Toward a practicable model of cartographic generalisation.

A system for fully automatic cartographic generalisation must be able to deal with severe reductions in map scale by preserving the salient characteristics of geographic features while avoiding overlapping symbols. An 'intelligent database' treating features as assemblages of landscape units can further reduce the complexity of the generalisation algorithm. A hybrid data structure is needed to allow the algorithm to track features and detect overlap. A supervised approach affords a practicable means of choosing options and setting tolerances. -from Author