An Approach Based on Video Analysis for Real-Time Monitoring of Passenger Flow in Subway Station

In urban rail transit system, The passenger flow of urban rail transit continues to grow. Monitoring and analyzing passenger flow density within subway stations has become a key requirement for ensuring operational safety and optimizing service quality. Traditional manual statistics and fixed sensor monitoring have problems such as low efficiency, multiple blind spots in coverage, and insufficient real-time performance. This article proposes a real-time perception method for subway station passenger flow based on video analysis, aiming to achieve intelligent processing of passenger flow quantity statistics, behavior trajectory tracking, and abnormal event warning in complex scenarios. This article focuses on the video surveillance network of key areas such as subway platforms, turnstiles, and escalators, and constructs a passenger flow video monitoring method based on the YOLOv5 optimization algorithm. The prediction accuracy and box effect of the YOLOv5 algorithm are improved through loss function optimization, achieving individual feature extraction and cross camera trajectory correlation in high-density populations, and solving the problem of target missed detection and false tracking under complex conditions such as occlusion and lighting changes. Experiments have shown that this method achieves an accuracy rate of 86.82% in object detection in dense passenger flow scenarios, with an average detection time of 0.026 s, meeting real-time requirements.