Research On Urban Spatial Pattern Supported By Building Footprint Data

After rapid development,there would be many ecological and environmental issues for cities,such as heat islands and urban sprawl.The understanding of urban space is the only way to prevent and solve these problems fundamentally.Although existing research try to find spatial knowledge with regularity from urban space and analyze specific application requirements.The analysis results such as spatial distribution obtained are unstructured information that are not readable for computers.The latent interaction and relationship between spatial objects are difficult to share with other fields.In order to assist urban construction and urban governance,it is necessary to reorganize the domain knowledge that reflects underlying rules in urban space.For this reason,this paper come up with the concept of urban spatial pattern for expressing the relationship of urban constituent elements based on 2D spatial vector data.In the field of geographic information science,scholars have explored urban spatial knowledge from different types of spatial data.Since building footprint data conveys the accurate geometric information of objects and the precise spatial location information of objects,it has become an important part of the 2D geospatial vector database.Although patterns are generated in the process of visual perception with certain subjectivity,the rule-based approach to explore urban spatial patterns driven by vector spatial data can reduce the uncertainty caused by individual consciousness.The spatial pattern can be expressed in a relatively uniform way by combining different perspectives of research objects and different spatial scales.This paper focus on the extraction and analysis of urban spatial patterns from both micro and macro perspectives.From the micro viewpoint,as an important component of a city,buildings are the main place of human activity and living area.In the process of extracting a spatial pattern of buildings,the basic contents should be selected to meet the purpose of cartographic.The typical map elements should be retained or highlighted.From the macro viewpoint,the urban functional areas are the main embodiments of an urban attributes.When analyzing the spatial pattern of functional areas,we should first define the research unit and consider the heterogeneity of social functions within a research unit.Secondly,the interaction between functional areas should be quantified.To fix the research gaps of the existing methods of extracting and expressing urban spatial patterns,this paper proposes a method for automatically extracting urban space patterns and quantifying and formalizing the extracted patterns.The main work is as follows:(1)The scientific issues of urban spatial pattern are discussed from two aspects: research field and research object.The different understandings of the pattern among various disciplines are summarized.To seek the consensus and combine the characteristics of urban spatial information,the definition of urban spatial pattern is proposed.Based on the theory of spatial cognition,the mathematical model supporting the mining of urban spatial patterns are reviewed.The mainstream methods of describing the positional relationship of spatial objects,such as adjacency relationship,topological relationship and direction relationship,are discussed.(2)According to Gestalt psychology's "comparison and judgment" theory on the level of visual cognition,the contour diffusion method based on multi-scale features is proposed to calculate shape similarity of polygons.Combining the visual habits from global to local,the statistical grids for three scales are constructed.The contour information of a polygon is projected into the statistical grid to fuse the contour information with the regional information.By the use of the context information of the adjacent space of the statistical unit,the sparse context information is compressed by the convolution kernel.The multi-level texture feature tensor is then extracted from condensed grid context information.The correlation coefficients between the texture feature tensors of two polygons is calculated to obtain the shape similarity.(3)Based on the shape similarity and adjacency relations of the building,an efficient self-organizing clustering algorithm is proposed.The main idea is to use a similar and adjacent polygon for the search center as the next center of search process iteratively.The clustered buildings will no longer participate in the subsequent clustering process.This paper defines a building cluster as a variable area unit(VAU).The straight-line pattern and grid pattern are extracted from the VAU by designed rules.For the two modes,three spatial morphological metrics are designed to quantify the differences between building patterns of a city.Aiming at the domain knowledge contained in the spatial pattern of buildings,the ontology is used to formally express the spatial knowledge.The knowledge graph of the real scene is constructed to prepare for data enrichment and knowledge mining.(4)Different from the micro-scale building space pattern,the spatial pattern of social function areas based on the macro-scale.Social function areas need to be quantified firstly before extracting the space pattern of the social function zone.According to the spatial morphological characteristics of buildings with different social functions in VAU,five VAU-based spatial metrics are designed.In view of the mixture of social functions in urban blocks,three functional likelihood models are constructed based on the five spatial metrics proposed in this paper and the existing spatial metrics.Geometric centers of urban block are identified as the node and three social function likelihood as the properties of the adjacent edges.Two methods of quantifying the relationship are proposed: difference minimum spanning tree(DMIST)and difference maximum spanning tree(DMAST).Taking buildings and functional areas in Munich as an example,this article discusses the expression and extraction methods of spatial patterns at both microscopic and macroscopic spatial scales.The quantitatively expressed building spatial pattern initially has the ability to horizontally compare the spatial feature building spatial pattern.Formally expressed knowledge can be used to construct the knowledge graph of the actual scene.It is a foundation for applying the extracted knowledge to solve urban problems in the next step.Based on the urban spatial pattern,spatial knowledge can be effectively shared with other research fields.There is still a lot of potential for the development of urban spatial pattern in related research.