Labeling Points of Interest in Dynamic Maps using Disk Labels

Abstract

Dynamic maps which support panning, rotating and zooming are available on every smartphone today. To label geographic features on these maps such that the user is presented with a consistent map view even on map interaction is a challenge. We are presenting a map labeling scheme, which allows to label maps at an interactive speed. For any possible map rotation the computed labeling remains free of intersections between labels. It is not required to remove labels from the map view to ensure this. The labeling scheme supports map panning and continuous zooming. During zooming a label appears and disappears only once. When zooming out of the map a label disappears only if it may overlap an equally or more important label in an arbitrary map rotation. This guarantees that more important labels are preferred to less important labels on small scale maps. We are presenting some extensions to the labeling that could be used for more sophisticated labeling features such as area labels turning into point labels at smaller map scales.
The proposed labeling scheme relies on a preprocessing phase. In this phase for each label the map scale where it is removed from the map view is computed. During the phase of map presentation the precomputed label set must only be filtered, what can be done very fast. We are presenting some hints that allow to efficiently compute the labeling in the preprocessing phase. Using these a labeling of about 11 million labels can be computed in less than 20 minutes. We are also presenting a datastructure to efficiently filter the precomputed label set in the interaction phase.