Super-resolution of PROBA-V images using convolutional neural networks
deep learning, convolutional neural network, (CNN), super-resolution imaging, remote sensing, Earth observation (EO)
European Space Aqency (ESA)’s PROBA-V Earth observation (EO) satellite enables us tomonitor our planet at a large scale to study the interaction between vegetation and climate,and provides guidance for important decisions on our common global future. However, theinterval at which high-resolution images are recorded spans over several days, in contrastto the availability of lower-resolution images which is often daily. We collect an extensivedataset of both high- and low-resolution images taken by PROBA-V instruments duringmonthly periods to investigate Multi Image Super-resolution, a technique to merge severallow-resolution images into one image of higher quality. We propose a convolutional neuralnetwork (CNN) that is able to cope with changes in illumination, cloud coverage, andlandscape features which are introduced by the fact that the different images are takenover successive satellite passages at the same region. Given a bicubic upscaling of lowresolution images taken under optimal conditions, we find the Peak Signal to Noise Ratio ofthe reconstructed image of the network to be higher for a large majority of different scenes.This shows that applied machine learning has the potential to enhance large amounts ofpreviously collected EO data during multiple satellite passes.
Tsinghua University Press
Marcus Märtens, Dario Izzo, Andrej Krzic et al. Super-resolution of PROBA-V images using convolutional neural networks.Astrodyn.2019, 3(4): 387–402.