Sea surface wind field prediction method

Abstract

The present invention relates to a method for predicting sea surface wind field, comprising: obtaining current wind field information of a target sea area by inverting a joint multi-beam GNSS-R and a multi-polarization SAR, wherein the current wind field information includes wind speed information and wind direction information; obtaining wind field correlation information of the target sea area, and determining the wind field of an adjacent area of ​​the target sea area according to the wind field correlation information and the current wind field information. The wind field inversion method of the present invention has high precision, high calculation efficiency, high resolution, and can predict the wind field of an adjacent sea area, and has high timeliness.

Description

[0001] This application is a divisional application with the application date of June 18, 2020, application number 202010561649.6, and the invention name: Multi-beam GNSS-R system, sea surface wind field inversion method and prediction method. Technical Field

[0002] The present invention relates to the field of radar technology, and in particular to a method for predicting sea surface wind fields. Background Art

[0003] When detecting ships on the sea, the target intensity and sea clutter intensity are affected by the ocean environment and observation mode respectively. To improve the detection probability of the target, the space-based intelligent radar can adjust the radar observation mode according to the current environment, increase the target intensity, and reduce the sea clutter intensity. Among them, an important part of obtaining the ocean environment is to obtain the sea surface wind field. Due to the characteristics of ship targets and the operation mode of space-based radars, in target detection, the sea surface wind field data is required to have high accuracy, high resolution and good real-time performance.

[0004] The existing methods for acquiring sea surface wind field data include field measurement methods such as buoys and survey ships, as well as microwave remote sensing methods such as scatterometers, altimeters, radiometers, GNSS-R (Global Navigation Satellite System-Reflectometry) and SAR (Synthetic Aperture Radar). However, field measurement methods have few sampling points, high maintenance costs, and are difficult to use on a large scale. Scatterometers, altimeters, radiometers and GNSS-R cannot meet the needs of high-resolution wind field data. Conventional co-polarization SAR wind field inversion requires the use of external data such as numerical forecasts to obtain wind direction, which is difficult to achieve real-time. The signal-to-noise ratio of sea surface cross-polarization SAR is lower than that of co-polarization SAR. Using this model to invert wind speed will result in large deviations. Moreover, under low wind speed conditions, the radar backscatter cross section is not only related to wind speed, but also to other factors. The acquired data has low accuracy and is difficult to serve for real-time acquisition of accurate wind field information near ship targets. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-beam GNSS-R system, a sea surface wind field inversion method and a prediction method to solve the problems of low resolution, poor real-time performance and low accuracy in the sea surface wind field inversion method in the prior art.

[0006] In order to achieve the above-mentioned object of the invention, the technical solution of the present invention is:

[0007] A first aspect of an embodiment of the present invention provides a multi-beam GNSS-R system, comprising: a receiving antenna array, a sampling quantization module, a multi-beam forming module and a position estimation module;

[0008] The receiving antenna array is used to receive sea surface reflection signals within a preset width range;

[0009] The sampling and quantization module is used to perform low-noise amplification, filtering and sampling and quantization on the sea surface reflection signal in sequence to obtain a multi-channel digital domain GNSS-R signal;

[0010] The position estimation module is used to receive the direct signal of the GNSS-R satellite and determine the reflection area position corresponding to the sea surface reflection signal according to the direct signal;

[0011] The multi-beam forming module is used to obtain sea surface wind speed information of multiple subspaces according to the reflection area position and the multi-channel digital domain GNSS-R signal.

[0012] Furthermore, the receiving antenna array includes a plurality of antenna array elements, and the array element spacing is half the wavelength of the GNSS-R signal.

[0013] Furthermore, the multi-beam forming module is specifically used for:

[0014] Based on a beamforming algorithm, a plurality of narrow beams pointing to the reflection area position are generated according to the reflection area position and the multi-channel digital domain GNSS-R signal;

[0015] Acquire a reflection signal at the position of the reflection area received by the receiving antenna array;

[0016] Based on the geophysical model, the sea surface wind speed information of multiple subspaces is inverted according to the reflection signals at the reflection area positions.

[0017] A second aspect of an embodiment of the present invention provides a method for inverting a sea surface wind field, comprising:

[0018] Acquire sea surface wind speed information of multiple subspaces of a target sea area detected by a multi-beam GNSS-R system, and determine spatially continuous wind speed information of the target sea area based on the multiple subspaces of the sea surface wind speed information;

[0019] Establish a VH polarization wind speed inversion model;

[0020] Acquire a VH polarization image and a VV polarization image of the target sea area scanned by a multi-polarization SAR, and update the VH polarization wind speed inversion model according to the spatial continuous wind speed information and the VH polarization image to obtain the wind speed information of the target sea area;

[0021] The wind direction information of the target sea area is determined according to the wind speed information and the VV polarization image.

[0022] Further, determining the spatial continuous wind speed information of the target sea area according to the plurality of sea surface wind speed information includes:

[0023] Counting wind field correlation information between multiple sub-areas, and determining correlation factors corresponding to a preset number of directions one by one according to the wind field correlation information;

[0024] Calculating the full-space sea surface wind speed information in each direction according to the preset number of correlation factors and the sea surface wind speed information of the multiple sub-areas;

[0025] A full-space wind speed model within the target sea area is established according to the full-space sea surface wind speed information in each direction, and the spatial continuous wind speed information is determined according to the full-space wind speed model.

[0026] Further, updating the VH polarization wind speed inversion model according to the spatial continuous wind speed information and the VH polarization image to obtain the wind speed information of the target sea area includes:

[0027] Obtaining a normalized backscattering cross-sectional area according to the VH polarization image;

[0028] Establishing a linear function of the spatially continuous wind speed information and the normalized backscattering cross-sectional area;

[0029] Solving the linear function to obtain model parameters corresponding to the plurality of wind speed ranges;

[0030] The VH polarization wind speed inversion model of the corresponding wind speed range is updated according to the model parameters to obtain the wind speed information of the target sea area.

[0031] Further, after the VH polarization wind speed inversion model corresponding to the wind speed range is updated according to the model parameters, it also includes:

[0032] Determining whether the wind speed segmentation points of adjacent wind speed ranges are equal;

[0033] If they are equal, the wind speed information of the target sea area is obtained according to the VH polarization wind speed inversion model;

[0034] If they are not equal, the wind speed information of the target sea area closest to the spatially continuous wind speed information is selected.

[0035] Further, determining the wind direction information of the target sea area according to the wind speed information and the VV polarization image includes:

[0036] Based on the geophysical model, determining the fuzzy wind direction of the target sea area according to the wind speed information and the VV polarization image;

[0037] The wind direction information of the target sea area is determined according to the polarization correlation coefficient and the fuzzy wind direction.

[0038] A third aspect of an embodiment of the present invention provides a method for predicting a sea surface wind field, comprising:

[0039] Obtain the current wind field information of the target sea area;

[0040] The wind field correlation information of the target sea area is obtained, and the wind field in the adjacent area of ​​the target sea area is determined according to the wind field correlation information and the current wind field information.

[0041] Further, obtaining wind field correlation information of the target sea area, and determining a wind field in an adjacent area of ​​the target sea area according to the wind field correlation information and the current wind field information, includes:

[0042] Acquire historical wind field information of the target sea area, and obtain wind field correlation information of the target sea area according to the historical wind field information;

[0043] Based on a long short-term memory unit recurrent neural network, the wind field in the adjacent area of ​​the target sea area is determined according to the wind field correlation information and the current wind field information.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] According to the scheme of the present invention, the multi-beam GNSS-R system can obtain high-precision sea surface wind speed information of multiple subspaces; then the present invention provides a large-width sea surface wind field rapid inversion and prediction method combining multi-beam GNSS-R and multi-polarization SAR, and performs continuous sea surface wind speed inversion through sea surface wind speed information of multiple subspaces to obtain full-space sea surface wind speed information matching the sea surface observation range of the multi-polarization SAR, with high wind speed inversion accuracy and high calculation efficiency; then the full-space wind speed information is combined with the VH polarization image to optimize the VH polarization SAR wind speed inversion model, and the wind direction is inverted according to the VV polarization image to obtain high-precision and high-resolution sea surface wind field information; further combined with the wind field correlation information, the wind field in the adjacent area is predicted based on the target sea area. Compared with the traditional sea surface wind field inversion of traditional scatterometers, SAR, GNSS-R, etc., the present invention has the advantages of high timeliness, high resolution, and high precision, and does not require external information input, and can independently provide support for the real-time information acquisition and prediction of future sea surface wind fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] 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 are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0047] Figure 1 A schematic diagram showing a structure of a multi-beam GNSS-R system provided by an embodiment of the present invention;

[0048] Figure 2 A schematic diagram schematically shows wind field detection using a multi-beam GNSS-R system provided by an embodiment of the present invention;

[0049] Figure 3 A flowchart schematically showing an implementation of a sea surface wind field inversion method provided by an embodiment of the present invention;

[0050] Figure 4 Schematic representation Figure 3 Specific implementation flow chart of step S301;

[0051] Figure 5 Schematic representation Figure 3 Specific implementation flow chart of step S303;

[0052] Figure 6 Schematic representation Figure 3 Specific implementation flow chart of step S304;

[0053] Figure 7 A flowchart schematically showing an implementation of a method for predicting sea surface wind fields provided by an embodiment of the present invention;

[0054] Figure 8 A schematic diagram schematically showing correlation factors in eight directions provided by an embodiment of the present invention;

[0055] Fig. 9 A schematic diagram schematically showing the distribution of resolution units and known wind field points in the reflection area provided by an embodiment of the present invention;

[0056] Fig.10 A schematic diagram schematically showing the calculation of sea surface wind speed information at each point in direction 1 provided by an embodiment of the present invention;

[0057] Fig.11 A schematic diagram schematically showing the sea surface wind speed information of each point in the calculation direction 2 provided by an embodiment of the present invention;

[0058] Fig.12 A schematic diagram schematically showing the fusion of spatial continuous wind speed information and VH polarization image provided by an embodiment of the present invention;

[0059] Fig.13 A schematic diagram schematically illustrates the corresponding relationship between the wind speed range provided by an embodiment of the present invention and the VH polarization wind speed inversion model. DETAILED DESCRIPTION

[0060] The description of the embodiments of this specification should be combined with the corresponding drawings, which should be considered as part of the complete specification. In the drawings, the shape or thickness of the embodiments may be enlarged and indicated for simplification or convenience. Furthermore, the parts of each structure in the drawings will be described separately. It is worth noting that the elements not shown in the drawings or not described in words are in a form known to ordinary technicians in the relevant technical field.

[0061] The description of the embodiments herein and any reference to directions and orientations are only for the convenience of description and are not to be construed as any limitation on the scope of protection of the present invention. The following description of the preferred embodiments may involve combinations of features, which may exist independently or in combination, and the present invention is not particularly limited to the preferred embodiments. The scope of the present invention is defined by the claims.

[0062] See also Figure 1 This embodiment provides a multi-beam GNSS-R system, which mainly includes: a receiving antenna array 10, a sampling quantization module 20, a multi-beam forming module 30 and a position estimation module 40. The receiving antenna array 10, the sampling quantization module 20 and the multi-beam forming module 30 are connected in sequence, the position estimation module 40 is connected to the sampling quantization module 20, and the receiving antenna array 10 can also be connected to the multi-beam forming module 30.

[0063] Since GNSS-R does not require a transmitter, but uses the globally shared GNSS constellation as a multi-source microwave signal reflection source, and has low power consumption and cost, the wind speed inversion method is simple and mature, has strong real-time performance, and the inversion result has high accuracy, so this embodiment adopts a multi-beam GNSS-R system.

[0064] Specifically, the receiving antenna array 10 receives multi-channel sea surface reflection signals within a preset width range and sends them to the sampling and quantization module 20; the sampling and quantization module 20 performs low-noise amplification, filtering and sampling and quantization on the multi-channel sea surface reflection signals in sequence to obtain multi-channel digital domain GNSS-R signals, and sends the multi-channel digital domain GNSS-R signals to the multi-beam forming module 30; the position estimation module 40 receives the direct signal of the GNSS-R satellite, and determines the reflection area position corresponding to the sea surface reflection signal according to the direct signal; the multi-beam forming module 30 obtains the sea surface wind speed information of multiple subspaces according to the reflection area position and the multi-channel digital domain GNSS-R signals.

[0065] Optionally, the receiving antenna array 10 of this embodiment includes a plurality of antenna array elements 101, the spacing between adjacent antenna array elements 101 is half the wavelength of the GNSS-R signal, and the plurality of antenna array elements 101 receive sea surface reflection signals in a wide range in parallel. It should be understood that this embodiment does not limit the number of antenna array elements 101. Figure 1 , the number of antenna array elements 101 may be N×M, where N and M are both positive integers greater than 0.

[0066] For example, see Figure 1 and Figure 2 The satellite-borne multi-beam GNSS-R system receives and samples quantizes GNSS-R signals in parallel within a wide bandwidth through a receiving antenna array 10 with N×M antenna elements, performs multi-beam forming in the digital domain, restores the subspace sea surface scattering information corresponding to each narrow beam, and inverts the sea surface wind speed information of each subspace according to the corresponding geophysical model.

[0067] Furthermore, the multi-beam forming module 30 is specifically used to: based on the beam forming algorithm, generate multiple narrow beams pointing to the reflection area position according to the reflection area position and the multi-channel digital domain GNSS-R signal; obtain the reflection signal of the reflection area position received by the receiving antenna array; based on the geophysical model, invert the sea surface wind speed information of multiple subspaces according to the reflection signal of the reflection area position.

[0068] Specifically, the multi-beam GNSS-R system receives GNSS-R signals within a wide bandwidth in parallel through a receiving antenna array 10 having N×M antenna elements, and the sampling and quantization module 20 performs low-noise amplification, bandpass filtering and sampling and quantization on the N×M GNSS-R signals in sequence to obtain N×M digital domain GNSS-R signals; then the position estimation module 40 receives the direct signal of the GNSS satellite, extracts the synchronization signal, estimates the corresponding satellite position, and calculates the corresponding sea surface reflection area position, guiding the multi-beam forming module 30 to perform high-gain narrow beam synthesis on multiple reflection areas; the multi-beam forming module 30 generates K narrow beams pointing to the reflection area position according to the estimated reflection area positions of K GNSS satellites and the N×M digital domain GNSS-R signals through a digital beam forming method, thereby restoring the sea surface reflection signals of the K reflection areas, and based on the geophysical model, inverts the sea surface wind speed information of multiple subspaces according to the reflection signals of the regional positions.

[0069] Furthermore, since the reflection area and number formed by GNSS satellites are limited, in order to collect as many GNSS-R sea surface reflection signals as possible, the multi-beam GNSS-R system of this embodiment can simultaneously receive sea surface reflection signals within and outside the detection area, and invert the sea surface wind speed in the corresponding area, such as Figure 2As shown, more sea surface wind speed data are provided for the subsequent full-space sea surface wind speed inversion in the detection area.

[0070] The multi-beam GNSS-R system of the above-mentioned embodiment realizes wide-width and multi-space sea surface wind speed detection, and inverts the spatial high-precision wind speed. Compared with the traditional remote sensing wind field detection technology, this technology has the advantages of high timeliness and high precision, and does not require external information input. It can independently provide support for the near-real-time information acquisition and prediction of future sea surface wind fields.

[0071] This embodiment also provides a method for inverting the sea surface wind field. Figure 3 , is a schematic diagram of an implementation flow of a method for inverting a sea surface wind field according to this embodiment, which is described in detail as follows:

[0072] Step S301, obtaining sea surface wind speed information of multiple subspaces of a target sea area detected by a multi-beam GNSS-R system, and determining spatially continuous wind speed information of the target sea area based on the multiple subspaces of the sea surface wind speed information.

[0073] In one embodiment, see Figure 4 The specific implementation process of determining the spatial continuous wind speed information of the target sea area according to the plurality of sea surface wind speed information in step S301 includes:

[0074] Step S401 : collecting statistics on wind field correlation information between multiple sub-areas, and determining correlation factors corresponding to a preset number of directions one by one according to the wind field correlation information.

[0075] Step S402: Calculate the full-space sea surface wind speed information in each direction according to the preset number of correlation factors and the sea surface wind speed information of the multiple sub-areas.

[0076] Step S403, establishing a full-space wind speed model in the target sea area according to the full-space sea surface wind speed information in each direction, and determining the spatial continuous wind speed information according to the full-space wind speed model.

[0077] This embodiment statistically analyzes the spatial correlation of the sea surface wind field based on historical sea surface wind field information to obtain a set of correlation factors related to the direction and with a certain range of values; based on the obtained correlation factors, the wind speed information of multiple subspaces known in the target sea area is calculated along each direction to obtain continuous full-space sea surface wind speed information of the entire detection area (target sea area); then a mathematical model of the full-space wind speed in the detection area is established, and finally, the distribution of the full-space wind speed in the entire detection area is obtained through least squares optimization.

[0078] For example, by statistically analyzing the spatial correlation of the sea surface wind field, a correlation factor R corresponding to each direction is obtained. k, where k = 1, 2, ..., 8, see Figure 8 Here, 8 directions with an angle of 45° are selected, that is, the preset number of directions is 8. The relationship between the wind speeds of two adjacent resolution units (the wind speed information of two adjacent full-space sea surfaces) can be expressed as v i+p,j+q =R k v ij , where v ij is the wind speed of the i-th row and j-th column resolution unit, i = 1, 2, ... I, j = 1, 2, ... J, see Fig. 9 , there are I×J resolution units in the entire target sea area, p and q∈(-1,0,1).

[0079] Optionally, based on the correlation analysis of a large amount of wind field data, the value range of the correlation factor in the target sea area is: 0.8 <R k <1.2, spatially continuous wind speed information can be obtained.

[0080] Furthermore, according to the obtained correlation factor R k As well as the sea surface wind speed information of multiple sub-areas randomly distributed and known in the target sea area, the full-space sea surface wind speed information in all directions in the entire target sea area is obtained. Fig.10 , first along direction 1, using the correlation formula v i,j+1 =R1v ij After one calculation, the wind field of a point on the right side of the known point is obtained. By calculating the obtained points multiple times, the wind fields of all points on the right side of the known point can be obtained. Then the other points are supplemented with random values ​​to obtain the wind field distribution of the entire target sea area in direction 1. Similarly, see Fig.11 , along direction 2, the correlation formula v is used to calculate i+1,j+1 =R2v ij After multiple calculations, the wind field at each point in this direction is obtained, such as Figure 7 , and then fill in the other points with random values ​​to get the wind field distribution of the entire target sea area in direction 2. Similarly, calculations can be made along 8 directions respectively, and finally 8 groups of wind field distribution values ​​in the target sea area are obtained, that is, the full-space sea surface wind speed information in 8 directions.

[0081] Furthermore, a full-space wind speed model in the target sea area is established based on the full-space sea surface wind speed information in each direction, and spatial continuous wind speed information is determined based on the full-space wind speed model to obtain an accurate wind field distribution of the entire target sea area. Exemplarily, 8 sets of full-space sea surface wind speed information are averaged to obtain an average wind field distribution of:

[0082]

[0083] The standard deviation of the 8 wind field distributions is obtained as follows:

[0084]

[0085] By minimizing the sum of the standard deviations of the wind fields at each point in the detection area:

[0086]

[0087] The wind field of each resolution unit in the detection area is optimized, and finally the accurate wind field distribution of the entire target sea area is obtained, that is, spatially continuous wind speed information is obtained.

[0088] Step S302: establishing a VH polarization wind speed inversion model.

[0089] In one embodiment, the specific implementation process of step S302 includes:

[0090] Establish an initial multi-polarization wind speed inversion model; obtain historical wind speed information of the target sea area, and determine applicable factors of the multi-polarization wind speed inversion model in multiple wind speed ranges and historical wind speed information corresponding to the multiple wind speed ranges according to the historical wind speed information; update the multi-polarization wind speed inversion model according to the applicable factors in the multiple wind speed ranges and the historical wind speed information corresponding to the multiple wind speed ranges, and obtain the VH polarization wind speed inversion model corresponding to each wind speed range.

[0091] Specifically, this embodiment first constructs a VH polarization wind speed inversion piecewise function according to the prior information and the scene wind speed range, that is, the initial multi-polarization wind speed inversion model. The linear parameters in the existing multi-polarization wind speed inversion model will change with the SAR working mode, observation scene, wind speed range, etc. Therefore, before using the model for wind speed inversion, this embodiment first accurately estimates the linear parameters in the model. First, based on the statistical analysis of historical data, the applicability of the multi-polarization wind speed inversion model in different wind speed ranges and the wind speed in different wind speed ranges are obtained, and the segmentation method of the VH polarization wind speed inversion model is determined. Specifically, it can be obtained by

[0092]

[0093] The VH polarization wind speed inversion model in different wind speed ranges is obtained Where i = 1, 2, ..., n, n is the total number of wind speed ranges, v1, v2, ..., v n It is the segmentation node of adjacent wind speed range.

[0094] Step S303, obtaining a VH polarization image and a VV polarization image of the target sea area scanned by a multi-polarization SAR, and updating the VH polarization wind speed inversion model according to the spatial continuous wind speed information and the VH polarization image to obtain the wind speed information of the target sea area.

[0095] This embodiment adopts a wind field inversion method combined with a multi-polarization SAR and a multi-beam GNSS-R system. The multi-beam GNSS-R system can be carried on a SAR satellite at the same time. It has low cost, a mature wind speed inversion method, and can obtain wind speed quickly and accurately. Compared with the scatterometer working mode, wind speed inversion is more real-time and can provide SAR observations with real-time and high-precision wind speed information in the same area. The combined SAR data and wind field inversion model can obtain high-precision, high-resolution, and high-timeliness wind field information, and the wind field information includes wind speed information and wind direction information of the target sea area. At the same time, this embodiment can also use the wind field information obtained in the observed area, combined with the spatiotemporal correlation characteristics of the wind field, to predict the wind field of the area to be observed, and can guide the space-based intelligent radar to make intelligent adjustments according to the environment to improve the target detection performance.

[0096] In one embodiment, see Figure 5 The specific implementation process of updating the VH polarization wind speed inversion model according to the spatial continuous wind speed information and the VH polarization image in step S303 to obtain the wind speed information of the target sea area includes:

[0097] Step S501, obtaining a normalized backscattering cross-sectional area according to the VH polarization image.

[0098] Step S502: establishing a linear function of the spatially continuous wind speed information and the normalized backscattering cross-sectional area.

[0099] Step S503, solving the linear function to obtain model parameters corresponding to the multiple wind speed ranges.

[0100] Step S504: updating the VH polarization wind speed inversion model corresponding to the wind speed range according to the model parameters to obtain the wind speed information of the target sea area.

[0101] This embodiment optimizes the VH polarization wind speed inversion model of different wind speed ranges based on the obtained spatial continuous wind speed information and the normalized backscatter cross-sectional area of ​​the SAR image in the corresponding range, and can calculate the high-resolution wind speed. Exemplarily, a linear equation of the normalized backscatter cross-sectional area and the spatial continuous wind speed information is constructed using multiple GNSS-R wind speed points, as shown below:

[0102]

[0103]

[0104] The subscript represents the wind speed segment of the model parameter, for example, the first wind speed segment has n samples and the second wind speed segment has m samples. The model parameters a1, b1, a2, b2, and L of different wind speed segments are solved to obtain a high-precision VH polarization segmented wind speed inversion model for the current working mode and detection range.

[0105] Furthermore, the updated VH polarization wind speed inversion model is used to invert high-resolution wind speed information. Specifically, on the basis of obtaining a high-precision VH polarization segmented wind speed inversion model, the wind speed inversion is performed on the SAR observation scene with a higher resolution than the GNSS-R wind speed, such as using a 5km resolution grid to calculate the normalized backscattering cross-sectional area, bringing it into the VH polarization wind speed inversion model to calculate the wind speed, and obtaining the wind speed information of the target sea area, such as Fig.12 shown.

[0106] Optionally, after step S504, the inversion method of this embodiment may further include:

[0107] Determine whether the wind speed segmentation points of adjacent wind speed ranges are equal; if they are equal, obtain the wind speed information of the target sea area according to the VH polarization wind speed inversion model; if they are not equal, select the wind speed information of the target sea area that is closest to the spatial continuous wind speed information.

[0108] Specifically, this embodiment can also select the wind speed inversion result. Since the linear relationship between the wind speed and the VH polarization wind speed inversion model of the corresponding wind speed range is obtained by segmenting the wind speed, the linear models at the wind speed segmentation points may overlap. Fig.13 , so that there may be multiple corresponding wind speeds in the VH polarization wind speed inversion model corresponding to the same wind speed range. When high-resolution wind speed inversion encounters this problem, this embodiment needs to use GNSS-R to guide the selection and select a high-resolution wind speed value whose inversion value is close to the GNSS-R wind speed to ensure that the inversion value is more accurate.

[0109] Step S304: determining the wind direction information of the target sea area according to the wind speed information and the VV polarization image.

[0110] In one embodiment, see Figure 6 The specific implementation process of step S304 may include:

[0111] Step S601, based on a geophysical model, determining the fuzzy wind direction of the target sea area according to the wind speed information and the VV polarization image.

[0112] Step S602: determining the wind direction information of the target sea area according to the polarization correlation coefficient and the fuzzy wind direction.

[0113] This embodiment obtains fuzzy wind direction through geophysical model CMOD. Specifically, CMOD is a function of normalized radar cross section and three parameters of wind direction, wind speed and radar incident angle. The VV polarization image obtained by multi-polarization SAR and the wind speed information of the target sea area obtained by the above inversion method are brought into the model to obtain fuzzy wind direction. Since the normalized radar cross section of the VV polarization image and the wind direction are approximately cosine function relations, there are multiple values ​​of the wind direction inverted thereby. Then this embodiment presents an odd symmetry law with the change of wind direction according to the polarization correlation coefficient, determines the wind direction information of the target sea area in multiple inverted wind directions, and obtains the complete wind field information of the target sea area in combination with the wind speed information obtained by S303.

[0114] The inversion method in the above embodiment provides a rapid inversion method for a wide sea surface wind field by combining multi-beam GNSS-R and multi-polarization SAR. The sea surface continuous wind speed inversion is performed through the sea surface wind speed information of multiple subspaces, and the full-space sea surface wind speed information matching the sea surface observation range of the multi-polarization SAR is obtained. The wind speed inversion accuracy is high and the calculation efficiency is high. Then, the full-space wind speed information is combined with the VH polarization image to optimize the VH polarization SAR wind speed inversion model, and the wind direction is inverted according to the VV polarization image to obtain high-precision and high-resolution sea surface wind field information.

[0115] This embodiment also provides a method for predicting sea surface wind field. Figure 7 , is a schematic diagram of an implementation flow of a method for predicting a sea surface wind field according to this embodiment, which is described in detail as follows:

[0116] Step S701, obtaining current wind field information of the target sea area.

[0117] Step S702: Obtain wind field correlation information of the target sea area, and determine a wind field in an adjacent area of ​​the target sea area according to the wind field correlation information and the current wind field information.

[0118] Optionally, the specific implementation process of step S702 includes:

[0119] Acquire historical wind field information of the target sea area, and obtain wind field correlation information of the target sea area based on the historical wind field information; based on a long short-term memory unit recurrent neural network, determine the wind field in an adjacent area of ​​the target sea area based on the wind field correlation information and the current wind field information.

[0120] This embodiment can first use the existing numerical forecast data to conduct a large number of statistical studies on the spatial correlation of wind fields in different regions and under different sea conditions, and further adopt the time series prediction method of the long short-term memory unit recurrent neural network (Long Short-Term Memory, LSTM) to perform real-time high-precision prediction of adjacent areas.

[0121] Specifically, a large number of statistical studies are conducted on the spatial correlation of wind fields in different regions and under different sea conditions through historical numerical forecast data, and a wind field spatial correlation database is established. In this embodiment, based on wind field information and its wind field spatial correlation database, a long short-term memory unit recurrent neural network is used to perform real-time, high-precision prediction of adjacent areas.

[0122] The vector of the wind field can be expressed as a sequence in is a two-dimensional coordinate vector, the subscript represents its dimension, the superscript represents the number of sequences, and the distance between every two adjacent detection points (detection areas) is Δd. Considering that when the prediction of the wind field vector after (d-1)·Δd is performed, a more accurate prediction is obtained based on the wind field data at the previous distance n, the input long short-term memory unit model should be:

[0123]

[0124] ...

[0125]

[0126] The corresponding output is An input sequence and a corresponding output sequence constitute a sample. Here, this embodiment designs an LSTM neural network for wind farm prediction, including an input layer, a hidden layer and an output layer. For example, the input layer can have 2 neurons, that is, the dimension of each element in the input sequence is 2n, the hidden layer can be composed of 20 long and short-term memory units, and the output layer can have 2 neurons, that is, the dimension of each element in the output sequence is 2.

[0127] In specific applications, when training the LSTM neural network, all network parameters are initialized to random numbers between 0 and 1, and then optimized using training samples. For each training sample, each element in the sequence is read one by one by the LSTM neural network, and an output vector is obtained after the hidden layer and the output layer, which is then compared with the label, and the error is back-propagated through a time-based back-propagation algorithm, and the model testing process in this embodiment follows the same method as the training process. Then, the wind field correlation information and the current wind field information are input into the trained model, and the wind field in the adjacent area of ​​the target sea area is obtained, with high prediction accuracy, high resolution, and good timeliness.

[0128] The prediction method in the above embodiment combines the high-precision and high-resolution wind field information and wind field correlation information of the target area to predict the wind field in the adjacent area based on the target sea area. Compared with traditional sea surface wind field inversion such as scatterometers, SAR, GNSS-R, etc., the present invention has the advantages of high timeliness, high resolution and high precision, and does not require external information input. It can provide near real-time sea surface wind field information for future on-orbit applications such as space-based radar intelligent detection.

[0129] The serial numbers of the above-mentioned steps involved in the method of the present invention do not mean the order of execution of the method. The execution order of each step should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the implementation mode of the present invention.

[0130] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting sea surface wind fields, comprising: The current wind field information of the target sea area is obtained by combining multi-beam GNSS-R and multi-polarization SAR inversion, wherein the current wind field information includes wind speed information and wind direction information; Acquire wind field correlation information of the target sea area, and determine a wind field in an adjacent area of ​​the target sea area according to the wind field correlation information and the current wind field information; The current wind field information of the target sea area is obtained by inverting the multi-beam GNSS-R and the multi-polarization SAR. The current wind field information includes wind speed information and wind direction information, including: Acquire sea surface wind speed information of multiple subspaces of a target sea area detected by a multi-beam GNSS-R system, and determine spatially continuous wind speed information of the target sea area based on the multiple subspaces of the sea surface wind speed information; Establish a VH polarization wind speed inversion model; Acquire a VH polarization image and a VV polarization image of the target sea area scanned by a multi-polarization SAR, and update the VH polarization wind speed inversion model according to the spatial continuous wind speed information and the VH polarization image to obtain the wind speed information of the target sea area; The wind direction information of the target sea area is determined according to the wind speed information and the VV polarization image.

2. The method according to claim 1, characterized in that The multi-beam GNSS-R system comprises: a receiving antenna array (10), a sampling quantization module (20), a multi-beam forming module (30) and a position estimation module (40). The receiving antenna array (10) is used to receive sea surface reflection signals within a preset width range; The sampling and quantization module (20) is used to sequentially perform low-noise amplification, filtering and sampling and quantization on the sea surface reflection signal to obtain a multi-channel digital domain GNSS-R signal; The position estimation module (40) is used to receive a direct signal from a GNSS-R satellite and determine the position of a reflection area corresponding to the sea surface reflection signal according to the direct signal; The multi-beam forming module (30) is used to obtain sea surface wind speed information of multiple subspaces according to the reflection area position and the multi-channel digital domain GNSS-R signal.

3. The method according to claim 2, characterized in that The receiving antenna array (10) comprises a plurality of antenna array elements (101), and the array element spacing is half the wavelength of a GNSS-R signal.

4. The method according to claim 2, characterized in that: The multi-beam forming module (30) is specifically used for: Based on a beamforming algorithm, a plurality of narrow beams pointing to the reflection area position are generated according to the reflection area position and the multi-channel digital domain GNSS-R signal; Acquire a reflection signal at the position of the reflection area received by the receiving antenna array; Based on the geophysical model, the sea surface wind speed information of multiple subspaces is inverted according to the reflection signals at the reflection area positions.

5. The method according to claim 1, characterized in that Determining the spatial continuous wind speed information of the target sea area according to the plurality of sea surface wind speed information includes: Counting wind field correlation information between multiple sub-areas, and determining correlation factors corresponding to a preset number of directions one by one according to the wind field correlation information; Calculating the full-space sea surface wind speed information in each direction according to the preset number of correlation factors and the sea surface wind speed information of the multiple sub-areas; A full-space wind speed model within the target sea area is established according to the full-space sea surface wind speed information in each direction, and the spatial continuous wind speed information is determined according to the full-space wind speed model.

6. The method according to claim 1, characterized in that Updating the VH polarization wind speed inversion model according to the spatial continuous wind speed information and the VH polarization image to obtain the wind speed information of the target sea area includes: Obtaining a normalized backscattering cross-sectional area according to the VH polarization image; Establishing a linear function of the spatially continuous wind speed information and the normalized backscattering cross-sectional area; Solving the linear function to obtain model parameters corresponding to the plurality of wind speed ranges; The VH polarization wind speed inversion model of the corresponding wind speed range is updated according to the model parameters to obtain the wind speed information of the target sea area.

7. The method according to claim 6, characterized in that After the VH polarization wind speed inversion model corresponding to the wind speed range is updated according to the model parameters, the method further includes: Determining whether the wind speed segmentation points of adjacent wind speed ranges are equal; If they are equal, the wind speed information of the target sea area is obtained according to the VH polarization wind speed inversion model; If they are not equal, the wind speed information of the target sea area closest to the spatially continuous wind speed information is selected.

8. The method according to claim 1, characterized in that Determining wind direction information of the target sea area according to the wind speed information and the VV polarization image includes: Based on the geophysical model, determining the fuzzy wind direction of the target sea area according to the wind speed information and the VV polarization image; The wind direction information of the target sea area is determined according to the polarization correlation coefficient and the fuzzy wind direction.

9. The method according to claim 1, characterized in that: Acquiring wind field correlation information of the target sea area, and determining a wind field in an adjacent area of ​​the target sea area according to the wind field correlation information and the current wind field information, including: Acquire historical wind field information of the target sea area, and obtain wind field correlation information of the target sea area according to the historical wind field information; Based on a long short-term memory unit recurrent neural network, the wind field in the adjacent area of ​​the target sea area is determined according to the wind field correlation information and the current wind field information.

Images