Guided Unsupervised Learning by Subaperture Decomposition for Ocean SAR Image Retrieval

Type Article
Date 2023
Language English
Author(s) Ristea Nicolae-Cătălin1, Anghel AndreiORCID1, Datcu MihaiORCID2, Chapron BertrandORCID3
Affiliation(s) 1 : Research Center for Spatial Information (CEOSpaceTech) and the Department of Telecommunications, University Politehnica of Bucharest, Bucharest, Romania
2 : Research Center for Spatial Information (CEOSpaceTech), University Politehnica of Bucharest, Bucharest, Romania
3 : Laboratoire d'Ocanographie Physique et Spatiale (LOPS), Plouzané, Ifremer, France
Source Ieee Transactions On Geoscience And Remote Sensing (0196-2892) (Institute of Electrical and Electronics Engineers (IEEE)), 2023 , Vol. 61 , N. 5207111 , P. 11p.
DOI 10.1109/TGRS.2023.3272279
WOS© Times Cited 1
Keyword(s) Radar polarimetry, Image retrieval, Doppler effect, Synthetic aperture radar, Transformers, Task analysis, Sea surface, Doppler centroid estimation (DCE), image retrieval, ocean imagery, remote sensing (RS), subapertures decomposition, synthetic aperture radar (SAR), unsupervised learning

A spaceborne synthetic aperture radar (SAR) can provide accurate images of the ocean surface roughness day-or-night in nearly all-weather conditions, being a unique asset for many geophysical applications. Considering the huge amount of data daily acquired by satellites, automated techniques for physical features extraction are needed. Even if supervised deep learning methods attain state-of-the-art results, they require a great amount of labeled data, which are difficult and excessively expensive to acquire for ocean SAR imagery. To this end, we use the subaperture decomposition (SD) algorithm to enhance the unsupervised learning retrieval on the ocean surface, empowering ocean researchers to search into large ocean databases. We empirically prove that SD improves the retrieval precision with over 20% for an unsupervised transformer autoencoder network. Moreover, we show that SD brings an important performance boost when Doppler centroid images are used as input data, leading the way to new unsupervised physics-guided retrieval algorithms.

Full Text
File Pages Size Access
Author's final draft 9 29 MB Open access
11 5 MB Access on demand
Top of the page