Abstract: In order to promptly identify potential safety hazards and abnormal events in railway passenger transport sites and intervene proactively, this paper designed a passenger flow analysis and abnormal alarm system for passenger stations based on image analysis. The paper utilized existing high-definition cameras, intelligent sensors, and other devices in passenger stations to obtain real-time images of passenger transport areas, combined image analysis technology and algorithms, continuously optimized object detection and deep learning models based on the specific needs of passenger stations, applied complex feature extraction methods, optimized abnormal classification alarm thresholds, and implemented real-time monitoring, analysis, prediction, and abnormal alarm of passenger flow, passenger crossing or falling down, public fire safety and other safety hazards in passenger stations. The pilot operation based on a certain station shows that the system can accurately identify important events such as passenger flow density, abnormal personnel behavior, and fire early warning, effectively improves the safety early warning capability of the station, as well as the level of station safety management and transportation services.