A Formal Approach to the Automated Labelling of Groups of Features
1Reimer, A.; 2Van Goethem, A.; 3Rylov, M.; 4Van Kreveld, M.; 5Speckmann, B.
A variety of recurring geographic entities form collections of point, line or area features. Examples are groups of islands (Archipelago), relief features in deserts, periglacial lakes or geomorphological forms, such as drumlins and sinkholes. All these groups of features may be best identified with a single label. Surprisingly, the (automated) labeling of groups of features has received little attention so far. This is at least partially caused by the lack of cartographic principles that can be found in the literature on this subject. Though extensive guidelines for map labeling have been given in seminal works by Imhof (1962/1975), Wood (2000) and Brewer (2005), information on the labeling of feature groups is sparse. Following Imhof (1962/1975), the fundamental function of any label is an unambiguous and clear relationship to the feature it belongs as well as to other labels and features. This function is naturally harder to achieve for object groups, which are potentially close to other object groups, than for the simple object-label pairs in point labeling. A single label might even be all that ties the objects together visually. Group labels hence have a constructive function in addition to the descriptive function that all labels carry. We investigated a corpus of printed small-scale maps in order to generate an overview of the pertinent labeling strategies employed in manual labeling. We conclude that there are indeed variegated placement choices that would warrant a different automatization process each. The main categories we identify are placement within the group or outside, and either following the coordinate grid or the shape of the group of features. These are further varied by the geometry used for the label baseline itself be it straight, circular or curved like a Bézier spline. We propose a framework to determine formal measures that are –possibly subconsciously– used by cartographers when labeling features. Our framework gives rise to a large variety of geometrically optimal labels. We list the optimal label positions and shapes for which labels can already be computed using existing algorithms. However, in many of the cases computing optimal label placements is still an open algorithmic question which can readily be investigated in future work. Once the necessary algorithms are developed, our framework provides an objective basis to investigate the geometric measures used by cartographers to label groups of features. We preliminarily explore the applicability of our framework using one of the geometric optimality choices.