Visualization of Traffic Data: A Survey of Methods and Datasets

Abstract

With the rapid development of cities, massive and complex traffic data is being generated and collected. The traffic data is not intuitive and cannot highlight key information about urban traffic conditions. However, traffic data visualization can directly correlate users with the data, and support users to interact with data in a convenient and visual way. Then realize the feedback of blending user wisdom and machine intelligence. This paper investigates a structured survey of the state of the art in the visualization of traffic data. First, we reviewed five representative traffic data visualization methods including WebVRGIS based traffic analysis and visualization system, TripMiner, IoV distributed architecture, SMASH architecture, and LDA-based topic modelling. Meanwhile, we analyzed the traffic datasets that applied in each method. Then we summarize these methods from seven aspects: scalability, data storage, data update, interactivity, reliability, data anomaly detection, and spatiotemporal visualization. In addition, we make a detailed comparative analysis of the key capabilities of five representative traffic data visualization methods in processing traffic big data. Finally, we conclude that the SMASH architecture performs better in processing high speed and large flow traffic data. Moreover, we propose a novel direction for optimizing traffic data visualization techniques.