A Method for Road Network Selection Considering the Traffic Flow Semantic Information

Road traffic flow is an important feature to measure the importance of the road. Based on the GPS trajectory data and road network data, this paper proposed a novel method for the selection of road network considering both the commonly used geometric features (i.e. length, topological connection, distribution density, closeness and betweenness) and the sematic information—long-term week average traffic flow of the roads. Firstly, in order to ensure the continuity of the road network at small scale, the strokes of the original road network are obtained by using the continuity and direction constraints of Gestalt laws. Then, the geometric features and the traffic flow information of each strokes are computed. Further, a global measure is defined based on the geometric features and traffic flow information, and the weights between the length, topological connection, distribution density, closeness, betweenness and traffic flow are determined using the modified entropy-weight method. Finally, the proposed method of this paper is verified by using Wuhan road network data and taxi GPS trajectory data. The experimental results show that the proposed method can maintain the connectivity of the original road network and tend to select the important roads that are consistent with the human perception. On the other hand, this method can be used to discover the roads that are designed as high-level roads but in contrary rarely used in reality for applications such as urban road network planning and reconstruction.