Method for inverting ocean surface flow velocity based on satellite remote sensing data

By using a feedforward neural network combined with multi-factor method in the equatorial sea area, the problem of hypothetical sea water balance and missing features in the prior art is solved, and the accuracy of sea surface flow velocity inversion in the equatorial sea area is significantly improved.

CN119066963AActive Publication Date: 2024-12-03GUANGDONG OCEAN UNIVERSITY
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202411121027.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-12-03
Estimated Expiration
2044-08-15

AI Technical Summary

Technical Problem

The prior art in the inversion of sea surface flow velocity in the equatorial sea area is assumed to be in equilibrium and the lack of important features, resulting in the inversion effect being unsatisfactory and cannot meet the needs of high temporal resolution and multi-factor influence.

Method used

Using a feedforward neural network method, combining satellite remote sensing data, reanalysis data and derived data, including sea surface temperature, salinity, wave elements, wind stress and other factors, to build an equatorial sea surface flow velocity inversion model, and through training data The set is trained to achieve the determination of nonlinear relationships.

Benefits of technology

Improve the accuracy of the inversion of the sea surface in the equatorial sea area, reduce the error of the inversion of the flow velocity, increase the correlation coefficient of the zonal flow velocity by 66%, decrease the root mean square error by 69%, increase the correlation coefficient of the meridian flow velocity by 340%, and the root mean square error Reduced by 52%.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066963A_ABST
    Figure CN119066963A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a method for inverting ocean surface flow velocity based on satellite remote sensing data, which comprises the following steps: acquiring satellite remote sensing data and reanalysis data of an equatorial sea area, and acquiring derivative data of the satellite remote sensing data and the reanalysis data; the satellite remote sensing data, the reanalysis data and the derivative data are input into an equatorial sea surface flow velocity inversion model, the equatorial sea area sea surface flow velocity is obtained and subjected to inverse normalization processing, the final equatorial sea area sea surface flow velocity is obtained, and the equatorial sea surface flow velocity inversion model is constructed based on a feedforward neural network and is obtained through training of a training data set. The training data set comprises historical satellite remote sensing data, historical reanalysis data, historical derivative data and a corresponding field observation flow rate data set. The invention provides a simple and effective high-precision sea surface flow velocity inversion method, and the inversion effect of the equator flow velocity is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method for inverting ocean surface current velocity based on satellite remote sensing data. Background Art

[0002] The ocean surface velocity in the equatorial region is one of the key local ocean variables. The high-precision ocean surface velocity inversion based on satellite remote sensing data is of great significance for climate change prediction, maritime navigation safety, marine ecological and dynamic disaster warning, etc. However, due to the special geographical location of the equator, the latitude is 0 and the Coriolis parameter is 0, the geostrophic balance theory and Ekman balance theory cannot be directly applied. The inversion of the sea surface velocity in the equatorial sea area has always been a major problem in the inversion of ocean data. The current technology for inverting the equatorial sea surface velocity based on satellite remote sensing data is mainly to add the geostrophic current calculated by the β balance after the geostrophic balance is corrected in the equatorial sea area and the wind-driven current calculated by adding the amplitude coefficient after the statistical correction of the Ekman balance.

[0003] In order to verify the accuracy of the sea surface velocity obtained by the current inversion algorithm, the sea surface velocity in the equatorial sea area was inverted based on the current inversion algorithm, AVISO (Archiving, Validation, and Interpretation of Satellite Oceanographic Data), satellite remote sensing sea surface height dataset and ERA5 (ECMWF Reanalysis-5) ocean 10m wind speed dataset, and compared with the TAO (Tropical Atmosphere Ocean) ocean velocity observation dataset in this sea area. The comparison results are shown in the figure below. Figure 2 As shown in (a)-(b), the results show that the effect of velocity inversion based on the algorithm is not ideal, where the correlation coefficient between the inverted zonal velocity and the observed zonal velocity is only 0.38, and the root mean square error can reach 0.97m / s; the correlation coefficient between the inverted meridional velocity and the observed meridional velocity is only 0.1, and the root mean square error reaches 0.44m / s.

[0004] The main problems of current ocean current inversion algorithms are as follows: (1) The formula for inverting geostrophic and wind-driven currents is based on the assumption that the seawater is in equilibrium, ignoring key physical processes such as advection and mixing. It is derived using a simplified equation of seawater motion. This equation is only applicable to sea surface velocity inversion with a coarse temporal resolution of more than one month and a large spatial scale that conforms to geostrophic equilibrium. However, the temporal resolution of current satellite remote sensing data has reached 1 day or even higher. The inversion calculation method based on the equilibrium assumption cannot meet the current inversion needs. (2) In the velocity inversion process, only two factors, sea surface height and sea surface wind speed, are considered. In nature, sea surface velocity is also affected by sea surface temperature, salinity and waves. Ignoring these factors will greatly reduce the accuracy of the inversion results. Summary of the invention

[0005] The purpose of the present invention is to provide a method for inverting ocean surface current velocity based on satellite remote sensing data. The method mainly aims at the problem that the inversion effect of sea surface current velocity in the equatorial sea area is poor due to the assumption that the sea water is in a state of equilibrium and the lack of important characteristics in the traditional algorithm. The sea surface temperature, salinity and wave elements are integrated into the artificial intelligence neural network to determine the nonlinear relationship between the sea surface current velocity and the satellite remote sensing characteristics.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A method for inverting ocean surface current velocity based on satellite remote sensing data, comprising:

[0008] Obtaining satellite remote sensing data and reanalysis data of the equatorial sea area, and obtaining derivative data of the satellite remote sensing data and the reanalysis data;

[0009] The satellite remote sensing data, the reanalysis data and the derived data are input into an equatorial sea surface velocity inversion model to obtain the equatorial sea area sea surface velocity and perform inverse normalization processing to obtain the final equatorial sea area sea surface velocity, wherein the equatorial sea surface velocity inversion model is constructed based on a feedforward neural network and obtained through training with a training data set, and the training data set includes historical satellite remote sensing data, historical reanalysis data, historical derived data and corresponding field observation velocity data sets.

[0010] Optionally, the reanalysis data includes: reanalyzed sea surface temperature, salinity, wave elements, wind stress data, and sea surface density.

[0011] Optionally, the derived data include: the longitude and latitude gradients of the satellite remote sensing data, the longitude and latitude gradients and the wave direction sine and cosine of the reanalysis data.

[0012] Optionally, obtaining the training data set includes:

[0013] Using the spatiotemporal points of the field observed velocity data set as a benchmark, interpolating the historical satellite remote sensing data, the historical reanalysis data, and the historical derivative data;

[0014] The field observed velocity data set is used as an inversion label, and the historical satellite remote sensing data, the historical reanalysis data, the historical derivative data and the corresponding inversion label are used as the training data set, and the training data set is normalized.

[0015] Optionally, constructing the equatorial sea surface current velocity inversion model based on the feedforward neural network includes:

[0016] According to the surface current velocity characteristics of the Pacific Ocean, the preset division longitudes are used to delineate the east and west basins of the Pacific Ocean, and the feedforward neural networks with different structures are used to construct the equatorial sea surface current velocity inversion model according to the division results, wherein a smoothing coefficient is added within a preset range of the preset division longitudes.

[0017] Optionally, the feedforward neural networks with different structures are feedforward neural networks with different numbers of hidden layers and different numbers of neurons in the hidden layers.

[0018] Optionally, training the equatorial sea surface velocity inversion model includes:

[0019] The historical satellite remote sensing data, the historical reanalysis data, and the historical derivative data are used as inputs of the equatorial sea surface velocity inversion model, and the corresponding field observed velocity data are used as outputs. When a preset training stop condition is reached, the equatorial sea surface velocity inversion model training is completed to obtain the trained equatorial sea surface velocity inversion model.

[0020] Optionally, during the training of the equatorial sea surface velocity inversion model, the hidden layer activation function uses an asymmetric Sigmoid function activation, and the output layer uses linear activation.

[0021] The beneficial effects of the present invention are:

[0022] The present invention proposes a method for inverting the surface current velocity in the equatorial sea area using a neural network. This method reasonably avoids the assumption in the traditional algorithm that the seawater is in a state of equilibrium, avoids the large error in the velocity inversion caused by the unreasonable approximation of the fluid motion equation in the equatorial sea area, and effectively improves the inversion accuracy of the sea surface current velocity in the equatorial sea area.

[0023] The present invention constructs an inversion method for surface current velocity in the equatorial sea area with multi-factor feature input, and uses a reasonable and effective combination of multi-satellite remote sensing parameters to make up for the deficiency of using only two parameters, sea surface height and wind stress, in traditional algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 A flow chart of a method for inverting ocean surface current velocity based on satellite remote sensing data according to an embodiment of the present invention;

[0026] Figure 2 The figures are the effect diagrams of inverting latitudinal and meridional flow velocities by the conventional algorithm and the method according to the embodiment of the present invention, wherein (a) is the effect diagram of inverting latitudinal flow velocity by the conventional algorithm, (b) is the effect diagram of inverting meridional flow velocity by the conventional algorithm, (c) is the effect diagram of inverting latitudinal flow velocity by the method, and (d) is the effect diagram of inverting meridional flow velocity by the method;

[0027] Figure 3 The time-longitude diagram of the latitudinal inversion velocity and latitudinal wind stress of an embodiment of the present invention, wherein (a) is a diagram showing the latitudinal velocity in the equatorial sea area inverted by the method and its changes after 150 days of low-pass filtering, and (b) is a diagram showing the latitudinal wind stress and its changes after 150 days of low-pass filtering;

[0028] Figure 4 This is a 13-month sliding mean zonal wind stress effect diagram of an embodiment of the present invention;

[0029] Figure 5 The time-longitude diagram of the meridional inversion velocity and meridional wind stress of the embodiment of the present invention, wherein (a) is the meridional velocity of the equatorial sea area inverted by the method and the change diagram after 150 days of low-pass filtering, and (b) is the meridional wind stress and the change diagram after 150 days of low-pass filtering;

[0030] Figure 6 This is a 13-month sliding mean meridional wind stress effect diagram of an embodiment of the present invention;

[0031] Figure 7 This is a schematic diagram of the 5-month moving average Nino3.4 index of an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] like Figure 1 As shown, this embodiment provides a method for inverting ocean surface current velocity based on satellite remote sensing data, including:

[0035] Obtain satellite remote sensing data and reanalysis data of the equatorial sea area, and obtain derivative data of the satellite remote sensing data and reanalysis data, wherein the satellite remote sensing data is sea surface height data;

[0036] The satellite remote sensing data, reanalysis data and derived data are input into the equatorial sea surface velocity inversion model to obtain the sea surface velocity in the equatorial sea area and perform inverse normalization processing to obtain the final sea surface velocity in the equatorial sea area. The equatorial sea surface velocity inversion model is constructed based on a feedforward neural network and obtained through training data sets. The training data sets include historical satellite remote sensing data, historical reanalysis data, historical derived data and corresponding field observation velocity data sets.

[0037] Specifically, a feedforward neural network is selected as the basic framework of the inversion algorithm, with interpolated satellite remote sensing data and reanalysis data and their derivative data as input features, and the corresponding velocity labels as output, forming the basic structure of the equatorial sea surface velocity inversion algorithm.

[0038] Specifically, the inverse normalization process includes: inverse normalizing the obtained normalized equatorial sea surface velocity, inverting the equatorial sea surface velocity to obtain the true latitudinal and longitudinal velocities, that is, converting the dimensionless normalized velocity into a dimensional sea surface velocity (in m / s). The specific method of inverse normalization is: in is the equatorial velocity vector, u is the zonal velocity, v is the meridional velocity, i is the index of the data point, is the unit vector in the longitude direction, is the unit vector in the latitude direction; is the normalized dimensionless equatorial velocity vector, is the maximum flow rate, is the minimum flow rate.

[0039] Furthermore, the reanalysis data include: reanalyzed sea surface temperature, salinity, wave elements, wind stress, and sea surface density data.

[0040] Specifically, the surface density of seawater is calculated as follows:

[0041]

[0042] Among them, ρ(S, T, P) is the density of seawater under salinity S, temperature T and pressure P, ρ(S, T, P) is the density of seawater under standard atmospheric pressure (P = 0), and K(S, T, P) is the secant bulk modulus of seawater. The density of seawater can be obtained using temperature and salinity data.

[0043] Furthermore, the derived data include: longitude and latitude gradients of satellite remote sensing data, longitude and latitude gradients and wave direction sine and cosine of reanalysis data.

[0044] Specifically, the derivative data of both satellite remote sensing data and reanalysis data are calculated. The derivative data mainly include longitude and latitude gradients and wave direction sine and cosine. The longitude and latitude gradients are calculated as follows:

[0045]

[0046] Among them, f is a physical quantity, is the latitudinal gradient, is the meridional gradient, is the unit vector in the longitude direction, is a unit vector in the latitude direction, x is longitude, y is latitude, is the Hamiltonian operator.

[0047] The wave direction sine and cosine are calculated as follows:

[0048] wd x =cos(wd) (3)

[0049] wd y = sin(wd) (4)

[0050] Among them, wd x and wd y are the latitudinal and longitudinal components of the wave direction respectively, cos and sin are the cosine and sine functions respectively, and wd is the wave direction.

[0051] Furthermore, constructing a training data set includes: taking the spatiotemporal points of the field observed velocity data set as a benchmark, interpolating historical satellite remote sensing data, historical reanalysis data, and historical derived data; taking the field observed velocity data set as an inversion label, obtaining a training data set corresponding to the historical satellite remote sensing data, historical reanalysis data, and historical derived data and the inversion label, and normalizing the training data set.

[0052] Specifically, we construct a multi-source large-area data set of equatorial sea area and a sea surface field observation velocity data set. We collect historical satellite remote sensing data of the Pacific equatorial sea area and historical reanalyzed sea surface temperature, salinity, wave elements, wind stress and sea surface density data to form a multi-source large-area data set and sea surface observation velocity (TAO) data. After quality control of the data set, a large data set is formed. The reanalyzed data is obtained by downloading from the Copernicus Data Center. The quality control process of the data set includes the removal of meaningless extreme values ​​of field observed velocity and the removal of abnormal instrument status observation data.

[0053] Then, based on the spatiotemporal points of the field observed velocity dataset, the historical satellite remote sensing data, historical reanalysis data and their historical derivative data are interpolated. The field observed velocity is used as the inversion label, and the historical satellite remote sensing data, historical reanalysis data and their historical derivative data are used as features to form a mapping dataset with one-to-one correspondence between features and labels.

[0054] All variables of the mapping data set are normalized separately, and the normalization method is: Among them, X i is a dimensionless physical quantity before normalization, i is the index of the data point, nin(X i ) is the minimum value of a dimensioned physical quantity, max(X i ) is the maximum value of the dimensionless physical quantity, X norm The dimensionless physical quantity is normalized to [-1,1] and redivided according to the location of the observation site. Finally, the mapping data sets of different sites are divided into 7:3, of which 70% of the data is used for training and 30% of the data is used to test the algorithm performance.

[0055] Furthermore, the equatorial sea surface velocity inversion model based on the feedforward neural network includes:

[0056] According to the characteristics of the Pacific Ocean surface current velocity, the preset division longitudes are used to delineate the East and West Basins of the Pacific Ocean. According to the division results, feedforward neural networks with different structures are used to construct the equatorial sea surface current velocity inversion model, wherein a smoothing coefficient is added within the preset range of the preset division longitudes.

[0057] Specifically, the basin division and optimization are carried out according to the surface current velocity characteristics of the Pacific Ocean, and the east and west basins of the Pacific Ocean are delineated. Different basins are trained using different algorithms. For example, in this embodiment, the basins are divided into the east and west basins at 165°E for training, and a smoothing coefficient is added to the preset range of longitude for dividing the eastern basin and the western basin, so that the flow velocity can smoothly transition within the preset range of the dividing longitude without generating singular values.

[0058] The specific smoothing process is: multiply the flow velocity by the smoothing coefficient. For example, smoothing from 165°E to 5° east means multiplying the flow velocity at 165°E by the smoothing coefficient a, and then adding the flow velocity of one data resolution unit east of 165°E multiplied by the smoothing coefficient b, where a+b=1. At 165°E, a=1, and gradually decreases to 0 as smoothing moves eastward.

[0059] Furthermore, feedforward neural networks with different structures are feedforward neural networks with different numbers of hidden layers and different numbers of neurons in the hidden layers.

[0060] Specifically, the structure of the adjustment algorithm includes: selecting different numbers of hidden layers (the number of hidden layers ranges from 1 to 3) and different numbers of neurons in the hidden layers (the number of neurons ranges from 1 to 10), training algorithms with different structures, and during training, the hidden layer activation function uses an asymmetric Sigmoid function activation, and the output layer uses a linear activation. The initial guess parameters are determined using the Levenberg-Marquardt method. In order to avoid overfitting of the algorithm, the following exit conditions are used: (1) The root mean square error no longer decreases after 20 consecutive optimizations; (2) The root mean square error reaches 5×10 -4 ; (3) The number of iterations exceeds 1000. The training results are brought into the test data set to test the effects of different structural algorithms, and the structure and parameters with the best effect are selected as the equatorial sea surface velocity inversion algorithm, where the best effect is the feedforward neural network corresponding to the maximum correlation coefficient. The algorithm structures of different sea basins may be different, but the input data are the same, that is, the inputs are satellite remote sensing data and reanalysis data and their derivative data. Finally, one sea basin area corresponds to one equatorial sea surface velocity inversion model.

[0061] Based on the above method, this embodiment establishes and verifies the effect of the intelligent inversion algorithm for surface current velocity in the equatorial sea area. The effect of inverting latitudinal and meridional current velocities by the traditional algorithm and the method of this embodiment is as follows: Figure 2 As shown in (a)-(d), the horizontal axis is the TAO observation data, the vertical axis is the inversion data, and the color-filled graph is the normalized probability density. The larger the value, the more distributed data there is in the range. The black dotted straight line is the perfect inversion line, that is, the inversion value is completely equal to the observation value. The closer the distribution of the inverted and observed flow velocities is to this straight line, the better the inversion effect. Figure 2 (a)-(d) show that, compared with the traditional algorithm, the inversion algorithm of this embodiment effectively improves the inversion effect of the equatorial current velocity, wherein the correlation coefficient of the inverted zonal current velocity is increased from 0.38 of the traditional algorithm to 0.63, an increase of 66%, and the root mean square error is reduced from 0.97 m / s to 0.3 m / s, with an error reduction of 69%; the correlation coefficient of the inverted meridional current velocity is increased from 0.1 of the traditional algorithm to 0.44, an increase of 340%, and the root mean square error is reduced from 0.44 m / s to 0.21 m / s, with an error reduction of 52%.

[0062] In order to further verify the inversion effect of this embodiment on the equatorial sea surface velocity, the inverted equatorial sea surface velocity is applied to the seasonal and interannual variation analysis of the equatorial current velocity. The analysis results are as follows: Figure 3-Figure 7 As shown, Figure 3 (a)- Figure 3 (b) and Figure 5 (a)- Figure 5 The black contour lines in (b) are the result of low-pass filtering of the current velocity for 150 days, indicating the annual variation of the inverted current. Figure 7 The light-colored part is less than -0.5℃, which means La Niña occurs, and the dark-colored part is greater than 0.5℃, which means El Niño occurs. The results show that: (1) For the zonal velocity, the positive velocity anomaly generated in the sea area east of 170°W will propagate westward at a speed of about 0.88m / s, which is comparable to the magnitude of the wind-generated Rossby wave. In El Niño years, the Kelvin wave excited by the westerly wind burst in the East China Sea Basin will propagate eastward to about 140°W at a speed of 2.63m / s. (2) For the meridional velocity, its seasonal variation in the East China Sea Basin is stronger, and the seasonal variation becomes stronger in La Niña years and weakens in El Niño years. In the western basin, the positive velocity anomaly propagates eastward as far as about 160°E, which is consistent with the seasonal outbreak of southerly winds. In the eastern basin, the velocity anomaly propagates westward at a speed of about 0.84m / s. This westward propagation is associated with the southerly wind anomaly and weakens in El Niño years.

[0063] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for inverting ocean surface current velocity based on satellite remote sensing data, characterized in that: include: Obtaining satellite remote sensing data and reanalysis data of the equatorial sea area, and obtaining derivative data of the satellite remote sensing data and the reanalysis data; The satellite remote sensing data, the reanalysis data and the derived data are input into an equatorial sea surface velocity inversion model to obtain the equatorial sea area sea surface velocity and perform inverse normalization processing to obtain the final equatorial sea area sea surface velocity, wherein the equatorial sea surface velocity inversion model is constructed based on a feedforward neural network and obtained through training with a training data set, and the training data set includes historical satellite remote sensing data, historical reanalysis data, historical derived data and corresponding field observation velocity data sets.

2. The method for inverting ocean surface current velocity based on satellite remote sensing data according to claim 1, characterized in that: The reanalysis data include: reanalyzed sea surface temperature, salinity, wave elements, wind stress data, and sea surface density.

3. The method for inverting ocean surface current velocity based on satellite remote sensing data according to claim 1, characterized in that: The derived data include: the longitude and latitude gradients of the satellite remote sensing data, the longitude and latitude gradients and the wave direction sine and cosine of the reanalysis data.

4. The method for inverting ocean surface current velocity based on satellite remote sensing data according to claim 1, characterized in that: Acquiring the training data set includes: Using the spatiotemporal points of the field observed velocity data set as a benchmark, interpolating the historical satellite remote sensing data, the historical reanalysis data, and the historical derivative data; The field observed velocity data set is used as an inversion label, and the historical satellite remote sensing data, the historical reanalysis data, the historical derivative data and the corresponding inversion label are used as the training data set, and the training data set is normalized.

5. The method for inverting ocean surface current velocity based on satellite remote sensing data according to claim 1, characterized in that: Constructing the equatorial sea surface velocity inversion model based on the feedforward neural network includes: According to the surface current velocity characteristics of the Pacific Ocean, the preset division longitudes are used to delineate the east and west basins of the Pacific Ocean, and the feedforward neural networks with different structures are used to construct the equatorial sea surface current velocity inversion model according to the division results, wherein a smoothing coefficient is added within a preset range of the preset division longitudes.

6. The method for inverting ocean surface current velocity based on satellite remote sensing data according to claim 2, characterized in that: The feedforward neural networks with different structures are feedforward neural networks with different numbers of hidden layers and different numbers of neurons in the hidden layers.

7. The method for inverting ocean surface current velocity based on satellite remote sensing data according to claim 1, characterized in that: Training the equatorial sea surface velocity inversion model includes: The historical satellite remote sensing data, the historical reanalysis data, and the historical derivative data are used as inputs of the equatorial sea surface velocity inversion model, and the corresponding field observed velocity data are used as outputs. When a preset training stop condition is reached, the equatorial sea surface velocity inversion model training is completed to obtain the trained equatorial sea surface velocity inversion model.

8. The method for inverting ocean surface current velocity based on satellite remote sensing data according to claim 7, characterized in that: In the process of training the equatorial sea surface velocity inversion model, the hidden layer activation function uses an asymmetric Sigmoid function activation, and the output layer uses linear activation.

Citation Information

Patent Citations

  • Sea surface flow velocity inversion method based on synthetic aperture radar

    CN112162282A

  • Method for inverting ocean three-dimensional temperature and salt field by using satellite remote sensing data

    CN115712807A

  • Deep learning-based equatorial Indian ocean surface ocean current velocity prediction method

    CN117271979A

  • Global underwater temperature-salinity profile intelligent reconstruction method based on multi-source satellite remote sensing data

    CN117763451A

  • Global marine environment forecasting method based on multilevel feature aggregation

    CN117849903A