交通视觉语言大模型

Search documents
公路重大异常“秒捕捉” “鹰眼守护”预警技术撑起出行“安全伞”
Ke Ji Ri Bao· 2025-09-29 09:48
Core Insights - The "Eagle Eye Guardian" early warning technology, developed by the China Academy of Safety Science and Technology in collaboration with Amap, aims to enhance emergency response efficiency for major road incidents and ensure public safety during travel [1][2]. Group 1: Technology Overview - The "Eagle Eye Guardian" system integrates multi-source spatiotemporal data, including map navigation and mobile positioning, to achieve real-time perception, warning, and alerting for road and bridge disaster events [1]. - This system is an upgraded version of Amap's "lane-level safety warning" capability, utilizing vast amounts of anonymized vehicle data to analyze abnormal driving behaviors and provide timely alerts to drivers [1][2]. Group 2: System Capabilities - The system can detect collective abnormal vehicle behaviors, significantly improving the effectiveness of warnings compared to individual vehicle detection [2]. - It employs the "TrafficVLM" traffic visual language model to convert extensive traffic data into dynamic twin video streams, allowing for real-time identification of anomalies and trend analysis [2]. Group 3: Testing and Implementation - Real-world testing conducted on August 23 showed that vehicles could receive alerts within seconds during significant abnormal events, with a coverage distance of nearly 1 kilometer and an accuracy rate exceeding 90% [2]. - The "Eagle Eye Guardian" system is now operational nationwide, covering 17 common driving scenarios, enhancing safety for users who opt into the system [3].