Abstract:

To thoroughly investigate the traffic state of highways, Jiqing Highway was selected as the study case. By mining the gantry data, a two-stage clustering algorithm combining k-means and density-based special clustering of applications with noise (DBSCAN) algorithms were proposed. The method was used to identify vehicles entering the service area and driving abnormally. Subsequently, the filtered vehicle records were extracted to realize a traffic state index weighted by the vehicle type to analyze the traffic state of the highway in terms of spatiotemporal dimensions. Results indicate that the two-stage clustering algorithm performs very well in the identification. The traffic state index indicated three periods when the highway is defined as congested during 7:00—20:00. Furthermore, it accurately identifies the congested sections of the highway. Moreover, it shows out that the mixed rate of large vehicles and the degree of traffic congestion in a section have a close positive correlation. Finally, according to the evaluation index, the traffic state of the Jiqing Highway is divided into four levels, which provides technical support for the traffic authorities to evaluate and manage the highway sections.

Key words: gantry data, traffic condition, clustering algorithm, traffic state index, mixed rate of large vehicles