地球信息科学学报 ›› 2019, Vol. 21 ›› Issue (11): 1802-1810.doi: 10.12082/dqxxkx.2019.190041

• 遥感科学与应用技术 • 上一篇    下一篇

基于珞珈一号夜间灯光数据的广州市建设用地识别

李翔, 朱江*(), 尹向东, 姚江春, 黄嘉玲, 李密滔   

  1. 广州市城市规划勘测设计研究院,广州 510060
  • 收稿日期:2019-01-24 修回日期:2019-07-08 出版日期:2019-11-25 发布日期:2019-12-11
  • 通讯作者: 朱江 E-mail:zhujiang_gzpi@163.com
  • 作者简介:李 翔(1991-),男,河北邯郸人,硕士,主要从事土地利用变化、城市扩张监测方面研究。
  • 基金资助:
    广州市城市规划勘测设计研究院年基金项目(No.2018-60)

Mapping Construction Land of Guangzhou based on Luojia No.1 Nightlight Data

LI Xiang, ZHU Jiang*(), YIN Xiangdong, YAO Jiangchun, HUANG Jialing, LI Mitao   

  1. Guangzhou Urban Planning and Design Survey Research Institute, Guangzhou 510060, China
  • Received:2019-01-24 Revised:2019-07-08 Online:2019-11-25 Published:2019-12-11
  • Contact: ZHU Jiang E-mail:zhujiang_gzpi@163.com
  • Supported by:
    Youth Fund Sustentation Project ofGuangzhou Urban Planning and Design Survey Research Institute(No.2018-60)

摘要:

夜间灯光数据和人类活动密切相关,可用于识别城市建设用地。目前主要利用DMSP/OLS和NPP-VIIRS夜间灯光数据进行建设用地识别,由于数据质量原因,这两类数据的识别结果精度较差。珞珈一号夜间灯光数据与比以往夜间灯光数据相比,时间分辨率、空间分辨率和光谱分辨率明显提升,是进行建设用地提取的更理想的数据源。本研究首先对珞珈一号夜间灯光数据进行辐射和影像配准,提高数据质量,然后利用支持向量机(Support Vector Machine, SVM)影像分类方法对广州市2017 年建设用地分区识别,并利用Kappa系数分区、分地类评价识别结果精度。研究发现:① 利用珞珈一号夜间灯光数据识别建设用地的精度明显优于利用DMSP/OLS和NPP-VIIRS夜间灯光数据识别结果的精度;② 广州市中心城区辖区的建设用地识别结果精度较高,识别结果Kappa系数均在0.9以上;外围辖区识别结果精度相对较低,识别结果Kappa系数为0.85左右;③ 城市、建制镇等单个地块面积较大、灯光亮度较高的地类识别结果精度较高,识别结果Kappa系数均在0.9以上;村庄用地、铁路公路用地由于单个地块面积小、布局比较分散、部分路段无照明条件等原因,识别结果Kappa系数相对较低,为0.85左右;采矿、风景及特殊用地夜间基本无人类活动,缺少夜间灯光,难以用夜间灯光数据识别,Kappa系数为0.45左右。本研究证明了利用珞珈一号夜间灯光数据能有效识别建设用地,同时丰富了珞珈一号夜间灯光数据的应用场景。

关键词: 珞珈一号卫星, 夜间灯光数据, 支持向量机, 建设用地识别, 广州市

Abstract:

Nightlight data is closely related to human activities, which can be used for mapping construction land. For now, DMSP/OLS (Defense Meteorological Satellite Program, Operational Linescan System) and NPP-VIIRS (National Polar-orbiting Partnership, Visible Infrared Imaging Radiometer Suite) are the most widely-used data for construction land mapping. However, because of low data quality, the mapping results of these two kinds of data are not good enough. Luojia No.1 nightlight has an obvious quality improvement than existing nightlight data. Using Luojia No.1 nightlight data, we can map construction land more precisely. In this research, we first conducted georeferencing and radiometric correction to Luojia No.1 nightlight data to improve data quality further, then used the Support Vector Machine (SVM) classification method to map the construction land of Guangzhou in 2017 at the district level, and we used Kappa coefficient to check the mapping results. Results show that: (1) Better mapping results can be obtained by using Luojia No.1 nightlight data than the DMSP/OLS and NPP-VIIRS nightlight data. (2) Central districts like Haizhu and Tianhe's Kappa coefficients of mapping results are over 0.9. However, remote districts like Conghua and Nansha had relatively low Kappa coefficients of about 0.85. (3) Construction land like city, town, and airport had high Kappa coefficients usually over 0.9, because the nightlight emitted by these lands is easily to capture. In comparison, rural construction land’s single massif was small, and some sections of road and railway lacked lighting condition; thus, these lands are a little harder to recognize in the nightlight data. Correspondingly, their Kappa coefficients were about 0.85. Besides, land of the mining industry and scenic tourism seldom emit light at night, it is hard to identify these lands in Luojia No.1 nightlight data. Therefore, the Kappa coefficients of these lands are pretty low, about 0.45. Our findings suggest that Luojia No.1 nightlight data's great potential in mapping construction land.

Key words: Luojia No.1 satellite, Nightlight data, Support vector machine, Construction land mapping, Guangzhou