Research On Typical Spatial Pattern Recognition Methods Of Urban Road Networks

Spatial pattern is a kind of advanced spatial structure knowledge.Spatial patterns are composed of basic geographical elements,and they can be defined as spatial objects with certain meaning or specific arrangements with a certain name.Spatial pattern recognition is one of the main research contents of spatial data mining and knowledge discovery.It is an important means to transform data into knowledge,and can provide effective support for map generalization and multi-scale representation,urban analysis and modeling,and geographic phenomenon and process simulation.Spatial structure abstraction,spatial conceptual modeling and mixed pattern recognition are the difficulties in the representation and description of geospatial features.Road networks are the basic framework of a city.The rich and multi-level spatial patterns of road networks affect people's perception of cities in different scales.Taking the urban road networks as an example,the theory and methods of road network spatial pattern recognition are studied in depth for automatic map generalization and urban structure analysis.The main work and innovations of this paper include:(1)The topic background and significance of this study are clarified,and the necessity of spatial pattern recognition is analyzed from three aspects: intelligent development of cartographic generalization,rational planning of urban space and geospatial pattern analysis.The research status at home and abroad on spatial pattern recognition of point group targets,polyline group targets and polygon group targets are summarized,and the research deficiencies of road network spatial pattern recognition are analyzed.Based on this,the research contents and techniques of this paper are determined,and the organization structure of the paper is given.(2)Basic theory of road network distribution pattern recognition.On the basis of combing the concept of spatial distribution pattern,the basic concepts of road network spatial pattern are given.Combining with the hierarchy of visual perception,a hierarchical classification model of road network spatial pattern is proposed.The manifestations of morphological pattern and semantic pattern at different scales,such as macro,meso and micro scales,are expounded.The relationships between typical spatial patterns are analysed,and the recognition sequence of the complex distribution patterns are presented.Based on Gestalt theory,the visual perception organization of road network spatial pattern is analyzed,and how to select structural parameters and establish recognition model is explained.(3)Micro-ring pattern recognition based on combination strategy.Referring to the principle of circular target detection in raster images and taking roundabout as the research object,a recognition method for vector ring pattern based on linear tessellation is proposed.From the perspective of road meshes,the simple ring can be recognized based on the improved shape angle indicator,and the accuracy is effectively improved in the case of reducing the number of parameters.From the perspective of road lines,a recognition method based on improved Hough transform is proposed,which solves the identification problem of complex ring,and makes up for the deficiency that the mesh method is disabled to identify the ring with the road crossing.The combination of the two methods can not only improve the accuracy of circular feature description,but also overcome the interference of other roads in the ring,which is more applicable for the actual road networks.The structure description of roundabouts identified based on this method is reasonable and can be directly used for the next complex intersection typification.(4)Grid pattern recognition with road skeleton constraints.In the context of cartographic generalization,grid pattern recognition should take into account the constraints of the overall shape of the road networks.Aiming at the skeleton pattern of road networks,a stroke generation method considering road corner is proposed,which can extract the stroke pattern that match the visual coherence feature.Then a skeleton pattern recognition strategy based on the comprehensive index of stroke importance is designed,which effectively avoids the disadvantage that single index is insufficient to maintain spatial features.Aiming at grid pattern,under the constraints of road skeleton,a grid recognition strategy based on mesh rectangularity and adjacent search is proposed,which solves the inherent defects of road mesh in grid pattern recognition to a certain extent.A multi-attribute clustering method is further proposed to identify grid groups.The results describe the grid features of sub-region division,and can provide reasonable basis for block merging combination.(5)Urban main center pattern recognition using density-based clustering with fuzzy neighbourhood.There are always dense urban road networks in the urban centers.By extracting the main center pattern of the urban road networks,the range of the main urban center area can be delineated.From the perspective of the main center pattern recognition,the fuzzy membership function is used to replace the binary membership,and density-based clustering with fuzzy neighbourhood is proposed based on the principle of DBSCAN algorithm to extract the main cluster in the road nodes.The coverage area of the main cluster is constructed considering the influence range of the node,and the complete center areas with clear geographical meaning are generated according to the area ratio of the road mesh and the coverage area,which have higher overlapped degree with the planning center.This research is not only a useful attempt to model spatial concepts represented by natural language,but also can provide support for complex pattern recognition.(6)Urban ring-radiation pattern recognition supported by graph theory.The ring-radiation pattern is a mixed distribution pattern formed by the gradual expansion from the urban center.According to the characteristics of road segment adjacency and the basic characteristics of urban rings,based on the urban main center recognition results,a multi-center and multi-level urban ring pattern identification method is proposed.Next,the road rays are extracted by taking into account the central ring,and the ray measurement method is proposed based on the characteristics of ring connection and ray extension characteristics,thus to recognize the radiation pattern whose angle and length distribution is more uniform.The principle and parameters of this method are clear,and the recognition results have high consistency with regional overall planning.This study provides ideas for quantitative expression of macroscopic and fuzzy distribution patterns,and also provides a feasible case for mixed distribution pattern recognition.