铁路客运车站旅客异常行为智能识别和监测方法研究

Research on intelligent recognition and monitoring method of railway station passenger abnormal behavior

摘要:

目前国内多数车站仍采用视频回放方式识别旅客异常行为，无法保障效率和精确度，影响车站安全稳定运营。为提升车站旅客异常行为监测的智能化水平，采用CNN算法，选取站台区域越界旅客为重点研究对象，对旅客异常行为进行智能识别。选取候车大厅、检票口等区域，采用MCNN算法对人群密度进行识别和监控。借助仿真平台和车站现场数据模拟验证，结果显示距站台第一、二边界内的人数均占当前视频画面总人数12%左右，识别率达90%，该结果可为车站安全保障业务和客运组织优化提供作业指导，以保障车站安全稳定运营。

Abstract:

At present, most stations in China still use video playback to identify abnormal behavior of passengers, with low efficiency and accuracy. In order to improve the intelligence of station passenger abnormal behavior monitoring, we adopt Convolutional Neural Network(CNN) algorithm, select the cross-border passengers in the platform area as the research object, and conduct intelligent recognition for the abnormal behavior to intelligently identify their abnormal behaviors. The Multi-Column Convolutional Neural Network(MCNN) algorithm is used to identify and monitor the crowd density in the waiting hall, ticket gate and other areas. Finally, we build a simulation platform, using the real video monitoring data to verify the models. With the help of simulation platform and station field data simulation verification, the results show that the number of people within the first and second boundaries of the platform accounts for about 12% of the total number of current video images, and the recognition rate is up to 90%. The results can provide operation guidance for the station safety guarantee business and passenger transport organization optimization, so as to ensure the safe and stable operation of the station.