Acquisition of traffic flow density using multi-source data fusion

Traffic flow density is one of the most important parameters in traffic flow theory. It provides basic data for urban traffic management and control. However, it is highly difficult to directly collect real-time traffic flow density. The existing methods are simple in form, which are not suitable for complex traffic situations. The main purpose of this paper is to studying acquisition method of real-time traffic flow density by fusing Floating Car Data (FCD) and geomagnetic detection data. Least Squares Support Vector Regression (LS-SVR), which has properties such as global convergence and strong generalization capacity, is introduced to accomplish multi-source data fusion. The experimental result indicated that our approach has higher estimation accuracy than the traditional models.