Research group on remote sensing
- Keywords: Keywords: deep learning, remote sensing, image processing, satellite imaging
Research group activities
Recognition in remote sensing images. The group has developing the new techniques for the recognition of objects in remote sensing images. The techniques are based on the use of deep learning and convolutional and transformer neural networks.
Different spacial-temporal image processing. The group has been developing new techniques to deal with the special-temporal images of different satellites and different resolutions. The data fusion is also a topic of interest, especially the fusion of images of hyperspectral and multispectral satellite data.
Creation of new machine learning algorithms. The algorithms based on multi-task, transfer learning problems. The shortage of in-situ data is a problem in remote sensing. The group have proposed new algorithms to estimate the chlorophyll-a concentration and estimate the blooming from multispectral data.
Change Detection in Remote Sensing: The research group focuses on advanced methodologies for detecting temporal and spatial changes in remote sensing imagery. Utilizing hyperspectral and multispectral satellite data, our efforts aim to quantify and characterize dynamic shifts in terrestrial features, enabling precise monitoring of environmental and anthropogenic changes.
Semantic Information Extraction & Agent-Based Modeling: Our research delves into the extraction of semantic information from remote sensing datasets. This extracted data is subsequently integrated into agent-based mathematical models. These models simulate and analyze interactions of distinct entities, such as land use patterns and vegetation dynamics, providing insights into complex system behaviors and potential future trajectories.
Proposal
Remote sensing research group supports:
- Cooperation in join projects on remote sensing from hyper/multispectral data.
- Creation of novel machine learning algorithms for automate predictions over in-situ measuments.
Meet our team
Lead researcher: Assoc. Prof. Linas Petkevicius
Research Group:
Assoc. Prof. Valentas Gruzauskas; Partn. Prof. Jonas Matuzas; Prof. Sebastien Gadal; Dalia Grendaite; Dalia Breskuviene; Thomas Gloaguen; Marius Somka.
Research outcomes
Fyleris, T., Krisciunas, A., Gruzauskas, V., Calneryte, D., & Barauskas, R. (2022). Urban change detection from aerial images using convolutional neural networks and transfer learning. ISPRS International Journal of Geo-Information, 11 (4), 246.
Grendaite, D., & Stonevicius, E. (2022). Uncertainty of atmospheric correction algorithms for chlorophyll a concentration retrieval in lakes from Sentinel-2 data. Geocarto International, 37(23), 6867-6891.
Bagdonavicius, V., & Petkevicius, L. (2020). A new multiple outliers identification method in linear regression. Metrika, 83(3), 275-296.
Gadal, Sebastien, Mounir Oukhattar, and Solomon Abody Otobo. "Multi-temporal recognition of built-up area and land cover changes using machine learning approach in the Metropolis of Aix-Marseille-Provence in France." 2023 Joint Urban Remote Sensing Event (JURSE). IEEE, 2023.
Gadal, S., & Gloaguen, T. (2021). Environmental issuesin the coastal regions of the south-eastern Baltic Sea: A sensitive natural environment in the face of increasing anthropic pressures. Baltica: An International Yearbook for Quaternary Geology and Palaeogeography, Coastal Morphology and Shore Processes, Marine Geology and Recent Tectonics of the Baltic Sea Area, 34(2), 203-215.
Resources
Segmentation and detection compute resourses, machine learning compute pipelines.
Contacts
Assoc. Prof. Linas Petkevicius
Phone: +370 603 18707
E-mail:
More about the faculty: https://mif.vu.lt/lt3/en/
Department for Research and Innovation
E-mail:
More information: https://www.vu.lt/en/business