An Optimisation Approach to Cartographic Generalisation

Cartographic generalisation is the process of simplifying a representation to suit the scale and purpose of a map. As such, generalisation has always been a central part in map production. This thesis is a contribution to the ongoing work on automating cartographic generalisation. Two topics are treated: optimisation methods in graphic generalisation, and propagation of updates between cartographic data sets of different scales. To enable an optimisation approach to graphic generalisation several topics are studied. Firstly, requirements are stated for defining a good map. These requirements can act as constraints and control the graphic generalisation process. In this thesis, ten analytical constraints for graphic generalisation are proposed, some for single objects and some for groups of objects. Rules are also established regarding when these constraints should be set up. Secondly, the least-squares method is studied as an approach to finding the optimal solution according to the constraints. Related to this, weighting strategies for the constraints are also discussed. Thirdly, computationally efficient methods are tested, which are especially important for establishing spatial relationships between objects and solving the large normal equation systems. Finally, the optimisation approach has been evaluated visually and quantitatively in mid-scale applications. Mapping organisations have to maintain cartographic data sets of different scales (resolutions). Ideally, mapping organisations should only be concerned with updating the most detailed data set and the updates should automatically be propagated to all other data sets. In this study, a prototype system was created for automatic propagation of updates. This system is integrated in an object-oriented map production software, in which the data sets are stored in a multiple representation database. The rules implemented in the prototype system were gathered from data set specifications and by studying printed map series. The propagation of updates performed by the prototype system was compared with manual propagation performed by a cartographer.