Automated Generalization Of Roads Ana Buildings In Urban Map Representation

Map generalization or cartographic generalization is being used for many years by the National Mapping Agencies(NMAs)for producing various scale maps.Major changes occurred in the field of mapping and cartography around 1960s when digital cartography was first introduced and in the last two decades with the wide spread of internet,map generalization is no more limited for the production of multi-scale maps.Nowadays tools and techniques of generalizations are employed widely for bringing clarity in big data processing,transforming data into information,reducing data volume for swift web transfer and building multi scale spatial databases.Three strategies are available for efficient map generalization in digital form,i.e.representation oriented,process oriented and derivation oriented.In representation-oriented view,the pre-prepared generalized versions or different scale maps are accessed on user demand to meet the desired scale requirement.In Process-oriented view,the generalization process is performed on the basis of user requirements based on single detailed database.While the derivation oriented view is the combination of former two,in which different level of details are prepared based on a single detailed database in advance and suitable LOD(Level of Detail)is used to generate the result.The process oriented strategy plays an important role in the automation of map generalization.This study will focus on this question exploring the data model,algorithm and analysis of process oriented generalizationThe objective data of this study is within urban data including mainly built-up area and street features.Urban data is a very special type of spatial data,complexity,volatility and huge number of users are its main characteristics in comparison with the rural spatial data.As a man-made feature,the urban data has to consider the human-culture,architecture pattern and other social characteristics in map generalization.To satisfy the large number of users from diverse views,the generalization process on one hand should provide the tailored information,on the other the method should be efficient enough to finish the process at real time.Roads and buildings are two primary manmade features in urban morphology that directly affects every living being in the urban space.The requirements from the public,urban management field and special domains are all based on multi-scale representation of urban data.However,to finish this task is not an easy process,requiring to consider vital properties in preservation of distribution patterns and integration of geometric and semantic information.In most cases the distribution of these two features is not uniform.A single generalization operation cannot guarantee the pattern preservation for different situations,leading to the wrong decision making based on such false representation.Similarly,if the generalization only considers the geometric properties but ignores the semantics or other human-culture characteristics,the generalized result may be geometrically correct but transfer the wrong semantic information in such as land-use identification,road class assignment,land-scape recognition and others.To satisfy the two main objectives,preservation of urban pattern and consideration of semantics of building and road features in the automatic generalization,this study designs a process divided in four main phases,i.e.pre-processing,geometry extraction,pattern analysis and generalization operation.(1)At the Preprocessing stage,geometrical and topological errors of the data was removed and it was transformed in the shape suitable for performing further operations.Attribute class of the name was completed using different open online sources including Google map,OSM and World street.(2)At geometry extraction stage,the geometry of the features was organized in arc-node structure in the form of graph.(3)At the third stage,graph was utilized for generating strokes in case of roads and a Modified Voronoi Diagram(MVD)in case of buildings.These two-basic structures enabled to perform different analysis like data density,adjacent objects distances,directions and classifying them into different classes based on objects semantics.(4)At the last stage the rule was defined by using five constraints:graphical(specify the size based on the graphical limit),topological(take care of the relationship between the objects),structural(defines the structure both spatial and semantic),Gestalt(ensures the aesthetic and apparent logic),and process(defines how operators are chosen and linked in a sequence).Aiming at building feature and road feature respectively,the study establishes corresponding operators to finish the scale transformation.Formalization of the generalization process in the above stated manner ensures the urban pattern preservation with the help of pre-process analysis of building cluster and street stroke.Through the result evaluation,we find the building cluster identification by Delaunay triangulation analysis contributes to the urban data generalization.The method in this study keeps the main distribution pattern of building feature well.The generalization reduces the number of building object through aggregation operation,however remaining the relative distribution density.As for the road generalization,the presented method applies the stroke handling guaranteeing the complete structure of road network.The main road and high-class road playing connection role in the whole network,has been selected while the small and subsidiary roads were removed while maintaining the connectivity.Due to the limitation of time and scope this work was limited to the independent analysis and generalization of only roads/streets and buildings in the urban map morphology.Further when it comes to the semantics only type and name attributes were considered in the process of generalization.Similarly,only three subsequent representations were generated and analyzed.The future works mainly includes two extensions,(1)considering the relationship between building and road features to establish the collaboration generalization strategy,to consistently process the conflicts between two features.(2)Involves more semantics in the process of generalization.