铁路货车装载状态标准图像智能识别技术体系设计研究  

Design of Intelligent Recognition Technology System for Standard Images of Wagon Loading Status

作　　者：石磊 SHI Lei(Freight Transport Department,China State Railway Group Co.,Ltd.,Beijing 100844,China)

机构地区：[1]中国国家铁路集团有限公司货运部,北京100844

出　　处：《铁道货运》2022年第11期34-39,48,共7页Railway Freight Transport

基　　金：中国国家铁路集团有限公司科技研究开发计划课题(N2021X005)。

摘　　要：铁路货车装载状态安全是确保行车安全、货运安全,进而确保铁路运输安全的重要基础,是推动铁路货运高质量发展的前提保障。针对铁路货车装载状态标准图像智能识别训练集数据规模不足的问题,按照“识别需求-识别技术-识别应用”的体系构建思路,设计铁路货车装载状态标准图像智能识别技术体系架构,构建“车种车型-货物类型-检查项点”的铁路货车装载状态检查项点关联模型,明确目标检测、实例分割、数据集扩充和模型轻量化等关键技术难点,提出完善视频设备配置、加强标准图像评价、快速积累负类样本、统一图像识别模型等配套措施,为铁路货车装载状态检查提供可靠技术支撑。The safety of wagon loading status is an important basis to ensure the safety of driving,freight,and railway transportation,and it is a premise to promote the high-quality development of railway freight.Since the intelligent recognition training set for standard images of wagon loading status has insufficient data scale,this paper followed a system construction idea featuring demand recognitiontechnology recognition-application recognition and designed an intelligent recognition technology architecture for standard images of wagon loading status.In addition,the paper constructed a correlation model for check points of wagon loading status,which features vehicle type-cargo type-check points.Furthermore,key technical difficulties including object detection,instance segmentation,data set expansion,and model lightweight were clarified,and supporting measures were put forward,such as improving video equipment configuration,strengthening standard image evaluation,accumulating negative samples rapidly,and unifying image recognition models,which provide reliable technical support for checking wagon loading status.

关 键 词：铁路货运 货检站 货车装载状态 图像识别 目标检测 

分 类 号：U298.1[交通运输工程—交通运输规划与管理]