Object Selection in Map Generalization Using Geosocial Network Data: A Case Study in Wuhan, China

Abstract

In map generalization, determining which objects are to be selected is a challenging task in the selection operation. Decisions surrounding object selection depend on object semantics and spatial context. The advent of geosocial network services, such as Foursquare and Twitter, provides an evaluation system for physical space. These services have generated a wealth of geographic information that can reflect human mobility and urban dynamics. The purpose of this paper is to use the geographic information from geosocial networks as a reference for map generalization and subsequently transform the semantic-based selection operator into a simple statistical method. Analysing this geographic information quantitatively and acquiring the value of each object is the key of this research. The present study has organized an investigation using data from a geosocial network service, named Jiepang in Wuhan, China. The experiment was conducted by adopting spatial analysis and heavy-tailed distribution, which consisted of the following steps: first, semi-variance modelling was conducted for the distance threshold, to measure the magnitude of the effect of spatial dependence; a buffer analysis was then completed using a radius of this distance threshold and inverse distance weighting was applied to calculate the value of the geographic object; last, a head/tail division rule, a new principle based on heavy-tailed distribution, was used to select objects at different levels of detail. Though it is difficult to evaluate map generalization, the result demonstrates a different map. Furthermore, for a better understanding of the result, the method used in this study is compared to OpenStreetMap and essential differences between the two are discussed. This study successfully demonstrates that geosocial network data can be used as important criteria for object selection.