Abstract
Accurate cloud detection using medium-resolution multispectral satellite imagery (such as Landsat and Sentinel data) is always difficult due to the complex land surfaces, diverse cloud types, and limited number of available spectral bands, especially in the case of images without thermal bands. In this paper, a novel classification extension-based cloud detection (CECD) method was proposed for masking clouds in the medium-resolution images. The new method does not rely on thermal bands and can be used for masking clouds in different types of medium-resolution satellite imagery. First, with the support of low-resolution satellite imagery with short revisit periods, cloud and non-cloud pixels were identified in the resampled low-resolution version of the medium-resolution cloudy image. Then, based on the identified cloud and non-cloud pixels and the resampled cloudy image, training samples were automatically collected to develop a random forest (RF) classifier. Finally, the developed RF classifier was extended to the corresponding medium-resolution cloudy image to generate an accurate cloud mask. The CECD method was applied to Landsat-8 and Sentinel-2 imagery to test the performance for different satellite images, and the well-known function of mask (FMASK) method was employed for comparison with our method. The results indicate that CECD is more accurate at detecting clouds in Landsat-8 and Sentinel-2 imagery, giving an average F-measure value of 97.65% and 97.11% for Landsat-8 and Sentinel-2 imagery, respectively, as against corresponding results of 90.80% and 88.47% for FMASK. It is concluded, therefore, that the proposed CECD algorithm is an effective cloud-classification algorithm that can be applied to the medium-resolution optical satellite imagery.
Highlights
Optical remote sensing data ranging from the visible to shortwave infrared are widely used for surface cover mapping, surface parameters estimation, and ecosystem monitoring [1,2,3,4]
A large proportion of snow, bright bare lands, and bright impervious surfaces are shown in Figure 4b–e,g,h,k, but there is no overestimation of clouds in the corresponding classification extension-based cloud detection (CECD) results
Since each scene in the L8_Biome dataset has a manually labeled cloud mask, we measured the reliability of the training samples in Landsat-8 imagery by comparing the training samples collected in the 14 selected L8_Biome images with their corresponding cloud masks. We found that these training samples achieved overall accuracies of 96.97% and 93.58% for clouds and non-cloud surfaces in the Landsat-8 scenes, respectively
Summary
Optical remote sensing data ranging from the visible to shortwave infrared are widely used for surface cover mapping, surface parameters estimation, and ecosystem monitoring [1,2,3,4]. The use of thermal infrared bands can improve the accuracy of cloud detection, this generally only works well for thick, cold clouds and the commission error in high mountain regions may be high [14,15]. For those optical sensors that do not have thermal infrared bands (such as Sentinel-2, SPOT4-HRVIR), effectively mitigating the disturbances caused by bright non-cloud surfaces remains a challenge [1,15]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
