Generalization of Road Networks for Mobile Map Services: an Information Theoretic Approach

Automated generalization of road networks has become an urgent task in the cartographic utilization of mobile services. The drawing area of the hand held phones or units is small and the resolution is limited. Therefore, the information presented on these visual display units must be highly dedicated to the need of information, i.e., adaptive generalization must be considered. The use of precompiled map sets may to some degree satisfy the need for adaptability, but the strategy considered lacks the flexibility often required when we have changing context, e.g., when a person is walking or driving a car. The present paper introduces a new algorithm for the elimination of arcs in road maps. The algorithm is based on information theory and some case studies illustrate how the algorithm works. The perceptual properties of the map reader and the resolution of the display unit are brought into the algorithm by a similarity function. The new generalization algorithm computes how many roads that can be presented in a map by maximizing the amount of useful information, and it can be described as: (1) select an appropriate similarity function, (2) introduce topological constraints, (3) define the weights of the road hierarchy, (4) start the arc elimination and eliminate the most conflicting arc until the useful information reaches a maximum value.