基于单景遥感影像的去云处理研究

去云处理是遥感图像处理以及大气纠正的重要步骤。常规的去云处理算法会随云的覆盖类型的不同而不同,如同态滤波或时间平均法,这些算法在去除云对影像影响的同时,往往会伴随地物信息的丢失。提出了基于单景遥感影像的去云处理算法———基于遥感影像分类结果及云检测结果的去云处理算法,目的是去除影像中云的散射影响,恢复地物的光谱信息。算法是针对局部有云的单景Landsat7 ETM+影像进行的。根据Landsat7 ETM+波段4,5,7对影像进行聚类分析,确定不同地物的覆盖类型;利用波段1,2,3及波段6划分出影像中的无云区以及不同覆盖厚度的云层;按照相同地物覆盖类型对非云区与不同云区进行平均反射率匹配,以达到去云的效果。结果表明,经过去云处理的影像,在分类运算中能够明显地提高分类精度,能够很好地恢复地物的光谱信息。

Cloud removing based on single remote sensing image

Landsat7 ETM+ imagery are widely used in many fields,such as agriculture,environment investigate and geology survey. But few satellite images obtained are all clear sky and they are always contaminated by cloud or thick aerosol more or less. In order to extract objective information more effective,cloud can be removed form the remote sensing image. There are many algorithms for cloud removal including filtering algorithm,multi-temporal cloud removing algorithm and so on. Those algorithms can rernove cloud but at the same time renwve or distort some useful spectrum information from image. A new algorithm——cloud removal algorithm based on land cover types and cloud detection results is introduced. All pixels of image are classified to specific cover types through Landsat7 ETM+ band 4,5,7 because this three bands imagery are much less contaminated by most cloud and aerosols. Then band 1,2,3,6 imagery are used to classify the clear region and different levels cloud regions. The mean reflectance matching is performed in the first three bands separately according to different cover types in two regions (clear region and cloud regions). Results show that this algorithm not only can effectively remove cloud and doud shadow from imagery but also can recover the spectrum information effectively. It can lay a good foundation for follow step of image processing.