Abstract

Smart traffic management and monitoring is one of the key aspects of the modern smart city. Traffic flow estimation is crucial for sustainable traffic planning in the city. A requirement for successful planning and optimization of traffic is vehicle counting on the streets. Surveillance video is a suitable data source for precise vehicle counting. A solution for real-time vehicle traffic monitoring, tracking and counting is proposed in Jelgava city, Latvia. It is based on motion detection using background modeling, which is enhanced by statistical analysis. Two-phase assessment is utilized: motion blobs are detected and tracked using custom state machine implementation, then tracking results are passed through number of statistical filters to eliminate false positive detections. The system demonstrates good performance and acceptable accuracy on given test cases (about 97% accuracy for regular traffic conditions).