The microwave scatterometer is one of the most effective instruments in ocean remote sensing, which urges the need for some theoretical models to accurately estimate the scattering coefficient of the sea surface. For the simulation of the scattering from an ocean surface, the sea spectrum, or its inverse Fourier transform, autocorrelation function is essential. Currently, many sea spectral models have been proposed for describing sea waves. However, which spectrum should be adopted during electromagnetic (EM) computations? A systematic comparison of these models is needed to evaluate their accuracies. In this paper, we focus on numerical simulations of scattering from a rough sea surface in monostatic and bistatic configurations by using six different sea spectral models and the first-order small slope approximation (SSA-1). First, sea spectral models proposed by Elfouhaily et al., Hwang et al., Romeiser et al., Apel et al., Fung et al., and Pierson et al., are compared with each other from different points of view, e.g., the omnidirectional parts, the angular spreading functions, the autocorrelation functions, and the slope variances. We find that the spectra given by Elfouhaily and Hwang could reflect realistic wind sea waves more accurately. Then, the scattering coefficients are simulated in fully monostatic and bistatic configurations. Regarding the monostatic scattering, the results simulated using EM scattering models are compared with those obtained from the measured UAVSAR data in the L band and the empirical model CMOD5 in the C band. Comparisons are made for various incident angles, wind speeds, and wind directions. Meanwhile, special attention is paid to low to moderate incident angles. The comparisons show that, it is difficult to find one certain spectral model to simulate scattering coefficient accurately under all wind speeds or wind directions. Accurate estimations will be obtained using different methods according to different situations.

Based on the simulated signal, a numerical simulation method of synthetic aperture radar (SAR) imaging for time-varying sea surfaces is proposed, which is helpful to study the SAR imaging mechanism of time-varying sea surfaces so as to better extract ocean wave parameters from SAR images. Not only are the modulation of ocean waves, speckle noise, and temporal decorrelation of the small-scale waves considered, but the velocity bunching (VB) effect caused by the motion of large-scale waves is also effectively added to the simulation of the SAR echo signal. To verify the reliability of the simulation method, the simulated SAR images using the parameters of the RADARSAT-2 SAR, the corresponding wind wave information measured by an in-situ buoy, and the reanalysis wave spectra have been compared with the actual RADARSAT-2 SAR images. The comparisons demonstrate that the characteristics of simulated SAR images, such as the intensity distribution and the image spectra, are consistent with those of actual RADARSAT-2 SAR images. Based on the numerical simulation method proposed by us, SAR images of ocean waves for different marine environments and radar platform parameters are simulated. The imaging results indicate that the texture feature of the wind waves would be severely damaged due to the VB effect, while the texture of swells in the simulated SAR images may not be damaged or even becomes clearer. From the simulated SAR image spectrum, it can be found that the azimuth wavenumber is cut off when the VB effect is considered in the simulation process, and the azimuth cut-off wavelength increases with the range-to-velocity ratio.

The interferometric radar altimeter (IRA) is one of the main payloads of the “Guanlan” ocean science satellite proposed by the National Laboratory for Marine Science and Technology of China. To evaluate the effectiveness and accuracy of the IRA in retrieving the ocean dynamic parameters, such as sea surface height (SSH), ocean wave spectrum, and wind speed, two airborne IRA experiments were carried out at Qingdao Xiaomaidao (XMD) sea area on March 31, 2019, and Rizhao sea area on November 16, 2020. In the present work, wave-induced sea surface elevation (SSE) and its spectrum have been retrieved based on the interferograms acquired by the airborne IRA. To suppress the random phase noise, a mean filtering algorithm has been used in the multilook process of calculating the complex IRA images. The results show that the size of the filter window has a significant effect on the retrieved SSE. If the size of the filter window along THE range direction is too large, the flat earth would cause the spectral density of the retrieved ocean wave to be higher. In addition, the comparisons of the retrieved spectra with the buoy measurements demonstrate that the swell can be well-retrieved by IRA images at low sea-state conditions with significant wave height (SWH) less than 0.7 m. However, for wind wave, because of the effect of the velocity bunching along the azimuth direction, the wind wave spectrum can be extracted only when it propagates approximately along the ground-range direction of the IRA images.

The tilt modulation is one of the key factors affecting sea wave inversion accuracy based on some microwave sensors such as SAR, microwave spectrometer, and so on. Up to now, the traditional tilt modulation-transfer functions (MTFs) based on the Bragg scattering model and Phillips spectrum have been used in the sea wave spectrum retrieval algorithm for microwave sensors operating at medium incident angles. The traditional tilt MTFs are only related to radar polarization and incident angle, but not affected by wind speed and wind direction. However, the values of the tilt MTFs evaluated on the basis of the C-band empirical geophysical model function (GMF) CMOD5.na demonstrate that wind speed and wind direction also have significant effects on the tilt MTFs. In addition, the comparisons show that the values of the traditional tilt MTFs for both HH (horizontal) and VV (vertical) polarization cases are obviously smaller, which indicates that the traditional tilt MTFs would underestimate the influence of the local incident angle change on the scattering fields. On the other hand, the difference between the horizontal and the vertical polarization tilt MTFs, however, is overestimated by the traditional tilt MTFs. Finally, by fitting the data evaluated based on the C-band empirical GMF, we propose new tilt MTFs, which can overcome the shortcomings of the traditional tilt MTFs.

Due to the interaction between floating weak targets and sea clutter in complex marine environments, it is necessary to distinguish targets and sea clutter from different dimensions by designing universal deep learning models. Therefore, in this paper, we introduce the concept of multimodal data fusion from the field of artificial intelligence (AI) to the marine target detection task. Using deep learning methods, a target detection network model based on the multimodal data fusion of radar echoes is proposed. In the paper, according to the characteristics of different modalities data, the temporal LeNet (T-LeNet) network module and time-frequency feature extraction network module are constructed to extract the time domain features, frequency domain features, and time-frequency features from radar sea surface echo signals. To avoid the impact of redundant features between different modalities data on detection performance, a Self-Attention mechanism is introduced to fuse and optimize the features of different dimensions. The experimental results based on the publicly available IPIX radar and CSIR datasets show that the multimodal data fusion of radar echoes can effectively improve the detection performance of marine floating weak targets. The proposed model has a target detection probability of 0.97 when the false alarm probability is ${10}^{-3}$ under the lower signal-to-clutter ratio (SCR) sea state. Compared with the feature-based detector and the detection model based on single-modality data, the new model proposed by us has stronger detection performance and universality under various marine detection environments. Moreover, the transfer learning method is used to train the new model in this paper, which effectively reduces the model training time. This provides the possibility of applying deep learning methods to real-time target detection at sea.

To assist in the detection of ship targets in complex sea conditions, a numerical simulation method is proposed to obtain synthetic aperture radar (SAR) images of time-varying ocean ship wakes under various radar, ship, and sea surface parameters. This method addresses the limitations of recent simulations, which failed to simultaneously incorporate different types of time-varying ship wakes, simulate based on the echo data, and discuss the velocity bunching (VB) effect on the image results. To address these issues, firstly, the time-varying wave height and velocity fields of the sea surface, Kelvin wakes, and turbulence wakes are simulated using the linear filtering method, classic fluid dynamics models, and attenuation function method, respectively. Secondly, raw data of the ocean ship wakes are obtained by calculating the backscattering fields using geophysical model functions (GMFs), as well as by determining the changing slant range varying with the elevation and velocity fields. Thirdly, by applying the Range-Doppler algorithm (RDA) for pulse compression and range cell migration correction (RCMC) on the echo data, SAR images with and without the VB effect are generated. Our simulation also accounts for the influence of speckle noise. The SAR imaging results indicate that whether the VB effect is considered or not, the radar electromagnetic wavebands, polarization modes, wind speeds, and the relative wind directions have distinct impacts on the SAR image intensity, and the texture and morphology of ship wakes vary significantly with the wind speeds, ship speeds, and the relative radar looking directions. When considering the VB effect, the azimuthal offset and blur in the images caused by the more intense wave motion also increase with the wave speeds.