|Ph.D Thesis||Department of Civil and Environmental Engineering|
|Supervisor:||Prof. Doytsher Yerach|
|Full Thesis text - in Hebrew|
Automating the map generalization process has traditionally been a major focus of research in cartography and the GIS environment. Even though automation of cartographic generalization has been researched extensively, useful holistic understanding of the generalization methods is still lacking.
The developed model examines the behaviour of map objects and the interactions between them in order to gain an understanding of the generalization process. It is suggested that the electric field theory be employed, assuming that the map generalization process will be based on a pseudo-physical model leading to “powers” of map features affecting each other. The developed model, based on several predefined parameters and a spatial analysis of the cartographic reality, determines a "power" for each object. A neural network sub-model was set to determinate for each object its relative importance as a pseudo-physical power. The developed sub-model is flexible enough to enable defining the objects' powers to be modified by each user.
Interactions between map objects are expressed by actions of the forces constructed due to cartographic constraints and affected by several parameters as function of the properties of the surrounding objects. Returning to the electric field theory, each object has its "electric charge", with attraction and/or repulsion forces controlling its movements relatively to its neighbour objects expressed as a circumscribing “effective hulls” for the objects. These forces enable to apply the cartographic generalizations operators - delete, aggregate, displace or rotate of the objects - in order to solve spatial conflicts. Spatial conflict between map objects is detected when an object penetrates the other object’s "effective hull". The method assures that no new conflicts are being added during the adjustment process due to “alert hulls”. Each object is handled just once during the generalization process. The process is completed after passing over all objects on the map in a predefined sophisticated sequence, where the attraction forces is applied from the weakest to strongest object, while repulsion forces are applied from the strongest to the weakest object.
The suggested algorithm successfully generalized a layer of the map, taking into account and retaining it relations with another layer. This achievement can be considered as a significant step toward a holistic process that will handle and generalize at once all layers of the cartographic map. Such holistic generalization process would be a major contribution to the mapping community in general and to the GIS producers and users in particular.