Cloud Removal from Sentinel-2 Images with SAR data
Cloud Removal from Sentinel-2 Images with SAR data
Requirements
- M.Sc. in Machine Learning, Data Science, Computer Science, Mathematics, Telecommunications, or similar
- Knowledge of Python
- Software development skills
- Basic concepts of image processing
- Basic concepts of data science, concerning data analysis, processing and machine learning
Description
Clouds and atmospheric phenomena can significantly affect the quality and utility of satellite images, including Sentinel-2 optical images. However, Synthetic Aperture Radar (SAR) images are not affected by clouds and can be used to complement optical images for cloud removal in areas where clouds are prevalent. Therefore, developing an efficient and accurate technique able to reconstruct cloud-covered information while preserving originally cloud-free details from Sentinel-2 images, using both optical and SAR data, is crucial for various applications, including land cover classification, burned area delineation, and natural resource management. The proposed thesis aims to develop an efficient and accurate deep learning technique for clouds removal from Sentinel-2 images using SAR data. The candidate will explore available datasets (SEN12MS-CR, SEN12MS-CR-TS) and current state-of-the-art methods (GLF-CR, DSen2-CR, …) to understand the possible solutions to the task and develop a new deep learning-based algorithm. The thesis will include the following activities: (i) dataset and models exploration, (ii) development and implementation of a deep learning model, (iii) model evaluation and comparison, (iv) result visualization.