| Settlement is one of the six basic features on general maps.It is also important content that is of great concern for map readers.On large-scale maps,polygonal residential areas occupy a considerable proportion of the map load;on medium-and small-scale maps,only important and larger area settlements can be represented in polygonal forms after map generalization based on the original large-scale maps.Therefore,the automated generalization for settlements has always been the core in scholars’ research;and after years of study,great achievements have been made.However,the automation of residential area generalization is still a major issue because the essence of automated map generalization is a kind of similar transformation.But few automated generalization methods based on the theory of spatial similarity relationship have been proposed.This situation hinders the automation of map generalization because similarity is subjective,but automation requires objective(quantitative)measures.In recent years,with the development of theory,technology and the expansion of data acquisition methods,there have been some studies on the spatial similarity relationship in the fields of cartography and geographic information science,including one or more kinds of spatial similarity relations,e.g.,topology,direction,distance,location,semantics,shape.As well as scene similarity,overall similarity,etc.However,it is desiderating to explore automated generalization methods for settlements based on the spatial similarity relationship theory and its quantification,so as to help address the automation of algorithm threshold determination,process control,and result quality evaluation for map generalization.Aiming at this problem,the study introduces automated generalization methods for settlements considering geometric and semantic similarity based on the theories of automated settlement generalization and spatial similarity.Spatial similarity in the study can be divided into two categories: the horizontal spatial similarity and the vertical spatial similarity.The horizontal spatial similarity is relation between map objects at the same scale;while the vertical spatial similarity is relation between map objects at different scales.The main creative methods in the dissertation are as follows:(1)Aiming at problems of irregular large-scale city settlements classification and landmark buildings extraction,fuzzy clustering analysis of individual building and statistical method based on city blocks are applied respectively.Here,the ‘landmark buildings’ refer to the buildings that are quite different and less similar to their surrounding buildings in appearance on maps.The validity of different geometric similarity quantification indices is tested;also,the classification results obtained from different clustering methods and different geographic contexts are compared and analyzed.Furthermore,the cognitive laws of readers when they read maps are discussed.These methods can be used to the geometric similarity measurement of residential areas at the same scale.They play the role of knowledge acquisition before automated settlement generalization.(2)Aiming at the problem of residential areas generalization from 1:10,000 to 1: 50,000,a hybrid approach that combines two existing methods is developed.The two existing methods are the Boffet’s method for free space acquisition and kernel density analysis(KDE)for city hotspot detection.Using both methods,the proposed approach follows a knowledge-based framework by implementing map analysis and spatial similarity measurements in a multi-scale map space.The knowledge-based framework was proposed by cartographers Brassel and Weibel,it is a widely accepted conceptual framework of automated map generalization.In accordance with the framework,the approach proposed in this dissertation is applied in four steps which include structure recognition,process recognition,process modelling and process execution.It realizes the automated generalization from 1:10,000 to 1:50,000 for map features which are represented as large-scale buildings on the original map and meso-scale residential area polygons on the target map.In this generalization process,geometric similarity plays three roles: threshold value determination,result verification and analysis of geometric similarity changing law in the multi-scale map space.This method outperforms the traditional manual settlements aggregation methods in many aspects such as data consistency between different scale data and objectification of urban boundaries.(3)Aiming at the amalgamation of large-scale residential areas within city blocks and the semantic similarity calculation during this process,an amalgamation method under the constraints of semantic functional areas is proposed.In addition,a semantic similarity measure approach during this generalization is developed which is applicable for maps at scales from1:1750 to 1: 14000.This semantic constrained generalization method is proposed based on the considering of map readers’ habit of paying more attention to the settlements’ semantic information when they are reading maps.Therefore,the semantic similarity on maps is of importance.In the quantification of semantic similarity,the matching distance model(MD model)is used to construct the semantic ontology network.The semantic hierarchical network tree is constructed based on basic geographic information data coding and urban comprehensive functional unit data in national geographical conditions census data.Experiments are carried out to analyze the features of similarity values at different scales and under different generalization methods with or without the constraints of the semantic functional units.Using this method,map better plays the role of information transmission carrier.In this research,the semantic similarity is used in the following aspects: quality evaluation and semantic similarity changing analysis for residential areas in the multi-scale map space.(4)Building simplification is one of the classic map generalization problems.Most of the existing building simplification methods focus on one or two aspects of the three principles:area preservation,shape characteristics enhancement and drawing rectangles if possible.But in practice,map generalization is a process of mutual restriction and trade-off of various influencing conditions,cartographic experts consider all the three aspects and balance them to reach reasonable results,sometimes they delete the small details on building’s outlines,but it results in the area difference,sometimes they displace some points of the edges so that the area stays unchanged,but the vertices’ location changes.It depends on the short-edge structures and distributions.By imitating this thinking and operation process of cartographers,a method based on short-edge structure recognition and progressive simplification is proposed.First,the building outline is preprocessed;then the contour segments are grouped,and the edge structure is identified and analyzed.After this,some specific situations are analyzed in detail,and the calculation region(the sequence of vertices associated with the simplification of a set of edges less than the tolerance)is used to determine the optimal simplification method.Lastly,the area difference is distributed to the whole building so that the building’s area difference is reduced while the shape is maintained.Open Street Map(OSM)buildings are selected in the experiment,the results show that the proposed method can effectively reduce the number of graphic vertices,enhances the shape characteristics like parallel,right angle,etc.Also,the approach can control the area differences after generalization.The method gives consideration to and balances the three aspects of the building simplification principles,which provides a new way for building simplification.In the simplification method,three principles:area maintenance,shape maintenance and drawing right angles as possible reflect the preservation of geometric similarity,and the vertices position maintenance reflects the preservation of position similarity.By considering settlements classification and generalization such as aggregation,amalgamation and simplification within certain scales as a case,this dissertation proposes automated generalization methods for settlements considering geometric and semantic similarity.In the future,deeper research should be carried out to study and quantify the spatial similarity relation so as to achieve a more automated map generalization. |
