| 67 | 0 | 157 |
| 下载次数 | 被引频次 | 阅读次数 |
为提升城轨列车智能化监测水平，突破单模态模型对多源数据复杂关联捕捉不足的局限，提出了一种基于多模态数据融合的城轨列车监测系统。该系统作为城轨智能化转型的核心驱动力，灵活集成了多个关键监测单元，实现了对列车运行状态的全方位、多层次监测。该系统创新性地引入了多模态数据的融合技术，有效地挖掘了各模态数据间的互补优势，深入剖析了数据融合算法的三大层次——数据层、特征层与决策层融合，并针对多模态的监测场景加以应用。这一解决方案在高环境复杂度时，识别准确率有了一定提升，响应时间有显著提高；系统硬件显著精简，建设成本节约25%,安装空间减小60%。结论表明，该系统为城轨列车智能化运维提供了高效技术支撑，有效推动城轨智能化进程，具备一定的应用价值。
Abstract：To enhance the intelligent monitoring level of urban rail trains and overcome the limitation of single-modal models in capturing complex correlations within multi-source data, a monitoring system for urban rail trains based on multimodal data fusion is proposed. As a core driver for the intelligent transformation of urban rail, this system flexibly integrates multiple key monitoring units, achieving comprehensive and multi-level monitoring of train operational status. The system innovatively introduces multimodal data fusion technology, effectively leveraging the complementary advantages among different data modalities. It conducts an in-depth analysis of the three levels of data fusion algorithms-data-level, feature-level, and decision-level fusion-and applies them to multimodal monitoring scenarios. This solution achieves a certain degree of improvement in recognition accuracy and a marked improvement in response time under high environmental complexity. Furthermore, it results in a significant simplification of system hardware, yielding a 25% cost saving in construction and a 60% reduction in installation space. The conclusion indicates that this system provides efficient technical support for the intelligent operation and maintenance of urban rail trains, effectively promotes the process of urban rail intelligence, and demonstrates considerable application value.
[1]中国城市轨道交通协会.中国城市轨道交通智慧城轨发展纲要[J].城市轨道交通，2020(4):8-23.China Association of Metros.Development outline for intelligent urban rail transit in China [J].Urban Rail Transit,2020(4):8-23.
[2]邢智明.建设融合城轨谱写城轨交通可持续高质量发展新篇章 《中国城市轨道交通融合城轨发展指南》解读[J].城市轨道交通，2024 (8):21-25.XING Z M.Building integrated urban rail,writing a new chapter for sustainable and high-quality development of urban rail transit:an interpretation of the integrated urban rail development guide of China [J].Urban Rail Transit,2024(8):21-25.
[3]赵亮.多模态数据融合算法研究[D].大连：大连理工大学，2018.ZHAO L.Research on multimodal data fusion algorithms [D].Dalian:Dalian University of Technology,2018.
[4]李学龙.多模态认知计算[J].中国科学：信息科学，2023,53(1):1-32.LI X L.Multimodal cognitive computing [J].Science China:Information Sciences,2023,53(1):1-32.
[5]刘通，高思洁，聂为之.基于多模态信息融合的多目标检测算法[J].激光与光电子学进展，2022,59(8):339-348.LIU T,GAO S J,NIE W Z.Multi-target detection algorithm based on multimodal information fusion [J].Laser & Optoelectronics Progress,2022,59(8):339-348.
[6]吴志才，魏冠军，党空雁，等.基于改进双边滤波的轨道点云去噪算法[J].应用激光，2024,44(11):174-182.WU Z C,WEI G J,DANG K Y,et al.Track point cloud denoising algorithm based on improved bilateral filtering [J].Applied Laser,2024,44(11):174-182.
[7]JUN Q C,DEEPU R,NICHOLAS V.Multimodal few-shot classification without attribute embedding [J].EURASIP Journal on Image and Video Processing,2024 (4):1-15.
[8]LI X L,ZHANG H Y,ZHANG R.Adaptive graph auto-encoder for general data clustering [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2021,44:9725- 9732.
[9]司坤宇，牛春晖.基于混合差分卷积和高效视觉Transformer网络的三重多模态图像融合算法[J].红外与激光工程，2024,53(11):331-345.SI K Y,NIU C H.Triple multi-modal image fusion algorithm based on hybrid differential convolution and efficient visual transformer network[J].Infrared and Laser Engineering,2024,53(11):331-345.
[10]侯高鹏.基于三维激光扫描的隧道变形点云数据处理方法研究[D].成都：西南交通大学，2021.HOU G P.Research on tunnel deformation point cloud data processing method based on 3D laser scanning [D].Chengdu:Southwest Jiaotong University,2021.
[11]胡思宇，周远昌，赵曈，等.基于层重组扩展卡尔曼滤波的神经网络力场训练[J].软件学报，2025,36(9):4094-4110.HU S Y,ZHOU Y C,ZHAO T,et al.Neural network force field training based on layer reconstruction extended kalman filtering [J].Journal of Software,2025,36(9):4094-4110.
[12]王一朵.基于PCA、LDA与SVM相结合的人脸图像识别应用研究[D].兰州：兰州交通大学，2023.WANG Y D.Research on application of face image recognition combining PCA,LDA,and SVM [D].Lanzhou:Lanzhou Jiaotong University,2023.
[13]李振中，应梦飞.基于主成分分析法和支持向量机算法的驾驶人疲劳检测方法[J].专用汽车，2023(5):74-77.LI Z Z,YING M F.Driver fatigue detection method based on principal component analysis and support vector machine algorithms [J].Special Purpose Vehicles,2023 (5):74-77.
[14]赵晶晶.基于模糊贝叶斯网络的故障诊断方法研究及其在列控系统中的应用[D].北京：北京交通大学，2013.ZHAO J J.Research on fault diagnosis method based on fuzzy Bayesian network and its application in train control systems [D].Beijing:Beijing Jiaotong University,2013.
[15]刘素艳，乔一鸣，董一林，等.改进D-S证据理论的高冲突多信息融合方法研究[J].铁道科学与工程学报，2024,21(1):1-12.LIU S Y,QIAO Y M,DONG Y L,et al.Research on high-conflict multi-information fusion method based on improved D-S evidence theory [J].Journal of Railway Science and Engineering,2024,21(1):1-12.
基本信息:
DOI：10.13291/j.cnki.djdxac.2026.03.016
中图分类号:U284.48
引用信息:
[1]吴英帅,刘嘉琛,杨守君,等.城轨列车监测系统关键技术研究与应用[J].大连交通大学学报,2026,47(03):134-142+151.DOI:10.13291/j.cnki.djdxac.2026.03.016.
基金信息:
中国中车城轨车辆智能融合检测集成技术研究项目
2026-06-15
2026-06-15
2026-06-15