The paper discusses the global algorithm of broken line simplification, that: does not depend on parameters set by a map editor and maintains the accuracy of the 0-1 Instruction (General principles of surveying practice) of the Head Office of Geodesy and Cartography, Poland, for each map scale (smaller than the source map scale). In the discussed process of line simplification parameters depend on the map scale and on the smallest length of an elementary triangle (this length is a measure of the ability of the drawing recognition). In the process of simplification performed with the use of the discussed algorithm, the same shape of a line is ensured (maintaining the ability of the drawing recognition), since generalised data differ with the bigger or smaller range of scales from the source data. Besides, limits of intervals of generalisation thresholds have been specified, which are required for the process of automated selection of cartographic presentation methods exhibiting the results of line simplification.

The issue of line simplification is one of the fundamental problems of generalisation of geographical information, and the proper parameterisation of simplification algorithms is essential for the correctness and cartographic quality of the results. The authors of this study have attempted to apply computational intelligence methods in order to create a cartographic knowledge base that would allow for non-standard parameterisation of WEA (Weighted Effective Area) simplification algorithm. The aim of the conducted research was to obtain two independent methods of non-linear weighting of multi-dimensional regression function that determines the “importance” of specific points on the line and their comparison to each other. The first proposed approach consisted in the preparation of a set of cartographically correct examples constituting a basis for teaching a neural network, while the other one consisted in defining inference rules using fuzzy logic. The obtained results demonstrate that both methods have great potential, although the proposed solutions require detailed parameterisation taking into account the specificity of geometric variety of the source data.