Research On Automatic Selection Methods For Urban Road Networks Considering Semantic Information

Map generalization refers to the process of selecting and generalizing map data while converting maps from large scale to small scale.At present,the existing cartographic theory and methods are unable to completely achieve automatic generalization,which turns out to be a very challenging and innovative research field in modern cartography.Many researchers around the world keep conquering this difficult problem.Urban road networks are closely connected with human activities,supporting the socio-economic development of cities and people's daily lives.Road generalization thus has always been the focus and challenge of map generalization research.As prerequisites for road generalization,road network selection is a complex and comprehensive process,in which many factors require to be considered.Most of the existing studies use a single selection model such as stroke,mesh,graph theory,etc.Under certain conditions,these selection results can fulfill the current needs,but it is difficult to cover the global and local structure of the road network at the same time.Besides,the existing studies do not give enough consideration to the semantic information of the road network selection.Therefore,to address the problem of single road selection method and insufficient consideration of semantic information in existing studies,we propose an hybrid automatic road selection method that takes semantic information into consideration,where the POI(Point of Interest)data and taxi track data are incorporated into the selection process.Firstly,we construct the road mesh and the road stroke;secondly,road mesh model combining POI data and stroke model combining trajectory data are constructed respectively;finally,the selection results of the mesh model and stroke model,which integrate the semantic information from multiple sources,are fused to obtain the final road network selection results.The main innovations of the paper are as follows:(1)Introducing POI data and taxi trajectory data into the road network selection process compensates for the lack of consideration of semantic information in existing studies.POI data records the location and attribute information of important facilities on both sides of the road,which can be treated as proper semantic information of the road network.By counting the information of the POI data in the road network,the threshold condition for measuring the semantic feature of the road network is placed.Moreover,as an important part of urban internal transportation,taxis play an important role in transportation for citizens due to its convenience and efficiency.Accordingly,the road traffic flow formed by taxi trajectory data is able to reflect the importance of roads to a certain extent.Therefore,road traffic flow is considered as an important factor in our road selection method.(2)A hybrid road network selection method combining mesh model and stroke model is proposed,which improves the defect that the existing single selection methods are partial to the global structure or the local structure.The advantage of mesh model selection is that it can maintain the overall structure of the road network,while the advantage of stroke model selection is that it can maintain the local structure of the road network.Therefore,the combination of mesh model and the stroke model takes full advantage of two models,which leads to largely preserve both global and local features.This thesis conducts a comparative experiment using road networks in various scales from different cities.The results show that:(1)by considering the semantic information of the road network through the POI data and taxi track data,the probability of selecting important road networks is greatly increased and the main road networks are able to be preserved;(2)through the combination of the two selection models,this hybrid model shows the capability of maintaining the overall structure and the local structure of the road network.