GroupRugs: Visual Summaries for Groups in Collective Movement Data

As more and more moving objects are tracked, the amount and variety in trajectory data is ever increasing. Visual summaries provide an at-a-glance overview of such trajectories and are therefore a useful tool for exploring large trajectory collections. Typically, such a visual summary visualizes the spatial positions of moving entities using a one-dimensional representation and combines such representations by placing them in temporal order along a time line. However, existing summaries are generally not tailored to specific patterns that arise in the trajectory data. The formation of groups is a quintessential pattern that emerges in the collective motion of many types of tracked objects, such as humans, birds, and other animals. Our main contribution is GroupRugs, a visual summary technique for collective movement data that highlights the structure of the emerging groups and their evolution over time, while still summarizing the spatial relations in the data. Specifically, we introduce two methods to produce GroupRugs: a naive baseline approach, and a pipeline that optimizes several aspects of GroupRugs. The quality of a visual summary is usually assessed via two main criteria: spatial quality, which measures how well the one-dimensional representations capture the structure of the data points at each time step, and stability, which captures the coherence of consecutive one-dimensional representations over time. For GroupRugs, we additionally care for how well the structure of the emerging groups is expressed in the visualization. In extensive computational experiments, we quantitatively evaluate GroupRugs against state-of-the-art techniques for summarizing trajectories, using well-established metrics for the three important quality criteria. Our evaluation shows that GroupRugs greatly outperform existing techniques in expressing the structure of emerging groups, while sacrificing only little spatial quality or stability.