LINE STRUCTURE IN GRAPHIC AND GEOGRAPHIC SPACE (COMPUTER CARTOGRAPHY, ARTIFICIAL INTELLIGENCE, GENERALIZATION).

摘要

The research reported in this dissertation has been based on the idea that a cartographic line is a probabilistic representation of the geographic feature which it symbolizes. Numeric parameters have been measured for two orders of structural relationships, and these parameters have been shown to provide significant distinctions between categories of cartographic line structure. The categories which have been developed are not intended as an exhaustive typology of line structure, but rather to demonstrate that meaningful categories of graphic structure can be defined numerically, and statistically verified.; Categories for both orders of structure have been summarized graphically, as structure signatures, and digitally, by storing parameters for each category as a computer look-up table. Structure signatures can be applied to cartographic line generalization in several ways which utilize the digital look-up tables. One application involves generating lines of predictable graphic structure, by stochastic modelling techniques. The other application does not serve to generate line structures, but to identify them, to provide a means by which threshold criteria may be automatically set and modified during computer generalization.; Line identification proceeds by matching measured parameters against parameters stored in the look-up tables. A possible problem arises when a line is identified which does not match any of the existing structure categories. An algorithm is presented which has the flexibility to incorporate new structures into an existing knowledge base, in effect, to learn new structures, and to become more proficient in line identification over time. Intelligent algorithms have been developed for pattern recognition by other authors, but the contribution of this research is to provide an intelligent algorithm for a specifically cartographic task, the automated modification of tolerance criteria during line generalization.