高速铁路线路环境异物入侵视频检测系统研制  

Development of Video Detection System for Foreign Objection Intrusion in High-speed Railway Line Environments

作　　者：王浩然 戴鹏 刘俊博 时菁 宋浩然 顾子晨 WANG Haoran;DAI Peng;LIU Junbo;SHI Jing;SONG Haoran;GU Zichen(Postgraduate Department,China Academy of Railway Sciences,Ltd.,Beijing 100081,China;Infrastructure Inspection Research Institute,China Academy of Railway Sciences Corporation Ltd.,Beijing 100081,China;INMAI Railway Technology Co.,Ltd.,Beijing 100081,China)

机构地区：[1]中国铁道科学研究院研究生部,北京100081 [2]中国铁道科学研究院集团有限公司基础设施检测研究所,北京100081 [3]北京铁科英迈技术有限公司,北京100081

出　　处：《计算机测量与控制》2024年第10期86-91,共6页Computer Measurement &Control

基　　金：国家自然科学基金(52272427);中国国家铁路集团有限公司科技研究开发计划重大课题(K2021T015);中国铁道科学研究院集团有限公司院基金课题(2022YJ256)。

摘　　要：针对智能高铁2.0体系中智能装备建设要求,研制了高速铁路线路环境视频检测系统;该系统采用高实时、双模态补偿技术进行动态成像,克服了高速运行条件下的运动模糊及开放式场景下的环境光干扰问题;基于Faster-RCNN和YOLO v8模型开发了异物入侵智能识别算法,实现了基于高速动车组平台的列车运行环境状态异常在线检测;研发了线路环境视频动态检测应用软件,实现了列车运行环境异常的动态实时监测和异常数据管理;试验验证表明,系统可满足最高450 km/h运行条件下的线路环境高清成像和异物入侵检测,缺陷检出率≥90%。Aiming at the requirements of intelligent equipment construction in the intelligent high-speed railway system V2.0,a high-speed railway line environment video detection system is developed.The system uses high real-time and dual-mode compensation technology for dynamic imaging,which overcomes motion blur under high-speed operation conditions and ambient light interference in open scenes.An intelligent recognition algorithm for foreign object intrusion is presented based on the faster region-based convolutional neural networks(Faster-RCNN)framework and YOLO v8,achieving the online detection of abnormal train operating environment status for the high-speed train platform.The application software of video dynamic detection of line environments is developed,which realizes the dynamic real-time monitoring and abnormal data management of abnormal train operation environments.Experimental results show that the system can meet the high-definition imaging and foreign object intrusion detection of the line environment with a velocity of up to 450 km/h,and the accuracy of defect detection is greater than or equal to 90%.

关 键 词：高速铁路 异物入侵 视频检测系统 智能识别 数据管理 

分 类 号：U216.3[交通运输工程—道路与铁道工程]