Zobrazit v EDS

Automatic detection of floating aquatic vegetation from remote sensing data.

Uloženo v:
Podrobná bibliografie
Název: Automatic detection of floating aquatic vegetation from remote sensing data.
Autoři: Soria, Juan Miguel1 (AUTHOR), Soria, Esteve2 (AUTHOR), Molner, Juan Víctor1 (AUTHOR) juan.molner@uv.es
Zdroj: International Journal of Remote Sensing. Oct2025, Vol. 46 Issue 20, p7824-7846. 23p.
Témata: *AQUATIC plants, *WATER quality monitoring, *SUPERVISED learning, *REMOTE sensing, *IMAGE retrieval, *REMOTE-sensing images, *INTRODUCED species
Abstrakt: The Copernicus program is an Earth observation component of the European Union Space Program. One of the missions inside the program is Sentinel-2 with 2 satellites orbiting around the earth. The two satellites, Sentinel-2A and Sentinel-2B, can provide images in 13 different bands with a resolution of up to 10 m per pixel. Bands range from infrared through the visible spectrum to short wave infrared. The program, with its policy of free access, allows scientists and researchers to obtain data for several research fields such as cropland, glacier or lake monitoring. In this work we focus our attention on water quality monitoring. During the study of several lakes in the Spanish territory, patches of an invasive species were found floating in the water. Finding out when, where and why these plants appear is of great interest for researchers and environmental managers. Its detection is a manual process through the use of common remote-sensing indexes such as Normalized Difference Vegetation Index and water segmentation. The problem of this approach is its robustness out of the parameters under different locations and settings. The use of vegetation indices requires an adequate atmospheric correction, as well as the implementation of the index that is most useful for each case study. This work proposes an automatic search system for these forms of life with a small amount of labelled training data. By applying self-supervised learning, we are able to generate a model which can be fine-tuned with little data providing similar accuracy as other works in the same field. The model without fine-tuning provides image retrieval capabilities without the need of manual selection of visual features. [ABSTRACT FROM AUTHOR]
Databáze: Academic Search Index
Popis
Abstrakt:The Copernicus program is an Earth observation component of the European Union Space Program. One of the missions inside the program is Sentinel-2 with 2 satellites orbiting around the earth. The two satellites, Sentinel-2A and Sentinel-2B, can provide images in 13 different bands with a resolution of up to 10 m per pixel. Bands range from infrared through the visible spectrum to short wave infrared. The program, with its policy of free access, allows scientists and researchers to obtain data for several research fields such as cropland, glacier or lake monitoring. In this work we focus our attention on water quality monitoring. During the study of several lakes in the Spanish territory, patches of an invasive species were found floating in the water. Finding out when, where and why these plants appear is of great interest for researchers and environmental managers. Its detection is a manual process through the use of common remote-sensing indexes such as Normalized Difference Vegetation Index and water segmentation. The problem of this approach is its robustness out of the parameters under different locations and settings. The use of vegetation indices requires an adequate atmospheric correction, as well as the implementation of the index that is most useful for each case study. This work proposes an automatic search system for these forms of life with a small amount of labelled training data. By applying self-supervised learning, we are able to generate a model which can be fine-tuned with little data providing similar accuracy as other works in the same field. The model without fine-tuning provides image retrieval capabilities without the need of manual selection of visual features. [ABSTRACT FROM AUTHOR]
ISSN:01431161
DOI:10.1080/01431161.2025.2561125