遥感影像的云及其阴影覆盖区光谱重构

Spectral Reconstruction of Remote Sensing Image's Clouds and Its Shadow Coverage

以往遥感影像的云及其阴影去除方法侧重于受遮盖地物的几何特征恢复,不利于基于光谱特征的信息提取。为此,提出一种基于地物类别特征的光谱重构方法,采用两期相邻年份相近物候期的遥感数据,借助归一化差值和决策树分类的方法,检测厚云及其阴影,并提取相应区域地表覆盖类型信息,对两时相遥感影像各波段被遮挡区的每种地物类别做线性回归分析,获取被遮挡区的重构数据。实验结果表明,该方法在恢复受云影响地物几何信息的同时,较好地实现了遮盖区地物光谱信息模拟。

Previous method of cloud and its shadow removal in the remote sensing image focus on the geometrical characteristic recovery of coverd terrain,this method is against the information extraction based on the spectral characteristics. A method based on the spectral characteristics of covered terrain category reconstruction was proposed. Adopt two adjacent years phenophase remote sensing data, use normalized difference value and decision tree classification method to detect the thick cloud and its shadow, and ex- tract corresponding regional surface coverage type information, in remote sensing image of two various bands, do the linear regression analysis to each surface feature of the covered terrain, to obtain the recon- structed data of the covered terrain. The experimental results show that the method can recover the geometry information of the covered terrain, and it also realizes its spectral information simulation.