ABSTRACT
At present, the computation of the urban road traffic operation index is mainly based on the GPS data of floating cars in China. Because the number of the floating cars is not enough to cover the whole road network efficiently and the GPS data is affected by the positioning error, the road traffic operation index based on the GPS data only of the floating cars is not sufficiently accurate, which inevitably influences the validity and authority of the index. The key aim of the multi-resource traffic data fusion is to provide online, dynamic and accurate data for the computation of the index. In this paper, a segment of an urban expressway equipped with the geomagnetic detectors and the floating cars in Shenzhen is selected as the scenario for the data collection. Three data fusion methods are proposed, respectively based on the confidence in traffic sensors’ property, the artificial neural network and the least squared nonlinear support vector machine (abbreviated as LS-SVM). The method based on confidence is simple and straightforward and its computing speed is very fast, but the confidence level of the traffic sensor is required according to its operating performance. The accuracy and reliability of the method based on the artificial neural network relies on the structure and the size of the training sample for the neural network. The method based on LS-SVM shows the best data fusion accuracy.
