US Ignite: Focus Area 1: Fast Autonomic Traffic Congestion Monitoring and Incident Detection through Advanced Networking, Edge Computing, and Video Analytics

点击翻译按钮获取中文摘要

Video-based traffic monitoring systems have been widely used for traffic management, incident detection, intersection control, and public safety operations. Current designs pose critical challenges. First, it relies heavily on human operators to monitor and analyze video images. Second, commercially available computer vision technologies cannot satisfactorily handle severe conditions, such as weather and glare, which significantly impair video image quality. Third, the simultaneous transmission of numerous video signals to a central facility creates extreme demands on the communications network, which can lead to jamming. This project presents a novel approach that incorporates wireless sensor networks, hierarchical edge-computing, and advanced computer vision technology. The methods can be expanded to address a wide spectrum of potential applications including wrong-way driving alerts, congestion detection under bad weather conditions, accident scene management support, suspect vehicle tracking, wildfire detection and alert, and emergency evacuation, which could save lives and hundreds of billions of dollars annually. It also aligns with the smart city initiative.By using bluetooth/WiFi detection technology, the trajectories and speeds of vehicles equipped with such devices will be collected. This information, along with the captured video data, will be analyzed by the proposed computer vision software, installed at the edge of the network on cloudlets, to perform fast detection and prioritization of the video streams from different cameras. The proposed hierarchical edge-computing paradigm will not only enable real-time big data analysis at the edge but will also be demonstrated and actualized to perform timely efficient video analytics. Depending on the weather conditions, different detection and prioritization algorithms will be activated. Video coding will then be implemented to transmit the selected video streams to the central back-end system for further processing. If an incident is detected by the algorithm either at the edge or at the back-end, a necessary feedback action will be taken, such as calling an emergency group, the highway safety dispatch, or the police. Under a technical partnership with New Jersey Department of Transportation, multiple pilot tests of the proposed system will be implemented on selected highway corridors designated by the department.