DMSP/OLS夜间灯光影像中国区域的校正及应用

Correction of DMSP/OLS Night-time Light Images and Its Application in China

美国国防气象卫星搭载的业务型线扫描传感器(DMSP/OLS)获取的夜间灯光影像,可客观地反映人类开发建设活动强度,其广泛应用于城市遥感的多个领域。但该数据缺少星上的辐射校正,下载的原始影像数据集不能直接用于研究,需进行区域校正。长时间序列的DMSP/OLS夜间灯光影像数据集主要存在2个问题需在校正过程中解决:(1)原始影像数据集中的影像是非连续性的;(2)数据集中的每一期影像都存在着像元DN值饱和的现象。针对这2个问题,本文提出了一种不变目标区域法的影像校正方法,对提取出来的每一期中国区域的夜间灯光影像进行了校正,该校正方法包括相互校正、饱和校正和影像间的连续性校正。最后,为了检验校正方法的合理性与可靠性,本文将校正前后中国夜间灯光影像与GDP和电力消耗值,分别进行回归分析评价表明,校正后的影像更客观合理地反映区域经济发展的差异。

DMSP/OLS（Defense Meteorological Satellite Program Operational Linescan System） night-time light images can objectively reflect the intensity of human activities; therefore they were widely used in a variety of fields for urban remote sensing. However, the raw night-time dataset cannot be used directly in these researches due to the lack of inflight calibration, thus it needs to be further corrected. There are two problems existed in the long-time series of DMSP/OLS night-time light image dataset that should be addressed in the image correction procedure. First, every image in the raw night-time light image dataset cannot directly compare with each other due to the issue of discontinuity; second, there is a pixel saturation phenomenon existed in every image of the raw night-time light image dataset. In order to solve these problems, a method based on invariant region was proposed. This method included the intercalibration, the saturation correction, and the continuity correction procedures among all the images from the raw images dataset. All the night-time light images of China, which were extracted from the raw images dataset, were corrected using this method. Finally, this correction method was evaluated by analyzing the relationships between the night-time light images and the corresponding gross domestic product（GDP） data and the corresponding electric power consumption data respectively. Through the analysis toward the evaluated results, two main conclusions were acquired. One was that this method had solved the problem of discontinuity in the raw image dataset; the other one was that this method could reduce the pixel saturation phenomenon that existed in every images of the raw night-time light image dataset. However, this method has not completely solved the problem of pixel saturation. How to perfectly solve this problem is the core issue for future research on night-time light data application.