Abstract:

In allusion to a typical pattern of urban space features, from its semantic description, geometry, and many other features are summarized and analyzed. Based on the above, POI data are classified, and the algorithm based on planar kernel density estimation, spatial clustering and double density factor analysis are designed. As the example of POI data of Nanjing city, by the means of kernel density estimation algorithm, doing the raster overlay calculation, extraction by mask, rendering classification, and raster reclassification, to construct city spatial pattern and to testify the accuracy and practicability of the methods of extracting features. Therefore, the administration of planning can make urban plans better, to improve the city development. The result reveals the relationship between the spatial relationship and the attribute of the city. The urban commercial center and the urban commercial center, effectively reflect the regional distribution of the overall urban spatial pattern, and the Nanjing city commercial center is basically in line with the overall planning of the city of Nanjing for 2000-2020 years. However, it can't directly reflect the positioning of the city's commercial centers. Besides the main city zone is the commercial center of Nanjing, county towns, development zones and high-tech zones, have their own commercial center , which conforms to the multi business mode of the city. Compared with the traditional urban commercial space research, the quantitative research based on POI data has a more detailed understanding of the overall urban pattern and commercial space, which is helpful for the identification of commercial centers and the analysis of industry characteristics. The government departments are convenient to adjust the development direction of the existing planning functional areas, so as to make the spatial structure and characteristics of different functional areas more reasonable.

Key words: urban spatial pattern, spatial analysis, features characteristics’ extraction, POI data, Nanjing

摘要：

从语义属性、几何形态等方面对典型城市空间格局要素进行定量化抽取及特征分析,总结归纳出城市空间格局要素间的关联特征。在此基础上,将POI数据进行分类,设计了基于平面核密度估计算法、空间聚类与双重密度因子分析分析算法,并以南京市为例,通过栅格叠加计算，提取边界，分类渲染以及栅格重分类等步骤,实现南京城市空间格局及商业中心抽取。结果发现城市商业中心有效地反映了城市空间总体格局的区域分布，提取出的南京市商业中心基本符合2000—2020年的南京市总体规划，但文本并不能直接体现城市各商业中心的定位，城市中除了主城区这个集聚且组团的单中心商业模式，各个县城、开发区、高新区都有各自形成的商业中心，符合城市的多商业模式。较传统的城市商业空间研究来说，基于POI数据的定量研究对城市总体格局及商业空间的认知更为精细，有助于商业中心的识别及行业特征的分析，便于政府部门对现有规划功能区的发展方向进行调整，使得不同功能区空间结构与特征关系更加合理。

关键词: 城市空间格局, 空间分析, 要素特征抽取, POI数据, 南京市