Simulation Study Of Sea Surface SAR Images Based On Bilinear Model

Improving the accuracy of maritime ship detection algorithms requires a large amount of ocean SAR data,which is difficult or expensive to obtain from real measurements,so it is important to conduct the research of sea surface SAR image simulation.An important element of simulated sea surface SAR images is to build a 3D model of the sea surface that better matches the actual sea conditions.One of the most commonly used methods for 3D modeling of the sea surface is to use the sea spectrum to invert the sea surface,and this method is based on linear stochastic wave theory.However,the actual sea surface is not a completely linear phenomenon,and the ability of nonlinear models to characterize nonlinear features is insufficient.In order to construct a 3D sea surface that is more consistent with the real sea surface,this thesis proposes a bilinear model with mean sea level as the reference and Creamer model as the constraint.The main research of this thesis is as follows:(1)By studying the difference between different wave spectra,the appropriate wave spectra are selected to invert the linear stochastic sea surface by linear superposition method and linear filtering method.In response to the problem that the linear model and nonlinear model do not sufficiently characterize the nonlinear features of the actual sea surface,a model based on the mean sea level to judge the location of the point-the bilinear model is proposed.By establishing the one-dimensional sea surface maps of the linear model,Creamer model,and bilinear model,the experimental comparison shows that the bilinear model is closer to the real sea surface than the Creamer model.(2)The slope of the sea surface is crucial to calculate the electromagnetic scattering coefficient of the sea surface.In order to characterize the closeness of the linear sea surface,Creamer sea surface,and bilinear sea surface to the real sea surface,this thesis proposes to use the slope difference index and the mean value index for evaluation.From the slope difference diagram,it can be observed that the bilinear model is more accurate than the Creamer model,so the slope change of the bilinear sea surface is closer to the real sea surface.From the mean value table,we can get that the sea surface generated by the Creamer model and the bilinear model are both asymmetric,but the asymmetry of the bilinear model is stronger.Therefore,the bilinear model is closer to the real sea surface than the Creamer model in terms of the slope variation of the sea surface and the asymmetry of the wave surface.(3)To address the situation that the K distribution does not always fit the sea clutter data well in high sea state,this thesis verifies the statistical properties of the simulated images with a more suitable model according to different sea states.The simulation data are fitted with K distribution at low and medium wind speeds and K+R distribution at high sea state.The fitting results show that the simulated data have a better fit with the K distribution at low and medium wind speeds.At high wind speeds,the K+R distribution has a better fit than the K distribution.The fitting results show that the statistical properties of the simulated images are consistent with those of the actual sea clutter SAR images.Therefore,the method proposed in the thesis for simulating sea surface SAR images is feasible.