Intelligent vision based personnel intrusion prevention system for railway bridge
-
摘要: 铁路桥梁监测是保障铁路运输安全的重要手段。为提升现有监测系统对铁路大桥人员入侵的检测能力,设计了基于智能视觉的铁路大桥人员入侵防护系统,该系统由视频平台、智能视觉平台及业务管理平台组成。采用YOLOv5目标检测模型进行人员入侵检测;同时,采用多种图像数据增强技术,扩增训练数据集,进一步提升目标检测模型的泛化能力和场景适应能力。在包神铁路集团有限公司万南站区黄河大桥对该系统进行了部署和测试。测试结果表明,该系统对人员入侵检测的准确率为95.3%,检测实时性为2 ms;人员入侵检测的准确率与实时性均满足实际应用要求。Abstract: Railway bridge monitoring is an important means to ensure transportation safety. To enhance the detection capability of existing monitoring systems for railway bridge personnel intrusion, this paper designed a personnel intrusion prevention system for railway bridge that included a video platform, an intelligent visual platform, and a business management platform. The paper adopted the YOLOv5 object detection model for personnel intrusion detection, and adopted multiple image data enhancement technologies to expand the training dataset, further improve the generalization ability and scene adaptation ability of the object detection model. The system was deployed and tested at the Yellow River Bridge in the Wannan Station area of Baoshen Railway Group Limited Liability Company. The experimental results show that the detection accuracy of the system for personnel intrusion is 95.3%, and the real-time detection performance is 2 ms. The accuracy and real-time performance of personnel intrusion detection meet the practical application requirements.
-
-
表 1 铁路大桥人员入侵防护系统服务器参数
处理器 Intel Xeon Silver 4214R 2.4G/12Core/16.5M/100W/Tray 内存模块 32GB DDR4-2933 ECC REG 存储模块 Seagate ST2000NM0008 2TB/128M/7200RPM/SATA/3.5" GPU NVIDIA Tesla T4 GPU 75W 16GB LP Passive 900-2G183-0000-00 网络 双万兆以太网端口 操作系统 Ubuntu18.04 表 2 实验环境
环境配置 名称 信息 硬件配置 GPU NVIDIA TITAN V CPU Intel Core i7-11800H 内存 16G 显存 12G 软件环境 操作系统 Ubuntu18.04 Python 3.8.0 Pytorch 1.8.0 CUDA 11.1 cuDNN 8.1.0 表 3 超参数设置
名称 数值 输入图像分辨率 640×640×3 迭代运行次数 300 批处理大小 4 优化器 SGD 初始学习率 0.01 周期学习率 0.01 学习率动量 0.99 权重衰减系数 0.0001 -
[1] 王 俊,王江丽. 高速铁路防灾安全监控系统设计 [J]. 中国安全科学学报,2018,28(S1):39-45. [2] 贾利民,陈熙元,马小平,等. 基于云边交互机制的自主式高速铁路防灾系统架构 [J]. 中国铁道科学,2022,43(5):165-176. [3] Xu H, Qiao J, Zhang J G, et al. A high-resolution leaky coaxial cable sensor using a wideband chaotic signal [J]. Sensors, 2018, 18(12): 4154. DOI: 10.3390/s18124154
[4] Catalano A, Bruno F A, Pisco M, et al. Intrusion detection system for the protection of railway assets by using fiber Bragg grating sensors: a case study[C]//Proceedings of 2014 Third Mediterranean, 7-9 May, 2014, Trani, Italy. New York: IEEE, 2014: 1-3.
[5] McConnell P R H, Scragg R A. Audio railway crossing detector: USA, 5910929[P]. 1999-06-08.
[6] Garcia J J, Hernandez A, Urena J, et al. Low cost obstacle detection for smart railway infrastructures[C]//Proceedings of the IEEE Intelligent Vehicles Symposium, 14-17 June, 2004, Parma, Italy. New York: IEEE, 2004: 670-675.
[7] Oh S, Park S, Lee C. Vision based platform monitoring system for railway station safety[C]//Proceedings of the 7th International Conference on its Telecommunications, 6-8 June, 2007, Sophia Antipolis, France. New York: IEEE, 2007: 1-5.
[8] Delgado B, Tahboub K, Delp E J. Automatic detection of abnormal human events on train platforms[C]//Proceedings of the IEEE National Aerospace and Electronics Conference, 24-27 June, 2014, Dayton, OH, USA. New York: IEEE, 2014: 169-173.
[9] Shinoda N, Takeuchi T, Kudo N, et al. Fundamental experiment for utilizing LiDAR sensor for railway [J]. International Journal of Transport Development and Integration, 2018, 2(4): 319-329. DOI: 10.2495/TDI-V2-N4-319-329
[10] 王泉东,杨 岳,罗意平,等. 铁路侵限异物检测方法综述 [J]. 铁道科学与工程学报,2019,16(12):3152-3159. [11] 张腾云,荆 涛,霍 炎. 神朔铁路智能视频入侵检测系统设计 [J]. 铁路计算机应用,2012,21(12):1-3. DOI: 10.3969/j.issn.1005-8451.2012.12.001 [12] 杨 栋,黄文政,张秋亮,等. 基于Faster-RCNN的站台端部人员入侵检测研究 [J]. 铁路计算机应用,2020,29(2):6-11. [13] 傅荟瑾,史天运,王 瑞,等. 基于深度学习的京张高速铁路周界图像智能识别系统构建研究 [J]. 铁道运输与经济,2022,44(5):64-72. -
期刊类型引用(0)
其他类型引用(1)