Improving visualization of trajectories by dataset reduction and line simplification

Abstract

When visualizing a dataset of trajectories, whether it’s hurricane paths, pedestrians or car movements, overlapping and very similar trajectories cause the visualization to become obstructed and features of the overall set can go unnoticed by analysts. This clutter characteristic is common in data representation but hasn’t been consistently defined and quantified. The current thesis approaches the definition and measurement of clutter in the context of representing spatial trajectories and presents two methods of lowering the clutter values to a set degree. The first method investigates removing trajectories from the dataset in such a way as to maximize the improvement in the clutter measure. The second one is an adaptation of the Imai-Iri algorithm for line simplification, geared towards minimizing the intersection points within the dataset, thus lowering clutter. The two methods, applied on a real dataset of human movement trajectories, result in significant improvements of the clutter measure relative to an established goal. Both methods are interpreted with regard to human perception and a conclusion is drawn on which method is more efficient depending on the goal of the user; further improvements to the method are discussed in the future work section.