Digital twin technology offers tremendous potential for transforming intelligent transportation systems (ITS) by providing a high-fidelity virtual representation of the physical traffic environment. However, a major challenge is efficiently acquiring accurate spatiotemporal data to ensure dynamic consistency between the physical and virtual domains. This study proposes a vision-based, real-time vehicle risk assessment method that utilizes monocular roadside cameras without the need for additional sensors. This method integrates camera auto-calibration, vehicle speed measurement, and risk assessment into a unified framework, enabling low-cost and reliable data acquisition. This method provides a real-time, efficient, and low-cost approach for constructing digital twins and supports their application in dynamic safety monitoring and predictive assessment. Simulation experiments in traffic monitoring scenarios demonstrate the effectiveness of this method in capturing vehicle motion and performing real-time risk assessment, providing valuable support for advancing the development of transportation digital twin systems.
Miao et al. (2026) studied this question.