Stroke Selection Based On Geometric And Structural Properties Of Roads

Urban road networks give support to social and economic development,as well as humans' everyday life.It is a basic structure for urban population and material flows.As a result of urbanization,urban road networks have gained amazing scale expansion,and have become increasingly complicated.In addition,users put forward higher real-time and customized requirements for maps.Accordingly,maps must be updated more rapidly and multi-scale maps must be provided.Multi-scale presentation and automatic generalization of maps have been one of the hot topics in interdisciplinary study.In the multi-scale presentation of road maps,important roads are retained,but unimportant ones deleted.It is required that the final road network keeps the basic structure and model of the original road network as far as possible.Specifically,the keys are the definition of road importance and the structural function description of road networks.As a common presentation method of road networks,the stroke follows "goodness of continuity" in the Gestalt Principles.It has been intensively used in the cartographic generalization of road networks.Scholars find that the selection of the deflection angle threshold affects finally generated strokes and networks and the optimal deflection angle threshold range is 40° to 60°.However,intersection angle distribution of road networks and applicable deflection angle thresholds vary with cities.The questions remain:How can a proper deflection angle threshold range be defined based on road networks to build strokes?What is the relationship between the deflection angle threshold range and intersection angle distribution?Few researches are related to these topics.This paper introduces information content into stroke building,provides changing curves of stroke quantity,geometric information and structural information with the deflection angle threshold,and selects a proper deflection angle threshold range based on the information content stability.Moreover,the relationship between the deflection angle threshold range and intersection angle distribution is revealed.Building proper strokes is the first step in multi-scale presentation of road networks.Later,important strokes should be defined and selected.Currently,stroke importance can be defined by two methods:single-index model and multi-index model.The linear correlation model is a common multi-index model.This paper focuses on five common indexes,including length,degree centrality,betweenness centrality,closeness centrality and clustering coefficient.These indexes are combined to form eight linear correlation models.Stroke importance values are calculated through these models and used to rank strokes.Some strokes are deleted based on priority,and the changing curve of XXX with the percentage of deleted strokes is generated.According to the structural feature of road networks and curve stability distribution,the linear correlation model combined by length,degree centrality and betweenness centrality is the optimal one.This model is used to select strokes and compare selected strokes with those selected based on the single-index(length,degree centrality or betweenness centrality)model.The selection results are evaluated based on visual judgment,average connectivity and the power-law distribution coefficient,and show that the linear correlation model combined by length,degree centrality and betweenness centrality is more reliable.