Method for retrieving solitary wave propagation speed in ocean from single optical remote sensing image
By combining the least squares method and support vector machine, the velocity of isolated waves in the ocean is retrieved using single-scene optical remote sensing images, solving the problem of difficulty in obtaining instantaneous velocity in existing technologies and realizing efficient and economical velocity measurement.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- OCEAN UNIV OF CHINA
- Filing Date
- 2022-08-24
- Publication Date
- 2026-06-12
AI Technical Summary
Existing technologies make it difficult to invert the instantaneous velocity of isolated waves in the ocean from a single remote sensing image. Field measurement methods are costly and difficult to apply on a large scale. Multi-time image methods require at least two images and are difficult to obtain remote sensing images at any time, thus failing to obtain instantaneous velocity.
A method combining least squares and support vector machine was used to invert the velocity of isolated waves in the ocean based on single-scene optical remote sensing images. Instantaneous velocities were inverted by constructing a dataset, training a model, and identifying feature parameters.
A method for retrieving instantaneous velocities of isolated waves in the ocean from single-scene remote sensing images has been developed, which improves the convenience and accuracy of velocity measurement, reduces costs, and enables the acquisition of instantaneous velocities over a wide area.
Smart Images
Figure CN115408938B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of marine remote sensing technology, and particularly relates to a method for retrieving the propagation velocity of solitary waves in the ocean from a single scene of optical remote sensing image based on the least squares method and support vector machine. Background Technology
[0002] Internal ocean solitary waves are nonlinear waves generated in stable, stratified oceans, with maximum amplitude occurring within the ocean's interior. They are a common oceanic phenomenon. Internal ocean solitary waves can damage exploration and extraction facilities for offshore oil and gas resources, alter underwater acoustic signal propagation paths, and affect submarine navigation. Furthermore, they influence ocean temperature variations, significantly impacting the marine ecosystem. Therefore, studying the characteristics of internal solitary waves has significant scientific importance and promising application prospects.
[0003] There are many methods for studying internal solitary waves, among which satellite remote sensing imagery is a crucial one. Optical remote sensing, with its advantages of high temporal and spatial resolution and abundant imagery, is frequently used by researchers to study the spatiotemporal distribution and propagation characteristics of internal solitary waves. The characteristic parameters of ocean internal solitary waves mainly include half-width, amplitude, and velocity. The magnitude of these parameters determines the magnitude of the effect of internal solitary waves in ocean mixing; therefore, the inversion of internal solitary wave parameters from remote sensing images has attracted researchers' attention. The velocity of internal solitary waves, as an important propagation characteristic, has been studied by many researchers using various methods. Currently, the common methods for measuring the propagation velocity of internal solitary waves are mainly through in-situ measurements or calculations using remote sensing images. In-situ measurements can be performed by observing internal solitary waves from two or more mooring devices to calculate the propagation velocity; however, this method is difficult, expensive, and large-scale measurements are challenging to implement.
[0004] Remote sensing methods for measuring the velocity of internal solitary waves primarily utilize the multi-time image method. This involves using remote sensing images of the same internal solitary wave taken at different times on the same day, and estimating the propagation velocity using the ratio of propagation path to propagation time. While the multi-time image method is more convenient than field measurements, it requires at least two images to calculate the average velocity, and it is difficult to obtain remote sensing images of a specific sea area at any given time. These are limitations of this method. Therefore, both field measurements and the multi-time image method can only obtain the average velocity of internal solitary waves, not the instantaneous velocity. There is an urgent need for a method to invert the instantaneous velocity of ocean internal solitary waves using a single image, in order to address the problems existing in current techniques for studying ocean internal solitary waves. Summary of the Invention
[0005] To address the problems existing in the prior art, the purpose of this invention is to provide a method for retrieving the instantaneous velocity of internal solitary waves in the ocean from a single scene of optical remote sensing image based on the least squares method and support vector machine, thus providing a new method for the study and acquisition of internal solitary wave velocities.
[0006] To achieve the above-mentioned technical objectives, this invention provides an inversion method for the velocity of isolated waves in the ocean based on the least squares method and support vector machines, comprising the following steps:
[0007] A dataset for model training was constructed by collecting feature parameters and hydrological parameters from single-scene optical remote sensing images.
[0008] Based on the least squares method, a first inversion model is constructed for the first characteristic of the velocity of isolated waves in the ocean.
[0009] Based on support vector machines, a second inversion model is constructed to retrieve the second feature of the velocity of isolated waves in the ocean.
[0010] Extract the first data from the dataset for the first feature, and train the first inversion model.
[0011] Extract the second data from the dataset for the second feature, and train the second inversion model.
[0012] Based on the trained first and second inversion models, the instantaneous velocity of isolated waves in the ocean is inverted by identifying the characteristics of isolated waves in the ocean.
[0013] Preferably, in the process of acquiring a single-scene optical remote sensing image, the evolution process of the internal solitary wave is simulated by the gravity collapse method, the simulated optical remote sensing image and waveform image of the internal solitary wave are acquired, the data parameters used to construct the dataset are acquired, and after dimensionless processing, the dataset is generated.
[0014] Preferably, in the process of dimensionless processing of data parameters, the expression for dimensionless processing is:
[0015]
[0016]
[0017]
[0018]
[0019]
[0020] In the formula, h1 and h are the upper water depth and total water depth of the water tank experiment, respectively; ρ1 and ρ2 are the densities of the upper and lower water layers, respectively; d, G1, and G2 are the brightness-dark spacing, maximum gray value, and minimum gray value of the isolated wave stripes in the simulated remote sensing image, respectively; v is the propagation speed of the internal isolated wave; and c0 is the linear phase velocity.
[0021] Preferably, in the process of constructing the first inversion model, H, Δρ, D, and ΔG of the dimensionless dataset are used as independent variables of the model, and V is used as a dependent parameter. The first inversion model is constructed by fitting the data using the least squares method.
[0022] Preferably, during the fitting process using the least squares method, the first inversion model is constructed using the quadratic and cubic regression equations fitted by the least squares method.
[0023] Preferably, during the fitting of the quadratic regression equation, the quadratic regression equation is expressed as:
[0024] y = 1.84 - 6.17H + 9.08H 2 +4.68Δρ - 51.48Δρ 2 +0.29D -0.02D 2 -0.30ΔG + 0.64ΔG 2 .
[0025] Preferably, during the fitting of the cubic regression equation, the cubic regression equation is expressed as:
[0026] y = 1.71 - 6.84H + 12.26H 2 -4.26H 3 +41.53Δρ - 1896.90Δρ 2 +26826.00Δρ 3 +0.51D -0.11D 2 +0.01D 3 -0.54ΔG +5.40ΔG 2 -24.08ΔG 3 .
[0027] Preferably, in the process of constructing the second inversion model, based on the SVR model, the H, Δρ, D, and ΔG of the dimensionless dataset are used as the model input parameters and V as the output parameter to train the model and construct the second inversion model.
[0028] Preferably, in the process of inverting the velocity of isolated waves in the ocean, the characteristics of isolated waves in the ocean include a first feature and / or a second feature. The first feature and the second feature are feature data of the same attribute but different ranges, and the two feature data may or may not intersect. When they intersect, the feature data corresponding to the maximum ratio of the data in the intersection to each feature data is used as the inversion standard of the intersection data. When they do not intersect, other feature data that exist between the two feature data are deleted, and the two feature data are spliced together before inversion.
[0029] Preferably, the inversion system for implementing the inversion method includes:
[0030] The data acquisition module is used to construct a dataset for model training by collecting feature parameters and hydrological parameters of single-scene optical remote sensing images;
[0031] The first feature inversion module is used to construct a first inversion model for inverting the first feature of the velocity of isolated waves in the ocean based on the least squares method, and to extract the first data of the dataset for the first feature to train the first inversion model.
[0032] The second feature inversion module is used to construct a second inversion model for inverting the second feature of the velocity of isolated waves in the ocean based on a support vector machine, and to extract the second data of the dataset for the second feature to train the second inversion model.
[0033] The internal solitary wave velocity inversion module is used to invert the velocity of internal solitary waves in the ocean by identifying the characteristics of internal solitary waves based on the trained first and second inversion models.
[0034] The present invention discloses the following technical effects:
[0035] This invention obtains a large amount of experimental data by simulating optical remote sensing imaging of internal solitary waves. It establishes a model for retrieving internal solitary wave velocity from single-scene optical remote sensing images using two methods: least squares method and support vector regression. This provides a new method for the study and acquisition of internal solitary wave velocity and promotes the research on the propagation characteristics of internal solitary waves in the ocean. Attached Figure Description
[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0037] Figure 1This is a schematic diagram of the internal solitary wave optical remote sensing simulation platform structure described in this invention;
[0038] Figure 2 This is a schematic diagram of the parameters of the isolated wave in the laboratory according to the present invention, wherein (a) and (b) are simulated optical remote sensing images of the internal isolated wave and their grayscale profiles, and (c) is a waveform diagram of the internal isolated wave;
[0039] Figure 3 This is a scatter plot of the test results of the least squares regression equation inversion described in this invention;
[0040] Figure 4 This is a scatter plot of the test results for the SVR solitary wave velocity inversion model described in this invention.
[0041] Figure 5 This is a schematic diagram of the multi-time image method described in this invention, wherein (a) is a GF4 optical remote sensing image of a certain sea area; and (b) is a MODIS optical remote sensing image of a certain sea area.
[0042] Figure 6 The above is a scatter plot of the accuracy test of the least squares method internal solitary wave velocity inversion model described in this invention. In this plot, (a) shows the measured value and the inversion value after the correction of the constant term of the quadratic regression equation from 1.25 to 1.84. In this plot, (b) shows the measured value and the inversion value after the correction of the constant term of the cubic regression equation from 0.93 to 1.71.
[0043] Figure 7 This is a comparison chart of the accuracy of the internal solitary wave velocity inversion model described in this invention;
[0044] Figure 8 This is a flowchart illustrating the method described in this invention. Detailed Implementation
[0045] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0046] like Figure 1-8As shown, the inversion method for the velocity of isolated waves in the ocean based on the least squares method and support vector machine provided by this invention specifically includes the following process:
[0047] 1. Experimental Design and Data Processing:
[0048] 1.1 Experimental Design:
[0049] The experiment was conducted in a three-dimensional long straight water tank with dimensions of 5m × 0.8m × 0.35m. To simulate stratification in the ocean, based on the principle of fluid dynamics similarity, two layers of water were laid in the tank: the lower layer simulated high-salinity seawater, and the upper layer simulated low-salinity seawater. The long straight water tank was made of glass, allowing observation of the evolution of internal solitary waves within the water body from the side, thereby obtaining parameters such as the half-wave width, amplitude, and velocity of the internal solitary waves. To simulate optical remote sensing imaging, the water surface in the tank was considered as the sea surface. An LED surface light source was placed above one end of the tank, adjusted at an angle to emit white light to simulate sunlight; a CCD1 was placed above the other end of the tank, simulating a visible light sensor from a satellite. The light emitted by the LED light source shone on the water surface, and after reflection, formed a simulated optical remote sensing image of internal solitary waves on the CCD1. A CCD2 was placed on the side of the tank, at the same height as the stratification of the upper and lower water layers, to collect and record the internal solitary wave images and obtain the characteristic parameters of the internal solitary waves. The two CCDs have identical field-of-view boundaries and are connected to a computer for synchronous image acquisition, resulting in simulated optical remote sensing images and waveform images that correspond one-to-one with the internal solitary wave. The optical remote sensing image parameters perfectly match the characteristic parameters of the internal solitary wave. The experimental platform structure is as follows: Figure 1 As shown.
[0050] For oceans with different hydrological parameters and stratification, a series of experiments were designed based on the principle of fluid dynamics similarity, with different water depths, different stratifications, different density differences, and different collapse heights, as shown in Table 1.
[0051] Table 1: Experimental Design for Internal Solitary Wave Optical Remote Sensing Simulation
[0052]
[0053] 1.2 Data Processing:
[0054] Time series plots were created at different locations within the acquired images. The time series plot of the CCD1 acquired images simulated optical remote sensing images, from which remote sensing image feature parameters such as brightness-dark distance and grayscale difference could be extracted. The time series plot of the CCD2 acquired images could extract wave element parameters such as internal solitary wave propagation velocity, half-wave width, and amplitude. When extracting parameters from the same set of experiments, sampling lines corresponding to the surface and side images were selected according to the scale, and time series plots were created using MATLAB, as follows. Figure 2 As shown. Figure 2 (a) is a simulated optical remote sensing image of an internal solitary wave. Figure 2 (b) shows the waveform of the internal solitary wave, from which parameters such as amplitude and half-wave width can be extracted. Figure 2 (c) is a grayscale profile of an internal solitary wave simulated optical remote sensing image, from which remote sensing image parameters such as grayscale difference and brightness-dark distance can be extracted.
[0055] A total of 1440 sets of experimental data were extracted. Each set of data included parameters such as water depth, density, amplitude, velocity, half-wave width, brightness-darkness spacing, and grayscale difference. The 1440 data points were then processed to be dimensionless. The dimensionless parameters are defined as follows:
[0056]
[0057]
[0058]
[0059]
[0060]
[0061] h1 and h represent the upper and total water depths in the water tank experiment, respectively; ρ1 and ρ2 represent the densities of the upper and lower water layers, respectively; d, G1, and G2 represent the brightness spacing, maximum gray value, and minimum gray value of the isolated wave fringes in the simulated remote sensing image, respectively; v represents the propagation velocity of the internal isolated wave; and c0 represents the linear phase velocity. A dimensionless laboratory internal isolated wave data sample library was established, and the data in the sample library was used to train the internal isolated wave velocity inversion model.
[0062] 2. Internal solitary wave velocity inversion model:
[0063] This invention utilizes single-scene optical remote sensing images to invert the velocity of internal solitary waves. Therefore, the input parameters include optical remote sensing image feature parameters related to the amplitude of internal solitary waves: brightness interval and grayscale difference. These two remote sensing image parameters reflect the effect of the internal solitary wave amplitude. Thus, when establishing the internal solitary wave velocity inversion model, water depth, water stratification, density difference, and the brightness interval and grayscale difference of the remote sensing image are used as independent variables. The following sections describe the establishment of internal solitary wave velocity inversion models using both least squares and machine learning techniques.
[0064] 2.1 Least Squares Method for Inverting the Velocity of Internal Solitary Waves:
[0065] This invention utilizes a large amount of experimental data on internal solitary waves, employing the least squares method to find the optimal regression equation and establish an inversion model for internal solitary wave velocity based on the least squares method. H, Δρ, D, and ΔG from a dimensionless database are used as independent variables in the model, and V is used as the dependent parameter. Considering that the four independent variables are mutually independent, they are fitted as independent terms in the regression equation. According to theories such as the KdV equation, velocity has a nonlinear relationship with the independent variables, and the regression equation should include higher-order terms of the independent variables. Therefore, the quadratic and cubic regression equations fitted by the least squares method are as follows:
[0066] y = 1.25 - 6.17H + 9.08H 2 +4.68Δρ - 51.48Δρ 2 +0.29D -0.02D 2 -0.30ΔG + 0.64ΔG 2
[0067] y = 0.93 - 6.84H + 12.26H 2 -4.26H 3 +41.53Δρ - 1896.90Δρ 2 +26826.00Δρ 3 +0.51D -0.11D 2 +0.01D 3 -0.54ΔG +5.40ΔG 2 -24.08ΔG 3
[0068] The results of the regression equation inversion value were tested using experimental data, as follows: Figure 3 As shown, Figure 3 (a) shows the results of the quadratic regression equation. The correlation coefficient between the inverted values and the experimental values is 0.927, the mean absolute error is 1.64 cm / s, and the mean relative error is 8.25%. Figure 3 (b) shows the results of the regression equation with the highest power of 3. The correlation coefficient between the inverted values and the experimental values is 0.928, the mean absolute error is 1.61 cm / s, and the mean relative error is 8.13%. As can be seen from the figure, both regression equations are effective, and their correlation coefficients and error magnitudes are similar and not significantly different.
[0069] 2.2 Support Vector Regression Intra-Solitary Wave Velocity Inversion Model:
[0070] Support Vector Regression (SVR) in Support Vector Machines is a machine learning method applied to function regression prediction. Compared with other machine learning methods, SVR is more suitable for regression prediction with small samples and can handle the case where the sample data is linearly inseparable. Therefore, this invention uses the SVR method to train the internal solitary wave velocity inversion model. A Gaussian radial basis function (RBF) with strong locality and the ability to map the feature space of the training samples from low dimension to infinite dimension is selected as the model kernel function.
[0071] Similar to the least squares method, H, Δρ, D, and ΔG from the dimensionless database are used as input parameters for the model, and V as the output parameter. First, all data is input into the SVR program and run. After removing data with large errors, the remaining 1383 samples are randomly sorted and divided into a training set and a test set in approximately a 4:1 ratio, with 1100 samples in the training set and 283 samples in the test set. Then, the data is input into the SVR program again for model training. After obtaining the model, the values of the penalty coefficient c and the kernel function parameter g are recorded; c is set to 1, and g is set to 16. Figure 4 The scatter plot shows the experimental and inversion data of the test set of the solitary wave velocity inversion model in SVR. The correlation coefficient between the inversion values and the experimental values is 0.888, the mean absolute error is 2.10 cm / s, and the mean relative error is 10.28%, indicating that the model is effective.
[0072] 3. Results and Analysis:
[0073] 3.1 Optical Remote Sensing Image Download and Processing:
[0074] Optical remote sensing images and hydrological data of a specific sea area were selected for model accuracy verification. The required MODIS remote sensing images were downloaded from NASA's official remote sensing image data download website LADDSDAAC, and optical remote sensing images from the Gaofen-4 satellite were downloaded from the Resource Satellite Application Center. The remote sensing images were then screened and processed, parameters were extracted, and the model accuracy was verified.
[0075] The velocity of internal solitary waves was calculated using a multi-time image method. Images from different times of the same day were selected from remote sensing images containing internal solitary waves to locate the same internal solitary wave. The propagation distance was calculated based on the location of the internal solitary wave in different remote sensing images. The time interval between two remote sensing images was taken as the propagation time of the internal solitary wave, and the ratio of propagation distance to propagation time was the propagation velocity. Simultaneously, remote sensing image feature parameters such as the brightness-darkness spacing and grayscale difference of the internal solitary wave were extracted using ENVI software. Figure 5This is a schematic diagram of the multi-time image method. Generally, when using the multi-time image method to calculate the propagation velocity of the inner solitary wave, the time interval between the two images is about 1-3 hours. Since the time interval is relatively short, the calculated average velocity is approximated as the propagation velocity of the inner solitary wave in the two remote sensing images.
[0076] The brightness spacing and grayscale difference of inner solitary wave fringes are extracted from optical remote sensing images. The data is then dimensionless, and the input parameters are substituted into the inversion model to calculate the inner solitary wave velocity. The model accuracy is verified by comparing the calculated inner solitary wave propagation velocity with that obtained using a multi-time image method.
[0077] 3.2 Model accuracy verification:
[0078] When using the least squares internal solitary wave velocity inversion model to invert remote sensing data, the inversion values from both models are smaller than the measured values, indicating a systematic bias. Figure 6 The scatter plots are shown in the middle circle. The difference between the experimental simulation and the real ocean may be the cause of systematic bias during validation. To eliminate systematic bias, the model was corrected to minimize systematic bias and other errors. The corrected model is as follows:
[0079]
[0080]
[0081] Equation (1) is the corrected least squares inversion model of the quadratic regression equation. The constant term is corrected from 1.25 to 1.84. The corrected measured values and inversion values are as follows: Figure 6 (a) The triangle scatter plot is shown. Equation (2) is the corrected least squares inversion model of the 3rd regression equation. The constant term is corrected from 0.93 to 1.71. The corrected measured values and inversion values are shown below. Figure 6 (b) As shown in the triangle scatter plot. From Figure 6 As can be seen, the corrected model eliminates the systematic bias and has a good inversion effect, so the corrected model is used as the final inversion model.
[0082] Following the method in 3.1, 44 internal solitary wave velocity data points from optical remote sensing observations of a certain sea area were obtained. The obtained remote sensing data were divided into four ranges according to water depth: 400-1200 meters, 300-399 meters, 200-299 meters, and 83-199 meters. Three models were used to invert the optical remote sensing data. The inversion results are shown in Table 2 and... Figure 7 As shown.
[0083] Table 2: Accuracy Verification Table of Internal Solitary Wave Velocity Inversion Model
[0084]
[0085] Note: V represents the remotely sensed velocity. Z2 V represents the inversion value of the velocity inversion model of the least squares quadratic regression equation. Z3 V represents the inversion value of the velocity inversion model of the least squares cubic regression equation. SVR This represents the inversion value from the SVR velocity inversion model. AE Z2 V represents Z2 The absolute error of V, AE Z3 V represents Z3 The absolute error of V, AE SVR V represents SVR The absolute error with respect to V.
[0086] Figure 7 This chart compares the mean absolute errors of the three models at different water depths. It shows that the solitary wave velocity inversion model within the SVR has higher accuracy than the two least squares models in the water depth ranges of 400-1200 meters and 83-299 meters. In the water depth range of 300-399 meters, the least squares solitary wave velocity inversion model has the highest accuracy, with the third-order regression equation model exhibiting the smallest error and the highest accuracy. Figure 7 It can also be seen that the accuracy of the internal solitary wave velocity inversion model is relatively small for the least squares quadratic regression equation and the cubic regression equation at various water depths. Therefore, the internal solitary wave velocity inversion equation can be described by the quadratic equation when the accuracy requirements are met.
[0087] This invention establishes an internal solitary wave optical remote sensing simulation platform in the laboratory and conducts a series of simulation experiments. Experimental data are extracted and processed to create a dataset. Internal solitary wave velocity inversion models are established using both least squares and machine learning SVR techniques. Three inversion models are established: a least squares third-order regression equation internal solitary wave velocity inversion model, a least squares second-order regression equation internal solitary wave velocity inversion model, and an SVR internal solitary wave velocity inversion model.
[0088] Validation using optical remote sensing data from different water depths shows that both the least squares internal solitary wave velocity inversion model and the SVR velocity inversion model have high accuracy. The least squares internal solitary wave velocity inversion model provides a regression equation for inversion, making the physical meaning more intuitive; the SVR internal solitary wave velocity inversion model has high accuracy across all water depths. Both models can be applied to the inversion of internal solitary wave velocities in the aforementioned water depths.
Claims
1. A method for retrieving the propagation velocity of isolated waves in the ocean from a single optical remote sensing image, characterized in that, Includes the following steps: By collecting feature parameters and hydrological parameters from single-scene optical remote sensing images, a dataset for model training is constructed. Based on the least squares method, a first inversion model is constructed for the first feature of the velocity of isolated waves in the ocean, and the first data of the dataset for the first feature is extracted to train the first inversion model. Based on support vector machines, a second inversion model is constructed for the second feature of the velocity of isolated waves in the ocean, and second data for the second feature is extracted from the dataset to train the second inversion model. Based on the trained first inversion model and second inversion model, the instantaneous velocity of the isolated wave in the ocean is inverted by identifying the characteristics of the isolated wave in the ocean. In the process of inverting the velocity of isolated waves in the ocean, the characteristics of the isolated waves include the first feature and / or the second feature. The first feature and the second feature are feature data of the same attribute but different ranges, and the two feature data may or may not intersect. When they intersect, the feature data corresponding to the maximum ratio of the data in the intersection to each feature data is used as the inversion standard of the intersection data. When they do not intersect, other feature data that exist between the two feature data are deleted, and the two feature data are spliced together before inversion.
2. The method for retrieving the propagation velocity of isolated waves in the ocean from a single-scene optical remote sensing image according to claim 1, characterized in that: In the process of acquiring a single-scene optical remote sensing image, simulated optical remote sensing images and waveform images of the internal solitary wave are acquired, data parameters for constructing the dataset are acquired, and after dimensionless processing, the dataset is generated.
3. The method for retrieving the propagation velocity of isolated waves in the ocean from a single-scene optical remote sensing image according to claim 2, characterized in that: In the process of dimensionless processing of data parameters, the expression for dimensionless processing is: ； In the formula, h1 and h are the upper water depth and total water depth of the water tank experiment, respectively; ρ1 and ρ2 are the densities of the upper and lower water layers, respectively; d, G1, and G2 are the brightness-dark spacing, maximum gray value, and minimum gray value of the isolated wave stripes in the simulated remote sensing image, respectively; v is the propagation speed of the internal isolated wave; and c0 is the linear phase velocity.
4. The method for retrieving the propagation velocity of isolated waves in the ocean from a single-scene optical remote sensing image according to claim 3, characterized in that: In constructing the first inversion model, the H, Δρ, D, and ΔG of the dimensionless dataset are used as independent variables of the model, and V is used as a dependent parameter. The model is then fitted using the least squares method to construct the first inversion model.
5. The method for retrieving the propagation velocity of isolated waves in the ocean from a single-scene optical remote sensing image according to claim 4, characterized in that: During the fitting process using the least squares method, the first inversion model is constructed by fitting the second-order and third-order regression equations using the least squares method.
6. The method for retrieving the propagation velocity of isolated waves in the ocean from a single-scene optical remote sensing image according to claim 5, characterized in that: During the process of fitting the quadratic regression equation, the quadratic regression equation is expressed as: 。 7. The method for retrieving the propagation velocity of solitary waves in the ocean from a single-scene optical remote sensing image according to claim 6, characterized in that: In the process of fitting the cubic regression equation, the cubic regression equation is expressed as: 。 8. The method for retrieving the propagation velocity of isolated waves in the ocean from a single-scene optical remote sensing image according to claim 7, characterized in that: In the process of constructing the second inversion model, based on the SVR model, the H, Δρ, D, and ΔG of the dimensionless dataset are used as the model input parameters and V as the output parameter to train the model and construct the second inversion model.
9. An inversion system for implementing the method of inverting the propagation velocity of isolated waves in the ocean from a single-scene optical remote sensing image as described in claim 1, characterized in that: include: The data acquisition module is used to construct a dataset for model training by acquiring single-scene optical remote sensing images and hydrological parameters; The first feature inversion module is used to construct a first inversion model for inverting the velocity of isolated waves in the ocean based on the least squares method, and to extract the first data of the dataset for the first feature to train the first inversion model. The second feature inversion module is used to construct a second inversion model for inverting the second feature of the velocity of isolated waves in the ocean based on a support vector machine, and to extract the second data of the dataset for the second feature to train the second inversion model. The internal solitary wave velocity inversion module is used to invert the velocity of the internal solitary wave in the ocean by identifying the characteristics of the internal solitary wave, based on the trained first inversion model and the second inversion model.