| KEYWORD |
Proba-V super-resolution in combination with Sentinel-2
Thesis in external company
keywords MACHINE LEARNING, ARTIFICIAL NEURAL NETWORKS
Reference persons ENRICO MAGLI
Research Groups CCNE - COMMUNICATIONS AND COMPUTER NETWORKS ENGINEERING, ICT4SS - ICT FOR SMART SOCIETIES, Image Processing Lab (IPL)
Thesis type RESEARCH ORIENTED
Description The Proba-V pushbroom multi-spectral radiometer [1] was launched in May 2013 by the European Space Agency (ESA) to ensure global daily monitoring of land and coastal areas with main focus on vegetation monitoring. A recent inter-comparison exercise was launched by ESA to study the possibility to enhance spatial resolution of Proba-V imagery [2]. The outcome of this study was that deep learning methods have the potential to effectively enhance spatial resolution of Proba-V by merging several low-resolution images to reconstruct a super-resolved scene.
The aim of the proposed study is to build on the lessons learned during the ESA Super-resolution experiment [3], while tackling new challenges, in particular, using Sentinel-2 [4] 10 m reference imagery to further enhance Proba-V 100 m spatial resolution. Sentinel-2 data are sufficiently consistent with Proba-V in terms of overpass time, radiometry and spectral coverage in the visible and near-infrared range, ensuring smooth fusion of the different data streams. Atmospherically corrected surface reflectances from Proba-V and Sentinel-2 mission can be openly downloaded from ESA/Copernicus archives. The algorithm development and feasibility analysis will be performed on a set of reference images, with the long-term objective to prepare the ground for the deployment in an open source platform, where the full mission archive is directly accessible on-line, such as the Terrascope platform [5]. The expected outcomes of the study are: a Technical Note describing the new algorithm and the validation results over the considered regions of interest.
The proposed study has extreme relevance for ESA, since a new Proba-V-like Companion Cubesat mission (PV-CC) is currently in preparation, due for launch in Nov/Dec 2021. Re-use and further development of the algorithm prototyped within this study will play a crucial role in supporting the exploitation of this future PV-CC satellite in synergy with Sentinel-2 mission.
The activity is expected to last up to 9 months and it will be carried out under the umbrella of an existing contract with ESA/ESRIN awarded to Serco UK Company. Considering the nature of the study, mostly involving development of code, the envisaged work could be conducted remotely, although, regular progress meetings, both teleconferences and face-to-face meetings (Covid-19 permitting), will be planned with Serco UK responsible.
Serco UK in the frame of the RedLab initiative [6] will refund expenses for the whole project’s duration (up to 9 months); this includes a monthly allowance of: 400 Euro/month.
[1] Dierckx, W., Sterckx, S., Benhadj, I., Livens, S., Duhoux, G., Van Achteren, T., Francois, M., Mellab, K. and Saint, G., 2014. PROBA-V mission for global vegetation monitoring: standard products and image quality. International Journal of Remote Sensing, 35(7), pp.2589-2614.
[2] Märtens, M., Izzo, D., Krzic, A. and Cox, D., 2019. Super-resolution of PROBA-V images using convolutional neural networks. Astrodynamics, 3(4), pp.387-402.
[3] https://kelvins.esa.int/proba-v-super-resolution/
[4] http://www.esa.int/Applications/Observing_the_Earth/Copernicus/Sentinel-2
[5] https://terrascope.be/en
[6] https://www.serco.com/eu/redlab-project
Required skills Python; basic knowledge of deep neural networks. Practical hands-on experience will be considered a plus.
Deadline 04/03/2022
PROPONI LA TUA CANDIDATURA