Crowdsourced Camera Data Fusion for Urban Traffic Estimation and Monitoring

Abstract

Data quality is paramount in crowd-sourced traffic information systems where expanding data fusion methods, ensuring accuracy and reliability are essential. Conventional traffic information systems collect data through multiple methods such as web-based forms, speech data, and GPS sensors on mobile devices. This paper enhances these methods by introducing new approaches to traffic data fusion from cameras that leverage existing city surveillance infrastructures to provide a continuous stream of real-time traffic data. We also devise a novel approach for background scheduling to update traffic status predicted from camera-based data applying AI models. These novelties are introduced to combat the current ITS systems’ struggle to maintain continuous flows of user-reported crowd-sourced data. Then we design necessary components to implement background data fetching scheme to insert traffic camera data into a traffic information platform to evaluate the proposed approaches in real-world scenarios. The results revealed the effectiveness and efficiency of the proposed mechanisms showing that they are ready to be applied in real-world applications.