Research on application of intelligent video analysis for high-speed train
-
摘要: 为提升高速列车智能化水平,搭建高速列车智能视频分析系统,主要由车载摄像机、视频监控服务器、智能分析主机等设备构成,可通过车厢控制器与其他车载系统和设备进行信息交互。该系统基于深度学习算法构建智能视频分析模型,实现车厢敏感人员识别、车厢乘客拥挤度检测、车厢遗留行李检测、车厢重点位置监控、司机疲劳驾驶监测等功能;当检测和识别出异常事件时,告警信息可在旅客信息系统的显示终端上自动显示,或由广播装置播放告警信息。Abstract: In order to improve the intelligent level of high-speed train, the intelligent video analysis system for high-speed train is built, which is mainly composed of on-board camera, video monitoring server, intelligent analysis host and can exchange information with other on-board systems and equipment through the carriage controller. The intelligent video analysis models are built based on deep learning algorithm, and the functions of sensitive personnel identification, passenger congestion detection, baggage detection, key location monitoring and driver fatigue monitoring are realized. Once abnormal events are detected and identified, the alarm information can be automatically displayed on the display terminals or played by the broadcast device in the passenger information system.
-
表 1 车厢拥挤度系数阈值表
车厢拥挤等级 车厢拥挤度系数阈值 不拥挤 $ c \leqslant 0.2 $ 轻度拥挤 $0.2 < c \leqslant 0.6$ 中度拥挤 $ 0.6 < c \leqslant 1 $ 重度拥挤 $ c>1 $ -
[1] 裴巧娜. 基于光流法的运动目标检测与跟踪技术[D]. 北京: 北方工业大学, 2009. [2] 丁雪梅,王维雅,黄向东. 基于差分和特征不变量的运动目标检测与跟踪 [J]. 光学精密工程,2007,15(4):570-576. [3] 罗会兰,陈鸿坤. 基于深度学习的目标检测研究综述 [J]. 电子学报,2020,48(6):1230-1239. doi: 10.3969/j.issn.0372-2112.2020.06.026 [4] 卢湖川,李佩霞,王 栋. 目标跟踪算法综述 [J]. 模式识别与人工智能,2018,31(1):61-76. doi: 10.16451/j.cnki.issn1003-6059.201801006 [5] 董建国. 基于形状的图像目标识别方法研究[D]. 西安: 西安电子科技大学, 2009. [6] 郑 胤,陈权崎,章毓晋. 深度学习及其在目标和行为识别中的新进展 [J]. 中国图象图形学报,2014,19(2):175-184. [7] 段旺旺,杜志强,许华胜. 基于视频监控的动车组车厢拥挤度分析 [J]. 智慧轨道交通,2022,59(4):61-66. [8] 冯 勇,许华胜,段旺旺,等. 计算机视觉技术在列车行李遗留检测上的应用 [J]. 铁道车辆,2020,58(9):15-17. [9] 刘长伟. 基于MTCNN和Facenet的人脸识别 [J]. 邮电设计技术,2020(2):32-38. [10] 邱天衡,王 玲,王 鹏,等. 基于改进YOLOv5的目标检测算法研究 [J]. 计算机工程与应用,2022,58(13):63-73.