基于光谱特征分析的MODIS影像去云算法的研究

Cloud Removing from MODIS Based on Spectrum Analysis

去云处理是遥感图像处理的重要步骤,提出了一种新的基于光谱特征分析的MODIS影像去云算法——首先分析云和云影地域的光谱特征,总结出云影响增强模型,计算一定周期内多时相遥感影像中云影响最小的影像的索引矩阵;为了进行图像之间的匹配,选择这些影像数据中云影响最小的一幅影像作为基准图像,对于无云的部分地域,对其余影像进行一元线性回归分析;利用提取的索引矩阵和图像之间的回归模型,通过像元替换方法,可以获得去除或者弱化云影响的图像。为了验证结果的准确性,通过该算法得到的250m空间分辨率的云检测图像与NASA提供的云检测图像进行对比分析,研究结果表明:提出的去云算法能消除或者大大减弱云对MODIS影像的影响。

Cloud removing is an important step of remote sensing image process. In this paper, the author proposes a new algorithm for cloud removal from MODIS images based on spectrum analysis. Firstly spectral characteristic of cloud and cloud shadow area is analyzed, and cloud area enhancement model is summed up. Using this model an index matrix is extracted, in which, each pixel value is the index of the image contaminated least by cloud among the multi-temporal remote sensing images in a certain period. Secondly, the image contaminated least by cloud is selected and used as a benchmark image for the purpose of an image match for the common parts of the cloud free area. Other images are transformed by linear regression model and the above benchmark image. Thirdly, the result image is acquired by replacing pixels. In order to test the accuracy of the result, it is compared with the cloud mask provided by NASA. The result shows that the algorithm can eliminate or significantly reduce the cloud effect from MODIS images.