Abstract: To deal with the key hazards existing in the working process of train drivers, an intelligent real-time monitoring system for train driver was studied and developed. By using deep learning technology and embedded video analysis devices, this system can effectively recognize unsafe behaviors such as deviated line of vision, abnormal specific gestures, and fatigue driving. Meanwhile, the health and sleep-before-shift data of the train drivers can be acquired in a real-time manner by using Internet of Things technology so that the train drivers' health troubles and sleep insufficiency can be timely discovered and alarms will be sent out immediately. This system can change the traditional mode of supervision on the train drivers by lagging manual selective examination on train drivers’ bad driving behaviors and health as well as accountability after an event and replace it with real-time automatic machine detection and timely alarming in a more accurate and time-effective manner.