Objective Metro video surveillance analysis suffers from a disconnect between intelligent analytics and field operations, delayed responses, and difficult closed-loop management. To shift the operational mode from passive inspection and delayed detection to proactive warning and collaborative handling, it is necessary to construct an integrated "sensing - transmission - application" architecture for metro video surveillance systems.

Method  An integrated "sensing - transmission - application" architecture is proposed. High-performance video analytics algorithms are developed and iteratively optimized using custom datasets to improve anomaly recognition accuracy. A semantic-segmentation-based compression encoding technique is applied to differentiate dynamic targets from static backgrounds for efficient compression. Station mobile agents are engineered to automatically convert AI （artificial intelligence） alerts into mobile tasks embedded with video, location, and contingency plans, establishing a closed loop of "alerting - handling - feedback - reporting."

Result & Conclusion Following the deployment of the integrated "sensing - transmission - application" architecture and station mobile agents, false negatives and false positives decrease significantly during the statistical period, and video storage achieves high-ratio compression. The mobile integrated handling app enables alert pushing and collaborative response. By integrating video analytics results with field workflows, the app automatically generates reports and assists in daily station inspections, raising operational efficiency by 50%. This resolves the disconnection between alert outputs and field operations, achieving process restructuring, service reshaping, and complete operational closed-loop management.