Abstract
A neural network called sinusoidal decomposition neural network (SDNN) is proposed to reconstruct the digital elevation model (DEM) and orbital velocity field (OVF) of sea surface. According to the linear wave theory, DEM can be regarded as the superposition of a series of sine waves, from which OVF can be obtained. The SDNN adopts a fully connected network (FCN) to fit the DEM, which is similar to the inverse discrete Fourier transform (IDFT) model and regression model. The two-dimensional and three-dimensional SDNN are introduced in detail and their validities are demonstrated. A major advantage of the SDNN is that it requires only one scene of the wave height to reconstruct the OVF. By the applications to wind-driven sea surface and ship wake, respectively, the correctness and efficiency of the reconstruction are verified.
Disclosure statement
No potential conflict of interest was reported by the author(s).