Frontiers in Earth Science (Jan 2023)
Evaluation of feature extraction algorithms for oceanic internal waves based on nighttime detection data of spaceborne low light imager
Abstract
The day/night band channel on the JPSS series of satellites can detect the light and dark fringes of oceanic internal waves due to the reflectivity difference caused by the roughness of the sea surface under moon flare conditions. After optical imaging of oceanic internal waves, three image processing algorithms, i.e., the two-dimensional S transform, windowed Fourier transform, and wavelet packet transform methods, can be used to extract the parameter features of horizontal wavelength and propagation direction. The wave domain with known parameters is established through data simulation, and both image quality and image resolution are analyzed to assess algorithm performance in terms of relative errors. Finally, the experimental conclusions are verified in two examples of satellite observations in the South China Sea in 2020. We found that the windowed Fourier transform and wavelet packet transform methods exhibit better noise immunity, and the two-dimensional S transform method exhibits less calculation error and is more applicable to cases with small wavelengths. For large wavelengths, the windowed Fourier transform method is more suitable for calculating the horizontal wavelength, and the wavelet packet transform method is more suitable for calculating the propagation direction. By evaluating the applicability of these algorithms, this study provides a theoretical basis to support the analysis and processing of internal wave characteristics in future.
Keywords
