地理研究

地理研究 ›› 2020, Vol. 39 ›› Issue (2): 243-256.doi: 10.11821/dlyj020181085

• 观点与争鸣 • 上一篇    下一篇

基于多源数据和深度学习的城市边缘区判定

刘星南1,2, 吴志峰1,2(), 骆仁波1,2, 吴艳艳3   

  1. 1. 广州大学地理科学学院,广州 510006
    2. 广东省地理国情监测与综合分析工程技术研究中心,广州 510006
    3. 广东财经大学地理与旅游学院,广州 510320
  • 收稿日期:2018-10-08 修回日期:2019-02-20 出版日期:2020-02-20 发布日期:2020-05-20
  • 通讯作者: 吴志峰
  • 作者简介:刘星南（1994- ）,男,广东汕尾人,硕士,研究方向为城市规划大数据与机器学习。E-mail: liuxingnan@e.gzhu.edu.cn
  • 基金资助:
    国家自然科学基金项目(41671430);国家自然科学基金项目(41801250);广东省科技创新战略专项资金项目(2018A030310069)

The definition of urban fringe based on multi-source data and deep learning

LIU Xingnan1,2, WU Zhifeng1,2(), LUO Renbo1,2, WU Yanyan3   

  1. 1. School of Geographical Sciences, Guangzhou University, Guangzhou 510006, China
    2. Guangdong Province Engineering Technology Research Center for Geographical Conditions Monitoring and Comprehensive Analysis, Guangzhou 510006, China
    3. School of Geography and Tourism, Guangdong University of Finance and Economic, Guangzhou 510320, China
  • Received:2018-10-08 Revised:2019-02-20 Online:2020-02-20 Published:2020-05-20
  • Contact: WU Zhifeng

RichHTML

PDF (PC)

赞

摘要：

城市边缘区的定量分析及判定,对城市发展评价和规划,或是城市空间结构研究都具有重要意义。然而现有研究的边缘区判定指标选择过于单一,判定结果过于破碎,城市预设边界、水体及城市绿地对边缘区判定结果干扰大。针对上述问题,从自然、人口、社会经济的视角出发,以遥感影像、人口数据、POI大数据为数据基础,结合深度学习技术,构建基于多源数据和深度学习的城市边缘区判定方法,进行广州市城市边缘区判定及城市结构空间分布特征分析。结果表明：① 此方法能将城市划分为核心区-边缘区-外缘区,判定结果不会受到预设边界范围的影响,且消除了城市内部水体和城市绿地所造成的破碎化;② 城市边缘区与路网耦合良好;③ 广州市的城市核心区空间分布合理。综上所述,此方法能有效检测城市边缘地带,且结果符合实际情况,能为城市规划、政府决策提供参考。

关键词: 城市边缘区判定, POI大数据, 深度学习, 广州

Abstract:

With the development of the economy, most cities will expand continuously to the surrounding areas, thus leading to the emergence of urban fringe areas with both urban and rural characteristics. The urban fringe area, located between urban and rural areas, is the most intense area of urban land use change and one of the most likely areas for urban construction land expansion in the future. How to identify urban fringe accurately and quantitatively is of great significance for urban planning and sustainable land use. However, most existing methods about the delineation of urban fringe area is just based on one or one type of indicators, and the judgment result is too fragmented to reflect the continuity of the urban spatial structure. What's more, the urban preset boundary range, the water body and the urban green space have great interference with the judgment results of urban fringe. In view of the above problems and from multi-perspective of nature, population and social economy, this paper defines urban fringe based on deep learning and multi-source data (remote sensing image, population density and POI big data). Furthermore, the proposed method has been used to detect the urban fringe area of Guangzhou city in our experiments. The results show that: (1) This method can divide the city into urban core area, urban fringe and rural area accurately without the impact of the preset boundary range. Eventually, this way can eliminate the fragmentation caused by the internal water and green space of urban areas. (2) The results of urban fringe area are well coupled with the road network. Network distribution of the urban core area is densest, followed by the urban fringe area. (3) The spatial distribution of urban core area of Guangzhou from the experiments is reasonable and consistent with the actual situation. All in all, the proposed method can consider comprehensively multi- perspective factors and detect urban fringe effectively, thus can provide better guidance for formulation of policies for urban development, such as urban planning, sustainable development, and urban statistical analysis.

Key words: definition of urban fringe area, POI big data, deep learning, Guangzhou