Systemy wnioskowania rozmytego (FIS) jako narzędzie nieliniowej generalizacji numerycznego modelu terenu

W artykule omówiono zagadnienie zastosowania systemów generalizacji danych przestrzennych opartych na logice rozmytej do modelowania rzeźby terenu na różnym poziomie uogólnienia.

Classic filtering methods of raster data (e.g. digital terrain model), such as median or gaussian filtering level the result surface, and consequently flatten the end results. A significant modification of results' range, understood as narrowing of the scope of relative altitudes in the test area, is not the only side effect of the process of DTM generalization. Gaussian filtering, and especially non-linear median filtering leads to non-linear morphometric modifications of generalized terrain relief. Structural forms common for high mountain relief, such as ridge lines and deeply cut river valleys are flattened more than other forms. In the article the author attempts to elaborate a non-linear method of raster data filtering by defining the objective generalization rules of local character. These rules determine the global process of cartographic generalization of raster-type data. In order to build a database which would enable the realization of the process of spatial data generalization, fuzzy inference systems (FIS) are applied. Application of fuzzy logic makes it possible to define generalization rules for non-linear filtering of a digital terrain model recorded in the form of an altitude matrix. In the discussed context FIS can be interpreted as a non-linear digital terrain model transformation. Compared to other non-linear modeling techniques FIS has many advantages: - it keeps the parameters of source data distribution (slant, range, etc.,) - enables open and easy to interpret definition of rules in the data base (in relation to scale, purpose, cartographic school, etc.), - it bases on linguistic variables, which facilitates the understanding of the generalization process, - it facilitates scalability of the results through parametrization of the membership function. Application of fuzzy logic and generalization systems using fuzzy inference makes it possible to automatize the generalization process while preserving subjectivity of cartographic generalization. The final effects depend on the FIS database created by the researcher.