SAR soil moisture inversion method based on spatial neighborhood optimization
By using a SAR soil moisture inversion method based on spatial neighborhood optimization, and by processing fully polarimetric synthetic aperture radar images and an improved Alpha approximation model, the single-temporal problem of soil moisture inversion in crop-covered areas was solved. This method enables quantitative inversion of soil moisture in vegetated areas without ground-based measured data, and improves the stability and applicability of the inversion results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA UNIV OF GEOSCIENCES (WUHAN)
- Filing Date
- 2025-12-29
- Publication Date
- 2026-05-12
AI Technical Summary
In crop-covered areas, existing technologies struggle to achieve quantitative inversion of soil moisture in vegetation-covered areas during the rapid growth period of crops, without ground-based measured data, especially since the assumption of short-term invariance of vegetation conditions is difficult to satisfy.
A SAR soil moisture inversion method based on spatial neighborhood optimization is adopted. By preprocessing fully polarimetric synthetic aperture radar images, a scattering model is constructed and the backscattering coefficient is optimized using the minimum residual power criterion. Combined with the improved Alpha approximation model and Dubois model constraints, the dielectric constant is inverted, and finally the soil moisture estimate is obtained by spatial averaging.
It effectively separates the effects of vegetation and surface roughness under single-phase conditions, and realizes quantitative inversion of soil moisture in vegetated areas without ground-based measured data support. The stability of the inversion results is improved, making it suitable for large-scale farmland monitoring.
Smart Images
Figure CN122017769A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of quantitative inversion of polarimetric radar remote sensing, and more specifically, to a SAR soil moisture inversion method based on spatial neighborhood optimization. Background Technology
[0002] In agricultural production, soil moisture directly determines the water absorption efficiency of crop roots and is a key factor in crop yield formation. At the same time, as a core indicator for drought monitoring and a scientific basis for precision irrigation management, it plays an irreplaceable role in ensuring food security and efficient use of water resources.
[0003] Synthetic Aperture Radar (SAR), as an active microwave remote sensing technology, combines penetrating power, all-weather observation, and high resolution. It achieves quantitative inversion by establishing the relationship between the radar backscattering coefficient and the soil dielectric constant. In crop-covered scenarios, radar signals are affected by radar system parameters, vegetation parameters, and surface parameters. How to separate the contribution of vegetation scattering and reduce the influence of surface roughness are current challenges in soil moisture inversion in crop areas.
[0004] Alpha change detection models, as an effective time-series inversion method to remove the influence of vegetation and surface roughness, have demonstrated rapid and efficient advantages in soil moisture retrieval in bare soil and low-vegetation areas. However, in crop-covered areas, especially during the rapid growth period of crops, the assumption of short-term temporal invariance of vegetation conditions is difficult to satisfy. Summary of the Invention
[0005] The purpose of this invention is to provide a SAR soil moisture inversion method based on spatial neighborhood optimization, which can realize quantitative inversion of soil moisture in vegetated areas in a single time phase without the support of ground measurement data.
[0006] This invention provides a SAR soil moisture inversion method based on spatial neighborhood optimization, comprising the following steps: S1: Acquire fully polarimetric synthetic aperture radar images during the crop growth cycle, preprocess the fully polarimetric synthetic aperture radar images, and obtain the observation coherence matrix. S2: Construct a scattering model based on the observed coherence matrix; optimize the solution using the minimum residual power criterion based on the scattering model to obtain the backscattering coefficient of the surface scattering; S3: Construct an improved Alpha approximation model based on the backscattering coefficient, and obtain the surface dielectric constant of each pixel based on the improved Alpha approximation model; S4: Perform water conversion and spatial averaging on the surface dielectric constant of each pixel to obtain the final soil moisture estimate for each pixel.
[0007] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described SAR soil moisture inversion method based on spatial neighborhood optimization.
[0008] Implementing the SAR soil moisture retrieval method based on spatial neighborhood optimization provided by this invention has the following beneficial effects: This invention addresses the limitation of traditional Alpha approximation models that do not adequately consider the coupling effect of vegetation canopy scattering and surface roughness parameters. First, it employs a generalized polarimetric SAR two-component decomposition method to separate and correct for vegetation influence, obtaining a pure surface scattering component and achieving accurate separation of the surface scattering component. Then, it constructs a spatial neighborhood optimization Alpha algorithm, inputting this component into the Alpha approximation model. A 3×3 sliding window is used to traverse the entire image pixel by pixel. Within each window, a system of linear equations is constructed based on the Alpha approximation model, establishing a constrained underdetermined system of equations. By introducing the physical constraint of dielectric constant, the least squares optimization method is used to solve the equation set to obtain the initial estimate of soil moisture for each pixel within the window. In order to generate spatially continuous inversion results, the algorithm designs an accumulation matrix and a counting matrix to aggregate the inversion results of all sliding windows, realize the estimation of soil volumetric water content, and finally output a spatially continuous soil moisture distribution product.
[0009] This invention transforms the temporal assumptions of traditional change detection methods into single-temporal spatial assumptions, eliminating the dependence on time-series data. By converting temporal assumptions into spatial assumptions, it overcomes the reliance of traditional methods on multi-temporal data, supports single-temporal data, and is suitable for the rapid growth period of vegetation. It improves inversion stability by solving underdetermined equations through a sliding window combined with physical constraints. It effectively suppresses the influence of vegetation and roughness by combining polarization decomposition and spatial optimization (joint solution of neighborhood pixels to improve noise resistance). It requires no ground measurement data, is fully automated, and is suitable for large-scale farmland monitoring. It enables quantitative inversion of soil moisture in vegetated areas (such as farmland) in a single temporal phase without the support of ground measurement data. Attached Figure Description
[0010] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a flowchart of the SAR soil moisture inversion method based on spatial neighborhood optimization provided by the present invention; Figure 2 This is a flowchart of the SAR soil moisture inversion method based on spatial neighborhood optimization provided by the present invention; Figure 3 This is a schematic diagram of the research area location provided by the present invention; Figure 4This is a scatter plot of the inversion results for different farmlands provided by the present invention. Detailed Implementation
[0011] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0012] Figure 1 A schematic diagram of the SAR soil moisture inversion method based on spatial neighborhood optimization in this embodiment is shown. In this embodiment, the SAR soil moisture inversion method based on spatial neighborhood optimization includes the following steps: S1: Acquire fully polarimetric synthetic aperture radar images during the crop growth cycle, preprocess the fully polarimetric synthetic aperture radar images, and obtain the observation coherence matrix. In one exemplary embodiment, the preprocessing includes: single-view complex transformation, coherence matrix extraction, polarization filtering, geocoding, and data cropping.
[0013] As an exemplary embodiment, in step S1, the input is the raw fully polarimetric SAR image (such as RADARSAT-2 data), and the main processing operations include: single-view complex transformation (converting the raw data into complex form), extraction of the observation scattering matrix (raw scattering signal), polarimetric filtering (reducing speckle noise), geocoding (correcting geometric deformation and matching geographic coordinates), and study area data cropping (limiting the spatial range of the study area); the output is the preprocessed observation coherence matrix (…). The size is the same as the image, and it is used for polarization decomposition as input for subsequent decomposition.
[0014] S2: Construct a scattering model based on the observed coherence matrix; optimize the solution using the minimum residual power criterion based on the scattering model to obtain the backscattering coefficient of the surface scattering; As an exemplary embodiment, in step S2, for each pixel in the original image, a model-based polarization two-component decomposition framework is established, assuming that dihedral scattering is negligible, and the vegetation-ground signal is decoupled by combining the X-Bragg surface scattering model and the SNVSM volume scattering model, and the backscattering coefficient of the ground surface is obtained by the minimum residual power criterion. As an exemplary embodiment, in step S2, the generalized volume scattering model SNVSM is used to describe the second-order statistical characteristics of the vegetation canopy; for each model, the directional randomness parameter is iteratively adjusted to finally generate the corresponding normalized volume coherence matrix; the volume scattering coefficient is calculated using the non-negative eigenvalue decomposition method. ; As an exemplary embodiment, in step S2, the radar backscattering signal mainly consists of signals from two parts: the surface soil and the vegetation layer, which are simulated by the X-Bragg surface scattering model and the SNVSM volume scattering model, respectively. In one exemplary embodiment, the scattering model is as follows:
[0015] in, Represents the observation coherence matrix; Represents the surface scattering coefficient. This represents the coherence matrix of the X-Bragg surface scattering model; Indicates the volume scattering coefficient; This represents the coherence matrix of the volume scattering model. Represents the residual matrix; In one exemplary embodiment, the minimum residual power criterion is as follows:
[0016] in, Indicates minimization; Represents residual power; Indicates the total scattered power; This represents the Bragg scattering coefficient under H and V polarizations.
[0017] As an exemplary embodiment, in step S2, in order to obtain the optimal solution of parameters, a lookup table is constructed for the directional randomness parameters of the volume scattering model and the total scattering power of the residual matrix; the criterion for finding the optimal volume coherence matrix is to minimize the total power of the residual matrix; the final solution set is obtained, and then the surface scattering coherence matrix is obtained, from which the copolarized backscattering coefficient can be calculated.
[0018] S3: Construct an improved Alpha approximation model based on the backscattering coefficient, and obtain the surface dielectric constant of each pixel based on the improved Alpha approximation model; As an exemplary embodiment, using The sliding window traverses the entire image pixel by pixel. Within each sliding window, the backscattering coefficient is input into the Alpha approximation model. Then, combined with the range of dielectric constants constrained by the Dubois model, the system of equations is solved by the least squares method to obtain the surface dielectric constant of each pixel. In one exemplary embodiment, step S3 specifically includes: S31: Utilize The sliding window scans the fully polarimetric synthetic aperture radar image pixel by pixel to obtain the ratio of the backscattering coefficients of two adjacent pixels; In one exemplary embodiment, the formula for calculating the ratio of the backscattering coefficients is:
[0019] in, and These represent two adjacent cells within the sliding window. and The backscattering coefficient; and These represent two adjacent cells within the sliding window. and Alpha coefficient; Indicates the radar incident angle; and These represent two adjacent cells within the sliding window. and The dielectric constant of the soil; S32: Construct the observation equation based on the ratio of the backscattering coefficients; S33: Based on the observation equation, construct an underdetermined system of equations using multiple observation data from a sliding window to obtain an improved Alpha approximation model; In one exemplary embodiment, the improved Alpha approximation model is:
[0020] in, This is a matrix representing the ratios of the backscattering coefficients. This is the Alpha parameter matrix; S34: Based on the improved Alpha approximation model, the range of dielectric constant is constrained by the Dubois model, and the least squares method is used to solve the problem to obtain the Alpha parameter matrix. The surface dielectric constant of each pixel is obtained based on the Alpha parameter matrix. As an exemplary embodiment, in step S3, based on the Dubois model, the constraint range of the surface dielectric constant is obtained by combining the extreme range of the decomposed surface backscattering coefficient with the range of surface roughness. The system of equations is then solved using the least squares method under constraints to obtain... The matrix provides the dielectric constant value for each pixel.
[0021] S4: Perform water conversion and spatial averaging on the surface dielectric constant of each pixel to obtain the final soil moisture estimate for each pixel; As an exemplary embodiment, in step S4, the surface dielectric constant is converted into soil volumetric water content using the Topp dielectric hybrid model; an accumulation matrix and a counting matrix are constructed, and the final soil moisture estimate for each cell is obtained by dividing the accumulation value by the counting value.
[0022] In one exemplary embodiment, the formula for calculating the moisture conversion is:
[0023] in, Soil moisture; The dielectric constant of the Earth's surface; As an exemplary embodiment, in step S4, two matrices of the same size as the original image are initialized: an accumulation matrix for storing the sum of all calculated dielectric constant solutions for each pixel location, and a counting matrix Count for recording the number of times each pixel location is calculated as the center of the sliding window; after traversing the entire image, the first... Line 1 The final soil moisture characterization values for each column location were obtained by spatial averaging; In one exemplary embodiment, the formula for calculating the spatial average is:
[0024] in, Indicates the first Line 1 The final soil moisture estimate for the cells at the column location; Indicates the first Line 1 The summation matrix of the dielectric constant solutions of the cells at column positions; No. Line 1 The number of times the cell at the column position is used as the center of the sliding window.
[0025] In some embodiments, the above-described SAR soil moisture retrieval method based on spatial neighborhood optimization can also be implemented in the following ways.
[0026] In this embodiment, the SAR soil moisture retrieval method based on spatial neighborhood optimization includes: Step 1: Acquire one or more fully polarimetric synthetic aperture radar images during the crop growth cycle, preprocess the original images, and obtain the observation coherence matrix. In step 1, the study area is selected, and full polarimetric synthetic aperture radar images covering the crop growth cycle of the study area in the time dimension are acquired. The preprocessing operations for the raw data include: single-view complex transformation, coherence matrix extraction, polarimetric filtering, geocoding, and data cropping, which transforms the observation scattering matrix into the observation coherence matrix.
[0027] Step 2: For each pixel in the original image, establish a model-based polarization two-component decomposition framework. Assuming that dihedral scattering is negligible, combine the X-Bragg surface scattering model and the SNVSM volume scattering model to decouple the vegetation-ground signal. Obtain the backscattering coefficient of the ground surface scattering through the minimum residual power criterion. In step 2, the generalized volumetric scattering model (SNVSM) is used to describe the second-order statistical properties of the vegetation canopy. For each model, the directional randomness parameter is iteratively adjusted to generate the corresponding normalized volumetric coherence matrix. The volumetric scattering coefficient is calculated using the nonnegative eigenvalue decomposition method. .
[0028] In step 2, the radar backscattering signal mainly consists of signals from two parts: the surface soil and the vegetation layer. These are simulated by the X-Bragg surface scattering model and the SNVSM volume scattering model, respectively. The observation coherence matrix is expressed as:
[0029] in, Represents the surface scattering coefficient. This represents the coherence matrix of the X-Bragg surface scattering model. Represents the volume scattering coefficient. This represents the coherence matrix of the volume scattering model. This represents the residual matrix.
[0030] In step 2, to obtain the optimal solution for the parameters, a lookup table is constructed for the directional randomness parameters of the volumetric scattering model and the total scattering power of the residual matrix. The criterion for finding the optimal volumetric coherence matrix is to minimize the total power of the residual matrix, i.e.:
[0031] The final solution set is obtained through retrieval, from which the surface scattering coherence matrix is derived. The co-polarized backscattering coefficients can then be calculated.
[0032] Step 3: Use a 3×3 sliding window to traverse the entire image pixel by pixel. Within each sliding window, input the backscattering coefficients into the Alpha approximation model, and then combine the range of dielectric constants constrained by the Dubois model. Solve the equations using the least squares method to obtain the surface dielectric constant of each pixel. In step 3, within each sliding window, the ratio of the radar backscattering coefficients acquired from two adjacent pixels is processed to eliminate the influence of vegetation and roughness. The ratio is then approximated as the soil dielectric function. Radar incident angle And the function of polarization mode P, specifically expressed as:
[0033] Based on nine SAR observations using a sliding window, an underdetermined system of eight observation equations can be constructed. The final improved Alpha approximation model takes the following form:
[0034] in, The backscattering coefficient ratio matrix, This is the Alpha parameter matrix.
[0035] In step 3, based on the Dubois model, the constraint range of the surface dielectric constant is obtained by combining the extreme value range of the decomposed surface backscattering coefficient with the range of surface roughness. The system of equations is then solved using the least squares method under constraints to obtain... The matrix provides the dielectric constant value for each pixel.
[0036] Step 4: Convert the surface dielectric constant into soil volumetric water content using the Topp dielectric mixture model. Construct an accumulation matrix and a counting matrix, and obtain the final soil moisture estimate for each cell by dividing the accumulation value by the count value.
[0037] In step 4, the surface dielectric constant is converted into soil volumetric water content using the Topp dielectric mixing model. The formula for the dielectric mixing model is:
[0038] Where MV represents soil moisture. is the dielectric constant of the soil surface.
[0039] Initialize two matrices of the same size as the original image: an accumulation matrix to store the sum of all calculated dielectric constant solutions for each pixel location, and a count matrix (Count) to record the number of times each pixel location is calculated as the center of the sliding window. After traversing the entire image, the final soil moisture characterization value at the i-th row and j-th column is obtained by spatial averaging. .
[0040] In some embodiments, the above-described SAR soil moisture retrieval method based on spatial neighborhood optimization can also be implemented in the following ways.
[0041] The embodiments of the present invention provide a polarimetric SAR soil moisture inversion method based on a spatial neighborhood optimized Alpha model, which is used to estimate soil moisture under vegetation cover during the crop growth cycle.
[0042] Please refer to Figure 2 , Figure 2This is a flowchart of this embodiment, which selects a wheat-growing area in an agricultural region as the study area. To verify the effectiveness and advancement of the method using more validation sample data, 12 C-band RADARSAT-2 fully polarimetric SAR images covering corn and wheat fields were selected. Field surveys were conducted on the dates corresponding to the image data acquisition, obtaining real soil moisture data from 360 field sampling points for verification of the inversion results. Figure 3 The image shows the location of the study area and the PauliRGB image from May 9, 2019, where the green box represents the maize study area and the yellow box represents the wheat study area. The specific implementation steps are as follows: Step 1: Polarimetric SAR Image Preprocessing A study area was selected, and fully polarimetric synthetic aperture radar (SAR) imagery covering the crop growth cycle over time was acquired. Preprocessing operations, including single-view complex transformation, coherence matrix extraction, polarimetric filtering, geocoding, and data cropping, were performed on the raw data to transform the observed scattering matrix into an observed coherence matrix. The observed scattering matrix is the observed value, directly provided by the satellite imagery.
[0043] Step 2: Polarization Two-Component Decomposition To effectively separate the contributions of surface soil and vegetation layer in radar backscattering, the X-Bragg model is used to describe surface scattering, and the SNVSM generalized volume scattering model is used to characterize the second-order statistical properties of the vegetation canopy. The observation coherence matrix is modeled as the sum of the weighted sum of the two and the residual matrix.
[0044]
[0045] The Bragg scattering model was rotated along the radar line of sight, and the depolarization rotation angle caused by the surface slope was compared with that of the model. By integrating the probability density function, we obtain an X-Bragg scattering model with a wider range of applicability to roughness, which effectively solves the depolarization problem and avoids non-zero cross-polarization power.
[0046]
[0047]
[0048] Based on this, the probability distribution of surface depolarization angle Using zero mean and variance The normal distribution function. Integral derivation yields the coherence matrices as follows:
[0049]
[0050]
[0051]
[0052] in, The surface scattering coherence matrix is constructed based on the X-Bragg model with zero mean normal distribution; It is related to the local angle of incidence. and the dielectric constant of the earth's surface The relevant scattering coefficients satisfy ; Indicates conjugate transpose; This represents the variance of the normal distribution related to surface roughness. and These are the Bragg coefficients under horizontal and vertical polarization, respectively.
[0053] In the volume scattering model, the canopy is modeled as a cloud of particles of arbitrary shape, and the scattering particles follow a circular Gaussian distribution. Particle anisotropy. Its modulus can describe the effective shape of the average scattering particles. When When the value approaches 0, it indicates that the shape of the scattering particles tends to be an isotropic sphere. A value approaching 1 indicates that the scattering particle shape tends towards a dipole. This introduces the scattering particle as a horizontal dipole (…). ) or vertical dipole ( Based on the assumptions of [previous scenario], a simplified Neumann volume scattering model (SNVSM) is proposed. The coherence matrix is represented as follows:
[0054]
[0055] in This indicates the degree of randomness in particle orientation; the larger the parameter, the stronger the randomness of the particle's orientation. It indicates a completely random orientation.
[0056] The volume scattering coefficients were calculated using the nonnegative eigenvalue decomposition method. The particle orientation randomness τ was set to 0–1, and the step size was 0.01. The coherence matrix data table of the SNVSM volume scattering model was obtained. The constructed equation set is shown below:
[0057] By iteratively adjusting the directional randomness parameters of the volume scattering model, a corresponding normalized volume coherence matrix is generated, and then the total power of the residual matrix is minimized. Using this as the criterion, an optimized search is performed in the preset parameter lookup table to finally obtain the optimal surface scattering coherence matrix and co-polarized backscattering coefficient.
[0058] Step 3: Alpha Algorithm Based on Spatial Neighborhood Under the assumption that surface roughness and vegetation parameters remain constant over a short time period, an Alpha approximation model is proposed. This model directly correlates the ratio of backscattering coefficients between two consecutive time phases with changes in soil moisture, thus enabling the calculation of absolute soil moisture.
[0059] polarization amplitude is the Fresnel reflection coefficient under different polarization modes, which is a function of the radar incident angle and the dielectric constant. Its expression is:
[0060]
[0061] The relationship between the radar backscattering coefficient and the polarization amplitude is as follows:
[0062]
[0063] Traditional change detection algorithms suffer from limitations in applicability because they fail to consider the impact of vegetation cover and its rapid short-term changes on soil moisture retrieval. Therefore, this paper assumes that adjacent pixels within a local window have similar vegetation cover and surface roughness at the same time phase, and that the spatial differences in backscattering mainly stem from variations in the soil dielectric constant.
[0064] A 3×3 sliding window is used, and the ratio of radar backscattering coefficients acquired from two adjacent pixels within the window is processed to eliminate the influence of vegetation (attenuation factor) and roughness. The ratio is then approximated as a function of soil dielectric function, radar incident angle, and polarization mode (P), specifically expressed as:
[0065] according to and Two pixels can construct an observation equation:
[0066] Furthermore, based on nine SAR observations using a sliding window, an underdetermined system of eight observation equations can be constructed. The final improved Alpha approximation model takes the following form:
[0067]
[0068]
[0069] Based on the Alpha approximation model and the optimal constraint condition of the dielectric constant, the Alpha coefficient of each pixel is obtained by solving the system of equations using the least squares method. Specifically, a physical constraint method for the dielectric constant based on the Dubois model is adopted. According to the physical range of soil roughness and the extreme value of the decomposed surface backscattering coefficient, the optimal range of the surface dielectric constant is obtained.
[0070] Step 4: Spatial Average Soil Moisture The dielectric mixing model converts the dielectric constant of the soil surface under vegetation cover into soil moisture content, thus achieving soil moisture inversion. The formula for the dielectric mixing model is:
[0071] Where MV represents soil moisture. is the dielectric constant of the soil surface.
[0072] Initialize two matrices of the same size as the original image: an accumulation matrix to store the sum of all calculated dielectric constant solutions for each pixel location. And a counting matrix used to record the number of times each cell position is counted as the center of the sliding window. After traversing the entire image, the final soil moisture characterization value at the i-th row and j-th column position is obtained by spatial averaging:
[0073] The inversion results were compared with the actual soil moisture measured at sampling points on five dates within the maize growth cycle and seven dates within the wheat growth cycle in the experimental area. Figure 4 Scatter plots show the measured and model-derived soil moisture results under HH and VV polarization in the two study areas. The results indicate that the overall root mean square error (RMSE) in the maize area reached 5.22 Vol.% and 4.48 Vol.% under HH and VV polarization models, respectively; while in the wheat area, the RMS errors were 5.64 Vol.% and 5.56 Vol.%. This demonstrates that efficient and accurate soil moisture retrieval under vegetation cover was achieved.
[0074] This embodiment provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described polarimetric SAR soil moisture inversion method.
[0075] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.
Claims
1. A SAR soil moisture retrieval method based on spatial neighborhood optimization, characterized in that, Includes the following steps: S1: Acquire fully polarimetric synthetic aperture radar images during the crop growth cycle, preprocess the fully polarimetric synthetic aperture radar images, and obtain the observation coherence matrix. S2: Construct a scattering model based on the observed coherence matrix; optimize the solution using the minimum residual power criterion based on the scattering model to obtain the backscattering coefficient of the surface scattering; S3: Construct an improved Alpha approximation model based on the backscattering coefficient, and obtain the surface dielectric constant of each pixel based on the improved Alpha approximation model; S4: Perform water conversion and spatial averaging on the surface dielectric constant of each pixel to obtain the final soil moisture estimate for each pixel.
2. The SAR soil moisture retrieval method based on spatial neighborhood optimization according to claim 1, characterized in that, The preprocessing includes single-view complex transformation, coherence matrix extraction, polarization filtering, geocoding, and data pruning.
3. The SAR soil moisture retrieval method based on spatial neighborhood optimization according to claim 1, characterized in that, The scattering model is as follows: ， in, Represents the observation coherence matrix; Represents the surface scattering coefficient. This represents the coherence matrix of the X-Bragg surface scattering model; Indicates the volume scattering coefficient; This represents the coherence matrix of the volume scattering model. This represents the residual matrix.
4. The SAR soil moisture retrieval method based on spatial neighborhood optimization according to claim 1, characterized in that, The minimum residual power criterion is as follows: ， in, Indicates minimization; Represents residual power; Indicates the total scattered power; This represents the Bragg scattering coefficient under H and V polarizations.
5. The SAR soil moisture retrieval method based on spatial neighborhood optimization according to claim 1, characterized in that, Step S3 specifically includes: S31: Utilize The sliding window scans the fully polarimetric synthetic aperture radar image pixel by pixel to obtain the ratio of the backscattering coefficients of two adjacent pixels; S32: Construct the observation equation based on the ratio of the backscattering coefficients; S33: Based on the observation equation, construct an underdetermined system of equations using multiple observation data from a sliding window to obtain an improved Alpha approximation model; S34: Based on the improved Alpha approximation model, and constraining the range of dielectric constants using the Dubois model, the least squares method is used to solve the problem to obtain the Alpha parameter matrix. The surface dielectric constant of each pixel is then obtained from the Alpha parameter matrix.
6. The SAR soil moisture inversion method based on spatial neighborhood optimization according to claim 5, characterized in that, The formula for calculating the ratio of the backscattering coefficients is: ， in, and These represent two adjacent cells within the sliding window. and The backscattering coefficient; and These represent two adjacent cells within the sliding window. and Alpha coefficient; Indicates the radar incident angle; and These represent two adjacent cells within the sliding window. and The dielectric constant of the soil.
7. The SAR soil moisture retrieval method based on spatial neighborhood optimization according to claim 5, characterized in that, The improved Alpha approximation model is as follows: ， in, This is a matrix representing the ratios of the backscattering coefficients. This is the Alpha parameter matrix.
8. The SAR soil moisture inversion method based on spatial neighborhood optimization according to claim 1, characterized in that, The formula for calculating the water conversion is: ， in, Soil moisture; is the dielectric constant of the Earth's surface.
9. The SAR soil moisture inversion method based on spatial neighborhood optimization according to claim 1, characterized in that, The formula for calculating the spatial average is: ， in, Indicates the first Line number The final soil moisture estimate for the cells at the column location; Indicates the first Line number The summation matrix of the dielectric constant solutions of the cells at column positions; No. Line number The number of times the cell at the column position is used as the center of the sliding window.
10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the SAR soil moisture inversion method based on spatial neighborhood optimization as described in any one of claims 1-9.