Abstract
INTRODUCTION: The rising incidence of road traffic crashes has emerged as a pressing global concern, representing a multifaceted challenge to public health, urban safety, and sustainable mobility. With rapid urbanization and increasing vehicular density, traffic environments have become highly dynamic, multi-agent systems characterized by complex interactions among heterogeneous road users, including vehicles, cyclists, and pedestrians. These interactions occur within high-dimensional and context-rich environments that are difficult to model using traditional approaches. Consequently, there is an urgent need for robust, data-driven frameworks capable of capturing the intricacies of real-world traffic scenarios. METHODS: In this study, we propose a novel computational framework designed to enhance the analysis of traffic safety through a multidisciplinary lens. Our approach integrates advanced video data analytics with artificial intelligence techniques, emphasizing the fusion of spatiotemporal modeling, behavioral analysis, and environmental context. RESULTS: By leveraging large-scale, in-situ video data from urban intersections and road networks, our method provides a granular understanding of risk factors and interaction patterns that precede collisions or near-miss events. DISCUSSION: This framework is tailored for the challenges posed by complex, heterogeneous traffic systems and aligns with current research interests in deploying AI for risk-sensitive, behavior-aware decision-making within urban mobility infrastructures.