摘要
本文提出一种面向轨道交通受电弓的LSTM增强型Deep-SVDD红外打弓智能检测方法。其主要利用安装于运营机车顶部红外相机连续采集的受电弓红外图像,稳定获取运行车辆的受电弓状态。以大批量连续拍摄的11帧正常图像序列为基础,采用LeNet对图像进行特征提取,有效挖掘其空间特征信息。输出的特征图序列化后由LSTM进一步捕捉图像序列的时间依赖性,提取关键特征。SVDD据此学习正常样本的特征分布,从而构建超球体。实验结果表明,该方法能够准确地检测出运行车辆受电弓打弓故障,为保障轨道交通供电系统的稳定运行提供有力技术支撑,具有显著的实际应用价值与广阔的推广前景。
This paper proposes an LSTM-enhanced Deep-SVDD intelligent detection method for infrared pantograph strike detection in rail transit.It primarily utilizes infrared images of the pantograph continuously captured by an infrared camera installed on the top of operating locomotives to stably obtain the status of the pantograph on moving vehicles.Based on a large batch of continuously captured 11-frame normal image sequences,LeNet is employed for feature extraction from the images,effectively mining their spatial feature information.The output feature maps are then serialized and further processed by LSTM to capture the temporal dependencies within the image sequence and extract key features.Subsequently,SVDD learns the feature distribution of normal samples to construct a hypersphere accordingly.Experimental results demonstrate that this method can accurately detect pantograph strike faults in moving vehicles,providing strong technical support for ensuring the stable operation of the rail transit power supply system.It holds significant practical application value and broad promotion prospects.
作者
薛振华
牛晓伟
李强
张墨涵
王烨堃
Xue Zhenhua;Niu Xiaowei;Li Qiang;Zhang Mohan;Wang Yekun(CHN ENERGY Institute of TransportationTechnology Research,Beijing,China)
出处
《科学技术创新》
2025年第17期51-54,共4页
Scientific and Technological Innovation
基金
国能朔黄铁路发展有限责任公司科技创新项目(SHTL-23-16)
