Sea surface wind speed inversion method, system, equipment and medium combining simulation model and satellite data advantages

Abstract

The invention relates to a sea surface wind speed inversion method, system, equipment and medium in combination with advantages of a simulation model and satellite data, a simulated time delay Doppler (DDM) image is generated by inputting parameters such as positions and speeds of a satellite and a receiver, gain of a receiving antenna and the like, satellite actual measurement data is expanded in combination with the simulated image, an actual measurement-simulation combined data set is constructed, and a sea surface wind speed inversion result is obtained. Extracting a leading edge slope and a trailing edge slope, and calculating to obtain a leading edge-trailing edge composite characteristic value to invert a sea surface wind speed value; the system, the equipment and the medium are used for implementing the method. According to the invention, the simulation of the time delay Doppler diagram is carried out through the position and speed of the satellite and the receiver, the wind speed, the GNSS signal frequency and the receiving antenna gain parameter, and the data sample can be expanded under the condition that the actually measured data sample of a specific sea area or sea condition is scarce. The method has the advantages that the applicability is wide, and sea surface wind speed inversion can be accurately achieved under the condition that samples are scarce.

Description

Technical Field

[0001] The present invention belongs to the technical field of ocean data processing, and in particular relates to a sea surface wind speed inversion method, system, equipment and medium combining the advantages of simulation models and satellite data. Background Art

[0002] Traditional sea surface wind speed monitoring methods (such as buoy measurements, radar measurements, and weather station observations) are difficult to implement over large areas due to their high costs and limited observation areas. Global Navigation Satellite System-Reflectometry (GNSS-R) technology has become an ideal means of monitoring sea surface wind speed due to its low cost, all-weather operation, and global coverage.

[0003] The patent application document with publication number CN114861537A discloses a GNSS-R sea surface wind speed inversion method and system based on CNN multi-information fusion. The sea surface wind speed is inverted through a deep learning method of CNN multi-information fusion. However, since the deep learning inversion method has a large data demand, the scarcity of measured data samples in the scenario of studying specific sea areas or sea conditions leads to a problem of reduced sea surface wind speed inversion accuracy.

[0004] Patent application document with publication number CN117872424B discloses a GNSS-R sea surface DDM data generation method and device. The method uses a flow model to mine the conversion relationship between the distribution of measured DDM data and the normal distribution, and uses normal distribution random sampling to obtain variables to generate DDM data. However, since the generated DDM data lacks a physical basis, the generated DDM depends on the input measured data and lacks real sample details. Summary of the Invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a sea surface wind speed inversion method, system, equipment and medium that combines the advantages of simulation models and satellite data. The delay Doppler map is simulated by the position and speed of the satellite and the receiver, the wind speed, the GNSS signal frequency and the receiving antenna gain parameters. The data samples can be expanded when the measured data samples of specific sea areas or sea conditions are scarce. It has the advantages of wide applicability and can achieve relatively accurate sea surface wind speed inversion even when samples are scarce.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A sea surface wind speed inversion method combining the advantages of simulation models and satellite data includes the following steps:

[0008] Step 1: Construct a delay-Doppler map simulation model;

[0009] Step 2: Process the satellite data and buoy wind speed data and perform spatiotemporal matching to preliminarily construct a GNSS-R measured data set. Based on the delay-Doppler map simulation model constructed in Step 1, input the satellite and receiver positions and velocities, wind speed, GNSS signal frequency, and receiving antenna gain parameters for simulation to construct simulated GNSS-R data. The GNSS-R measured data set is then combined with the simulated GNSS-R data to construct a joint measured-simulated data set.

[0010] Step 3: Extracting eigenvalues ​​related to wind speed from the delay-Doppler map in the measured-simulated joint data set constructed in step 2, wherein the eigenvalues ​​related to wind speed include: a leading edge slope, a trailing edge slope, and a delay-Doppler mean; combining the leading edge slope and the trailing edge slope to calculate a leading-trailing edge composite eigenvalue, thereby obtaining a measured-simulated joint data set containing the eigenvalues;

[0011] Step 4. Based on the deep neural network, a deep neural network model is constructed with the delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions as joint inputs. The measured-simulation joint data set containing the eigenvalues ​​constructed in step 3 is trained to obtain a trained sea surface wind speed inversion model. The delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions are input to realize the sea surface wind speed inversion.

[0012] The step 1 specifically includes:

[0013] Step 1.1, calculate the coordinates of the mirror reflection point;

[0014] The positions and velocities of the satellite and receiver in the Earth-centered, Earth-fixed coordinate system are input. The coordinates of the mirror reflection point are iteratively calculated using the shortest path method, based on the characteristics that the mirror reflection point is located on the Earth's surface, the path from the satellite to the receiver through the mirror reflection point is the shortest, and the angle of incidence at the mirror reflection point is equal to the angle of reflection.

[0015] Step 1.2: Based on the coordinates of the mirror reflection point calculated in step 1.1, construct the local coordinate system of the satellite and receiver in the YOZ plane with the mirror reflection point as the center;

[0016] First, the Earth-centered Earth-fixed coordinate system is converted to the Northeast Sky coordinate system by multiplying the position in the Earth-centered Earth-fixed coordinate system by the transformation matrix;

[0017] Then, convert the Northeast Sky coordinate system to the local coordinate system centered on the mirror reflection point; first calculate the angle between the y-axis of the Northeast Sky coordinate system and the local coordinate system, and the projection of any point m(x,y,z) in the Northeast Sky coordinate system on the XOY plane The calculation formula for the coordinates of the point in the local coordinate system after conversion is as follows:

[0018] m srf =[|m xy |sin(Δθ),|m xy |cos(Δθ),z]

[0019] Δθ=θ1-θ0

[0020] Where θ1 is the angle between the point and the y-axis of the northeast celestial coordinate system, θ0 is the angle between the point and the y-axis of the local coordinate system, and Δθ is the difference between θ1 and θ0;

[0021] Step 1.3: Construct a geometric model close to the real sea surface in the local coordinate system constructed in step 1.2, and calculate the delay-Doppler value of each bin;

[0022] In the constructed sea surface model, any scattering surface element S x,y The time delay and Doppler are expressed as:

[0023]

[0024] Where T and R are the positions of the satellite and receiver, c is the speed of light, and V T and V R is the velocity vector of the satellite and receiver, e i is the incident signal vector, e s is the scattered signal vector;

[0025] Step 1.4: Calculate the scattering coefficient of each bin based on the input GNSS signal frequency and satellite and receiver position parameters;

[0026] Step 1.5: Combine the receiving antenna gain, the transmitted signal wavelength, and the coherent integration time parameters, calculate the simulated delay-Doppler map using the bistatic scattering signal correlation power model based on the delay-Doppler value obtained in step 1.3 and the scattering coefficient obtained in step 1.4. The bistatic scattering signal correlation power model is expressed as:

[0027]

[0028] Among them, G R is the receiving antenna gain, T i is the coherent integration time, P T is the transmitted signal power, λ is the transmitted signal wavelength, σ 0 is the scattering coefficient, GT is the transmit antenna gain, R t is the distance from the bin to the satellite, R r is the distance from the bin to the receiver, Λ 2 (τ-τ x,y ) is the PRN code coherence function, |S(f c -f x,y )| 2 is the Doppler spread function.

[0029] The step 2 specifically includes:

[0030] Step 2.1: Filter the satellite data and buoy wind speed data to remove outliers with missing or negative wind speeds in the buoy wind speed data. Filter the satellite data for mirror reflection points located in non-land and near-shore areas. Select satellite data with a peak signal-to-noise ratio greater than 3 dB, a receiving antenna gain greater than 3 dB, and an RCG value greater than 5.

[0031] Step 2.2, matching the satellite data obtained in step 2.1 with the buoy wind speed data to obtain satellite-buoy wind speed matching data;

[0032] During the matching process, the spatial distance between the mirror reflection point of the satellite data and the buoy site location of the buoy wind speed data is less than 7.5 kilometers, and the satellite data and buoy wind speed data with an observation time difference of less than 15 minutes are selected for matching;

[0033] Step 2.3: De-noise the satellite-buoy wind speed data obtained in step 2.2 using a 3D matched filter method to obtain satellite-buoy wind speed matched denoised data;

[0034] Step 2.4: Normalize the satellite-buoy wind speed matching denoised data obtained in step 2.3 to preliminarily construct a GNSS-R measured data set.

[0035] Step 2.5: Based on the delay-Doppler map simulation model constructed in step 1, input the position and velocity of the satellite and receiver, wind speed, GNSS signal frequency, and receiving antenna gain parameters for simulation to obtain simulated GNSS-R data. Combine the measured GNSS-R data set with the simulated GNSS-R data to construct a joint measured-simulated data set.

[0036] The specific method of step 3 includes:

[0037] Step 3.1: Extract the leading edge slope LES, trailing edge slope TES, and delay Doppler mean eigenvalue DDMA from the delay-Doppler map in the measured-simulation joint data set constructed in step 2. The calculation formula is as follows:

[0038]

[0039] in, and Represents the leading edge waveform and trailing edge waveform, using the delay Doppler diagram at the center delay waveform of the Doppler frequency of 0 Hz, taking the delay of -1.25 to 0 chips and 0 to 1.25 chips respectively; P DDM (i, j) represents the value of the delay Doppler map at the position (i, j); N represents the number of delay Doppler map values ​​taken, n represents the number of delay waveform values ​​taken, τ i is the delay value at point i;

[0040] Step 3.2: Combine the leading edge slope LES and trailing edge slope TES extracted in step 3.1 to calculate the leading-trailing edge composite eigenvalue Eigen L-T The calculation formula is as follows:

[0041]

[0042] Among them, LES represents the leading edge slope, and TES represents the trailing edge slope;

[0043] The leading edge slope LES, trailing edge slope TES, delay Doppler mean DDMA calculated in step 3.1 and the leading-trailing edge composite eigenvalue Eigen calculated in step 3.2 are used to calculate the leading edge slope LES, trailing edge slope TES, delay Doppler mean DDMA and trailing edge composite eigenvalue Eigen calculated in step 3.2. L-T Add it to the measured-simulated joint dataset constructed in step 2 to obtain the measured-simulated joint dataset containing the eigenvalues.

[0044] The specific method of step 4 includes:

[0045] Step 4.1. Based on the CNN and CNN_1D models, a deep neural network model is constructed with the delay-Doppler map, eigenvalues, and other parameters, namely, signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions as joint inputs. The delay-Doppler map is input into a neural network composed of three layers of two-dimensional convolution kernels for training. The convolution kernel size is 3×3, and the number of convolution kernels is 256, 128, and 64, respectively. The training results are then flattened and passed through two layers of FCN layers to obtain a 128×1 The first set of data; the eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incidence angle, longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions, are input into a neural network consisting of three layers of one-dimensional convolutional kernels for training. The convolution kernel size is 3×1, and the number of convolution kernels is 512, 256, and 128, respectively. The training results are then flattened and passed through two layers of FCN layers to obtain the second set of 128×1 data. The two sets of data are then fused based on the attention mechanism and finally passed through two layers of FCN layers to obtain the sea surface wind speed inversion results.

[0046] Step 4.2: Input the measured-simulated joint data set containing the eigenvalues ​​constructed in step 3 into the deep neural network model constructed in step 4.1 for training to obtain the sea surface wind speed inversion model. Input the delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions to realize sea surface wind speed inversion. Finally, evaluate the sea surface wind speed inversion accuracy of the trained sea surface wind speed inversion model.

[0047] The present invention also provides a sea surface wind speed inversion system that combines the advantages of simulation models and satellite data, including:

[0048] A model building module is used to build a delay-Doppler map simulation model;

[0049] The joint measurement-simulation dataset construction module is used to process and spatially match satellite data and buoy wind speed data, initially construct a GNSS-R measurement dataset, and simulate the position and velocity of the satellite and receiver, wind speed, GNSS signal frequency, and receiving antenna gain parameters based on the delay-Doppler map simulation model to construct simulated GNSS-R data. The measured GNSS-R dataset is then combined with the simulated GNSS-R data to construct a joint measurement-simulation dataset.

[0050] A module for constructing a joint measured-simulated dataset containing eigenvalues ​​is used to extract eigenvalues ​​related to wind speed from the delay-Doppler map in the joint measured-simulated dataset. The eigenvalues ​​related to wind speed include: leading edge slope, trailing edge slope, and delay-Doppler mean. The module combines the leading edge slope and trailing edge slope to calculate the leading-trailing edge composite eigenvalue to obtain the joint measured-simulated dataset containing the eigenvalues.

[0051] The sea surface wind speed inversion module is used to implement a deep neural network-based model that constructs a delay-Doppler map, eigenvalues ​​and other parameters, namely, signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions as joint inputs. It trains the measured-simulation joint data set containing the eigenvalues ​​to obtain a trained sea surface wind speed inversion model, and inputs the delay-Doppler map, eigenvalues ​​and other parameters, namely, signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions to achieve sea surface wind speed inversion.

[0052] The present invention also provides a sea surface wind speed inversion device that combines the advantages of simulation models and satellite data, including:

[0053] Memory: a computer-readable device storing a computer program for the sea surface wind speed inversion method combining the advantages of simulation models and satellite data;

[0054] Processor: used to implement the sea surface wind speed inversion method combining the advantages of simulation model and satellite data when executing the computer program.

[0055] The present invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement the sea surface wind speed inversion method that combines the advantages of simulation models and satellite data.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. The present invention simulates the delay Doppler diagram through physical parameters such as the satellite and receiver position and speed, receiving antenna gain, transmission signal wavelength and coherent integration time, and combines theoretical models to simulate GNSS-R echo signals under different wind speeds and observation conditions. This method can expand the data samples to address the problem of scarcity of measured data samples when studying sea surface wind speed inversion based on GNSS-R in specific sea areas or sea conditions. Compared with traditional methods that rely on a large number of measured samples, the present invention has good versatility and adaptability. It can still invert sea surface wind speed more accurately under conditions of limited measured samples, thereby improving the reliability and application scope of the model.

[0058] 2. Based on the study of time-delay waveforms, the present invention extracts the leading slope, trailing slope, and delay Doppler mean eigenvalues. The combined effect of wind speed on the leading and trailing edge shapes of the time-delay waveform is used as a key inversion feature, introducing a leading-trailing composite eigenvalue parameter. This composite eigenvalue not only enhances the ability of wind speed changes to express the overall shape of the waveform, but also effectively reduces the error caused by fluctuations in a single feature. By constructing a composite eigenvalue, the modulation effect of wind speed changes on the scattering characteristics can be more accurately characterized, thereby improving the sensitivity and stability of sea surface wind speed inversion.

[0059] In summary, the present invention simulates the delay Doppler map through physical parameters such as the position and speed of the satellite and receiver, and constructs a characteristic value combination in combination with the leading slope and the trailing slope. It can expand the data samples when the measured data samples are scarce in specific sea areas or sea conditions, and better reflect the influence of wind speed. It has the advantages of wide applicability and can more accurately realize the inversion of sea surface wind speed even when samples are scarce. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 4 is a flow chart of a method for inverting sea surface wind speed according to an embodiment of the present invention.

[0061] Figure 2 This is a flow chart of a simulated delay-Doppler diagram according to an embodiment of the present invention.

[0062] Figure 3is a diagram showing the DDM calculation results for different wind speeds according to an embodiment of the present invention, wherein: Figure 3 (a) is the DDM diagram for a wind speed of 5 m / s. Figure 3 (b) is the DDM with a wind speed of 20 m / s, Figure 3 (c) is the delay waveform of two DDMs. Figure 3 (d) is the normalized delay waveform.

[0063] Figure 4 This is a flow chart of measured data matching according to an embodiment of the present invention.

[0064] Figure 5 is a comparison chart of measured and simulated DDM of different parameters of an embodiment of the present invention, wherein: Figure 5 (a) is the measured image of sample DDM1, Figure 5 (b) is the simulation diagram of sample DDM1, Figure 5 (c) is the measured image of sample DDM2. Figure 5 (d) is a sample DDM2 simulation diagram.

[0065] Figure 6 : is a comparison diagram of the measured and simulated normalized integrated delay waveforms of an embodiment of the present invention, wherein: Figure 6 (a) is the normalized integrated delay waveform of the sample DDM1 measured and simulated. Figure 6 (b) is the normalized integrated delay waveform of the measured and simulated sample DDM2.

[0066] Figure 7 is a comparison chart of the simulated and measured characteristic values ​​of an embodiment of the present invention, wherein: Figure 7 (a) is the distribution diagram of the composite characteristic value of the leading-trailing edge of the measured data at different wind speeds. Figure 7 (b) is the distribution diagram of the composite characteristic value of the front-back edge of the simulation data at different wind speeds. Figure 7 (c) is a comparison chart of the leading-trailing composite eigenvalues ​​of the measured data and the simulated data.

[0067] Figure 8 : is a comparison diagram of the inverted wind speed and the buoy measured wind speed of different data sets in the embodiment of the present invention, wherein: Figure 8 (a) is a comparison chart of the model inversion wind speed trained using measured data and the buoy wind speed. Figure 8 (b) is a comparison chart of the model inverted wind speed and the buoy wind speed trained using measured and simulated data.

[0068] Figure 9 : is a diagram of the root mean square error and mean absolute error of different wind speed ranges of the embodiment of the present invention, wherein, Figure 9 (a) is the root mean square error diagram for different wind speed ranges, Figure 9 (b) is the mean absolute error diagram for different wind speed ranges. DETAILED DESCRIPTION

[0069] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0070] The embodiments of the present invention mainly study the sea surface wind speed inversion method based on the measured data collected by the Fengyun-3E satellite (FY-3E) and the simulated GNSS-R data, providing theoretical and methodological support for applications such as marine environment monitoring and maritime target detection, and have important applied research significance.

[0071] The present invention starts from the geometric characteristics of the sea surface and the theory of sea surface electromagnetic scattering, realizes GNSS-R simulation modeling that takes into account the position and speed of the satellite and receiver, wind speed, GNSS signal frequency and receiving antenna gain parameters, and obtains a delay-Doppler map calculation result that is more in line with the real scene. The FY-3E satellite data and buoy wind speed data are processed, and then strict time-space matching is performed to preliminarily construct a GNSS-R measured data set. In response to the practical problem of scarcity of measured data samples in specific sea areas or sea conditions, a measured-simulation joint data set is constructed by combining measured data and simulation data. Afterwards, the eigenvalues ​​related to wind speed (leading slope, trailing slope and delay-Doppler mean) are extracted from the delay-Doppler map, and the leading-trailing composite eigenvalue is calculated by combining the leading slope and trailing slope. Finally, the measured-simulation joint data set is trained based on a deep neural network to obtain a sea surface wind speed inversion model, achieving efficient and widely applicable sea surface wind speed inversion.

[0072] See also Figure 1 , a sea surface wind speed inversion method combining the advantages of simulation model and satellite data, including the following steps:

[0073] Step 1: See Figure 2 , build a delay-Doppler map simulation model

[0074] The step 1 specifically includes:

[0075] Step 1.1, calculate the coordinates of the mirror reflection point;

[0076] Input the positions and velocities of the satellite and receiver in the Earth-centered, Earth-fixed coordinate system. Based on the characteristics that the mirror reflection point is located on the Earth's surface, the path from the satellite to the receiver through the mirror reflection point is the shortest, and the angle of incidence at the mirror reflection point is equal to the reflection angle, the coordinates of the mirror reflection point are iteratively calculated using the shortest path method. The formula for calculating the path length from the satellite to the receiver through the reflection point is as follows:

[0077] f(S)=|T ecef -S|+|R ecef -S|

[0078]

[0079] Among them, T ecef 、R ecef and S are the positions of the satellite, receiver and reflection point in the Earth-centered Earth-fixed coordinate system, respectively. The shortest path f(S sp )=|T ecef -S sp |+|R ecef -S sp |, the coordinates of the reflection point corresponding to the shortest path S sp It is the coordinate of the mirror reflection point;

[0080] Step 1.2: Based on the coordinates of the mirror reflection point calculated in step 1.1, construct the local coordinate system of the satellite and receiver in the YOZ plane with the mirror reflection point as the center;

[0081] First, the Earth-centered Earth-fixed coordinate system is converted to the Northeast Sky coordinate system by multiplying the position in the Earth-centered Earth-fixed coordinate system by the transformation matrix. The conversion process is as follows;

[0082]

[0083] Among them, λ s is the longitude of the specular reflection point, is the latitude of the specular reflection point;

[0084] Then, the Northeast Sky coordinate system is converted to the local coordinate system centered on the mirror reflection point. The difference between the Northeast Sky coordinate system and the local coordinate system is that the directions of the x and y axes are different. The y axis direction of the local coordinate system is the same as the direction of the projection of the vector from the receiver to the satellite on the XOY plane. The conversion process requires first calculating the angle between the y axis of the Northeast Sky coordinate system and the local coordinate system. The projection of any point m(x,y,z) in the Northeast Sky coordinate system on the XOY plane is The calculation formula for the coordinates of the point in the local coordinate system after conversion is as follows:

[0085] m srf =[|m xy |sin(Δθ),|m xy |cos(Δθ),z]

[0086] Δθ=θ1-θ0

[0087] Where θ1 is the angle between the point and the y-axis of the northeast celestial coordinate system, θ0 is the angle between the point and the y-axis of the local coordinate system, and Δθ is the difference between θ1 and θ0;

[0088] Step 1.3: Construct a geometric model close to the real sea surface in the local coordinate system constructed in step 1.2, and calculate the delay-Doppler value of each bin;

[0089] The sea surface geometric model is constructed using the linear filtering method. The formula of the linear filtering method is as follows;

[0090]

[0091] Among them, S(k) is the sea spectrum function, which is used to construct the sea surface model of different sea conditions. is the direction function, k is the wave number, δk x and δk y is the discrete interval of wave number in the x-direction and y-direction, ω is the angular frequency of the incident electromagnetic wave, and t is time;

[0092] In the constructed sea surface model, any scattering surface element S x,y The time delay and Doppler are expressed as:

[0093]

[0094] Where T and R are the positions of the satellite and receiver, c is the speed of light, and V T and V R is the velocity vector of the satellite and receiver, e i is the incident signal vector, e s is the scattered signal vector; step

[0095] 1.4. Calculate the scattering coefficient of each bin based on the input GNSS signal frequency and satellite and receiver position parameters. The scattering coefficient calculation formula is as follows:

[0096]

[0097] Where q is the scattering vector, q z is the z-component of the scattering vector, q ⊥ is the vertical component of the scattering vector, π is the circumference of the circle, is the Fresnel scattering coefficient, is the probability density function, which is calculated using the sea surface geometry model constructed in step 1.3. The calculation formula is as follows:

[0098]

[0099]

[0100] c 21 =0.01-0.0086·u 10 ,c 03 =0.04-0.0330·u 10

[0101] c 40 =0.40,c 22 =0.12,c04 =0.23

[0102] Among them, σ u and σ c are the mean square slope of the sea surface under headwind and crosswind conditions, u 10 is the wind speed at 10 meters above sea level, q x ,q y and q z are the components of the scattering vector in the x, y, and z directions;

[0103] Step 1.5: Combine the receiving antenna gain, the transmitted signal wavelength, and the coherent integration time parameters, calculate the simulated delay-Doppler map using the bistatic scattering signal correlation power model based on the delay-Doppler value obtained in step 1.3 and the scattering coefficient obtained in step 1.4. The bistatic scattering signal correlation power model is expressed as:

[0104]

[0105] Among them, G R is the receiving antenna gain, T i is the coherent integration time, P T is the transmitted signal power, λ is the transmitted signal wavelength, σ 0 is the scattering coefficient, G T is the transmit antenna gain, R t is the distance from the bin to the satellite, R r is the distance from the bin to the receiver, Λ 2 (τ-τ x,y ) is the PRN code coherence function, |S(f c -f x,y )| 2 is the Doppler spread function.

[0106] Step 2, see Figure 4 , perform data processing and spatiotemporal matching on the FY-3E satellite data and buoy wind speed data to preliminarily construct a GNSS-R measured dataset. Based on the delay-Doppler map simulation model constructed in step 1, input the satellite and receiver position and velocity, wind speed, GNSS signal frequency, and receiving antenna gain parameters for simulation to construct simulated GNSS-R data. Combine the GNSS-R measured dataset with the simulated GNSS-R data to construct a joint measured-simulated dataset.

[0107] The step 2 specifically includes:

[0108] Step 2.1: Filter the FY-3E satellite data and buoy wind speed data to remove outliers with missing or negative wind speeds in the buoy wind speed data. Filter the FY-3E satellite data with specular reflection points located in non-land and near-shore areas. Select FY-3E satellite data with a peak signal-to-noise ratio greater than 3 dB, a receiving antenna gain greater than 3 dB, and an RCG value greater than 5.

[0109] Step 2.2, matching the FY-3E satellite data obtained in step 2.1 with the buoy wind speed data to obtain satellite-buoy wind speed matching data;

[0110] During the matching process, it is necessary to ensure that the spatial distance between the mirror reflection point of the FY-3E satellite data and the buoy site location of the buoy wind speed data is less than 7.5 kilometers. Then, the FY-3E satellite data with an observation time difference of less than 15 minutes are selected for matching with the buoy wind speed data. The spatial distance and observation time difference can be determined based on the data characteristics (such as satellite and receiver speed, coherent integration time) and subsequent application requirements. Usually, the spatial distance is within 7 to 10 kilometers, and the time difference is within 10 to 60 minutes. The specific thresholds can be adjusted according to actual conditions or determined through experiments.

[0111] The delay-Doppler maps in the satellite-buoy wind speed matching data obtained in step 2.3 and step 2.2 contain noise caused by equipment and the environment. The satellite-buoy wind speed data obtained in step 2.2 are denoised using the 3D matched filter method to obtain satellite-buoy wind speed matching denoised data.

[0112] Step 2.4: To prepare for subsequent deep learning, normalize the denoised satellite-buoy wind speed matching data obtained in step 2.3 to preliminarily construct a GNSS-R measured data set.

[0113] Step 2.5: Based on the delay-Doppler map simulation model constructed in step 1, input the position and velocity of the satellite and receiver, wind speed, GNSS signal frequency, and receiving antenna gain parameters for simulation to obtain simulated GNSS-R data. Combine the measured GNSS-R data set with the simulated GNSS-R data to construct a joint measured-simulated data set.

[0114] Flowchart as Figure 4 As shown:

[0115] Step 3: Extracting eigenvalues ​​related to wind speed from the delay-Doppler map in the measured-simulated joint data set constructed in Step 2. The eigenvalues ​​related to wind speed include: leading edge slope, trailing edge slope, and delay-Doppler mean; combining the leading edge slope and trailing edge slope to calculate the leading-trailing edge composite eigenvalue, thereby obtaining the measured-simulated joint data set containing the eigenvalues;

[0116] The specific method of step 3 includes:

[0117] Step 3.1: Extract the leading edge slope LES, trailing edge slope TES, and delay Doppler mean eigenvalue DDMA from the delay-Doppler map in the measured-simulation joint data set constructed in step 2. The calculation formula is as follows:

[0118]

[0119]

[0120] in, and Represents the leading edge waveform and trailing edge waveform, using the delay Doppler diagram at the center delay waveform of the Doppler frequency of 0 Hz, taking the delay of -1.25 to 0 chips and 0 to 1.25 chips respectively; P DDM (i, j) represents the value of the delay Doppler map at the position (i, j); N represents the number of delay Doppler map values ​​taken, n represents the number of delay waveform values ​​taken, τ i is the delay value at point i;

[0121] Step 3.2: Combine the leading edge slope LES and trailing edge slope TES extracted in step 3.1 to calculate the leading-trailing edge composite eigenvalue; the calculation formula is as follows:

[0122]

[0123] Among them, LES represents the leading edge slope, and TES represents the trailing edge slope;

[0124] The leading edge slope LES, trailing edge slope TES, delay Doppler mean DDMA calculated in step 3.1 and the leading-trailing edge composite eigenvalue calculated in step 3.2 are added to the measured-simulated joint data set constructed in step 2 to obtain a measured-simulated joint data set containing eigenvalues.

[0125] Step 4. Based on the deep neural network, a deep neural network model is constructed with the delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions as joint inputs. The measured-simulation joint data set containing the eigenvalues ​​constructed in step 3 is trained to obtain a trained sea surface wind speed inversion model. The delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions are input to realize the sea surface wind speed inversion.

[0126] The specific method of step 4 includes:

[0127] Step 4.1. Based on the CNN and CNN_1D models, a deep neural network model is constructed with the delay-Doppler map, eigenvalues, and other parameters, namely, signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions as joint inputs. The delay-Doppler map is input into a neural network composed of three layers of two-dimensional convolution kernels for training. The convolution kernel size is 3×3, and the number of convolution kernels is 256, 128, and 64, respectively. The training results are then flattened and passed through two layers of FCN layers to obtain a 128×1 The first set of data; the eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incidence angle, longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions, are input into a neural network consisting of three layers of one-dimensional convolutional kernels for training. The convolution kernel size is 3×1, and the number of convolution kernels is 512, 256, and 128, respectively. The training results are then flattened and passed through two layers of FCN layers to obtain the second set of 128×1 data. The two sets of data are then fused based on the attention mechanism and finally passed through two layers of FCN layers to obtain the sea surface wind speed inversion results.

[0128] During the training process, the initial learning rate is set to 0.001, and the learning rate is dynamically adjusted as the training progresses. The random dropout is set to 0.1 to improve the generalization ability of the model.

[0129] Step 4.2: Input the measured-simulated joint data set containing the eigenvalues ​​constructed in step 3 into the deep neural network model constructed in step 4.1 for training to obtain the sea surface wind speed inversion model. Input the delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions to realize sea surface wind speed inversion. Finally, evaluate the sea surface wind speed inversion accuracy of the trained sea surface wind speed inversion model.

[0130] Example

[0131] A sea surface wind speed inversion method combining the advantages of simulation models and satellite data includes the following steps:

[0132] Step 1: Build a delay-Doppler map simulation model

[0133] Delay-Doppler maps are key to GNSS-R research. They contain a wealth of surface remote sensing information and can be used for sea surface wind speed inversion, sea ice detection, soil moisture detection, forest fire warning, and other applications. Therefore, the key to GNSS-R research lies in simulating Delay-Doppler maps under different parameter conditions. This paper, based on the geometric characteristics of the sea surface and the theory of sea surface electromagnetic scattering, implements Delay-Doppler map simulation modeling that takes into account parameters such as satellite and receiver position, velocity, and antenna gain.

[0134] Real-world ocean waves are complex and unpredictable. This paper uses ocean spectrum functions combined with sea surface parameters to model ocean surface geometry. Based on the frequency domain distribution characteristics of the waves, a Fourier transform is performed to produce an ocean surface model that more closely resembles real-world scenarios. When studying delay-Doppler maps for special sea conditions, such as mixed wind and swell waves, modifications to the sea surface simulation can be made to target specific sea conditions. This effectively overcomes the limitations of obtaining satellite-measured data for specific sea conditions.

[0135] Commonly used theoretical models for forward scattering from the ocean are mainly analytical approximation methods and numerical methods. The numerical method has a relatively high calculation accuracy, but the calculation is complex and the simulation efficiency is low. When faced with the problem of calculating large-scale ocean surface electromagnetic scattering, the consumption of computing resources is huge. The analytical approximation method uses a physical approximation method to reduce the consumption of computing resources. In the approximation process, the main physical mechanisms are considered according to the specific situation, so that the modeling also has good accuracy. The present invention adopts the Kirchhoff approximation-geometrical optics (KA-GO) method in the analytical approximation method to model ocean surface electromagnetic scattering.

[0136] In the satellite-measured GNSS-R data, the positions of the satellite and receiver are usually given in the Earth-centered Earth-fixed coordinate system (ECEF). Therefore, to compare the simulated delay Doppler map results with the measured data, it is necessary to start from the satellite and receiver positions in the ECEF coordinate system, calculate the coordinates of the mirror reflection point, and construct a local coordinate system centered on the mirror reflection point. In the local coordinate system, a geometric model that is closer to the real sea surface is constructed, the delay Doppler value of each surface element is calculated, and then the scattering coefficient of each surface element is calculated. Combined with the receiving antenna gain, the wavelength of the transmitted signal and the coherent integration time parameters, the dual-base scattered signal correlation power model is used to calculate the simulated delay Doppler map. The process of simulating the delay Doppler map is as follows: Figure 2 shown.

[0137] The following simulation uses the GNSS-R parameter settings for an incident angle of 20° and a receiver altitude of 500 km as an example to simulate delay-Doppler patterns at different wind speeds. The simulated wind speeds at 10 meters above sea level are 5 m / s and 20 m / s, and the GNSS signal frequency is 1.57 GHz. Figure 3 The calculation results of delay Doppler diagram and delay waveform under different wind speeds are given. Figure 3 (a) and (b) are two-dimensional top views of the delay Doppler map at wind speeds of 5m / s and 20m / s respectively. Figure 3 As can be seen in (a) and (b), the higher the wind speed, the lower the overall power of the delay-Doppler pattern, and the smaller the proportion of the power near the peak to the total power. This is because as wind speed increases, the roughness of the sea surface increases, and diffuse reflection strengthens, which reduces the intensity of the reflected signal near the specular reflection point. Figure 3(c) and (d) are the delay waveforms of the delay Doppler diagram at the Doppler frequency of 0 Hz. Figure 3 (d) shows the normalized delay waveform. As can be seen, the slower the wind speed, the faster the normalized delay waveform decreases. If the wind speed drops to 0 m / s, the sea surface becomes like a mirror, with only the reflection points reflecting the signal.

[0138] Step 2: Process the FY-3E satellite data and buoy wind speed data and perform spatiotemporal matching to preliminarily construct a GNSS-R measured data set. Based on the delay-Doppler map simulation model constructed in Step 1, input the satellite and receiver positions and velocities, wind speed, GNSS signal frequency, and receiving antenna gain parameters for simulation to construct simulated GNSS-R data. The measured GNSS-R data set is then combined with the simulated GNSS-R data to construct a joint measured-simulated data set.

[0139] The steps for constructing the GNSS-R measured data set include: (1) data filtering. Eliminate outliers and invalid data points and perform quality control; (2) time-space matching. Match the FY-3E satellite observation data with the buoy wind speed data to ensure that the time difference and spatial error between the two are within a reasonable range; (3) data denoising; (4) data normalization. The flowchart is as follows Figure 4 shown.

[0140] To address the scarcity of measured data samples in some research situations (such as specific regions, times, and sea conditions), simulations were considered to expand the dataset. Therefore, further verification of the accuracy of the simulated data was necessary. Because the Fengyun-3E satellite data is non-uniformly sampled in the time delay direction, the delay resolution near the specular reflection point is 0.125 chips, while the delay resolution in areas far from the specular reflection point is 0.25 chips. Specifically, the delay interval is 0.125 chips in the range -2.875 to 2.875 chips around the specular reflection point, and 0.25 chips elsewhere. Therefore, to compare the simulated data with the measured data, the simulated data must first be resampled to the delay interval of the Fengyun satellite data. Two DDMs were selected for comparison. The measured wind speeds for these two DDMs were 6.60 and 17.15 m / s, respectively, covering a wide and representative wind speed range. The simulation parameters are derived from the measured data. Table 1 gives the specific parameter settings, such as data detection time, the position and speed of the transmitting satellite and receiver, and the wind speed in the observation area.

[0141] Table 1 Simulation calculation parameter settings

[0142]

[0143] In order to facilitate the evaluation of the accuracy of the simulation results, the root mean square error (RMSE) and structural similarity (SSIM) of the simulated and measured data near the mirror reflection point of the delay-Doppler map are used to evaluate the accuracy of the simulation results. The calculation formula is as follows:

[0144]

[0145] If the root mean square error is smaller, the structural similarity value is closer to 100%, which means that the real delay Doppler image and the simulated delay Doppler image are consistent.

[0146] Table 2 shows the results of a comparison between simulated and measured delay-Doppler maps. The simulated delay-Doppler map is compared with the denoised measured satellite data. The RMSE values ​​between the simulated and measured data are both less than 0.06, and the SSIM values ​​are both greater than 80%, indicating a small difference. A comparison of the leading-edge slopes of the two reveals that the LES values ​​are essentially consistent, with little difference. Figure 5 This is a comparison chart of measured and simulated DDM with different parameters. Figure 5 (a) and (b) are the measured and simulated images of sample DDM1. Figure 5 (c) and (d) are the measured and simulated figures of sample DDM2. Figure 6 The figure below is a comparison of the measured and simulated normalized integrated delay waveforms of two DDM samples. Figure 6 (a) is the normalized integrated delay waveform of the sample DDM1 measured and simulated. Figure 6 (b) is the normalized integrated delay waveform of the DDM2 sample measured and simulated. Figure 5 (a) Figure 5 (b) Figure 5 (c) and Figure 5 (d) It can be seen that Figure 5 (a) and Figure 5 (c) The DDM of the measured data has a lot of noise interference, but the overall trend including the peak point, Doppler broadening and power value decrease trend are consistent with Figure 5 (b) and Figure 5 The simulation results are consistent with the measured data. The simulation results can be used to calculate the delay-Doppler map and construct a joint measurement-simulation data set.

[0147] Table 2 Comparison of simulation results and measured data

[0148]

[0149] Step 3: Extract eigenvalues ​​related to wind speed from the delay-Doppler map in the measured-simulated joint data set constructed in step 2. The eigenvalues ​​related to wind speed include: leading edge slope, trailing edge slope, and delay-Doppler mean; combine the leading edge slope and trailing edge slope to calculate the leading-trailing edge composite eigenvalue to obtain the measured-simulated joint data set containing the eigenvalues.

[0150] By further calculating the delay-Doppler map, the leading edge slope, trailing edge slope, and delay-Doppler mean are extracted, and then the leading edge slope and trailing edge slope are combined to calculate the leading-trailing edge composite eigenvalue. The calculation formula is as follows:

[0151]

[0152]

[0153] Figure 7 A comparison of the leading-trailing edge composite eigenvalues ​​calculated by simulation and actual measurement is given. Figure 7 (a) is the distribution diagram of the composite characteristic value of the leading-trailing edge of the measured data at different wind speeds. Figure 7 (b) is the distribution diagram of the composite characteristic value of the front-back edge of the simulation data at different wind speeds. Figure 7 (a) and Figure 7 (b) It can be seen that the front slope distribution trends of the simulated and measured data are the same, both decreasing with increasing wind speed. The simulated data are more concentrated, while the measured data are more discrete due to noise and other interference factors. However, they are generally distributed in the same area. Figure 7 (c) is a comparison chart of the leading-backward composite eigenvalues ​​of the simulated and measured data. It can be seen that the data points are basically distributed near the isovalue lines, which further illustrates that the simulated data and the measured data are consistent. When the measured data is insufficient, the simulated data can be used to expand the sample.

[0154] The calculated leading edge slope, trailing edge slope, delay Doppler mean, and leading-trailing edge composite eigenvalue are added to the measured-simulated joint data set constructed in step 2 to obtain a measured-simulated joint data set containing eigenvalues.

[0155] Step 4. Based on the deep neural network, a deep neural network model is constructed with the delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions as joint inputs. The measured-simulation joint data set containing the eigenvalues ​​constructed in step 3 is trained to obtain a trained sea surface wind speed inversion model. The delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions are input to realize the sea surface wind speed inversion.

[0156] GNNS-R, due to its low cost, all-weather capabilities, and global coverage, is suitable for global wind speed monitoring. Its use in both routine and military observations can effectively save manpower and material resources. Therefore, retrieving actual sea surface wind speed from time-delay Doppler images is of great research significance and value. Traditional wind speed inversion methods typically analyze wind speed using satellite-measured data, which requires large amounts of data and is difficult to obtain in some cases. Based on the combined measured and simulated dataset containing eigenvalues ​​constructed in Step 3, a deep learning neural network model is introduced to propose a sea surface wind speed inversion method with fewer restrictions and applicable to situations with fewer observation samples.

[0157] Based on the CNN and CNN_1D models, the present invention constructs a Ddm-Eigen model that takes as input the delay-Doppler map, eigenvalues, and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions. The Ddm-Eigen model uses a two-dimensional convolution kernel to extract the spatial features of the DDM power map and a one-dimensional convolution kernel to process the eigenvalues ​​and other parameters. Both are trained separately, and the training results are then flattened. After passing through two layers of FCN layers, two sets of data of the same dimensionality are obtained. The two sets of data are then feature-fused based on the attention mechanism, and finally, the sea surface wind speed inversion result is calculated through two layers of FCN layers.

[0158] First, we used a small sample of measured data for training to investigate the model's effectiveness in retrieving sea surface wind speeds when measured data is limited. We then added simulated data to the training set, leaving the test and validation sets unchanged, and used the model for further training. Table 3 provides a statistical comparison of the wind speed retrieval accuracy of the model using only measured data and using a simulated augmented dataset. Figure 8 This is a comparison chart of the model inversion wind speed and the buoy measured wind speed. Figure 8 (a) is a comparison chart of the model inversion wind speed trained using measured data and the buoy wind speed. Figure 8 (b) is a comparison of the wind speed inverted by the model trained using measured and simulated data and the buoy wind speed. Combined with Table 3, it can be seen that after the model is expanded with simulated data, the accuracy of the inverted sea surface wind speed is improved, the RMSE and MAE are reduced, and the proportion of samples with an absolute error of less than 1m / s and 2m / s between the inverted wind speed and the true wind speed is increased. Figure 8 (a) and (b) show that after adding the simulation data, the model-inverted wind speed is more distributed near the y=x reference line, and the accuracy of the inverted wind speed is improved.

[0159] Table 3 Accuracy of wind speed inversion from different datasets

[0160]

[0161] In order to compare the inversion accuracy of the model in different wind speed ranges, Figure 9 The root mean square error (RMSE) and mean absolute error (MAE) for different wind speed ranges are given. Figure 9 (a) is the root mean square error diagram for different wind speed ranges, Figure 9 (b) is the mean absolute error diagram for different wind speed ranges. Figure 9 As can be seen from (a) and (b), after the model uses the simulation to expand the data set, the root mean square error and mean absolute error in each wind speed range are reduced, which indicates that the inversion effect is improved, especially the inversion accuracy in the high wind speed range (15-20m / s) is improved most significantly.

[0162] The present invention proposes a sea surface wind speed inversion method that combines the advantages of simulation models and satellite data. The simulated time-delay Doppler map is introduced to expand the measured samples, which improves the limitations of traditional wind speed inversion methods based on measured GNSS-R data caused by the scarcity of measured data samples in specific sea areas or sea conditions.

[0163] The eigenvalues ​​used in the sea surface wind speed inversion method implemented in the present invention are different from the leading edge slope eigenvalues ​​used in traditional methods. They take into account the combined influence of wind speed on the leading and trailing edges of the time-delay waveform, and construct an eigenvalue combination by combining the leading edge slope and the trailing edge slope, which can better reflect the influence of wind speed.

[0164] The present invention also provides a sea surface wind speed inversion system that combines the advantages of simulation models and satellite data, including:

[0165] A model building module, used to implement the construction of the delay-Doppler map simulation model in step 1;

[0166] The module for constructing a joint measurement-simulation dataset is used to implement the data processing and spatiotemporal matching of the FY-3E satellite data and the buoy wind speed data in step 2, preliminarily construct a GNSS-R measurement dataset, and simulate the delay-Doppler map simulation model constructed in step 1 by inputting the position and velocity of the satellite and receiver, wind speed, GNSS signal frequency, and receiving antenna gain parameters to construct simulated GNSS-R data. The module then combines the measured GNSS-R dataset with the simulated GNSS-R data to construct a joint measurement-simulation dataset.

[0167] A module for constructing a joint measured-simulated data set including eigenvalues ​​is used to extract eigenvalues ​​related to wind speed from the delay-Doppler map in the joint measured-simulated data set constructed in step 2 in step 3, wherein the eigenvalues ​​related to wind speed include: a leading edge slope, a trailing edge slope, and a delay-Doppler mean; and to calculate a leading-trailing edge composite eigenvalue by combining the leading edge slope and the trailing edge slope to obtain a joint measured-simulated data set including eigenvalues.

[0168] The sea surface wind speed inversion module is used to implement the deep neural network in step 4, construct a deep neural network model with the delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions as joint inputs, train the measured-simulation joint data set containing the eigenvalues ​​constructed in step 3, obtain a trained sea surface wind speed inversion model, input the delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions, to realize sea surface wind speed inversion.

[0169] The present invention also provides a sea surface wind speed inversion device that combines the advantages of simulation models and satellite data, including:

[0170] Memory: a computer-readable device storing a computer program for the sea surface wind speed inversion method combining the advantages of simulation models and satellite data;

[0171] Processor: used to implement the sea surface wind speed inversion method combining the advantages of simulation model and satellite data when executing the computer program.

[0172] The present invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement the sea surface wind speed inversion method that combines the advantages of simulation models and satellite data.

Claims

1. A sea surface wind speed inversion method combining the advantages of simulation models and satellite data, characterized in that: The following steps are involved: Step 1: Construct a delay-Doppler map simulation model; Step 2: Process the satellite data and buoy wind speed data and perform spatiotemporal matching to preliminarily construct a GNSS-R measured data set. Based on the delay-Doppler map simulation model constructed in Step 1, input the satellite and receiver positions and velocities, wind speed, GNSS signal frequency, and receiving antenna gain parameters for simulation to construct simulated GNSS-R data. The GNSS-R measured data set is then combined with the simulated GNSS-R data to construct a joint measured-simulated data set. Step 3: extracting characteristic values ​​related to wind speed from the delay-Doppler map in the measured-simulation joint data set constructed in step 2, wherein the characteristic values ​​related to wind speed include: leading edge slope, trailing edge slope, and delay-Doppler mean; Combining the leading edge slope and the trailing edge slope, the leading edge-trailing edge composite eigenvalue is calculated to obtain a measured-simulated joint data set containing the eigenvalue; Step 4. Based on the deep neural network, a deep neural network model is constructed with the delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions as joint inputs. The measured-simulation joint data set containing the eigenvalues ​​constructed in step 3 is trained to obtain a trained sea surface wind speed inversion model. The delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, the longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions are input to realize the sea surface wind speed inversion.

2. The sea surface wind speed inversion method combining the advantages of simulation model and satellite data according to claim 1 is characterized in that: The step 1 specifically includes: Step 1.1, calculate the coordinates of the mirror reflection point; The positions and velocities of the satellite and receiver in the Earth-centered, Earth-fixed coordinate system are input. The coordinates of the mirror reflection point are iteratively calculated using the shortest path method, based on the characteristics that the mirror reflection point is located on the Earth's surface, the path from the satellite to the receiver through the mirror reflection point is the shortest, and the angle of incidence at the mirror reflection point is equal to the angle of reflection. Step 1.2: Based on the coordinates of the mirror reflection point calculated in step 1.1, construct the local coordinate system of the satellite and receiver in the YOZ plane with the mirror reflection point as the center; First, the Earth-centered Earth-fixed coordinate system is converted to the Northeast Sky coordinate system by multiplying the position in the Earth-centered Earth-fixed coordinate system by the transformation matrix; Then, convert the Northeast Sky coordinate system to the local coordinate system centered on the mirror reflection point; first calculate the angle between the y-axis of the Northeast Sky coordinate system and the local coordinate system, and the projection of any point m(x,y,z) in the Northeast Sky coordinate system on the XOY plane The calculation formula for the coordinates of the point in the local coordinate system after conversion is as follows: m srf =[|m xy |sin(Δθ),|m xy |cos(Δθ),z] Δθ=θ1-θ0 Where θ1 is the angle between the point and the y-axis of the northeast celestial coordinate system, θ0 is the angle between the point and the y-axis of the local coordinate system, and Δθ is the difference between θ1 and θ0; Step 1.3: Construct a geometric model close to the real sea surface in the local coordinate system constructed in step 1.2, and calculate the delay-Doppler value of each bin; In the constructed sea surface model, any scattering surface element S x,y The time delay and Doppler are expressed as: Where T and R are the positions of the satellite and receiver, c is the speed of light, and V T and V R is the velocity vector of the satellite and receiver, e i is the incident signal vector, e s is the scattered signal vector; Step 1.4: Calculate the scattering coefficient of each bin based on the input GNSS signal frequency and satellite and receiver position parameters; Step 1.5: Combine the receiving antenna gain, the transmitted signal wavelength, and the coherent integration time parameters, and use the bistatic scattering signal correlation power model to calculate the simulated delay-Doppler map based on the delay-Doppler value obtained in step 1.3 and the scattering coefficient obtained in step 1.

4. The bistatic scattering signal correlation power model is expressed as: Among them, G R is the receiving antenna gain, T i is the coherent integration time, P T is the transmitted signal power, λ is the transmitted signal wavelength, σ 0 is the scattering coefficient, G T is the transmit antenna gain, R t is the distance from the bin to the satellite, R r is the distance from the bin to the receiver, Λ 2 (τ-τ x,y ) is the PRN code coherence function, |S(f c -f x,y )| 2 is the Doppler spread function.

3. The sea surface wind speed inversion method combining the advantages of simulation model and satellite data according to claim 1 is characterized in that: The step 2 specifically includes: Step 2.1: Filter the satellite data and buoy wind speed data to remove outliers with missing or negative wind speeds in the buoy wind speed data. Filter the satellite data for mirror reflection points located in non-land and near-shore areas. Select satellite data with a peak signal-to-noise ratio greater than 3 dB, a receiving antenna gain greater than 3 dB, and an RCG value greater than 5. Step 2.2, matching the satellite data obtained in step 2.1 with the buoy wind speed data to obtain satellite-buoy wind speed matching data; During the matching process, the spatial distance between the mirror reflection point of the satellite data and the buoy site location of the buoy wind speed data is less than 7.5 kilometers, and the satellite data and buoy wind speed data with an observation time difference of less than 15 minutes are selected for matching; Step 2.3: De-noise the satellite-buoy wind speed data obtained in step 2.2 using a 3D matched filter method to obtain satellite-buoy wind speed matched denoised data; Step 2.4: Normalize the satellite-buoy wind speed matching denoised data obtained in step 2.3 to preliminarily construct a GNSS-R measured data set. Step 2.5: Based on the delay-Doppler map simulation model constructed in step 1, input the position and velocity of the satellite and receiver, wind speed, GNSS signal frequency, and receiving antenna gain parameters for simulation to obtain simulated GNSS-R data. Combine the measured GNSS-R data set with the simulated GNSS-R data to construct a joint measured-simulated data set.

4. The sea surface wind speed inversion method combining the advantages of simulation model and satellite data according to claim 1 is characterized in that: The specific method of step 3 includes: Step 3.1: Extract the leading edge slope LES, trailing edge slope TES, and delay Doppler mean eigenvalue DDMA from the delay-Doppler map in the measured-simulation joint data set constructed in step 2. The calculation formula is as follows: in, and Represents the leading edge waveform and trailing edge waveform, using the delay Doppler diagram at the center delay waveform of the Doppler frequency of 0 Hz, taking the delay of -1.25 to 0 chips and 0 to 1.25 chips respectively; P DDM (i, j) represents the value of the delay Doppler map at the position (i, j); N represents the number of delay Doppler map values ​​taken, n represents the number of delay waveform values ​​taken, τ i is the delay value at point i; Step 3.2: Combine the leading edge slope LES and trailing edge slope TES extracted in step 3.1 to calculate the leading-trailing edge composite eigenvalue Eigen L-T The calculation formula is as follows: Among them, LES represents the leading edge slope, and TES represents the trailing edge slope; The leading edge slope LES, trailing edge slope TES, delay Doppler mean DDMA calculated in step 3.1 and the leading-trailing edge composite eigenvalue Eigen calculated in step 3.2 are used to calculate the leading edge slope LES, trailing edge slope TES, delay Doppler mean DDMA and trailing edge composite eigenvalue Eigen calculated in step 3.

2. L-T Add it to the measured-simulated joint dataset constructed in step 2 to obtain the measured-simulated joint dataset containing the eigenvalues.

5. The sea surface wind speed inversion method combining the advantages of simulation model and satellite data according to claim 1 is characterized in that: The specific method of step 4 includes: Step 4.

1. Based on the CNN and CNN_1D models, a deep neural network model is constructed with the delay-Doppler map, eigenvalues, and other parameters, namely, signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions as joint inputs. The delay-Doppler map is input into a neural network composed of three layers of two-dimensional convolution kernels for training. The convolution kernel size is 3×3, and the number of convolution kernels is 256, 128, and 64, respectively. The training results are then flattened and passed through two layers of FCN layers to obtain a 128×1 The first set of data; the eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incidence angle, longitude and latitude coordinates of the mirror reflection point, and the satellite and receiver positions, are input into a neural network consisting of three layers of one-dimensional convolutional kernels for training. The convolution kernel size is 3×1, and the number of convolution kernels is 512, 256, and 128, respectively. The training results are then flattened and passed through two layers of FCN layers to obtain the second set of 128×1 data. The two sets of data are then fused based on the attention mechanism and finally passed through two layers of FCN layers to obtain the sea surface wind speed inversion results. Step 4.2: Input the measured-simulated joint data set containing the eigenvalues ​​constructed in step 3 into the deep neural network model constructed in step 4.1 for training to obtain the sea surface wind speed inversion model. Input the delay-Doppler map, eigenvalues ​​and other parameters, namely the signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions to realize sea surface wind speed inversion. Finally, evaluate the sea surface wind speed inversion accuracy of the trained sea surface wind speed inversion model.

6. A sea surface wind speed inversion system based on the method according to any one of claims 1 to 5, combining the advantages of simulation model and satellite data, characterized in that: include: A model building module is used to build a delay-Doppler map simulation model; The joint measurement-simulation dataset construction module is used to process and spatially match satellite data and buoy wind speed data, initially construct a GNSS-R measurement dataset, and simulate the position and velocity of the satellite and receiver, wind speed, GNSS signal frequency, and receiving antenna gain parameters based on the delay-Doppler map simulation model to construct simulated GNSS-R data. The measured GNSS-R dataset is then combined with the simulated GNSS-R data to construct a joint measurement-simulation dataset. A module for constructing a joint measurement-simulation dataset including eigenvalues ​​is used to extract eigenvalues ​​related to wind speed from the delay-Doppler map in the joint measurement-simulation dataset. The eigenvalues ​​related to wind speed include: leading edge slope, trailing edge slope, and delay-Doppler mean. Combining the leading edge slope and the trailing edge slope, the leading edge-trailing edge composite eigenvalue is calculated to obtain a measured-simulated joint data set containing the eigenvalue; The sea surface wind speed inversion module is used to implement a deep neural network-based model that constructs a delay-Doppler map, eigenvalues ​​and other parameters, namely, signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions as joint inputs. It trains the measured-simulation joint data set containing the eigenvalues ​​to obtain a trained sea surface wind speed inversion model, and inputs the delay-Doppler map, eigenvalues ​​and other parameters, namely, signal-to-noise ratio, receiving antenna gain, incident angle, longitude and latitude coordinates of the mirror reflection point, and satellite and receiver positions to achieve sea surface wind speed inversion.

7. A sea surface wind speed inversion device that combines the advantages of simulation models and satellite data, characterized in that: include: Memory: a computer-readable device storing a computer program for a sea surface wind speed inversion method combining the advantages of a simulation model and satellite data as described in any one of claims 1 to 5; Processor: used to implement the sea surface wind speed inversion method combining the advantages of simulation model and satellite data as described in any one of claims 1-5 when executing the computer program.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it can implement the sea surface wind speed inversion method that combines the advantages of simulation models and satellite data as described in any one of claims 1 to 5.

Images