Traffic Video and VANET data fusion algorithm

Modern Intelligent Transport Systems incorporate the traffic control strategies that are based not only on long term traffic analysis and forecasts, but also on the real time events detection like accidents or high congestion. The flexibility of these systems depends on accurate and precise data set describing the current state of road network. To estimate it, the data from various sources like: video surveillance, induction loops or vehicles itself (Vehicle to Infrastructure communication -V2I) is gathered. Excluding detection errors, the video surveillance data is a reliable source of general information about the traffic flow. On the other hand, the vehicle communication can provide less reliable, but more detailed information about a particular vehicle like: its engine state or planned manoeuvre. Unreliable or forged C2I information can be used to disturb traffic or to gain a higher priority on the road. The paper reviews the fusion algorithms that are used to merge data from video tracking algorithms and vehicular networks. Based on the survey, a weighted fusion algorithm is proposed that estimates the acquired data reliability. The algorithm uses the video surveillance data as a filter for C2I communication. Finally, applications for microscopic traffic models and safety issues are taken into consideration.