Cost-effective Computer Vision Solutions for Smart Transportation

Urban Work Zone Detection and Safety Analysis Using Near Miss Data

While computer vision technology is widely acknowledged, it remains underutilized for routine operations and traffic safety in complex urban environments. Our research develops a deep learning-based, automatic data acquisition and analytics approach utilizing vision-based sensors. This approach harnesses data from existing transportation infrastructure and connected and automated vehicles (CAVs), illustrating how computer vision can create fresh streams of mobility and safety information for strategic planning and operations. It presents two applications: WorkZoneX, leveraging 900+ traffic cameras in NYC for real-time urban work zone identification and size estimation, and SAFExMAP, a risk indicator scoring system using near-miss data from in-vehicle cameras. Findings suggest significant potential for computer vision in smart cities, offering cost-effective solutions to improve planning, operations and traffic safety analysis.

Presented by Dr. Jingqin Gao and Dr. Chuan Xu