Intelligent inversion method and system for sea surface cyclone wind speed of sar image

Abstract

This invention provides an intelligent inversion method and system for sea surface cyclone wind speed from SAR images, comprising: acquiring the backscattering coefficient, incident angle, and radar line of sight of the SAR image under test; acquiring ECMWF wind direction data spatiotemporally matched with the SAR image under test; calculating the relative wind direction; normalizing the relevant parameters of the SAR image under test; inputting the normalized parameters into an artificial intelligence model for sea surface cyclone wind speed inversion to obtain the sea surface cyclone wind speed; and denormalizing the inverted sea surface cyclone wind speed. The model is an SVR model trained using the normalized backscattering coefficient, incident angle, and relative wind direction parameters of the sample SAR image as input and ECMWF wind speed data spatiotemporally synchronized with the sample SAR image as labels. The SVR model of this invention fully utilizes the relationship between SAR image relevant parameters and wind speed for learning. The model is simple and fast, and can perform high-precision inversion of sea surface wind speed under cyclone conditions.

Description

Technical Field

[0001] This invention relates to the fields of Synthetic Aperture Radar (SAR) signal processing and marine remote sensing applications, specifically to a method and system for intelligent inversion of sea surface cyclone wind speed from SAR images. Background Technology

[0002] Cyclones, accompanied by extreme waves and heavy rainfall, often bring significant marine hazards to coastal waters. Due to thick cloud cover, extreme sea states, and extremely high risks, in-situ observation of cyclone wind fields is very difficult. Spaceborne satellites carrying microwave sensors, such as scatterometers, microwave radiometers, and altimeters, can be used for near-real-time monitoring of cyclone wind fields. However, these remote sensing products cannot meet the requirements for studying the fine-scale sea surface wind fields of cyclones.

[0003] Synthetic Aperture Radar (SAR) has advantages such as high spatial resolution, large spatial coverage, and all-weather and all-day operation, and can be well used for the study of wind speed in sea surface cyclones.

[0004] Geophysical Model Function (GMF) is commonly used to retrieve sea surface wind fields from SAR images. However, under cyclonic conditions, due to the saturation of drag coefficients, aerodynamic roughness, and sea surface roughness, the same-polarization NRCS (Noise, Radiation, and Cross-Screen) no longer increases and saturates as wind speed exceeds a certain threshold. This results in the current GMF, applicable to same-polarization SAR images, being inaccurate for retrieving high-wind-speed wind fields. Studies on cross-polarization SAR images show that cross-polarization signals have the ability to track wave breaking under moderate to strong wind conditions, exhibit a high signal-to-noise ratio, and are sensitive to sea surface wind speeds under cyclonic conditions. This overcomes the limitation of wind field saturation in SAR image retrieval. At wind speeds above 20 m / s, wind speeds retrieved from cross-polarization SAR images show significantly better performance than those retrieved from same-polarization images.

[0005] Due to the complexity of ocean motion and the limitations of theoretical models for sea surface electromagnetic scattering, the imaging mechanism of SAR for sea surface wind fields is quite complex. Acquiring SAR images from the sea surface is a complex nonlinear process. Furthermore, the GMF (Gross Wind Velocity) requires empirical or statistical determination of multiple nonlinear functions and sub-functions, making GMF adjustment for SAR images challenging. Machine learning methods are a valuable tool for handling nonlinear problems. However, training machine learning models effectively requires a large amount of high-quality data; currently, the availability of sea surface cyclone SAR data is limited, resulting in scarce research on sea surface cyclone wind speed inversion from SAR images.

[0006] Patent document CN111639746B (application number 202010406706.3) discloses a GNSS-R sea surface wind speed inversion method and system based on a CNN neural network. This invention's CNN neural network fully utilizes wind-speed-related physical quantities in the DDM image for feature learning, reducing computational load and shortening processing time while maintaining inversion accuracy. It features a simple model, speed, and high result accuracy. However, this method is based on GNSS-R for wind field inversion and is not suitable for SAR image wind field inversion.

[0007] Patent document CN111639747B (application number CN202010406708.2) discloses a GNSS based on a BP neural network. R. Sea Surface Wind Speed ​​Inversion Method and System. This invention utilizes a BP neural network to invert GNSS. The R-type sea surface wind speed model is simple, shortening modeling and inversion time and further improving inversion accuracy. However, this method, based on GNSS-R for wind field inversion, is not suitable for wind field inversion from SAR images.

[0008] Patent document CN113534159B (application number: CN202110775266.3) discloses a method and system for inverting sea surface wind speed based on SAR satellite remote sensing data. This method can achieve high-precision inversion of sea surface wind speed under swell conditions, effectively improving the accuracy of the inversion results. However, this method is not suitable for inverting wind fields from SAR images under high-wind-speed cyclone conditions.

[0009] Patent document CN119886224A (application number: CN202510365161.9) discloses a method for inverting tropical cyclone wind speeds based on physically constrained deep learning from a spaceborne microwave radiometer. This method utilizes the trajectory information of tropical cyclones to obtain brightness temperature data from a spaceborne microwave radiometer, and combines D-matrix equations with neural networks to construct a deep learning model incorporating physical information to obtain more accurate and precise tropical cyclone wind speed information. However, this method relies on spaceborne microwave radiometer data and does not involve cyclone wind speed inversion based on SAR data.

[0010] Patent document CN120470752A (application number: CN202510501012.0) discloses a method and system for retrieving sea surface wind speed based on CYGNSS data and the Willoughby model. This invention utilizes GNSS... R technology improves the reliability of sea surface wind speed monitoring under extreme weather conditions through the collaborative optimization of physical models and measured data. However, the model constructed by this method is a general extreme wind speed model and is not suitable for cyclone wind speed inversion based on SAR data.

[0011] The above analysis shows that existing geophysical function models for sea surface cyclone wind speed inversion are based only on fitting a small amount of SAR data and lack generality. Therefore, SAR sea surface cyclone wind speed inversion exhibits significant errors, and the inversion accuracy is low under cyclone conditions. Summary of the Invention

[0012] To address the shortcomings of existing technologies, the purpose of this invention is to provide an intelligent inversion method and system for sea surface cyclone wind speed in SAR images.

[0013] The intelligent inversion method for sea surface cyclone wind speed in SAR images provided by the present invention includes: Step S1: Obtain the normalized backscattering coefficient, incident angle, and radar line of sight of the SAR image under test; Step S2: Acquire ECMWF reanalysis wind direction data that is spatiotemporally matched with the SAR image to be tested; Step S3: Calculate the relative wind direction based on the radar line of sight and the ECMWF reanalysis wind direction data; Step S4: Normalize the normalized backscattering coefficient, incident angle, and relative wind direction to obtain normalized parameters; Step S5: Input the normalized parameters into the pre-trained AI model for sea surface cyclone wind speed inversion, and the AI ​​model outputs the normalized wind speed. Step S6: Perform denormalization on the normalized wind speed to obtain the wind speed of the sea surface cyclone.

[0014] Preferably, step S2 includes: matching SAR image pixels using ECMWF reanalysis wind direction data with timestamps and geographic coordinates, and performing spatiotemporal matching based on the ECMWF reanalysis vector wind field. Perform wind direction calculation:

[0015] Where u and v are the components of the ECMWF reanalysis vector wind field; Step S3 includes: according to Calculate relative wind direction ,in The ECMWF wind direction is relative to due north. This indicates the line of sight of the SAR satellite relative to true north.

[0016] Preferably, step S4 includes: normalizing the input parameters, that is:

[0017] in, It is the input data. and These are the maximum and minimum values ​​of the input data, respectively. These are normalized input values.

[0018] Preferably, step S5 includes: Step S5.1: The artificial intelligence model for inverting sea surface cyclone wind speed in SAR images is converted into an SVR model. The SVR input is established as follows:

[0019] Each sample It is a normalized SAR eigenvector; Step S5.2: Match ECMWF wind speed, select the normalized SAR image related parameters and the spatiotemporally synchronized ECMWF wind speed data to obtain the original sample set. Each sample set consists of SAR related parameters and ECMWF corresponding wind speed. Step S5.3: Forming the training set , The target wind speed is divided into training and test sets in an 8:2 ratio. Step S5.4: Train the SVR model using the training set, using a kernel function. The i-th input data Mapped into a high-dimensional space, where x is the input feature vector of the sample to be predicted; Step S5.5: Objective function of SVR for:

[0020] in, It is a weight vector, T is the transpose sign, and b is the bias term; So that under a given error tolerance Inside, training target value Compared with the predicted value Minimize the error between them; Step S5.6: This is achieved by solving the following optimization problem:

[0021] in, It is a slack variable. is the regularization parameter, and N is the total number of input data; Step S5.7: Solve the above optimization problem to obtain the model's weight vector. and bias terms .

[0022] Preferably, step S6 includes: inverse normalizing the retrieved sea surface cyclone wind speed, i.e.:

[0023] in, It is cyclone wind speed. and These are the maximum and minimum wind speeds in the training set, respectively. It is the normalized wind speed output from the sea surface cyclone wind speed inversion model.

[0024] The intelligent inversion system for sea surface cyclone wind speed based on SAR images provided by the present invention includes: Module M1: Acquires the normalized backscattering coefficient, incident angle, and radar line of sight of the SAR image under test; Module M2: Acquires ECMWF reanalysis wind direction data that is spatiotemporally matched with the SAR image under test; Module M3: Calculates the relative wind direction based on the radar line of sight and the ECMWF reanalysis wind direction data; Module M4: Normalizes the normalized backscattering coefficient, incident angle, and relative wind direction to obtain normalized parameters; Module M5: Inputs normalized parameters into a pre-trained AI model for sea surface cyclone wind speed inversion, and the AI ​​model outputs normalized wind speed. Module M6: Performs denormalization on the normalized wind speed to obtain the wind speed of the sea cyclone.

[0025] Preferably, module M2 includes: matching SAR image pixels with ECMWF reanalysis wind direction data using timestamps and geographic coordinates, and performing spatiotemporal matching based on the ECMWF reanalysis vector wind field. Perform wind direction calculation:

[0026] Where u and v are the components of the ECMWF reanalysis vector wind field; The module M3 includes: according to Calculate relative wind direction ,in The ECMWF wind direction is relative to due north. This indicates the line of sight of the SAR satellite relative to true north.

[0027] Preferably, the module M4 includes: normalizing the input parameters, that is:

[0028] in, It is the input data. and These are the maximum and minimum values ​​of the input data, respectively. These are normalized input values.

[0029] Preferably, the module M5 includes: Module M5.1: The artificial intelligence model for inverting sea surface cyclone wind speed in SAR images is an SVR model. The SVR input is established as follows:

[0030] Each sample It is a normalized SAR eigenvector; Module M5.2: Match ECMWF wind speed, select normalized SAR image related parameters and spatiotemporally synchronized ECMWF wind speed data to obtain the original sample set, each sample set consists of SAR related parameters and ECMWF corresponding wind speed; Module M5.3: Forming the training set , The target wind speed is divided into training and test sets in an 8:2 ratio. Module M5.4: Trains the SVR model using a training set and kernel functions. The i-th input data Mapped into a high-dimensional space, where x is the input feature vector of the sample to be predicted; Module M5.5: Objective function of SVR for:

[0031] in, It is a weight vector, T is the transpose sign, and b is the bias term; So that under a given error tolerance Inside, training target value Compared with the predicted value Minimize the error between them; Module M5.6: Implemented by solving the following optimization problem:

[0032] in, It is a slack variable. is the regularization parameter, and N is the total number of input data; Module M5.7: Solve the above optimization problem to obtain the model's weight vector. and bias terms .

[0033] Preferably, module M6 includes: inverse normalizing the retrieved sea surface cyclone wind speed, i.e.:

[0034] in, It is cyclone wind speed. and These are the maximum and minimum wind speeds in the training set, respectively. It is the normalized wind speed output from the sea surface cyclone wind speed inversion model.

[0035] Compared with the prior art, the present invention has the following beneficial effects: (1) The SVR model provided by this invention makes full use of the relationship between SAR image related parameters (normalized backscattering coefficient, incident angle, relative wind direction) and wind speed for learning. The model is simple and fast and can perform high-precision inversion of sea surface wind speed under cyclone conditions. (2) The present invention uses cross-polarization SAR image signals, which overcomes the limitation that the normalized backscattering coefficient of the same polarization SAR image saturates under high wind speed conditions of cyclones, and makes the inversion performance significantly better than the same polarization method when the wind speed is higher than 20m / s. (3) This invention uses the SVR model in machine learning to process the complex nonlinear problem of sea surface wind field inversion in SAR images, which avoids the challenge of traditional geophysical model functions requiring the determination of multiple nonlinear functions and sub-functions based on experience or statistics, and reduces the difficulty of model adjustment; (4) This invention achieves intelligent inversion of sea surface wind speed under cyclone conditions by spatiotemporally matching SAR image parameters with ECMWF reanalysis wind direction data and calculating relative wind direction, filling the gap of limited related research caused by the limited SAR data of sea surface cyclones in the past. Attached Figure Description

[0036] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating an intelligent inversion method for sea surface cyclone wind speed in SAR images, as proposed in an embodiment of the present invention. Figure 2 This is an SVR model structure diagram of an intelligent inversion method for sea surface cyclone wind speed in SAR images proposed in an embodiment of the present invention; Figure 3 This is a Sentinel-1 SAR image of a selected cyclone proposed in an embodiment of the present invention; Figure 4 This is the intelligent inversion result of SAR image sea surface cyclone wind speed proposed in one embodiment of the present invention. Detailed Implementation

[0037] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0038] Example 1 like Figure 1 As shown, this invention provides an intelligent inversion method for sea surface cyclone wind speed from SAR images, comprising: Step S1: Obtain the normalized backscattering coefficient, incident angle, and radar line of sight of the SAR image under test, such as... Figure 3 As shown; Step S2: Acquire ECMWF reanalysis wind direction data that is spatiotemporally matched with the SAR image to be tested; Step S3: Calculate the relative wind direction; Step S4: Normalize the relevant parameters of the SAR image to be tested; Step S5: Input the standardized relevant parameters of the SAR image to be tested into the artificial intelligence model for sea surface cyclone wind speed inversion to obtain the corresponding sea surface cyclone wind speed; Step S6: Inverse normalize the sea surface cyclone wind speed obtained from the inversion.

[0039] Preferably, step S1 includes: acquiring SAR images via RD, CS, The data can be obtained through SAR imaging algorithms or downloaded from publicly available data websites. The normalized backscattering coefficient, incident angle, and radar line of sight of the SAR image under test are then acquired.

[0040] Preferably, step S2 includes: matching ECMWF reanalysis wind direction data with timestamps and geographic coordinates to SAR image pixels. This involves spatiotemporally matched ECMWF reanalysis vector wind fields. Then, wind direction calculation:

[0041] Preferably, step S3 includes: according to Calculate relative wind direction ,in The ECMWF wind direction is relative to due north. This indicates the line of sight of the SAR satellite relative to true north.

[0042] Preferably, step S4 includes: normalizing the input parameters, that is:

[0043] in It is the input data. and These are the maximum and minimum values ​​of the input data, respectively. These are normalized input values.

[0044] Preferably, step S5 includes: Step S5.1: The artificial intelligence model for inverting sea surface cyclone wind speed in SAR images is the SVR model, such as... Figure 2 As shown, establish the SVR input:

[0045] Each sample It is a normalized SAR eigenvector.

[0046] Step S5.2: Match ECMWF wind speed, select the normalized SAR image related parameters and the spatiotemporally synchronized ECMWF wind speed data to obtain the original sample set. Each sample set consists of SAR related parameters and ECMWF corresponding wind speed. Step S5.3: Forming the training set , The target wind speed is divided into training and test sets in an 8:2 ratio. Step S5.4: Train the SVR model using the training set, using a kernel function. Input features Mapped to a higher-dimensional space; Step S5.5: Objective function of SVR for:

[0047] So that under a given error tolerance Inside, training target value Compared with the predicted value Minimize the error between them.

[0048] Step S5.6: This is achieved by solving the following optimization problem:

[0049] in, It is the model's weight vector. It is a slack variable. It is the regularization parameter.

[0050] Step S5.7: Solve the above optimization problem to obtain the model's weight vector. and bias .

[0051] Preferably, step S6 includes: inverse normalizing the retrieved sea surface cyclone wind speed, i.e.:

[0052] in, It is cyclone wind speed. and These are the maximum and minimum wind speeds in the training set, respectively. It is the normalized wind speed output from the sea surface cyclone wind speed inversion model, such as Figure 4 As shown.

[0053] Example 2 This invention also provides an intelligent inversion system for sea surface cyclone wind speed from SAR images, comprising: Module M1: Obtain the normalized backscattering coefficient, incident angle, and radar line of sight of the SAR image under test; Module M2: Acquires ECMWF reanalysis wind direction data that is spatiotemporally matched with the SAR image under test; Module M3: Calculates relative wind direction; Module M4: Normalizes the relevant parameters of the SAR image under test; Module M5: Inputs the standardized relevant parameters of the SAR image to be tested into the artificial intelligence model for sea surface cyclone wind speed inversion to obtain the corresponding sea surface cyclone wind speed; Module M6: Normalizes the sea surface cyclone wind speed obtained from the inversion.

[0054] Preferably, the module M1 includes: SAR image acquisition via RD, CS, The data can be obtained through SAR imaging algorithms or downloaded from publicly available data websites. The normalized backscattering coefficient, incident angle, and radar line of sight of the SAR image under test are then acquired.

[0055] Preferably, module M2 includes: matching ECMWF reanalysis wind direction data with timestamps and geographic coordinates to SAR image pixels. This involves spatiotemporally matched ECMWF reanalysis vector wind fields. Then, wind direction calculation:

[0056] Preferably, the module M3 includes: according to Calculate relative wind direction ,in The ECMWF wind direction is relative to due north. This indicates the line of sight of the SAR satellite relative to true north.

[0057] Preferably, the module M4 includes: normalizing the input parameters, that is:

[0058] in, It is the input data. and These are the maximum and minimum values ​​of the input data, respectively. These are normalized input values.

[0059] Preferably, the module M5 includes: Module M5.1: The artificial intelligence model for inverting sea surface cyclone wind speed in SAR images is an SVR model. The SVR input is established as follows:

[0060] Each sample It is a normalized SAR eigenvector.

[0061] Module M5.2: Match ECMWF wind speed, select normalized SAR image related parameters and spatiotemporally synchronized ECMWF wind speed data to obtain the original sample set, each sample set consists of SAR related parameters and ECMWF corresponding wind speed; Module M5.3: Forming the training set , The target wind speed is divided into training and test sets in an 8:2 ratio. Module M5.4: Trains the SVR model using a training set and kernel functions. Input features Mapped to a higher-dimensional space; Module M5.5: Objective function of SVR for:

[0062] So that under a given error tolerance Inside, training target value Compared with the predicted value Minimize the error between them; Module M5.6: Implemented by solving the following optimization problem:

[0063] in, It is the model's weight vector. It is a slack variable. It is a regularization parameter; Module M5.7: Solve the above optimization problem to obtain the model's weight vector. and bias .

[0064] Preferably, module M6 includes: inverse normalizing the retrieved sea surface cyclone wind speed, i.e.:

[0065] in, It is cyclone wind speed. and These are the maximum and minimum wind speeds in the training set, respectively. It is the normalized wind speed output from the sea surface cyclone wind speed inversion model.

[0066] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0067] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for intelligent inversion of sea surface cyclone wind speed from SAR images, characterized in that, include: Step S1: Obtain the normalized backscattering coefficient, incident angle, and radar line of sight of the SAR image under test; Step S2: Acquire ECMWF reanalysis wind direction data that is spatiotemporally matched with the SAR image to be tested; Step S3: Calculate the relative wind direction based on the radar line of sight and the ECMWF reanalysis wind direction data; Step S4: Normalize the normalized backscattering coefficient, incident angle, and relative wind direction to obtain normalized parameters; Step S5: Input the normalized parameters into the pre-trained AI model for sea surface cyclone wind speed inversion, and the AI ​​model outputs the normalized wind speed. Step S6: Perform denormalization on the normalized wind speed to obtain the wind speed of the sea surface cyclone.

2. The intelligent inversion method for sea surface cyclone wind speed in SAR images according to claim 1, characterized in that, Step S2 includes: matching SAR image pixels using ECMWF reanalysis wind direction data with timestamps and geographic coordinates, and performing spatiotemporal matching based on the ECMWF reanalysis vector wind field. Perform wind direction calculation: Where u and v are the components of the ECMWF reanalysis vector wind field; Step S3 includes: according to Calculate relative wind direction ,in The ECMWF wind direction is relative to due north. This indicates the line of sight of the SAR satellite relative to true north.

3. The intelligent inversion method for sea surface cyclone wind speed in SAR images according to claim 1, characterized in that, Step S4 includes: normalizing the input parameters, that is: in, It is the input data. and These are the maximum and minimum values ​​of the input data, respectively. These are normalized input values.

4. The intelligent inversion method for sea surface cyclone wind speed in SAR images according to claim 3, characterized in that, Step S5 includes: Step S5.1: The artificial intelligence model for inverting sea surface cyclone wind speed in SAR images is converted into an SVR model. The SVR input is established as follows: Each sample It is a normalized SAR eigenvector; Step S5.2: Match ECMWF wind speed, select the normalized SAR image related parameters and the spatiotemporally synchronized ECMWF wind speed data to obtain the original sample set. Each sample set consists of SAR related parameters and ECMWF corresponding wind speed. Step S5.3: Forming the training set , The target wind speed is divided into training and test sets in an 8:2 ratio. Step S5.4: Train the SVR model using the training set, using a kernel function. The i-th input data Mapped into a high-dimensional space, where x is the input feature vector of the sample to be predicted; Step S5.5: Objective function of SVR for: in, It is a weight vector, T is the transpose sign, and b is the bias term; So that under a given error tolerance Inside, training target value Compared with the predicted value Minimize the error between them; Step S5.6: This is achieved by solving the following optimization problem: in, It is a slack variable. is the regularization parameter, and N is the total number of input data; Step S5.7: Solve the above optimization problem to obtain the model's weight vector. and bias terms .

5. The intelligent inversion method for sea surface cyclone wind speed in SAR images according to claim 1, characterized in that, Step S6 includes: inverse normalizing the retrieved sea surface cyclone wind speed, i.e.: in, It is cyclone wind speed. and These are the maximum and minimum wind speeds in the training set, respectively. It is the normalized wind speed output from the sea surface cyclone wind speed inversion model.

6. A smart inversion system for sea surface cyclone wind speed from SAR images, characterized in that, include: Module M1: Acquires the normalized backscattering coefficient, incident angle, and radar line of sight of the SAR image under test; Module M2: Acquires ECMWF reanalysis wind direction data that is spatiotemporally matched with the SAR image under test; Module M3: Calculates the relative wind direction based on the radar line of sight and the ECMWF reanalysis wind direction data; Module M4: Normalizes the normalized backscattering coefficient, incident angle, and relative wind direction to obtain normalized parameters; Module M5: Inputs normalized parameters into a pre-trained AI model for sea surface cyclone wind speed inversion, and the AI ​​model outputs normalized wind speed. Module M6: Performs denormalization on the normalized wind speed to obtain the wind speed of the sea cyclone.

7. The SAR image sea surface cyclone wind speed intelligent inversion system according to claim 6, characterized in that, The module M2 includes: matching SAR image pixels using ECMWF reanalysis wind direction data with timestamps and geographic coordinates, and performing spatiotemporal matching of ECMWF reanalysis vector wind fields. Perform wind direction calculation: Where u and v are the components of the ECMWF reanalysis vector wind field; The module M3 includes: according to Calculate relative wind direction ,in The ECMWF wind direction is relative to due north. This indicates the line of sight of the SAR satellite relative to true north.

8. The SAR image sea surface cyclone wind speed intelligent inversion system according to claim 6, characterized in that, The module M4 includes: normalizing the input parameters, that is: in, It is the input data. and These are the maximum and minimum values ​​of the input data, respectively. These are normalized input values.

9. The intelligent inversion system for sea surface cyclone wind speed in SAR images according to claim 8, characterized in that, The module M5 includes: Module M5.1: The artificial intelligence model for inverting sea surface cyclone wind speed in SAR images is an SVR model. The SVR input is established as follows: Each sample It is a normalized SAR eigenvector; Module M5.2: Match ECMWF wind speed, select normalized SAR image related parameters and spatiotemporally synchronized ECMWF wind speed data to obtain the original sample set, each sample set consists of SAR related parameters and ECMWF corresponding wind speed; Module M5.3: Forming the training set , The target wind speed is divided into training and test sets in an 8:2 ratio. Module M5.4: Trains the SVR model using a training set and kernel functions. The i-th input data Mapped into a high-dimensional space, where x is the input feature vector of the sample to be predicted; Module M5.5: Objective function of SVR for: in, It is a weight vector, T is the transpose sign, and b is the bias term; So that under a given error tolerance Inside, training target value Compared with the predicted value Minimize the error between them; Module M5.6: Implemented by solving the following optimization problem: in, It is a slack variable. is the regularization parameter, and N is the total number of input data; Module M5.7: Solve the above optimization problem to obtain the model's weight vector. and bias terms .

10. The intelligent inversion system for sea surface cyclone wind speed in SAR images according to claim 6, characterized in that, The module M6 includes: inverse normalizing the retrieved sea surface cyclone wind speed, i.e.: in, It is cyclone wind speed. and These are the maximum and minimum wind speeds in the training set, respectively. It is the normalized wind speed output from the sea surface cyclone wind speed inversion model.

Images