Comparative Analysis of Methods Based on Semantic Segmentation for Cloud Detection in Remote Sensing Imagery

Cloud coverage estimation is a fundamental step in remote sensing imagery because it can be useful information to user using imagery. Recently, cloud detection using semantic segmentation is being actively studied to automatically detect cloud regions. However, there are many limitations to obtaining accurate cloud regions; therefore, the related methods need to be analyzed. Accordingly, this study aims to identify the best network architecture that is applicable for future works, and evaluates its performance using datasets created for this research, First, we classified multiple network architectures and public datasets for the semantic segmentation of clouds in remote sensing imagery. Next, we selected the best architecture by assessing the accuracy of each architecture and evaluate it using our datasets. We believe this work will be beneficial for future research on cloud detection in the field of remote sensing imagery.