乘务员值乘状态监测技术研究

Research on the Monitoring Technology of Locomotive Crew On-Duty State

利用多维度视频融合感知技术对机车乘务员值乘状态进行实时分析,通过乘务员前方图像采集设备和后上方图像采集设备实时获取视频,先对视频图像进行动态自适应校正,再利用mobilenet-yolov3轻量级网络检测视频中乘务员的人脸多维特征信息、玩手机以及驾驶室车门开闭等特征,采用Hu矩不变量特征辅助分析乘务员的盹睡状态,从而及时提醒司机规范驾驶行为,实时预警,对保障行车安全具有重大意义。

The multi-dimensional video fusion perception technology is used to conduct real-time analysis on the duty status of locomotive crew members. The video is obtained in real time through the image acquisition equipment in front of the crew members and the image acquisition equipment at the upper back. First, the video image is dynamically self adaptively corrected. Then, the mobile yolov3 lightweight network is used to detect the multi-dimensional feature information of the crew members’ faces in the video, mobile phones, cab door opening and closing and other features, Hu moment invariants are used to assist the analysis of the sleeping state of the stewards, so as to timely remind drivers to regulate driving behavior and give real-time warning, which is of great significance to ensure driving safety.