Research On Automatic Recognition Method Of Geographical Scene Based On ARG And Spatial Pattern Matching

Geographical scene is a geographical complex with specific structure and function formed by the interconnection and interaction of various natural and human elements within a certain area and different time and space.The formation of geographic scenes is the result of the comprehensive effects of various geographic processes under the constraints of geological laws,and presents obvious regional spatial structure,geometric forms,material composition,geographic functions and other characteristics.For a long time,experts and scholars in different fields of geoscience have carried out field identification,indoor interpretation,mapping and geoscientific analysis and application of different types of typical geographical scenes based on their unique structure,form and function characteristics.However,the traditional geographic scene recognition methods based on field survey and manual visual interpretation are not only inefficient,costly,but also highly subjective,and the interpretation quality varies from person to person.Therefore,on the basis of discussing the geographical scene pattern definition method,recognition unit division method and regional scene model modeling method,this paper uses ARG(attributed relational graph)model and spatial pattern matching method to study and form a geographical scene automatic recognition method based on spatial pattern matching.The method proposed in this paper has strong reusability and high degree of automation,which can provide method reference and data support for research on geographic scene recognition,cartography,analysis,modeling and simulation applications in multiple fields.The main conclusions and results of this paper are as follows:(1)Discussion on classification methods of geographic scenes and analysis of characteristics of typical geographic scenes.Based on the concept of geographical scene,this paper collects the typical geographical scenes from different fields of geoscience,and discusses the classification methods of geographical scenes from the aspects of geometric form,spatial structure,discipline and field.And the characteristics of different geographical scenes are analyzed.(2)Research on the definition method of scene mode based on ARG.Based on the feature analysis of geographical scene,this paper abstracts the geographical objects in the scene as nodes and the spatial structure features between objects as edges,thus realizing the structure definition method of geographical scene spatial pattern based on ARG.And the important eigenvalues contained in the geographical scene are stored in the attributes of nodes and edge elements,which realizes the semantic definition method of the spatial pattern of geographical scene.The method of schema definition realizes the integrated description of objects and relationships in the scene.(3)Recognition unit division and ARG model construction method.A method for dividing large-scale research area units based on watershed units and geomorphological feature lines is designed.Based on the methods of curve splitting and point element construction based on approximate linear segment sequence,edge element construction based on Delaunay triangulation network,and curve segment Angle calculation based on approximate linear segment,an automatic ARG model construction method for recognition unit was developed.(4)Research on the automatic recognition method of geographic scene based on spatial pattern matching.Based on the principle of isomorphic sub-graph matching,a method for matching the spatial pattern structure of geographic scenes is designed for ARG graphs.A method for semantic matching of spatial patterns of geographic scenes is realized based on semantic matching rules.Based on the construction of the geographic scene pattern library,a method of automatic identification of geographic scenes with good scalability is formed.(5)Three typical geographic scenarios,which are fault control water system,river capture and Sixiang Shanshui model,is selected to test the applicability of this method.Case studies demonstrate that the recognition accuracy of the proposed method is above 75%.The recognition efficiency is high.The scalability is good.The reusability is good.