Automated generation of adaptive schematic network maps

Abstract

Topographic maps are intended to faithfully represent the real world and thus contain various types of detailed information, with emphasis on accuracy (in terms of positional relationship among map features). However, for many purposes (e.g. personal navigation), accuracy is not of great concern to users and too much detailed information makes the map appear noisy, especially for small displays. Therefore, the simplification of map information is a must. In general, two types of simplification can be applied, i.e. content and graphics. The former means the elimination of less important details from the map and the latter means to improve the clarity of the map for easy reading and interpretation. Schematization is a process to make graphics simplified and the resultant maps are called schematic maps. Automation in the generation of schematic maps is always a dream of cartographer and a number of methods have been developed in the last decades for such a purpose. However, the clarity of maps for those areas with dense lines is not properly achieved with current methodologies although researchers have realized that it is necessary to present such areas at desirable larger scales, i.e. scale-adaptive deformation. Also there is no consideration of adaption to different displays. This study aims to tackle these problems by adopting adaptive deformation in the automated generation of schematic network maps. A three-step strategy is proposed, i.e. adaptive deformation of network maps, formation of strokes from network segments and automated generation of schematic representation. Based on the proposed strategy, a method has been developed to automatically generate schematic network maps with focus on the adaptive enlargement of those areas with highly dense lines, called congested areas in this study. To define such areas, a psychological investigation into the clarity of graphic representations is conducted to obtain a threshold. Upon this threshold, congested areas are first automatically detected, then an appropriate deformation method is selected to enlarge such areas to a desirable larger scale with proper selection of controls, and finally the schematic representations of the deformed network maps are generated. This method has been tested with two sets of real-life data and experimental results indicate that the proposed method is able to produce schematic maps with good clarity and aesthetics. On a map, line features may be unevenly distributed. It is possible that some areas may appear to be crowded (although not quite congested) while other areas may appear to be sparse. To improve the clarity of map, it is suggested to make the lines more evenly distributed by making the map scale varied. That is, a method has been developed to optimize the line density over the whole map by shrinking sparse areas to fit into a display size. Experiment testing has been carried out and results indicate that this method is able to produce schematic maps with great clarity. Although the use of the adaptive deformation can improve the clarity of resultant schematized results, the large orientation distortions arising from the network deformation may reduce readability (or recognition) of a map compared with its original shape. Therefore, in this study, an optimization method is proposed to minimize orientation distortions while the map clarity is sufficiently achieved. Experimental evaluation has been carried out and the results show that the proposed method is able to improve the readability (cognition) of schematized results while well preserving the map clarity. With the variable scale concept, an investigation is also carried out into the improvement of map readability when a combination of variable display and the generalization of map contents are employed to produce a clearer map for different display sizes. The variable scale is to optimize the line density of network maps for a given display size, and a selective omission process is employed to reduce the map contents. Experiment testing has also been carried out and the results show that this method is able to produce variable scale network maps with improved clarity and visual balance. In summary, this study aims to automatically generate schematic network maps with great clarity. To achieve this, several methods have been developed to adaptively enlarge a locally congested area or globally re-distribute line density over a map. Experiments have been carried out and results show that these methods work well. Of course, the automated generation of schematic network maps is a rather complex problem and further development is still needed.