A deep learning-based SAR one-dimensional sea wave spectrum inversion method, device and medium

Abstract

The application discloses a kind of based on deep learning SAR one-dimensional sea wave spectrum inversion method, device and medium, method includes: obtaining the SAR image data of sea wave, satellite data and sea wave spectrum observation data;Utilize the SAR image data, satellite data and sea wave spectrum observation data are simulated one-dimensional sea wave spectrum of WW3 mode;Based on simulation result constructs a kind of based on deep learning SAR one-dimensional sea wave spectrum inversion algorithm, and utilizes the SAR one-dimensional sea wave spectrum inversion algorithm obtains inversion result;Verify the accuracy of the inversion result.The method provided in the application has applicability and accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of synthetic aperture radar, and in particular to a SAR one-dimensional sea wave spectrum inversion method and device based on deep learning and a medium. BACKGROUND

[0002] Due to the importance of energy exchange at the sea-atmosphere interface, sea surface waves are an important aspect of oceanography. In-situ observation data from moored buoys and aircraft are the best source for wave research. However, these observation data are mostly obtained in coastal waters, and there are few measurement data in open seas and extreme conditions. So far, operational remote sensing products, i.e., wind speed measured by scatterometers and microwave radiometers, sea surface height and significant wave height measured by altimeters, and sea wave spectrum measured by the Surface Wave Investigation and Monitoring (SWIM), are generally used in large-scale oceanography research. The spatial resolution of these products is about 10×20 km, consistent with the satellite orbit, so they are rarely used for small-scale wave research.

[0003] Compared with the above-mentioned sensors, synthetic aperture radar (SAR) has a strong ability to synchronously detect sea surface wind waves, with a spatial resolution of 1 km and a scanning range of 500 km. By analyzing airborne data during the Seasat mission in 1978, it was found that the backscattering complex of synthetic aperture radar was refined into a normalized radar cross section (NRCS), which was exponentially linear with the increase of wind speed. The same is true for microwave sensors with an incidence angle greater than 20° due to the Bragg resonance mechanism. Therefore, the C-band geophysical model function (GMF) designed for scatterometer wind retrieval, such as the CMOD series, is also applicable to synthetic aperture radar.

[0004] Recently, with the accumulation of synthetic aperture radar data, specific co-polarization (vertical-vertical (VV) and horizontal-horizontal (HH)) GMFs for synthetic aperture radar wind retrieval have been developed, such as C-SARMOD for Sentinel-1 (S-1) and Gaofen-3 (GF-3), LMOD for ALOS, and XMOD for TerraSAR-X / TanDEM-X (TS-X / TD-X). The key to applying the synthetic aperture radar global mobility framework is to know the wind direction in advance. Although the pattern with a wavelength of about 1×3 km on the synthetic aperture radar image is parallel to the wind direction, it is necessary to use external signal sources, wavelet analysis, or polar coordinate features to eliminate the ambiguity of 180°. In recent studies, machine learning and convolutional neural networks have been used for synthetic aperture radar wind direction inversion without any prior information. Since the co-polarization backscattering coefficient appears to be saturated under strong wind conditions, the joint use of VV and vertical-horizontal (VH) polarized NRCS has good performance for synthetic aperture radar wind direction inversion during tropical cyclones (TCs).

[0005] The sea wave mechanism of synthetic aperture radar imaging is relatively complex, including fluid dynamic modulation, tilt modulation, and velocity bunching. The former two are linear modulation effects, while the velocity bunching is the key to synthetic aperture radar imaging of sea wave imaging. The nonlinear effect of the velocity beam makes it impossible to detect when the wavelength propagating in the flight direction is shorter than a certain value, which is called the azimuthal cutoff wavelength. In addition, under the principle of limited sea wind during tropical cyclones, typhoons are inverted using the azimuthal cutoff wavelength. Synthetic aperture radar sea wave inversion methods are divided into two categories: theoretical algorithms based on imaging mechanisms and empirical models that construct the actual function between sea wave parameters and synthetic aperture radar imaging variables. The first category includes theoretical sea wave algorithms, namely MPI, SPRA, PRSA, and PFSM methods. Simply put, the principle of the theoretical algorithm is to solve the SAR image spectrum by minimizing the complex cost function to invert the sea wave spectrum. In addition, since the characteristics of polarimetric SAR are determined only by the tilt, the sea wave spectrum is directly inverted by the difference between the NRCS of the co-polarization and cross-polarization channels. The second category commonly uses a multiple-order regression function, namely CWAVE_Envi for Envisat-ASAR, CWAVE_S1 for S-1, and CSAR_WAVE for GF-3. The technical level of the first and second categories is 15 to 20 years ago. In recent years, sea wave algorithm research has replaced regression fitting with deep learning for SAR sea wave parameter inversion. Although the SWH obtained through deep learning is quite reliable (i.e., ~0.3 m root mean square error (RMSE)), the accuracy of the SAR sea wave spectrum needs to be improved (i.e., the RMSE of the SWH obtained by SAR inversion through a theoretical algorithm is 0.5-0.6 m). In addition, due to the highly nonlinear velocity aggregation and wave breaking in extreme sea conditions, sea wave spectrum inversion during tropical cyclones is a challenge. SUMMARY

[0006] The present application provides a deep learning-based SAR one-dimensional sea wave spectrum inversion method, device and medium, aiming to effectively solve the above technical problems.

[0007] In a first aspect, the present application provides a deep learning-based SAR one-dimensional sea wave spectrum inversion method, which comprises:

[0008] Obtaining SAR image data, satellite data, and sea wave spectrum observation data of the sea wave;

[0009] Simulating a one-dimensional sea wave spectrum using the WW3 model with the SAR image data, satellite data, and sea wave spectrum observation data;

[0010] Constructing a deep learning-based SAR one-dimensional sea wave spectrum inversion algorithm based on the simulation results, and obtaining an inversion result using the SAR one-dimensional sea wave spectrum inversion algorithm;

[0011] Verifying the accuracy of the inversion result.

[0012]

[0012] Further, the SAR image data is image data obtained by a GF-3 satellite, the satellite data is data obtained by a HY-2 satellite, and the sea wave spectrum observation data is obtained by a SWIM wave spectrum instrument.

[0013] Further, the WW3 mode simulation one-dimensional sea wave spectrum by using the SAR image data, the satellite data and the sea wave spectrum observation data comprises:

[0014] The time range and the space range of the SAR image data are determined, the WW3 is used to simulate the sea wave spectrum matched with the time range and the space range, and the sea wave spectrum is taken as a simulation result.

[0015] Further, the SAR one-dimensional sea wave spectrum inversion algorithm based on deep learning is constructed based on the simulation result, and the SAR one-dimensional sea wave spectrum inversion algorithm based on deep learning comprises:

[0016] The SAR image data is preprocessed, and polarized spectrum calculation is performed on the preprocessed image data to obtain a calculation result.

[0017] The simulation result and the calculation result are cooperatively positioned to obtain positioned data, and the positioned data are used to train a pre-constructed one-dimensional sea wave spectrum model to obtain a trained one-dimensional sea wave spectrum model.

[0018] Further, the accuracy of the inversion result is verified, and the accuracy of the inversion result comprises:

[0019] The inversion result is verified by a HY-2 altimeter and a SWIM wave spectrum instrument respectively, the two verification data are matched in time and space, the matched data are screened through a preset error range, and the screened data are used to evaluate the accuracy of the inversion result.

[0020] In the second aspect, the application further provides a SAR one-dimensional sea wave spectrum inversion device based on deep learning, which comprises:

[0021] A data acquisition module is configured to acquire SAR image data, satellite data and sea wave spectrum observation data of sea waves.

[0022] A simulation module is configured to simulate a one-dimensional sea wave spectrum by using the SAR image data, the satellite data and the sea wave spectrum observation data.

[0023] A construction module is configured to construct a SAR one-dimensional sea wave spectrum inversion algorithm based on deep learning based on a simulation result, and obtain an inversion result by using the SAR one-dimensional sea wave spectrum inversion algorithm.

[0024] A verification module is configured to verify the accuracy of the inversion result.

[0025] In a third aspect, the present application also provides a storage medium, wherein a plurality of instructions are stored in the storage medium, and the instructions are suitable for being loaded by a processor to execute the method according to any one of the above aspects.

[0026] By the above-mentioned deep learning-based SAR one-dimensional sea wave spectrum inversion method, device and medium in the present application, at least the following technical effects can be achieved: the sea wave spectrum matched with the GF-3 image is simulated by using the third-generation numerical model WAVEWATCH-III (WW3), and the simulated significant wave height (SWH) is verified to obtain the root mean square error (RMSE), the correlation coefficient (r) and the scatter index (SI). The basic deep learning method for SAR sea wave spectrum inversion, i.e., the multilayer perceptron (MLP), is trained by 1300 wave pattern images. The co-polarization SAR sea wave spectrum ((VV) and HH)), the hydrodynamic modulation and two updated modulation transfer functions (MTF), i.e., the tilt modulation and the velocity bunching, wherein the wind speed obtained from the VV polarization image, are used as the input of the training process. The sea wave spectrum algorithm developed based on the MLP is applied to 3000 QPS images, and compared with the SWIM data sea wave spectrum. The method provided by the present application has applicability and accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0028] Figure 1 The flowchart of the deep learning-based SAR one-dimensional sea wave spectrum inversion method provided by the embodiments of the present application;

[0029] Figure 2 The schematic diagram of the two-dimensional sea wave spectrum of the SWIM data product provided by the embodiments of the present application;

[0030] Figure 3 a is the schematic diagram of the wind speed of the ECMWF ERA-5 data product provided by the embodiments of the present application;

[0031] Figure 3 b, Figure 3 c is the schematic diagram of the flow rate and the schematic diagram of the water level of the CMEMS data product provided by the embodiments of the present application, respectively;

[0032] Figure 3 d is the schematic diagram of the significant wave height simulated by the WW3 mode provided by the embodiments of the present application;

[0033] Figure 4An algorithm flowchart of the MLP provided by the embodiment of the present application;

[0034] Figure 5 Fig. a is a comparison chart of the one-dimensional sea wave spectrum inversion result of the MLP algorithm provided by the embodiment of the present application and the one-dimensional sea wave spectrum result simulated by the WW3 mode;

[0035] Figure 5 Fig. b is a comparison chart of the one-dimensional sea wave spectrum inversion result of the MLP algorithm provided by the embodiment of the present application and the SWIM data product result;

[0036] Figure 6 Fig. a, Figure 6 Fig. b, Figure 6 Fig. c is a comparison chart of the inversion significant wave height result and the SWIM data product, the HY-2 altimeter product and the WW3 simulation result. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative labor fall within the protection scope of the present application.

[0038] In the description of the present application, it should be noted that, unless otherwise explicitly specified and limited, the term "and / or" in this paper is only to describe the association relationship of the associated objects, which means that there can be three kinds of relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " in this paper generally represents an "or" relationship between the associated objects without special explanation.

[0039] Referring to Figure 1 The present application provides a deep learning-based SAR one-dimensional sea wave spectrum inversion method, which comprises the following steps:

[0040] S101, acquiring SAR image data, satellite data and sea wave spectrum observation data of sea waves.

[0041] The SAR image data is image data acquired by a GF-3 satellite, the satellite data is data acquired by a HY-2 satellite, and the sea wave spectrum observation data is obtained by a SWIM spectrometer.

[0042] Illustratively, 1300 WV mode and 3000 QPS mode GF-3 SAR images during 2017-2022, and the matching SWIM sea wave spectrum observation data and HY-2 data products are collected in this embodiment. Since the spatial resolution of SWIM is 18km, only 400 QPS images match the sea wave product results of SWIM. Figure 2 An example of the two-dimensional sea wave spectrum of the SWIM data product is given. It should be noted that there is an error of 180 degrees in the propagation direction of the wind, so the real wave energy is half of the wave spectrum measured by SWIM.

[0043] S102, using the SAR image data, satellite data and sea wave spectrum observation data to simulate one-dimensional sea wave spectrum by WW3 mode.

[0044] Specifically, the time range and spatial range of the SAR image data are determined, the WW3 (third generation numerical model WAVEWATCH-III) is used to simulate the sea wave spectrum matching the time range and spatial range, and the sea wave spectrum is taken as the simulation result.

[0045] Among them, the model forcing field includes 0.25° spatial resolution data of ECMWF (European Center for Medium-Range Weather Forecasts) every 1 hour and daily reanalysis (i.e. sea surface flow and sea level) published by CMEMS (Copernicus Marine Environment Monitoring Service) with 0.08° spatial resolution. The 0.01° spatial resolution water depth is from GEBCO (General Bathymetry Chart of Oceans). The spatial resolution of the simulation result is 0.05°, and the time resolution is 30min. Figure 3 a is the wind speed of the ECMWF ERA-5 data product, Figure 3 b, Figure 3 c is the flow rate and water level of the CMEMS data product respectively, Figure 3 d is the significant wave height simulated by the WW3 mode.

[0046] S103, constructing a SAR one-dimensional sea wave spectrum inversion algorithm based on deep learning based on the simulation result, and obtaining an inversion result by using the SAR one-dimensional sea wave spectrum inversion algorithm.

[0047] Specifically, the SAR image data is preprocessed, and the polarized spectrum of the preprocessed image data is calculated to obtain a calculation result.

[0048] The simulation result and the calculation result are cooperatively positioned to obtain positioned data, and the pre-constructed one-dimensional sea wave spectrum model is trained by using the positioned data to obtain a trained one-dimensional sea wave spectrum model.

[0049] The trained one-dimensional wave spectrum model was used to test the validation dataset to obtain the inversion results.

[0050] Schematic illustration: The one-dimensional wave spectrum model is a multi-layer perceptron (MLP), a type of feedforward neural network model widely used in various classification and regression tasks. In regression tasks, the MLP model learns and establishes a non-linear relationship between input features and continuous target variables to predict the target variable.

[0051] In an MLP model, neurons form a complex nonlinear mapping network. By adjusting the connection weights and biases between neurons, effective modeling and prediction of various complex input data can be achieved. Neurons are hierarchically organized, consisting of an input layer, hidden layers, and an output layer. The input layer receives the raw data features, while the output layer generates predictions from the model. The hidden layer, located between the input and output layers, is responsible for the nonlinear transformation of the input and feature extraction.

[0052] The gradient of the loss function on the model parameters is calculated using the backpropagation algorithm, and then optimally updated to minimize the model's prediction error. Figure 4 This is a flowchart of the deep learning algorithm for MLP. In this embodiment, the model input includes the wind speed retrieved from SAR, the SAR wave spectrum under VV and HH polarizations (MLP), tilted MTFs under VV and HH polarizations, hydrodynamic modulation and velocity convergence MTFs, and the one-dimensional wave spectrum simulated by WW3. The output is the one-dimensional wave spectrum retrieved from SAR.

[0053] S104, verify the accuracy of the inversion results.

[0054] Specifically, the inversion results are verified using a HY-2 altimeter and a SWIM spectrometer, and the two verification data are spatiotemporally matched. The matched data are then filtered using a preset error range, and the accuracy of the inversion results is evaluated using the filtered data.

[0055] Schematively, the one-dimensional spectral results of this method are verified using SWIM data products and WW3 model simulation results. From Figure 5 The two case studies clearly demonstrate that the one-dimensional wave spectrum retrieved from one-dimensional SAR is consistent with the one-dimensional wave spectra obtained from WW3 simulations and SWIM. The significant wave height was calculated using the one-dimensional wave spectrum results, and the inversion results were verified using SWIM data products, WW3 model simulation results, and HY-2 data products. The comparison results are as follows: Figure 6 a, Figure 6 b、 Figure 6 As shown in Figure c, the accuracy of the inversion algorithm is verified.

[0056] The application provides a deep learning-based SAR one-dimensional sea wave spectrum inversion method, a sea wave spectrum matched with a GF-3 image is simulated by using a third-generation numerical model WAVEWATCH-III (WW3), and the simulated significant wave height (SWH) is verified to obtain a root mean square error (RMSE), a correlation coefficient (r) and a scatter index (SI). A basic deep learning method for SAR sea wave spectrum inversion, i.e., a multilayer perceptron (MLP), is trained by 1300 wave pattern images. The co-polarization SAR sea wave spectrum ((VV) and HH)), hydrodynamic modulation and two updated modulation transfer functions (MTF), i.e., tilt modulation and velocity bunching, wherein the wind speed obtained from the VV polarization image is used as the input of the training process. The sea wave spectrum algorithm developed based on the MLP is applied to 3000 QPS images, and compared with the SWIM data sea wave spectrum. The method provided by the application has applicability and accuracy.

[0057] Based on any one of the above embodiments, another embodiment of the application further provides a deep learning-based SAR one-dimensional sea wave spectrum inversion device, which comprises:

[0058] A data acquisition module is configured to acquire SAR image data, satellite data and sea wave spectrum observation data of sea waves.

[0059] A simulation module is configured to simulate a one-dimensional sea wave spectrum by using the SAR image data, satellite data and sea wave spectrum observation data.

[0060] A construction module is configured to construct a deep learning-based SAR one-dimensional sea wave spectrum inversion algorithm based on the simulation result, and obtain an inversion result by using the SAR one-dimensional sea wave spectrum inversion algorithm.

[0061] A verification module is configured to verify the accuracy of the inversion result.

[0062] The deep learning-based SAR one-dimensional sea wave spectrum inversion device provided by the application corresponds to the deep learning-based SAR one-dimensional sea wave spectrum inversion method described above, and will not be described here.

[0063] Based on any one of the above embodiments, another embodiment of the application further provides an electronic device, which can comprise a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor can invoke the logical instructions in the memory to execute the above method.

[0064] In addition, the logic instructions in the above-mentioned memory can be realized in the form of a software function unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0065] In another aspect, the embodiments of the present application also provide a storage medium having a plurality of instructions stored thereon, the instructions being adapted to be loaded by a processor to execute the deep learning-based SAR one-dimensional sea wave spectrum inversion method provided by the above-mentioned embodiments.

[0066] In another aspect, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0067] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e., they can be located in one place or distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. Those skilled in the art can understand and implement it without creative labor.

[0068] Those skilled in the art can clearly understand the implementation of the various embodiments by means of software and necessary general hardware platforms through the description of the above embodiments, and of course, the implementation can also be through hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, and the computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method described in each embodiment or some parts of the embodiment.

[0069] In summary, although the present application has been disclosed as above with preferred embodiments, the above preferred embodiments are not intended to limit the present application, and those skilled in the art can make various modifications and decorations without departing from the spirit and scope of the present application. Therefore, the protection scope of the present application is subject to the scope defined by the claims.

Claims

1. A deep learning-based SAR one-dimensional sea wave spectrum inversion method, characterized in that, The application relates to a method for obtaining a sea wave spectrum based on SAR (Synthetic Aperture Radar) data. The method comprises the following steps: SAR image data, satellite data and sea wave spectrum observation data are acquired; a one-dimensional sea wave spectrum is simulated by using the SAR image data, the satellite data and the sea wave spectrum observation data through a WW3 model; a SAR one-dimensional sea wave spectrum inversion algorithm based on deep learning is constructed based on simulation results, and inversion results are obtained by using the SAR one-dimensional sea wave spectrum inversion algorithm; the accuracy of the inversion results is verified; wherein the SAR one-dimensional sea wave spectrum inversion algorithm based on deep learning is constructed based on simulation results, and the method comprises the following steps: the SAR image data is preprocessed, and polarized spectrum calculation is performed on the preprocessed image data to obtain calculation results; the simulation results and the calculation results are cooperatively positioned to obtain positioned data, and a one-dimensional sea wave spectrum model is trained by using the positioned data to obtain a trained one-dimensional sea wave spectrum model; 2. The method of claim 1, wherein, in the one-dimensional sea wave spectrum model training process, co-polarization SAR sea wave spectrum, hydrodynamic modulation and two updated modulation transfer functions, i.e. tilt modulation and velocity bunching, are used as inputs of the training process, wherein the wind speed obtained from the VV polarized image is included.

3. The method of claim 1, wherein, The SAR image data is image data acquired by a GF-3 satellite, the satellite data is data acquired by a HY-2 satellite, and the sea wave spectrum observation data is obtained by a SWIM wave spectrum instrument. The one-dimensional sea wave spectrum is simulated by using the SAR image data, the satellite data and the sea wave spectrum observation data through the WW3 model, and the method comprises the following steps:

4. The method of claim 1, wherein, the time range and the space range of the SAR image data are determined, the WW3 is used to simulate a sea wave spectrum which is matched with the time range and the space range, and the sea wave spectrum is used as the simulation result. the accuracy of the inversion results is verified, and the method comprises the following steps:

5. A device for retrieving one-dimensional sea wave spectrum based on deep learning SAR, characterized in that, the inversion results are verified by using a HY-2 altimeter and a SWIM wave spectrum instrument, the two verification data are matched in time and space, the matched data are screened through a preset error range, and the accuracy of the inversion results is evaluated by using the screened data. The application relates to a method for obtaining a sea wave spectrum based on SAR (Synthetic Aperture Radar) data. The method comprises the following steps: SAR image data, satellite data and sea wave spectrum observation data are acquired; a one-dimensional sea wave spectrum is simulated by using the SAR image data, the satellite data and the sea wave spectrum observation data through a WW3 model; a SAR one-dimensional sea wave spectrum inversion algorithm based on deep learning is constructed based on simulation results, and inversion results are obtained by using the SAR one-dimensional sea wave spectrum inversion algorithm; the accuracy of the inversion results is verified; wherein the SAR one-dimensional sea wave spectrum inversion algorithm based on deep learning is constructed based on simulation results, and the method comprises the following steps: the SAR image data is preprocessed, and polarized spectrum calculation is performed on the preprocessed image data to obtain calculation results; the simulation results and the calculation results are cooperatively positioned to obtain positioned data, and a one-dimensional sea wave spectrum model is trained by using the positioned data to obtain a trained one-dimensional sea wave spectrum model; in the one-dimensional sea wave spectrum model training process, co-polarization SAR sea wave spectrum, hydrodynamic modulation and two updated modulation transfer functions, i.e. tilt modulation and velocity bunching, are used as inputs of the training process, wherein the wind speed obtained from the VV polarized image is included. The SAR image data is image data acquired by a GF-3 satellite, the satellite data is data acquired by a HY-2 satellite, and the sea wave spectrum observation data is obtained by a SWIM wave spectrum instrument. The one-dimensional sea wave spectrum is simulated by using the SAR image data, the satellite data and the sea wave spectrum observation data through the WW3 model, and the method comprises the following steps: the time range and the space range of the SAR image data are determined, the WW3 is used to simulate a sea wave spectrum which is matched with the time range and the space range, and the sea wave spectrum is used as the simulation result. the accuracy of the inversion results is verified, and the method comprises the following steps: the inversion results are verified by using a HY-2 altimeter and a SWIM wave spectrum instrument, the two verification data are matched in time and space, the matched data are screened through a preset error range, and the accuracy of the inversion results is evaluated by using the screened data. In the training process of the one-dimensional sea wave spectrum model, the co-polarization SAR sea wave spectrum, hydrodynamic modulation, and two updated modulation transfer functions, i.e. tilt modulation and velocity bunching, are taken as inputs of the training process, wherein the wind speed obtained from the VV-polarization image is included.

6. A storage medium, characterized by The storage medium has stored therein a plurality of instructions adapted to be loaded by the processor to execute the method of any one of claims 1 to 4.

Images