Thesis Code: 24012

Thesis type: M.Sc. in Machine Learning, Data Science, Computer Science, Mathematics, or similar

Keywords: image processing, deep learning, computer vision

Requirements:

  • Knowledge of Python
  • Software development skills
  • Basic concepts of data science, concerning data analysis, processing and machine learning
  • Basic concepts of image processing, modern computer vision preferred but not required.

Description:

The proposed thesis focuses on the application of super-resolution techniques to images acquired by the Sentinel-3 satellite in order to improve their spatial resolution, with Sentinel-2 used as the reference ground truth. The main objective is to obtain daily maps of the Normalized Difference Vegetation Index (NDVI) in the agricultural context.

NDVI is a key measure for assessing crop health and vegetation coverage, and its daily availability is essential for more precise and timely agricultural management. Sentinel-3, with its daily acquisition frequency, offers a unique opportunity to monitor agricultural dynamics on a daily basis. However, its spatial resolution may not be sufficient to capture crucial details at the field level.

The proposed super-resolution approach in this work leverages the high-resolution information provided by Sentinel-2, a satellite known for its excellent spatial resolution. By using Sentinel-2 as ground truth, the aim is to train machine learning models capable of learning the complex relationships between the spectral characteristics of Sentinel-3 images and the high-resolution ones of Sentinel-2.

Expected results include daily NDVI maps obtained through the super-resolution process, which will allow for a more detailed and accurate assessment of vegetative conditions in agricultural areas. This work will significantly contribute to improving the informational resources available to farmers, facilitating more timely and targeted decisions, with positive impacts on crop management and agricultural sustainability. The proposed methodology could be extended to other spectral indices and geographic areas, providing a significant contribution to the scientific community in the field of agricultural satellite monitoring.

Contact: send a resume with attached the list of exams to federico.oldani@linksfoundation.com specifying the thesis code and title.