN网络,优化架空线绝缘子缺陷检测模型的图像感知能力,利用坐标注意力机制提高模型对于绝缘子缺陷特征的关注度,优化缺陷识别效果；利用CIoU-Loss函数生成用于缺陷标识的边界框,提高模型检测精度。实验表明,所提方法建立的YOLOv5架空线绝缘子缺陷检测模型具有较好的收敛能力,能够精准识别出不同类别的架空线绝缘子缺陷,有助于电力系统的安全稳定运行和高效维护,从而助力实现双碳目标和碳达峰目标。

英文摘要:

In the actual inspection process of overhead line insulators, drones need to process a large amount of different types of data in real time, which increases the difficulty of insulator defect detection. Therefore, a research on unmanned aerial vehicle intelligent detection of insulator defects on overhead lines based on multi-source data is proposed. Combining compressive sensing algorithm and wavelet transform algorithm, design pixel fusion rules for high-frequency and low-frequency regions of the image to be fused, and complete multi-source data fusion. By fusing different types of images, real-time and accurate comprehensive insulator information can be obtained. Add an MSRN network to the input of the traditional YOLOv5 network to optimize the image perception ability of the overhead line insulator defect detection model. Utilize coordinate attention mechanism to improve the model"s attention to insulator defect features and optimize defect recognition performance; Generate bounding boxes for defect identification using the CIoU Loss function to improve model detection accuracy. The experiment shows that the YOLOv5 overhead line insulator defect detection model established by the proposed method has good convergence ability and can accurately identify different types of overhead line insulator defects, which is helpful for the safe and stable operation and efficient maintenance of the power system, thereby helping to achieve the dual carbon and carbon peak goals.

查看全文查看/发表评论下载PDF阅读器

关闭