Evaluation of Labeling Strategies for Rotating Maps

Abstract

We consider the following problem of labeling points in a dynamic map that allows rotation. We are given a set of feature points in the plane labeled by a set of mutually disjoint labels, where each label is an axis-aligned rectangle attached with one corner to its respective point. We require that each label remains horizontally aligned during the map rotation, and our goal is to find a set of mutually nonoverlapping visible labels for every rotation angle α ∈ [0, 2π) so that the number of visible labels over a full map rotation of 2π is maximized. We discuss and experimentally evaluate several labeling strategies that define additional consistency constraints on label visibility to reduce flickering effects during monotone map rotation. We introduce three heuristic algorithms and compare them experimentally to an existing approximation algorithm and exact solutions obtained from an integer linear program. Our results show that on the one hand, low flickering can be achieved at the expense of only a small reduction in the objective value, and on the other hand, the proposed heuristics achieve a high labeling quality significantly faster than the other methods.