Dr. rer. nat. Dipl.-Inform. Michael Burch
Email: michael.burch@visus.uni-stuttgart.de


VISUS - Institut für Visualisierung und Interaktive Systeme - Stuttgart

20110221339142935291937101629163617371


Comparative Eye Tracking Study on Node-Link Visualizations of Trajectories

Paper pdf-Version
Examples of different node-link visualizations: they show data from animal movement ecology in the form of GPS tracks
of two oystercatcher birds (distinguished by red/blue color). The visualizations show zoomed-in views of a large data set. Due to
zooming, there are some links for which neither start nor end node are visible; here, the texture (right image) provides an indication
of movement direction that is completely missing with the standard arrow rendering (left image).

Examples of different node-link visualizations: they show data from animal movement ecology in the form of GPS tracks of two oystercatcher birds (distinguished by red/blue color). The visualizations show zoomed-in views of a large data set. Due to zooming, there are some links for which neither start nor end node are visible; here, the texture (right image) provides an indication of movement direction that is completely missing with the standard arrow rendering (left image).

We present the results of an eye tracking study that compares different visualization methods for long, dense, complex, and piecewise linear spatial trajectories. Typical sources of such data are from temporally discrete measurements of the positions of moving objects, for example, recorded GPS tracks of animals in movement ecology. In the repeated-measures within-subjects user study, four variants of node-link visualization techniques are compared, with the following representations of directed links: standard arrow, tapered, equidistant arrows, and equidistant comets. In addition, we investigate the effect of rendering order for the halo visualization of those links as well as the usefulness of node splatting. All combinations of link visualization techniques are tested for different trajectory density levels. We used three types of tasks: tracing of paths, identification of longest links, and estimation of the density of trajectory clusters. Results are presented in the form of the statistical evaluation of task completion time, task solution accuracy, and two eye tracking metrics. These objective results are complemented by a summary of subjective feedback from the participants. The main result of our study is that tapered links perform very well. However, we discuss that equidistant comets and equidistant arrows are a good option to perceive direction information independent of zoom-level of the display.