Wyniki wyszukiwania - BazTech

Ograniczanie wyników

1 Teledetekcja Środowiska

1 2016

Znaleziono wyników: 1

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: wskaźniki wodne

Sortuj według:

Ogranicz wyniki do:

Wykrywanie wody na zdjęciach optycznych Sentinel-2 na podstawie wskaźników wodnych

Robak A., Gadawska A., Milczarek M., Lewiński S.

Teledetekcja Środowiska

2016

T. 55

59--72

Copernicus Programme managed by the European Commission and implemented in partnership with i.a. the European Space Agency (ESA) provides free access to satellite data from Sentinel mission including Sentinel-2 high resolution optical satellite data. The aim of the research was to recognize opportunities of water detection on Sentinel-2 imagery. Satellite data was analyzed before and after atmospheric correction. A number of tests were carried out using indices selected from the literature. Based on the gained experience, a new index for water detection has been proposed, Sentinel Water Mask (SWM), specially adapted for Sentinel-2 images. Its construction is based on the highest difference between spectral values of water surface and other land cover forms. SWM provides quick and effective detection of water which is especially important in flood assessment for crisis management. Research was performed on unprocessed images of Sentinel-2 Level-1C and images after atmospheric correction (Level-2A). Water was detected with the use of threshold values determined by the visual interpretation method. The accuracy of the obtained water masks was assessed on the basis of validation points. The performed analysis allowed to indicate indices, which enable estimation of areas covered by water on Sentinel-2 images with high classification accuracy, this is: AWEInsh (Automated Water Extraction Index), MNDWI (Modified Normalized Difference Water Index), NDWIMcFeeters (Normalized Difference Water Index). Their application allowed for achievement of overall accuracy of water detection oscillating around 95% and high Kappa coefficient. The usage of the proposed SWM index leads to slightly better results (more than 96%). The sensitivity to the selection of threshold values of analyzed indices was assessed and then the optimal threshold ranges were determined. The optimal threshold value for NDWIMcFeeters should be included in the value range (0.1, 0.2), for MNDWI (0.2, 0.3) and for SWM (1.4, 1.6). The unambiguous threshold range for AWEInsh index was impossible to indicate due to the large range of values.