Cartographic Algorithms: Problems of Implementation and Evaluation and the Impact of Digitising Errors

Abstract

Cartographic generalisation remains one of the outstanding challenges in digital cartography and Geographical Information Systems (GIS). It is generally assumed that computerisation will lead to the removal of spurious variability introduced by the subjective decisions of individual cartographers. This paper demonstrates through an in-depth study of a line simplification algorithm that computerisation introduces its own sources of variability. The algorithm, referred to as the Douglas-Peucker algorithm in cartographic literature, has been widely used in image processing, pattern recognition and GIS for some 20 years. An analysis of this algorithm and study of some implementations in wide use identify the presence of variability resulting from the subjective decisions of software implementors. Spurious variability in software complicates the processes of evaluation and comparison of alternative algorithms for cartographic tasks. No doubt, variability in implementation could be removed by rigorous study and specification of algorithms. Such future work must address the presence of digitising error in cartographic data. Our analysis suggests that it would be difficult to adapt the Douglas-Peucker algorithm to cope with digitising error without altering the method.