Abstract:

By taking Tianjin urban area as an example, a method is proposed to extract urban expansion by combining multi-temporal Landsat TM/ETM+ images and DMSP/OLS nighttime light data, and validated and analyzed. First, the candidate built-up area extent were obtained from DMSP/OLS data. Multitemporal Landsat TM/ETM+ images and derived multivariate textures of the obtained urban extent were classified to extract built-up areas for different dates. Urban expansions for different time intervals were obtained by post-classification comparison method. In addition, the thresholding method was applied to multitemporal DMSP/OLS nighttime light data to extract urban extents for different years and urban expansions of different time intervals were produced. The urban expansion results from DMSP/OLS data and urban statistical data were used to verify the Landsat TM/ETM+ results. The results demonstrate that the combination of Landsat data and DMSP/OLS data effectively reduce the spectral confusion between bare land and urban area and thus produce higher accuracy than useing spectral data from Landsat TM/TM+ images alone in extraction of urban built-up area expansion. Combination of Landsat TM/ETM+ data and DMSP/OLS nighttime light data provides an effective method for urban expansion extraction in mega-city areas.

Key words: urban expansion, Landsat TM/ETM+ images, DMSP/OLS nighttime light data, change detection, image classification

摘要：

以天津市为例, 将多时相的Landsat TM/ETM+影像与DMSP/OLS夜间灯光数据结合起来提取城市建成区扩展信息, 并进行验证和分析。首先, 利用DMSP/OLS夜间灯光数据得到城市建成区的大致范围, 利用这个范围内的Landsat TM/ETM+数据和多波段纹理进行图像分类, 得到城市建成区的空间分布。然后, 利用分类后比较法得到城市建成区的变化信息, 将提取的建成区变化结果与利用 DMSP/OLS数据得到的城市建成区扩展信息以及基于统计数据的城市建成区变化趋势进行对比, 验证方法的有效性。结果表明, 结合DMSP/OLS数据和Landsat 数据的光谱特征以及提取的多波段纹理特征, 可有效地减少裸土与城市建成区的混淆, 与只利用光谱信息的建成区提取及变化检测相比, 得到更高的精度。因此, 结合 Landsat TM/ETM+数据和DMSP/OLS夜间灯光数据提供了一种进行大型城市建成区扩展检测的有效方法。

关键词: 城市扩展, Landsat TM/ETM+影像, DMSP/OLS 夜间灯光数据, 变化检测, 图像分类