WHU Cloud

我们手动编辑了一个用于云检测和去除的Landsat 8数据集，其中包含多云图像，相应的无云历史图像以及云和阴影遮罩。我们将其命名为WHU cloud数据集。整个数据集包括六个不同的区域，涵盖了森林，城市，海洋，湖泊和山脉等不同的土地类型。仅选择与Landsat 8图像的4、3、2波段相对应的红色，绿色和蓝色波段。空间分辨率为30m。我们选择了在大约1-6个月之前不受云影响的同一区域的图像作为多时数据。多云图像中的云约占5%。We manually curated a Landsat 8 dataset for cloud detection and removal, which includes cloudy images, corresponding cloud-free historical images, and cloud and shadow masks. We named this dataset the WHU Cloud Dataset. The entire dataset covers six distinct regions, encompassing various land types such as forests, urban areas, oceans, lakes and mountains. Only the red, green and blue bands corresponding to Bands 4, 3 and 2 of Landsat 8 imagery were selected, with a spatial resolution of 30 meters. We selected images of the same region that were unaffected by clouds approximately 1 to 6 months prior as the multi-temporal data. Clouds account for approximately 5% of the area in the cloudy images.

搜集汇总

背景与挑战

背景概述

WHU Cloud是一个由武汉大学于2018年发布的公开数据集，专门用于云检测和去除任务。它基于Landsat 8卫星图像，包含多云图像、对应的无云历史图像以及云和阴影遮罩，覆盖森林、城市、海洋等多种土地类型，空间分辨率为30米，多云图像中云量约占5%。该数据集适用于遥感图像处理和环境监测研究。