Study on Artificial Intelligence Edge System for Detection of Abnormal Behavior in Trains

In this paper, we propose an artificial intelligence edge System that supports operators to respond quickly by detecting abnormal behaviors such as fainting, assault, and not wearing a mask in real time in a train. The AI edge System mounted on the vehicle consists of a CCTV to acquire image data, an NVR that stores the acquired image data, and an edge Server that executes artificial intelhgence algorithms. The edge Server notifies the railroad control center and the train driver of the detected abnormal behavior in real time. The proposed System was tested on one train, and real-time operability and effectiveness were confirmed. If an edge System is added to CCTV to be installed in trains in the future, it is expected that it will be possible to actively respond to abnormal situations in trains and improve the safety of railroad passengers.