Multisource Traffic Data Fusion with Entropy Based Method

It is a crucial part for ATMS to accurately identify and forecast traffic state from real-time traffic data. To improve the identification rate of traffic state, multisource information should be used. The multisource information fusion method is important. Information fusion is divided into three levels, i.e. data level, feature level, and decision level. In traffic congestion identification, many means collected traffic data source can be used, such as induce loop vehicle detector, video detector, GPS floating car and so on. The traffic state can be identified according to each data source. For improving the identification rate, we develop a decision level multisource fusion method. In the method, Bayesian inference is used to obtain the traffic state in probability style according to each data source, and entropy based weighted method is used to fuse the result in decision level to improve the identification rate. The entropy based fusion model and algorithm is introduced and presented in this paper. Field data collected through loop vehicle detector and GPS floating car are analyzed with the proposed method.