Research On Typical Map Pattern Recognition In Urban Building Groups

The description,generalization,abstraction,and processing of the spatial distribution and relationship of geographical features is one of the core contents of automatic cartographic generalization,which plays a vital role in spatial data integration,storage,transmission,representation,analysis and derivation process.As a high-level knowledge of spatial relationship,map patterns in urban building groups reflect the interrelations between material forms of cities and their social and economic functions,as well as their spatial distribution structures,which have a great significance for cartographic generalization,multiscale representation and spatio-temporal data mining.With the perspective of spatial data enrichment,this research aims to deeply excavate the theories and methods of map pattern recognition in urban building groups.By the investigating and exploring the morphologic features and multi-scale heterogeneity of map patterns in urban building groups,this research interprets the spatial arrangement of urban buildings and enriches the spatio-temporal data mining theories and methods for cartographic generalization and multi-scale representation.The main contents and innovations of this thesis are as the followings:1)This research summarizes and analyzes the current status of map pattern recognition home and abroad.Map pattern conceptual model,scientific connotation,influential factors and classification system are investigated via analyzing the cognitive features and organizational laws of spatial distribution of urban buildings.Quantitative index system is constructed based on visual perception.The structural models of different pattern of buildings are built by coupling the spatial relation and parameters.Theories and models are proposed to support map pattern recognition.2)Linear pattern recognition in building groups.The cognitive characteristics of linear pattern are studied firstly and linear pattern is divided into single-connected and multi-connected.According to the similarity model,an adaptive clustering algorithm is proposed to cluster and group buildings.Then single-connected linear pattern is recognized based on the alignment model.For the phenomenon of intersection,the multi-connected linear pattern is recognized completely taking advantage of parameter discrimination.The effect of parameters on the recognized results is discussed by the comparison of multiple sets of contrast experiments.3)Grid pattern recognition in building groups.Firstly the definitions,structural features and cognitive characteristics of grid pattern are analyzed and studied.On the base of multi-connected linear pattern,the line pattern's intersection graph and similar orientation graph are formed.Finally,the candidate grid patterns are extracted through graph operation such as finding maximal complete sub-graph,join,intersection and post-processing.The possible influencingfactors are discussed by the comparison of contrast experiments.4)Typical letter-like pattern recognition in building groups.Firstly,eight kinds of typical letter-like pattern are counted,analyzed and studied.Pattern complexity score is defined to determine the sequence of each pattern.Then interested typical letter-like pattern templates are described by Attributed Relational Graph using attribute and structure parameters.Template library is established.Typical letter-like pattern is recognized by solving the imprecise sub-graph isomorphism problem with Ullman algorithm from the perspective of structure pattern recognition.