With the development of information technology, massive traffic data is exploited for traffic behavior analysis. In particular, taxi trajectory data plays an important role in traffic analysis. In order to better explore the traffic law of taxis in urban traffic operation, a graph based visual method is applied to the analysis of urban traffic flow in this paper. First, the road-level graph is converted into a region-level graph through clustering all road segments, then the node-line connection between regions presents the urban road network structure. Meanwhile, taking the regional traffic as the weight, the importance of the regions is analyzed by combining graph centrality. Finally, this approach is implemented and evaluated by using more than 100 million taxi trajectories data within the fourth ring of Beijing. The experiment results indicate that the proposed method can intuitively show the variation of taxi flow in different regions over time and reveal effectively the importance of different regions.