Moving objects are captured in multivariate trajectories, often large data with multiple attributes. We focus on vessel traffic as a source of such data. Patterns appearing from visually analyzing attributes are used to explain why certain movements have occurred. In this research, we have developed visualization methods that result in these patterns for large data sets of multivariate trajectories and aim to answer the following research question: How can analysts be supported to understand large amounts of trajectories with multiple attributes by using interactive, visual representations? We have taken two research directions: one based on density visualization and the other one on visual analytics. Our main contribution is a definition for a density field of trajectories that gives for each point in the field the fraction of the time an object stays in that point. We have used this density in three different ways. The basic approach, called vessel density, shows this density at different scales simultaneously and various (maritime) movement features become visible, such as sea lanes (maritime highways) and anchor areas (maritime parking places). The second approach, called density maps, allow us to use subsets described by attributes, such as a ship type, to show the various distributions for each individual subset. Density maps allows us to distinguish between various moments in time, to conduct basic anomaly detection, and to explore the various attributes available in the data. The third approach is called composite density maps. Based on experience and domain knowledge, it allows users to describe expressions for sophisticated density fields. With this approach it is possible to extract sea lanes, find objects moving in the wrong direction of a sea lane, and conduct risk assessment. Next to these three approaches, we have evaluated vessel density in comparison with other trajectory visualizations. It shows that for some maritimemovement features vessel density excels (stops) and for othermovement features (fastmovers and finding sea lanes) vessel density performs not significantly worse than the other visualizations. Next to the density approaches, we have created a visual analytics framework with integrated results fromthe Poseidon project that allows us to generate, simplify, enrich, and reason with trajectories, shown in an interactive matrix visualization with small renderings in the cells. The key running example shows that a ship enters a harbor where it is not expected to be. The results of this research can be used to improve situational awareness in, for instance, vessel traffic monitoring systems by including historical data in the analysis of live data. Since movement of objects is of interest in many different disciplines, such as in traffic monitoring systems, urban design, and animal migration, we expect the approaches to be useful for those application domains as well.
|Qualification||Doctor of Philosophy|
|Award date||21 Dec 2011|
|Place of Publication||Eindhoven|
|Publication status||Published - 2011|