Based on the RTMS (Remote Traffic Microwave Sensor) data, FCD (floating car data) and plate number data collected from urban expressway, a travel speed estimation method based on BP (back-propagation) neural network is presented in this study. According to the spatial and temporal characteristics of traffic data, three kinds of data complement methods are respectively presented first, and then six data fusion models are established for each data missing and complement status. The input data includes average travel speed from FCD and traffic volume, spot speed, time occupancy rate from RTMS, and the output is the average travel speed estimation. In the model training phase, the travel speed calculated from plate number data is viewed as the real value. Finally, the models are examined by realistic traffic data with two evaluation indicators. The result shows that the fusion models can provide more effective and more accurate traffic information.
展开▼
