Sea wave deviation correction method based on wave orbit motion and scattering characteristic analysis

Abstract

The invention belongs to the technical field of ocean surface flow velocity remote sensing observation, and particularly relates to a wave deviation correction method based on wave orbit motion and scattering characteristic analysis, which comprises the following steps of: establishing a wave deviation correction method driven by a physical mechanism in combination with wave orbit motion and Bragg wave scattering intensity change; generating simulation data through the M4S model to train the correction model; applying the trained model to actually measured data of an airborne Ka-band along-track interferometric synthetic aperture radar of the Jones Zhou straits of the South China Sea, and verifying the effectiveness of the model by comparing and analyzing the actually measured data with real data provided by a GNSS drifting buoy; the reduction amplitude of the radial surface flow velocity measurement error is calculated, and the experimental result shows that the error is averagely reduced by about 83.5%. The invention provides an efficient and innovative technical scheme for improving the remote sensing measurement accuracy of the ocean surface flow velocity, can obviously reduce the remote sensing measurement error of the ocean surface flow velocity, and is suitable for the application fields of ocean environment monitoring, climate change research and ocean engineering.

Description

Technical Field

[0001] The invention belongs to the technical field of ocean surface velocity remote sensing observation, in particular to a method for correcting ocean wave deviation based on wave orbit motion and scattering characteristic analysis. Background Art

[0002] Ocean surface currents (OSC) are important observation parameters for marine environmental monitoring, climate change research and marine engineering, which directly affect marine material transport, heat exchange and ecosystem dynamics. Obtaining high-precision and large-scale ocean surface current data is of great significance for understanding ocean dynamic processes and predicting changes in the marine environment. At present, ATI-SAR technology has become an important tool for observing ocean surface currents, with the advantages of high resolution and wide coverage. However, the existence of waves in actual observations seriously affects the accuracy and stability of the observation results.

[0003] During SAR observations, the relative motion between the orbital motion of ocean surface waves and the radar sensor will cause Doppler shift, which will cause deviations in OSC measurements. This deviation is called wave bias (WB). The current mainstream WB correction model takes into account the hydrodynamic modulation of Bragg wave scattering to a certain extent, but has great limitations in dealing with the mutual coupling relationship between wave orbital motion and non-uniform distribution of scattering intensity. In addition, under different wind speeds, wave heights and complex sea conditions, there are significant differences in the correction accuracy of existing models. Especially under medium and high wind speed conditions, the maximum deviation error can reach 1.5m / s, which seriously affects the reliability of observation data.

[0004] On the other hand, most of the existing WB correction models rely on theoretical assumptions and are not fully verified by observational data in actual ocean environments. This reliance on theoretical assumptions results in poor adaptability and generalization of the model when faced with complex changes in the ocean environment. In addition, current correction methods often ignore the non-uniform effect of long-wave modulation on the ocean surface on the distribution of Bragg wave scattering intensity, and fail to accurately quantify the nonlinear coupling mechanism between orbital motion and scattering characteristics, resulting in significant deviations in the correction effect under different sea conditions.

[0005] Current theoretical models (such as DopRIM, IDopRIM and DPDop) usually consider the additional Doppler shift caused by the Bragg wave phase velocity as the core assumption when explaining the OSC deviation of SAR measurements. However, when establishing the deviation correction mechanism, these models often ignore the actual contribution of the wave orbital motion on the scattering unit and do not fully consider the nonlinear coupling relationship between the orbital motion velocity and the backscattering intensity, which leads to unsatisfactory performance of the model in practical applications, especially in complex sea conditions. It is difficult to effectively improve the observation accuracy. Summary of the invention

[0006] In view of the above-mentioned defects in the prior art, the present invention provides a remote sensing current wave deviation correction method based on wave orbit motion and scattering characteristic analysis to improve the accuracy of remote sensing measurement of ocean surface current velocity.

[0007] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: a wave deviation correction method based on wave orbit motion and scattering characteristics analysis, comprising the following steps:

[0008] 1) Use the numerical model M4S to generate simulation data under different wind speeds, wave heights and ocean surface currents to construct a training data set;

[0009] 2) By analyzing the non-uniform distribution characteristics of Bragg wave scattering intensity caused by the orbital motion of ocean waves and long-wave modulation, the contribution mechanism of wave deviation to SAR measurement of ocean surface velocity is clarified, thereby constructing a wave deviation correction model;

[0010] 3) Input the simulation data into the wave deviation correction model, train the model through deep learning method, and obtain the optimal hydrodynamic modulation weight factor;

[0011] 4) Based on the optimized hydrodynamic modulation weight factor, an accurate quantitative expression of radial surface velocity deviation is established; and the simulation results of downwind and upwind observation directions are statistically analyzed for deviation correction under different wind speed conditions;

[0012] 5) Based on the measured data in the test area, calculate the deviation of the radial surface velocity measured by the uncorrected along-track interferometric synthetic aperture radar ATI-SAR;

[0013] 6) Verify the correction effect of the trained wave deviation correction model and its accuracy and stability under different sea conditions;

[0014] 7) Visualize the radial surface velocity data after deviation correction to display the radial surface velocity distribution diagram and the comparison results before and after deviation correction; quantify the optimization effect of radial surface velocity measurement deviation through comparison results of GNSS buoy real data.

[0015] The step 1) comprises the following steps:

[0016] S11: Acquire along-track interferometric synthetic aperture radar ATI-SAR data, that is, obtain interferometric phase data; the numerical model M4S calls the sea wave information in the simulated interferometric phase data, and analyzes the sea wave spectrum information to obtain the sea wave spectrum data;

[0017] S12: According to the zero-order moment integration of the wave spectrum data, the effective wave height value is calculated, that is:

[0018]

[0019] Among them, H S is the effective wave height, m 0 is the zero-order moment of the wave spectrum;

[0020] S13: Combining the dispersion relationship of deep water waves with the Toba 3 / 2 exponential law, the wavelength and period information are further derived, namely:

[0021]

[0022] Among them, u * is the friction velocity, g is the gravitational acceleration, T S is the effective period, U is the average wind speed at the sea surface at height z, κ is the von Karman constant, z 0 is the sea surface roughness;

[0023] S14: The M4S model is used to simulate the along-track interference phase of different observation directions with and without wind, thereby generating a training dataset for observing radial surface velocity.

[0024] The step 2) comprises the following steps:

[0025] S21: Based on the Doppler principle, orbital motion law and long-wave modulation mechanism, an initial deviation correction physical model is constructed to ensure that the model can accurately describe the interaction between the orbital motion of water particles and the Bragg wave scattering intensity;

[0026] S22: A hydrodynamic modulation weight factor is introduced into the model to quantify the non-uniform distribution of Bragg wave scattering intensity caused by long-wave hydrodynamic modulation within the wavelength range, providing accurate weight parameters for subsequent bias correction;

[0027] S23: Introduce the orbital velocity integral of water particles into the deviation correction model, combine the non-uniform distribution characteristics of Bragg wave scattering intensity within the wavelength range, improve the theoretical framework of deviation correction, and complete the construction of the deviation correction physical model.

[0028] The step S23 is to improve the theoretical framework of deviation correction, specifically:

[0029] a. Introducing the wave deviation correction model into the measurement of ocean surface velocity, the wave deviation calculation model is:

[0030]

[0031] Among them, u b is the wave deviation, L is the wavelength, x is the spatial position, u(x) is the horizontal velocity component of orbital motion, w(x) is the vertical velocity component of orbital motion, is the hydrodynamic modulation effect, γ(x) is the tilt modulation effect, and θ is the radar incident angle;

[0032] b. The expressions for the horizontal and vertical velocity components of orbital motion are:

[0033] u=aωsin(kx-ωt)

[0034] w=-aωcos(kx-ωt)

[0035] Among them, a is the amplitude, ω is the angular frequency, and k is the wave number;

[0036] c. The expression of the fluid dynamics modulation effect is:

[0037]

[0038] Among them, ρ x Represents the weight factor of the electromagnetic wave backscattering intensity at each position along the wavelength, ranging from 0 to 1, ρ min , ρ mid and ρ max correspond to the minimum, intermediate, and maximum weights of the electromagnetic backscatter intensity, respectively;

[0039] d. The expression of tilt modulation effect is:

[0040] γ(x)=cos(θ±arctan(ζ'))

[0041] Among them, ζ = asin(kx-ωt) is the wave function, the positive sign is applicable to upwind observations, and the negative sign is applicable to downwind observations; at zero incident angle, almost all electromagnetic wave energy will contact the surface, thereby minimizing the loss of backscattered signals.

[0042] The step 3) comprises the following steps:

[0043] S31: inputting the generated simulation data into the wave deviation correction model, and using the deep learning algorithm to perform model training to calculate the wave deviation;

[0044] S32: During the training process, the cross-validation method is used to evaluate the performance of the model to ensure the stability and generalization ability of the model parameters and reduce the risk of overfitting;

[0045] S33: After the model training is completed, an independent validation data set is used to test the model performance, analyze the model's prediction effect under different sea conditions, and verify its correction accuracy and stability for wave deviations.

[0046] The step S31 is specifically as follows:

[0047] The model constructed by deep learning algorithm for model training has three hidden layers, and each layer has 256, 128, and 64 neurons respectively. The Adam optimizer is used to train the parameters of the wave deviation calculation model that integrates the physical model and deep learning, and Relu is used as the activation function in each layer. The loss function uses the mean square error function.

[0048] Based on the measured wave deviation data, the Adam optimizer selects the optimal solution of the hydrodynamic modulation weight factor from the random array, and brings 90% of the optimal solution data into the DNNs neural network for training, and 10% of the data is used to verify the training results. After the model training is completed, the wave parameters are input to obtain the corresponding predicted weight factor, and the predicted weight factor is brought into the wave deviation calculation model established above to obtain the final wave deviation.

[0049] The step 4) is specifically:

[0050] S41: The along-track interferometry method is selected to verify the model performance, and the direct relationship between the radial surface velocity and the interference phase is expressed as:

[0051]

[0052] Among them, u surf is the radial surface velocity, λ is the wavelength of the electromagnetic wave, V p is the satellite flight speed, φ ATI is the interference phase, B is the baseline length, and θ is the incident angle;

[0053] S42: According to the radial surface speed u surf Wave deviation u b , and then the radial surface velocity after wave deviation correction by wave deviation calculation model integrating physical model and deep learning is obtained: u surf -u b .

[0054] The step 5) is specifically:

[0055] Experiments were carried out in the test area using an airborne Ka-band along-track interferometric synthetic aperture radar (ATI-SAR) to obtain measured data. The data were compared with the real data of ocean surface velocity provided by a synchronously placed GNSS drifting buoy, and the deviation of the radial surface velocity measured by the uncorrected ATI-SAR was obtained.

[0056] The step 6) is specifically:

[0057] S61: Use the trained wave deviation correction model to correct the measured data of the on-track interferometric synthetic aperture radar ATI-SAR, compare the corrected data with the measured data provided by the GNSS drifting buoy, verify the effectiveness of the model correction, and statistically analyze the improvement magnitude of the radial surface velocity deviation after correction;

[0058] S62: Statistical analysis of the corrected ocean surface velocity data was performed to calculate the reduction in radial velocity error, and the correction effect of the model was quantitatively evaluated by comparing the differences in the measurement results before and after correction;

[0059] S63: Verify the validity of the corrected radial velocity data by comparing and analyzing the measured data of the GNSS drifter point by point to ensure that the corrected data is highly consistent and conforms to the actual distribution characteristics of the ocean surface velocity; verify the consistency of the data under different sea conditions to ensure that the corrected data is highly consistent with the actual observed ocean surface velocity in spatial distribution, quantitatively evaluate the accuracy and stability of the corrected data, and finally verify the reliability of the model in the actual ocean environment.

[0060] The present invention has the following beneficial effects and advantages:

[0061] 1. The present invention provides an efficient and innovative technical solution for improving the accuracy of remote sensing measurement of ocean surface current velocity, which can significantly reduce the error of remote sensing measurement of ocean surface current velocity and is suitable for marine environment monitoring, climate change research and marine engineering application fields.

[0062] 2. The present invention generates simulation data under different conditions through the numerical model M4S to construct a training data set, providing a rich and diverse data basis for model training, so that the model can adapt to various sea conditions. The interferometric phase information is obtained by along-track interferometric synthetic aperture radar (ATI-SAR), and the wave spectrum is analyzed and the parameters are calculated to obtain the wave information, providing data support for the subsequent accurate correction of the wave deviation.

[0063] 3. The present invention deeply analyzes the non-uniform distribution characteristics of Bragg wave scattering intensity caused by the orbital motion of ocean waves and long-wave modulation, clarifies the contribution mechanism of wave deviation to SAR measurement of ocean surface velocity, and constructs an accurate wave deviation correction model, providing a solid theoretical basis and effective tools for correcting wave deviation.

[0064] 4. The present invention uses deep learning methods to train the model and obtain the optimal hydrodynamic modulation weight factor, which can more accurately quantify the non-uniform distribution of Bragg wave scattering intensity caused by long-wave hydrodynamic modulation within the wavelength range, thereby improving the correction accuracy of the model. The cross-validation method is used to evaluate the model performance to reduce the risk of overfitting, and an independent validation data set is used for performance testing to ensure the stability and generalization ability of the model under different sea conditions.

[0065] 5. Based on the optimized hydrodynamic modulation weight factor, the present invention establishes an accurate quantitative expression for radial surface velocity deviation, and performs deviation correction statistical analysis on simulation results of different wind speeds and observation directions, which helps to gain a deeper understanding of the correction effect of the model under different conditions and provide detailed data reference for practical applications.

[0066] 6. The present invention calculates the uncorrected radial surface velocity deviation through the measured data of the test area, and compares it with the real data of the GNSS buoy to verify the correction effect, accuracy and stability of the trained model, so that the reliability of the model is based on the actual data, ensuring the effectiveness of the model in the actual marine environment.

[0067] 7. The present invention visualizes the radial surface velocity data after deviation correction, displays the comparison results before and after deviation correction, and quantifies the optimization effect by comparing with the real data of GNSS buoy, intuitively and clearly presents the improvement brought about by model correction, and facilitates users to understand and evaluate model performance.

[0068] 8. The present invention calculates the reduction in radial surface velocity measurement error, and experimental results show that the error is reduced by about 83.5% on average. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a flow chart of a wave deviation correction model provided by an embodiment of the present invention and its application method in ocean surface current velocity measurement;

[0070] Figure 2 It is a schematic diagram showing the long wave orbit motion process provided by an embodiment of the present invention;

[0071] Figure 3a The embodiment of the present invention shows the non-uniform distribution of Bragg wave density within the wavelength and the change of orbital motion speed with the spatial position of the wavelength;

[0072] Figure 3b It is a principle diagram showing that the orbital velocity integral within the wavelength is not zero due to the long-wave modulation effect demonstrated in the implementation of the present invention;

[0073] Figure 4It is a system diagram of quantized optimal fluid dynamics modulation weight factors integrating physical model and deep learning provided by the implementation of the present invention;

[0074] Figure 5 It is a comparison diagram between the prediction and actual of the optimal fluid dynamics modulation weight factor by the fusion physical model and deep learning system demonstrated by the implementation of the present invention;

[0075] Figure 6 1. It is a statistical result diagram of wave deviation before and after correction provided by an embodiment of the present invention;

[0076] Figure 7 This is a diagram showing the correction of the AIT-SAR radial surface velocity measured in the Qiongzhou Strait of the South China Sea provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0077] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0078] At present, OSC inversion based on remote sensing technology faces great challenges in practical applications, among which the main problem is that the observation error is significant, especially the deviation (WB) caused by waves has become the main source of measurement accuracy. Although traditional deviation correction models, such as those based on the Doppler radar imaging model (DopRIM), explain the influence of waves on OSC inversion to a certain extent, they still have the problem of insufficient accuracy in practical applications. The DopRIM model mainly describes the generation mechanism of WB by weighted average of the phase velocity of Bragg waves. However, the actual accuracy of this model is low under medium and high wind speed conditions, and it is difficult to meet the needs of high-precision OSC measurement. In view of the above problems, the present invention proposes a new wave deviation correction model (WBCM). This model comprehensively considers the orbital motion of waves and the changes in the scattering intensity of Bragg waves caused by them, and effectively corrects the wave deviation in OSC inversion by combining numerical simulation with deep learning methods.

[0079] like Figure 1 As shown, it is a flow chart of a wave deviation correction model and its application method in ocean surface current velocity measurement provided by an embodiment of the present invention. A remote sensing current wave deviation correction method based on wave orbit motion and scattering characteristic analysis of the present invention comprises the following steps:

[0080] S1: Use the numerical model M4S to generate simulation data under different wind speeds, wave heights and ocean surface currents, build a training dataset covering typical sea conditions, and provide a high-quality input data source for the wave deviation correction model (WBCM);

[0081] Specifically, constructing a training dataset covering typical sea conditions includes the following steps:

[0082] S11: Acquire along-track interferometric synthetic aperture radar ATI-SAR data, that is, obtain interferometric phase data; the numerical model M4S calls the sea wave information in the simulated interferometric phase data, and analyzes the sea wave spectrum information to obtain the sea wave spectrum data;

[0083] S12: According to the zero-order moment integration of the wave spectrum data, the effective wave height value is calculated, that is:

[0084]

[0085] Among them, H S is the effective wave height, m 0 is the zero-order moment of the wave spectrum;

[0086] S13: Combining the dispersion relationship of deep water waves with the Toba 3 / 2 exponential law, the wavelength and period information are further derived, namely:

[0087]

[0088] Among them, u * is the friction velocity, g is the gravitational acceleration, T S is the effective period, U is the average wind speed at the sea surface at height z, κ is the von Karman constant, z 0 is the sea surface roughness;

[0089] S14: The M4S model is used to simulate the along-track interference phase of different observation directions with and without wind, thereby generating a training dataset for observing radial surface velocity.

[0090] S2: Based on the Doppler principle, orbital motion law and long-wave modulation mechanism, a physical model for deviation correction is constructed to ensure that the model can accurately describe the interaction between the orbital motion of water particles and the Bragg wave scattering intensity. A hydrodynamic modulation weight factor is introduced into the model to quantify the non-uniform distribution of the Bragg wave scattering intensity caused by long-wave hydrodynamic modulation within the wavelength range, providing accurate weight parameters for subsequent deviation correction;

[0091] S3: Introduce the orbital velocity integral of water particles into the deviation correction model, and combine the non-uniform distribution characteristics of Bragg wave scattering intensity within the wavelength range to improve the theoretical framework of deviation correction;

[0092] S31 constructs an initial deviation correction physical model based on the Doppler principle, orbital motion law and long-wave modulation mechanism to ensure that the model can accurately describe the interaction between the orbital motion of water particles and the scattering intensity of Bragg waves;

[0093] The hydrodynamic modulation weight factor is introduced into the S32 model to quantify the non-uniform distribution of Bragg wave scattering intensity caused by long-wave hydrodynamic modulation within the wavelength range, providing accurate weight parameters for subsequent deviation correction;

[0094] S33 introduces the orbital velocity integral of water particles into the deviation correction model, combines the non-uniform distribution characteristics of Bragg wave scattering intensity within the wavelength range, improves the theoretical framework of deviation correction, and completes the construction of the deviation correction physical model.

[0095] S4: Use deep learning methods and neural networks to train the bias correction model to obtain the optimal hydrodynamic modulation weight factor to ensure that the model has high prediction accuracy;

[0096] S5: Substitute the optimized hydrodynamic modulation weight factor into the deviation correction formula to establish an accurate quantitative expression of radial velocity deviation, providing mathematical support for subsequent deviation correction;

[0097] S6: Under low, medium and high wind speed conditions, the simulation results in the downwind and upwind observation directions are statistically analyzed for bias correction, and the stability and applicability of the model under different sea conditions are comprehensively evaluated;

[0098] S7: At this point, the model is established. The radial surface velocity is calculated using the along-track interferometry method. An imaging area is selected in the Qiongzhou Strait of the South China Sea (109.89°E, 20.05°N). The airborne Ka-band ATI-SAR system designed by the Beijing Radio Measurement Institute is used for experiments. The system includes Ka-band ATI-SAR, data acquisition unit and inertial navigation system. The imaging area is about 1.91km×4.15km, covering six GNSS drifting buoy paths. The sea surface effective wave height, wave period and ocean surface velocity data are obtained through GNSS drifting buoys, and the wind vector information of the experimental area is provided in combination with the ERA-5 reanalysis wind field data. The radial surface velocity data observed by ATI-SAR are preliminarily compared with the ocean surface velocity data measured by GNSS drifting buoys. The accuracy of the uncorrected ATI-SAR measured radial velocity is calculated, and the initial comparison results are obtained, which provide a reference benchmark for the subsequent verification of the wave bias correction model.

[0099] S8: Introduce WBCM to correct the radial surface velocity deviation, and compare and analyze the corrected radial velocity results with the measured data of GNSS drifting buoy to verify the accuracy and stability of the model under different sea conditions;

[0100] The trained WBCM is used to correct the ATI-SAR measured data. During the correction process, the measured data are compared and analyzed with the ocean surface velocity data provided by the GNSS drifting buoy. The deviation correction is performed based on the non-uniform distribution characteristics of the Bragg wave scattering intensity caused by wave orbital motion and long-wave modulation, and finally the corrected radial surface velocity data is obtained. By comparing the radial velocity data before and after correction, the applicability and stability of WBCM under different flow rates and wave conditions are verified.

[0101] S9: Visualize the radial velocity data after deviation correction, display the radial surface velocity distribution diagram and the comparison results before and after deviation correction, and intuitively present the correction effect of the model;

[0102] S10: Through the model validation results, the optimization effect of radial surface velocity measurement deviation is quantified to ensure that the model has high generalization ability and reliability in the actual marine environment.

[0103] The present invention proposes a wave deviation correction model that combines a physical model with artificial intelligence, which significantly improves the accuracy and efficiency of wave deviation calculation. In traditional methods, wave deviation correction models represented by DopRIM show obvious limitations under high wind speeds and complex sea conditions, resulting in insufficient correction accuracy. In order to solve the problem of unstable correction effect of existing methods in complex environments, the present invention integrates a wave deviation calculation model based on physical mechanisms and a deep learning algorithm to establish a wave deviation correction model (WBCM). The model introduces a hydrodynamic modulation weight factor to quantify the contribution of the non-uniform distribution characteristics of wave orbital motion and Bragg wave scattering intensity to wave deviation, and uses simulation data, ATI-SAR observation data and GNSS buoy measured data to complete the correction of radial surface velocity deviation.

[0104] Embodiment 1:

[0105] like Figure 1 As shown, the wave deviation correction method integrating physical model and deep learning provided by the embodiment of the present invention includes the following steps:

[0106] S1: Use the numerical model M4S to generate simulation data under different wind speeds, wave heights and ocean surface currents, build a training data set covering typical sea conditions, and provide a high-quality input data source for the bias correction model;

[0107] The environmental parameters include wind speed, wave height, period and wavelength information, and the ATI-SAR data is interferometric phase information. The relevant expressions are as follows:

[0108]

[0109] Among them, H S is the effective wave height, m0 is the zero-order moment of the wave spectrum, u * is the friction velocity, g is the gravitational acceleration, T S is the effective period, U is the average wind speed at the sea surface at height z, κ is the von Karman constant, z 0 is the roughness of the sea surface, which indicates the resistance of the sea surface to the wind;

[0110] According to formulas (1) to (3) combined with the dispersion relationship of deep water waves, complete wave data can be obtained. Based on this, the M4S model can simulate the interference phase information under a given flow velocity.

[0111] S2: Based on the Doppler principle, orbital motion law and long-wave modulation mechanism, a physical model for deviation correction is constructed to ensure that the model can accurately describe the interaction between the orbital motion of water particles and the Bragg wave scattering intensity. A hydrodynamic modulation weight factor is introduced into the model to quantify the non-uniform distribution of the Bragg wave scattering intensity caused by long-wave hydrodynamic modulation within the wavelength range, providing accurate weight parameters for subsequent deviation correction;

[0112] S3: Introduce the orbital velocity integral of water particles into the deviation correction model, and combine the non-uniform distribution characteristics of Bragg wave scattering intensity within the wavelength range to improve the theoretical framework of deviation correction;

[0113] The Doppler shift is caused by the relative motion between the radar sensor and the sea surface. In ocean waves, this shift is mainly caused by the sensor and the orbital motion of water particles ( Figure 1 The Doppler shift is caused by the relative velocity between the two water particles (green dots). By considering the orbital motion of each scattering point on the ocean wave, the overall Doppler shift can be determined by integrating the contribution of all water particles within a wavelength. If the intensity of the backscattered signal is uniformly distributed along the wavelength, or the sensor can fully detect the orbital velocity of all water particles within a wavelength, the resulting Doppler shifts will cancel each other out. However, surface wave modulation causes the intensity of the Bragg wave backscatter to vary unevenly at different locations on the wavelength. This uneven distribution causes the Doppler shift to not be completely canceled, resulting in a residual frequency shift, which introduces a wave bias in the OSC measurement. The wave bias can be accurately quantified using formula (4):

[0114]

[0115] Among them, u b is the wave deviation, L is the wavelength, x is the spatial position, u(x) is the horizontal velocity component of orbital motion, w(x) is the vertical velocity component of orbital motion, is the fluid dynamics modulation effect, γ(x) is the tilt modulation effect, and θ is the radar incident angle.

[0116] The expressions for the horizontal and vertical velocity components of orbital motion are:

[0117] u=aωsin(kx-ωt)(5)

[0118] w=-aωcos(kx-ωt)(6)

[0119] Where a is the amplitude, ω is the angular frequency, and k is the wave number.

[0120] The expression of the fluid dynamics modulation effect is:

[0121]

[0122] Among them, ρ x The weight factor representing the backscattering intensity of electromagnetic waves at each position along the wavelength is a dimensionless quantity ranging from 0 to 1. min , ρ mid and ρ max corresponds to the minimum, intermediate and maximum weights of the electromagnetic backscattering intensity, respectively. This spatially inhomogeneous backscattering driven by hydrodynamic modulation is Figure 3b As shown. The red curve represents the convergence zone, and the blue and green curves represent the divergence zone. In the convergence zone, the Bragg wave intensity is high, and the intensity gradually decreases from the peak to the trough, which is modeled as a downward-opening parabola. In contrast, in the divergence zone, the Bragg wave intensity decreases rapidly, which is modeled as an upward-opening parabola, reflecting the asymmetric distribution of the Bragg wave intensity, as shown in Figure 3a shown.

[0123] The expression of tilt modulation effect is:

[0124] γ(x)=cos(θ±arctan(ζ'))(8)

[0125] Among them, ζ = asin(kx-ωt) is the wave function, and the positive sign is applicable to upwind observations, while the negative sign is applicable to downwind observations. At zero angle of incidence, almost all electromagnetic wave energy will contact the surface, thereby minimizing the loss of backscattered signal. As the local angle of incidence increases, the intensity of backscattering decreases, and this relationship can be well described by the cosine function. Figure 3b The changes in backscatter weights under headwind and downwind conditions due to tilt modulation are also shown.

[0126] S4: Use deep learning methods and neural networks to train the bias correction model to obtain the optimal hydrodynamic modulation weight factor to ensure that the model has high prediction accuracy;

[0127] Specifically, the present invention is based on deep neural networks (DNNs), and on the basis of the wave deviation correction physical model, constructs a wave deviation correction model that integrates the physical model and the deep learning algorithm, and verifies the wave deviation correction result using the GNSS buoy measured data. By constructing the correction relationship between the actual observed value of the radial surface velocity deviation and the numerical simulation result, the problem of error between the model result and the actual observation in the deviation correction process is solved, and the high-precision correction of the radial surface velocity in the actual sea area is further realized.

[0128] DNNs structure used in the present invention Figure 4 As shown in Figure 2, based on the measured WB data, the Adam optimizer selects the optimal solution of the hydrodynamic modulation weight factor from the random array, and brings 90% of the optimal solution data into the DNNs neural network for training, and 10% of the data is used to verify the training results. The final training results are shown in Figure 2. Figure 5 After the model training is completed, the corresponding prediction weight factor can be obtained by inputting the wave parameters, and the prediction weight factor can be brought into the wave deviation calculation model established above (Formula 4) to obtain the final wave deviation (trained under the conditions of tailwind and headwind respectively).

[0129] The present invention constructs a wave deviation calculation model integrating physical model and deep learning, which specifically includes a fluid dynamics weight factor inversion module and a wave deviation artificial intelligence calculation module.

[0130] Exemplarily, the fluid dynamics weight factor inversion module in the wave deviation calculation model of the fusion physical model and deep learning constructed by the present invention can accurately invert the weight factor value based on the measured wave deviation data. The artificial intelligence correction module is trained based on the inverted weight factor value, and the neural network can learn the direct relationship between the weight factor and the wave information. During the correction process, the input and output of the wave deviation calculation model of the fusion physical model and deep learning correspond one to one in time. The hidden layer of the wave deviation calculation model of the fusion physical model and deep learning is 3 layers, and each layer is set with 256, 128, and 64 neurons respectively. The Adam optimizer is used to train the parameters of the wave deviation calculation model of the fusion physical model and deep learning, and each layer uses Relu as the activation function; the loss function uses the mean square error function.

[0131] The amplitude, wave number, wavelength and circular frequency of the ocean waves are used as the basic inputs of the wave deviation calculation model that integrates the physical model and deep learning. The forecast effect of the wave deviation calculation model that integrates the physical model and deep learning under different wind speeds and observation directions is as follows: Figure 5 shown.

[0132] like Figure 5As shown in the figure, it focuses on the verification of the model prediction effect. By exploring the influence of wind speed, observation direction and other inputs on the model prediction accuracy, it is finally verified that the model has strong universality. By comparison, it can be found that the wave deviation calculation model integrating physical model and deep learning can achieve high prediction accuracy under various conditions.

[0133] S5: Substitute the optimized hydrodynamic modulation weight factor into the deviation correction formula to establish an accurate quantitative expression of radial velocity deviation, providing mathematical support for subsequent deviation correction.

[0134] This embodiment uses the along-track interferometry method to verify the performance of the proposed model. In this context, the direct relationship between the radial surface velocity and the interference phase is expressed as:

[0135]

[0136] Among them, u surf is the radial surface velocity, λ is the wavelength of the electromagnetic wave, V p is the satellite flight speed, φ ATI is the interference phase, B is the baseline length, and θ is the incident angle.

[0137] At this time, formula (9) minus formula (4) is the radial surface velocity result after the wave deviation is corrected by the WB calculation model that integrates the physical model and deep learning.

[0138] S6: Under low, medium and high wind speed conditions, the simulation results in the downwind and upwind observation directions are statistically analyzed for bias correction, and the stability and applicability of the model under different sea conditions are comprehensively evaluated;

[0139] Directly calculated by formula 9, such as Figure 6 As shown, it is a statistical result diagram of the wave deviation before and after correction provided by the embodiment of the present invention. It can be seen that WBCM has a good wave deviation correction ability.

[0140] S7: At this point, the model is established, and the radial surface velocity is calculated using ATI-SAR data, the deviation distribution characteristics of each observation area are extracted, and the uncorrected observation deviation results are obtained;

[0141] Among them, the ATI-SAR data were obtained from observation experiments conducted in the Qiongzhou Strait of the South China Sea. The corresponding GNSS buoys in the imaging area provided real radial surface velocity data.

[0142] The error statistics of the corrected ocean surface velocity data were analyzed, and the reduction of the radial surface velocity error was calculated. By comparing the radial surface velocity data before and after correction, the error distribution and convergence characteristics of the model under different sea conditions were evaluated, and the correction effect and stability of the model were quantitatively verified.

[0143] S8 introduces the wave deviation correction model (WBCM) to correct the radial surface velocity deviation, and compares and analyzes the corrected radial velocity results with the measured data of the GNSS drifting buoy to verify the accuracy and stability of the model under different sea conditions;

[0144] The effectiveness of the model correction is verified by comparing the corrected radial velocity data with the measured data provided by the GNSS drifting buoy point by point. Data consistency verification is carried out under different sea conditions to ensure that the corrected data is highly consistent with the actual observed ocean surface velocity in spatial distribution, and the accuracy and stability of the corrected data are quantitatively evaluated. Finally, the reliability of the model application in the actual marine environment is verified.

[0145] S9: Visualize the radial velocity data after bias correction, display the radial surface velocity distribution diagram and the comparison results before and after bias correction, and intuitively present the correction effect of the model. The statistical parameter is the average deviation; through the model verification results, quantify the optimization effect of the radial surface velocity measurement deviation, and ensure that the model has high generalization ability and reliability in the actual marine environment.

[0146] S10: Integrate and optimize the high-precision ocean surface velocity data after WBCM correction to output high-precision velocity products that conform to the actual distribution characteristics of the ocean environment; this data can be used in many fields such as marine environment monitoring, marine engineering design, marine disaster warning, and marine climate change research, significantly improving the practicality and reliability of satellite remote sensing technology in ocean surface velocity measurement.

[0147] Embodiment 2:

[0148] Combined with the steps of Example 1, taking the airborne along-track interferometric SAR measurement in the Qiongzhou Strait area of ​​the South China Sea as an example; combining the GNSS buoy observation data deployed in the Qiongzhou Strait of the South China Sea and the airborne ATI-SAR data, the uncorrected ATI-SAR measured radial surface velocity and wave deviation distribution characteristics are obtained.

[0149] Six GNSS buoys were deployed in the ATI-SAR imaging area of ​​the Qiongzhou Strait to produce six different drift paths.

[0150] The wind speed at the experimental site was about 5 m / s, and the observation direction was against the wind.

[0151] The subsets corresponding to the drift paths of the six GNSS buoys in the imaging area were selected to calculate the radial surface velocity of the original ATI-SAR observations and the radial surface velocity after the wave deviation was corrected by WBCM in each subset. The correction results are shown in the figure below. Figure 7 shown.

[0152] The measured results are corrected for wave deviations using the trained WBCM model, and the deviations between the corrected results and the GNSS buoy observations are calculated and compared with the uncorrected results to verify the correction performance of the WBCM.

[0153] Finally, based on visualization processing, the average deviation distribution of radial velocity before and after correction is visualized to intuitively evaluate the correction effect. The wave deviation correction effect is verified based on the visualization results. When the error improvement exceeds 80% and the average deviation is less than 0.05m / s, it shows that WBCM has a high wave deviation correction capability.

[0154] It can be seen that the present invention can effectively correct the measurement error caused by ocean waves in OSC inversion, greatly improving the practicability of remote sensing technology in flow velocity measurement.

[0155] In summary, the experimental results of the above embodiments show that the method of the present invention overcomes the defects of poor calculation accuracy of the traditional deviation correction method and optimizes the physical mechanism. Compared with the traditional model, the model can accurately quantify the contribution of the wave orbit motion speed to the wave deviation, significantly improve the accuracy and efficiency of deviation correction, and provide an innovative solution for correcting the wave influence in current observation in actual marine environment. The present invention provides important technical support for marine observation, navigation safety and marine climate research, and has broad application prospects and important scientific significance.

[0156] The technical solution of the present invention effectively solves the long-standing problem of large errors in remote sensing measurements of ocean surface current velocity. Due to the limited calculation accuracy of the traditional wave deviation correction model, how to achieve high-precision inversion of surface current velocity has always been a problem that people are eager to solve. The present invention successfully makes up for this technical problem by integrating the physical model and the WB correction method of deep learning. This technical solution successfully quantifies the errors caused by the non-uniform distribution of orbital motion speed and backscattering intensity in remote sensing technology observations, and provides a breakthrough solution for improving the accuracy of sea surface current velocity measurements. While filling the technical gap, this achievement also provides the industry with advanced technical means that meet actual needs, marking a successful response to long-standing challenges in this field.

[0157] Those skilled in the art will appreciate that the above are only preferred embodiments of the present invention, and the various embodiments of the present disclosure and / or the features described in the claims may be combined or combined in various ways, even if such combinations or combinations are not explicitly described in the present disclosure. It is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art may still modify the technical solutions described in the aforementioned embodiments, or perform equivalent substitutions on some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

[0158] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A wave deviation correction method based on wave orbit motion and scattering characteristics analysis is characterized in that: The following steps are involved: 1) Use the numerical model M4S to generate simulation data under different wind speeds, wave heights and ocean surface currents to construct a training data set; 2) By analyzing the non-uniform distribution characteristics of Bragg wave scattering intensity caused by the orbital motion of ocean waves and long-wave modulation, the contribution mechanism of wave deviation to SAR measurement of ocean surface velocity is clarified, thereby constructing a wave deviation correction model; 3) Input the simulation data into the wave deviation correction model, train the model through deep learning method, and obtain the optimal hydrodynamic modulation weight factor; 4) Establish an accurate quantitative expression of radial surface velocity deviation based on the optimized hydrodynamic modulation weight factor; And under different wind speed conditions, the simulation results in the downwind and upwind observation directions are statistically analyzed for bias correction; 5) Based on the measured data in the test area, calculate the deviation of the radial surface velocity measured by the uncorrected along-track interferometric synthetic aperture radar ATI-SAR; 6) Verify the correction effect of the trained wave deviation correction model and its accuracy and stability under different sea conditions; 7) Visualize the radial surface velocity data after deviation correction to display the radial surface velocity distribution diagram and the comparison results before and after deviation correction; quantify the optimization effect of radial surface velocity measurement deviation through comparison results of GNSS buoy real data.

2. The method for correcting sea wave deviation based on wave orbit motion and scattering characteristics analysis according to claim 1, characterized in that: The step 1) comprises the following steps: S11: Acquire along-track interferometric synthetic aperture radar ATI-SAR data, that is, obtain interferometric phase data; the numerical model M4S calls the sea wave information in the simulated interferometric phase data, and analyzes the sea wave spectrum information to obtain sea wave spectrum data; S12: According to the zero-order moment integration of the wave spectrum data, the effective wave height value is calculated, that is: Among them, H S is the significant wave height, m0 is the zero-order moment of the wave spectrum; S13: Combining the dispersion relationship of deep water waves with the Toba 3 / 2 exponential law, the wavelength and period information are further derived, namely: Among them, u * is the friction velocity, g is the gravitational acceleration, T S is the effective period, U is the average wind speed at the sea surface at height z, κ is the von Karman constant, and z0 is the sea surface roughness; S14: The M4S model is used to simulate the along-track interference phase of different observation directions with and without wind, thereby generating a training dataset for observing radial surface velocity.

3. The method for correcting sea wave deviation based on wave orbit motion and scattering characteristics analysis according to claim 1, characterized in that: The step 2) comprises the following steps: S21: Based on the Doppler principle, orbital motion law and long-wave modulation mechanism, an initial deviation correction physical model is constructed to ensure that the model can accurately describe the interaction between the orbital motion of water particles and the Bragg wave scattering intensity; S22: A hydrodynamic modulation weight factor is introduced into the model to quantify the non-uniform distribution of Bragg wave scattering intensity caused by long-wave hydrodynamic modulation within the wavelength range, providing accurate weight parameters for subsequent bias correction; S23: Introduce the orbital velocity integral of water particles into the deviation correction model, combine the non-uniform distribution characteristics of Bragg wave scattering intensity within the wavelength range, improve the theoretical framework of deviation correction, and complete the construction of the deviation correction physical model.

4. The method for correcting sea wave deviation based on wave orbit motion and scattering characteristics analysis according to claim 3 is characterized in that: The step S23 is to improve the theoretical framework of deviation correction, specifically: a. Introducing the wave deviation correction model into the measurement of ocean surface velocity, the wave deviation calculation model is: Among them, u b is the wave deviation, L is the wavelength, x is the spatial position, u(x) is the horizontal velocity component of orbital motion, w(x) is the vertical velocity component of orbital motion, is the hydrodynamic modulation effect, γ(x) is the tilt modulation effect, and θ is the radar incident angle; b. The expressions for the horizontal and vertical velocity components of orbital motion are: u=aωsin(kx-ωt) w=-aωcos(kx-ωt) Among them, a is the amplitude, ω is the angular frequency, and k is the wave number; c. The expression of the fluid dynamics modulation effect is: Among them, ρ x Represents the weight factor of the electromagnetic wave backscattering intensity at each position along the wavelength, ranging from 0 to 1, ρ min , ρ mid and ρ max correspond to the minimum, intermediate, and maximum weights of the electromagnetic backscatter intensity, respectively; d. The expression of tilt modulation effect is: γ(x)=cos(θ±arctan(ζ')) Among them, ζ = asin(kx-ωt) is the wave function, the positive sign is applicable to upwind observations, and the negative sign is applicable to downwind observations; at zero incidence angle, almost all electromagnetic wave energy will contact the surface, thereby minimizing the loss of backscattered signals.

5. The method for correcting sea wave deviation based on wave orbit motion and scattering characteristics analysis according to claim 1, characterized in that: The step 3) comprises the following steps: S31: inputting the generated simulation data into the wave deviation correction model, and using the deep learning algorithm to perform model training to calculate the wave deviation; S32: During the training process, the cross-validation method is used to evaluate the performance of the model to ensure the stability and generalization ability of the model parameters and reduce the risk of overfitting; S33: After the model training is completed, an independent validation data set is used to test the model performance, analyze the model's prediction effect under different sea conditions, and verify its correction accuracy and stability for wave deviations.

6. The method for correcting sea wave deviation based on wave orbit motion and scattering characteristics analysis according to claim 5, characterized in that: The step S31 is specifically as follows: The model constructed by deep learning algorithm for model training has three hidden layers, and each layer has 256, 128, and 64 neurons respectively. The Adam optimizer is used to train the parameters of the wave deviation calculation model that integrates the physical model and deep learning, and Relu is used as the activation function in each layer. The loss function uses the mean square error function. Based on the measured wave deviation data, the Adam optimizer selects the optimal solution of the hydrodynamic modulation weight factor from the random array, and brings 90% of the optimal solution data into the DNNs neural network for training, and 10% of the data is used to verify the training results. After the model training is completed, the wave parameters are input to obtain the corresponding predicted weight factor, and the predicted weight factor is brought into the wave deviation calculation model established above to obtain the final wave deviation.

7. The method for correcting sea wave deviation based on wave orbit motion and scattering characteristics analysis according to claim 1, characterized in that: The step 4) is specifically: S41: The along-track interferometry method is selected to verify the model performance, and the direct relationship between the radial surface velocity and the interference phase is expressed as: Among them, u surf is the radial surface velocity, λ is the wavelength of the electromagnetic wave, V p is the satellite flight speed, φ ATI is the interference phase, B is the baseline length, and θ is the incident angle; S42: According to the radial surface speed u surf Wave deviation u b , and then the radial surface velocity after wave deviation correction by wave deviation calculation model integrating physical model and deep learning is obtained: u surf -u b .

8. The method for correcting sea wave deviation based on wave orbit motion and scattering characteristics analysis according to claim 1, characterized in that: The step 5) is specifically: Experiments were carried out in the test area using an airborne Ka-band along-track interferometric synthetic aperture radar (ATI-SAR) to obtain measured data. The data were compared with the real data of ocean surface velocity provided by a synchronously placed GNSS drifting buoy, and the deviation of the radial surface velocity measured by the uncorrected ATI-SAR was obtained.

9. The method for correcting sea wave deviation based on wave orbit motion and scattering characteristics analysis according to claim 1, characterized in that: The step 6) is specifically: S61: Use the trained wave deviation correction model to correct the measured data of the on-track interferometric synthetic aperture radar ATI-SAR, compare the corrected data with the measured data provided by the GNSS drifting buoy, verify the effectiveness of the model correction, and statistically analyze the improvement magnitude of the radial surface velocity deviation after correction; S62: Statistical analysis of the corrected ocean surface velocity data was performed to calculate the reduction in radial velocity error, and the correction effect of the model was quantitatively evaluated by comparing the differences in the measurement results before and after correction; S63: Verify the validity of the corrected radial velocity data by comparing and analyzing the measured data of the GNSS drifter point by point to ensure that the corrected data is highly consistent and conforms to the actual distribution characteristics of the ocean surface velocity; verify the consistency of the data under different sea conditions to ensure that the corrected data is highly consistent with the actual observed ocean surface velocity in spatial distribution, quantitatively evaluate the accuracy and stability of the corrected data, and finally verify the reliability of the model in the actual ocean environment.

Images