Real-time multi-source AI, computer vision, and GIS fusion for data-driven urban traffic management

• Real-time multi-source traffic monitoring framework integrating AI, computer vision, and GIS. • Fusion of video analytics and smartphone sensing for traffic parameter estimation and validation. • High vehicle detection performance achieved using YOLOv8 (precision 0.93, recall 0.89). • GIS-based dashboards enabled segment-level mapping and spatiotemporal traffic analysis. • Cost-efficient and operationally practical approach for traffic monitoring in resource-constrained cities. Urban traffic management in developing cities is often constrained by limited sensing infrastructure, delayed data availability, and fragmented analytical tools. This study presents, deploys, and validates a real-time, multi-source traffic monitoring framework that integrates AI-based computer vision, smartphone sensing, and geographic information systems (GIS) within a unified operational platform. The objective of this work is not to introduce new analytical algorithms, but to demonstrate the practical integration and field deployment of heterogeneous data sources for real-time traffic monitoring under real urban operating conditions. The framework was deployed across thirty heterogeneous traffic sites in Irbid, Jordan, including intersections, arterials, and mid-block segments, and evaluated under peak and off-peak conditions as well as daytime and nighttime scenarios. The detection and tracking pipeline achieved high accuracy (precision = 0.93, recall = 0.89, F1-score = 0.91), while speed estimates were in close agreement with radar-based ground-truth measurements, yielding a mean absolute error of approximately 2.1 km/h. Validation against independent manual field counts using paired statistical testing indicated consistent performance without systematic bias. Integration with a PostGIS-enabled GIS environment enabled segment-level speed mapping, congestion hotspot identification, and spatiotemporal assessment of traffic dynamics. These outputs were operationalized through real-time dashboards that provide continuous situational awareness and support data-driven traffic management decisions. The results indicate that the proposed framework provides a cost-efficient and operationally practical alternative to conventional traffic monitoring approaches, suggesting a transferable approach for real-time urban traffic monitoring in resource-constrained cities.