Highway Traffic Condition Detection with Data Fusion

With the popularity of Internet of Things (IoT), sensing and sharing various data become rapid and easy, which promotes the shift of the research hot spot to data mining and exploiting. For effectively utilizing the collected data to detect highway traffic flows, two challenges must be dealt with. One is the sparsity inherent in the sensory data of a single kind, and the other is the multiple goals expected by the detecting requirements. In order to detect traffic conditions of entire highways in a high-precision, low-cost and quasi real-time way, this paper presents a data fusion approach named Megrez, which takes advantage of the signaling data in a mobile communication network and the data from vehicle detectors. The Megrez approach not only mines the inherent features of the sensory data to reconstruct the missing data, but also incorporates with the characteristics of traffic flows while rectifying the reconstructed data. Since the Megrez approach works in a non-intrusive way, it can carry out the traffic monitoring with full road segment coverage at a very low cost. Meanwhile, with the support from the parallel linear algebra library, the implemented Megrez approach can estimate the vehicle speeds in a designated interval. Using the large-scale real-world data as input, this paper evaluates the Megrez approach from different perspectives. The experimental results show that the vehicle speeds estimated by the Megrez approach have high precision and the Megrez approach can accurately detect the traffic conditions on highways.