Abstract: The operation environment of railway passenger stations is characterized by high mobility and complexity, and the limitations of the traditional manual service mode easily lead to delayed service responses and supervision blind spots for key passengers. To address these problems, this paper investigated the intelligent perception and service closed-loop technology for key passengers at railway passenger stations, and proposed a cascaded network based on RailYAV-Net. This network adopted an improved YOLOv8 model integrated with a Transformer architecture to automatically identify key passengers. It further used VAE technology to track the full-domain trajectories of key passengers in real time and monitor whether there were staff members around them, so as to achieve a closed-loop guarantee of “detection identification service verification”. Field application results show that this intelligent perception and closed-loop service technology significantly improves the intelligence, accuracy and timeliness of services for key passengers at railway passenger stations, and provides technical support for ensuring their travel safety.