摘要
受电弓是连接列车和接触网的重要部件,其运行中的工作状态直接影响列车的运营安全,近年来结合车载视频监控系统从图像上进行受电弓安全状态的检测成为主流方法。针对使用传统图像识别方法对受电弓安全状态进行检测存在的精度低和实时性差的现状,通过自建受电弓安全运行条件数据集,结合深度学习理论对其进行安全运行状态在线检测。采用GAN神经网络提升了数据集的性能,弥补了数据种类失衡的现象;通过对YOLOV3算法进行改造,插入即插即用的注意力模块,优化小目标的检测效果;通过与自研硬件的结合实现多路受电弓监控实时视频流的分析,为车载受电弓视频实时智能分析提供新的思路和有力支撑。
The pantograph is a key component connecting the rolling stock and the contact network,so the safety status of the pantograph is crucial to the smooth and stable operation of the rolling stock.In recent years it has become a mainstream method to detect the safety status of pantograph from images in combination with onboard video monitoring systems.Aiming at the traditional image recognition method's low accuracy and poor real-time performance to detect the pantograph's safety state,the pantograph's safety state is detected by the self-built pantograph safety state data set and the deep learning technique.GAN neural network is used to enhance the data-set to solve the problem of unbalanced data categories;the YOLOV3 algorithm is modified by inserting a plug-andplay attention module to optimize the detection of small targets.Combining with self-researched hardware to realize the analysis of multiple pantographs monitoring real-time video streams provides new ideas and strong support for the real-time intelligent analysis of onboard pantograph video in China's railways.
作者
辛恩承
XIN Encheng(Beijing Zongheng Electro-Mechanical Technology Co.,Ltd.,Beijing 100094,China;Locomotive&Car Research Institute,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China)
出处
《铁道机车车辆》
北大核心
2024年第2期94-100,共7页
Railway Locomotive & Car
基金
北京纵横机电科技有限公司自主立项课题(2050ZH3602)。
关键词
受电弓
智能识别
视频监控
YOLOV3算法
pantograph
intelligent identification
video surveillance
YOLOV3 algorithm
