Trajectory data is a very common kind of spatio-temporal data, which keep track of the movement by sampling points. According the sampling rate, trajectory data set can be categorized into sparse trajectory and quasi-continous trajectory. Sparse trajectory records the movement in low sampling rate, such as the cell-based RFID data, social media data, etc. Quasi-continous trajectory records the moving in high sampling rate. One typical example of quasi-continous trajectory data is the city GPS traffic trajectory data, e.g. pedestrians, bicycles and various other vehicles. Effective visualization and analysis of those trajectories may help us better understand the urban transportation system, find out strategies to reduce the number of accidents and traffic jams, thus ensuring people's safety, facilitating people's trip and elevating the economic efficiency. In a series of projects, we studied in both two kinds of trajectories from several aspects, including: traffic density, traffic jams, traffic pattern correlation, travel behaviour and semantic derivation.
We develop an interactive visual analytics system to help users explore the trajectories derived from geo-tagged social media. An uncertainty model integrated with the system is proposed to help users find reliable patterns from such sparse trajectories.
We develop an interactive visual analytics system to help users explore individual trajectories derived from geo-tagged social media. Users can explore their own and friends' footprint with multi-level spatial temporal visualization. Many interesting individual patterns could be derived with such visualilzation for the mass.
We present a visual analysis system to help law enforcement department to detect and track the trajectories and movement patterns of fake mobile base stations based on the indirect information collected from victim mobile phones.
We show how to explore sparse traffic trajectory generated by urban transporation cells. We focus on macro-traffic analysis, including the traffic patterns on cells and links. We further study the correlation between those patterns.
We propose a visual analytics approach to help analysts more efficiently define and detect data quality problems in the raw trajectory data. Based on problematic trajectories identified by users, we automatically disclose more trajectories with similar problems. We also support users to improve the results until they feel satisfied.
We develop a visual method to explore the route choice behaviour among multiple routes. Given a pair of interested origin and destination, all feasible routes are extracted and compared. It supports users to develop and verify the hypothesis on the impacting factors of route choice.
We develop a visual method, OD-Wheel, to explore OD patterns of a central region. Given a central region, OD-Wheel allows users to explore the dynamic patterns of OD clusters, including the variation of traffic flow volume and travel time.
We develop a visual method, TrajRank, to study the travel behaviour of vehicles along one route. We focus on the spatial-temporal distribution of travel time, that is, the time spent on each road segment, and the travel time variation in rush/non-rush hours.
We show how 2D trajectories can be converted to timelines for easy perception of temporal information, and easy aligned comparison. Our timelines are not limited to time varying attributes of trajectories, but also spatial temporal features such as stop, turn, etc. We provide some use cases to show the benefit of our method.
We show the variation of traffic conditions for each road, and the propagation of traffic jams along the road network. We used the taxi GPS trajectories of Beijing, then automatically extracted the traffic jam information. We allow user to further explore the traffic jams through multilevel exploration and multi-faceted filtering.
We develop a bus visualization interface, Bus Vis, to explore urban bus data. Bus Vis allows users to explore the urban bus information, including traffic animation and passenger counts with time variation.
Traffic density rendering is important, because it gives an intuitive overview of massive trajectories. It is also related to the study of traffic jams, hot spots and people's behaviors. We use density map algorithm to generate pictures of traffic density in Beijing, with real taxi GPS data.
We built a system, TripVista, to explore the micro behaviors of trajectories in three linked perspectives: space, time and attribute. We applied the system to traffic trajectory data at a road intersection, and a global hurricane trajectory dataset. We are able to give an overview and discover some outliers.