一种面向区域尺度的DMSP/OLS夜间灯光影像校正方法

A Correction Method of DMSP/OLS Nighttime Light Images in Regional Scale

针对美国国防气象卫星计划(Defense Meteorological Satellite Program,DMSP)/线性扫描业务系统(Operational Linescan System,OLS)夜间灯光影像存在大量饱和像元的问题,提出一种基于不饱和像元校正的去饱和方法。以北京市为研究区,提取出待校正影像内不饱和灯光像元与参考影像相应像元拟合,通过幂函数回归模型将参考影像校正到各期待校正影像尺度,得到饱和像元的真实灯光值;建立相互校正模型将待校正影像统一校正到参考影像尺度上,使影像具有可比性;利用连续性校正获得北京市1992—2013年时间序列影像数据。结果表明,该校正方法更大程度地改善了DMSP/OLS像元饱和现象,与GDP和人口的拟合优度分别为0.716和0.812,相比不变目标区域校正方法(GDP:R^(2)=0.613;人口:R^(2)=0.713),相关性有明显提高。该校正方法能消除地域间的差异,提高了夜间灯光影像的数据质量,形成较为优质的影像序列数据集。

A desaturation method based on unsaturated pixels correction is proposed to address the problem of large number of saturated pixels in Defense Meteorological Satellite Program(DMSP)/Operational Linescan System(OLS)nighttime light images.In Beijing,the unsaturated light pixels in the image are extracted and fitting to the corresponding pixels in the reference image as the research objects,and the reference image is corrected to the scale of each image by a power function regression model to obtain the true light values of saturated pixels.The mutual correction model is established to uniformly correct the images to the scale of the reference image to make the images comparable.The continuity correction is used to obtain the time series image from 1992 to 2013 of Beijing.The results show that the correction method can improve the saturation phenomenon of DMSP/OLS image pixels,and the goodness of fit with GDP and population is 0.716 and 0.812,respectively,which has significantly improved the correlation compared with the invariant target area correction method(GDP:R^(2)=0.613;population:R^(2)=0.713).This correction method can eliminate the differences of regions,improve the data quality of nighttime light images,and form a better quality image sequence dataset.