Soil moisture retrieval methods, devices, and equipment based on multi-source remote sensing information fusion
By improving the multi-source remote sensing information fusion algorithm and machine learning model, the stability and applicability issues of remote sensing data sources were solved, achieving high accuracy and high stability in soil moisture retrieval, enhancing the moisture-sensitive information of remote sensing images, and suppressing spectral distortion and information loss.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NORTHWEST A & F UNIV
- Filing Date
- 2026-04-09
- Publication Date
- 2026-05-26
AI Technical Summary
In existing technologies, a single remote sensing data source is insufficient to ensure both the stability and applicability of soil moisture retrieval. Differences in optical and SAR imaging mechanisms lead to spectral distortion and spatial information loss in fused images. Furthermore, the lack of constraint mechanisms for soil moisture retrieval tasks affects the accuracy and stability of the retrieval.
A constrained improved fusion algorithm based on the HVCE algorithm was adopted. The H-SVD and H-DWT algorithms were improved by improving the SVD and DWT algorithms. Combined with the machine learning regression model, the soil moisture inversion model was constructed by using the spectral physical constraints and correlation analysis of satellite remote sensing image data.
It improves the accuracy and stability of soil moisture retrieval, enhances the moisture-sensitive information of remote sensing images, suppresses spectral distortion caused by the fusion of heterogeneous image data, and improves the correlation and accuracy of retrieval results.
Smart Images
Figure CN122090168A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of multi-source remote sensing information fusion technology, and in particular to a method, apparatus and equipment for soil moisture inversion based on multi-source remote sensing information fusion. Background Technology
[0002] Soil moisture is a crucial variable in the terrestrial water cycle and energy balance processes. Rapidly and accurately acquiring its spatiotemporal variations is essential for optimizing agricultural irrigation, drought early warning, and water resource management. However, while traditional point-based monitoring methods such as drying and time-domain reflectometry offer high precision, they are insufficient for large-scale, continuous, and dynamic monitoring. Remote sensing technology, due to its wide coverage and timely response, has become a vital tool for soil moisture retrieval. Among these, Synthetic Aperture Radar (SAR) possesses all-weather observation capabilities, but its backscattering is susceptible to interference from factors such as surface roughness and vegetation structure. Optical remote sensing can indirectly characterize moisture information through multispectral reflectance and vegetation indices, but it is significantly limited by cloud cover and lighting conditions, making it difficult for a single data source to simultaneously ensure stability and applicability. To overcome the aforementioned problems, multi-source remote sensing data fusion has been increasingly applied to soil moisture monitoring. However, current pixel-level fusion methods mostly aim at visual enhancement or fidelity indicators, lacking constraint mechanisms for soil moisture retrieval tasks. Furthermore, the differences between optical and SAR imaging mechanisms can easily lead to spectral distortion and spatial information loss in fused images, thus weakening their correlation with soil moisture. While feature-level and decision-level methods can be jointly modeled, they are highly dependent on samples and do not adequately utilize the spatial-spectral structure of images. Therefore, improving the accuracy and stability of soil moisture retrieval is crucial. Summary of the Invention
[0003] The purpose of this application is to provide a method, apparatus, and equipment for soil moisture retrieval based on multi-source remote sensing information fusion, which can improve the accuracy and stability of soil moisture retrieval.
[0004] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a soil moisture inversion method based on multi-source remote sensing information fusion, including: Acquire satellite remote sensing image data; the satellite remote sensing image data includes radar images and optical images of the same time phase; An auxiliary dataset is determined based on the satellite remote sensing image data; the auxiliary dataset is used as the spectral physical constraint for the HVCE algorithm. The satellite remote sensing image data is fused using a constrained improved fusion algorithm to obtain a fused image. The constrained improved fusion algorithm includes spectral physical constraints based on the HVCE algorithm, and constrained improvements to the Singular Value Decomposition (SVD) algorithm and Discrete Wavelet Transform (DWT) algorithm to form new fusion algorithms H-SVD algorithm and H-DWT algorithm. The soil moisture inversion results are determined based on the fused image using the soil moisture inversion model. The soil moisture inversion model is a machine learning regression model that compares the correlation between the fused image of known in-situ soil moisture and the corresponding satellite remote sensing image data, and uses the coefficient of determination and normalized root mean square error to quantitatively analyze the inversion accuracy.
[0005] Secondly, this application provides a soil moisture retrieval device based on multi-source remote sensing information fusion, comprising: The data acquisition module is used to acquire satellite remote sensing image data; the satellite remote sensing image data includes radar images and optical images of the same time phase. An auxiliary dataset determination module is used to determine an auxiliary dataset based on the satellite remote sensing image data; the auxiliary dataset is used as a spectral physical constraint for the HVCE algorithm. The fusion processing module is used to perform fusion processing on the satellite remote sensing image data using a constrained improved fusion algorithm to obtain a fused image. The constrained improved fusion algorithm includes spectral physical constraints based on the HVCE algorithm, and constrained improvements to the Singular Value Decomposition (SVD) algorithm and Discrete Wavelet Transform (DWT) algorithm to form new fusion algorithms H-SVD algorithm and H-DWT algorithm. The inversion module is used to determine the soil moisture inversion results based on the fused image according to the soil moisture inversion model. The soil moisture inversion model is a machine learning regression model that performs correlation comparison analysis between the fused image of known in-situ soil moisture and the corresponding satellite remote sensing image data, and uses the coefficient of determination and normalized root mean square error to quantitatively analyze the inversion accuracy.
[0006] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the soil moisture inversion method based on multi-source remote sensing information fusion described above.
[0007] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method, apparatus, and device for soil moisture retrieval based on multi-source remote sensing information fusion. The method involves acquiring satellite remote sensing image data; determining an auxiliary dataset based on the satellite remote sensing image data; and employing a constrained improved fusion algorithm to fuse the satellite remote sensing image data to obtain a fused image. This fused image enhances the moisture-sensitive information of the remote sensing images while suppressing spectral distortion caused by the fusion of heterogeneous image data. The soil moisture retrieval result is determined based on the fused image using a soil moisture retrieval model. The soil moisture retrieval model is a machine learning regression model that compares the correlation between the fused image with known in-situ soil moisture and the corresponding satellite remote sensing image data, and uses the coefficient of determination and normalized root mean square error to quantitatively analyze the retrieval accuracy. This improves the accuracy and stability of soil moisture retrieval. Attached Figure Description
[0008] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0009] Figure 1 This is a flowchart of a soil moisture retrieval method based on multi-source remote sensing information fusion; Figure 2 This is a schematic diagram of the optimal spectral index calculation results; Figure 3 A visual representation of RGB images before and after image fusion in a QLB-NET experimental site; Figure 4 This is a schematic diagram illustrating the changes in the correlation between the image fusion and the in-situ soil moisture before and after image fusion. Figure 5 A schematic diagram illustrating the accuracy of the soil moisture inversion model; Figure 6 This is a structural diagram of a soil moisture retrieval device based on multi-source remote sensing information fusion. Figure 7 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation
[0010] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0011] This application proposes a radar-optical image fusion technology solution for soil moisture retrieval tasks, which enhances moisture-sensitive information while suppressing spectral distortion and information loss caused by heterogeneous fusion, thereby improving the accuracy and stability of soil moisture retrieval.
[0012] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0013] In one exemplary embodiment, such as Figure 1 As shown, a soil moisture retrieval method based on multi-source remote sensing information fusion is provided, including: Step 100: Acquire satellite remote sensing image data. Satellite remote sensing image data includes radar images and optical images from the same time period.
[0014] The optical images include Sentinel-2 multispectral images; the Sentinel-2 multispectral images include at least a portion of the blue, green, red, red-edge, near-infrared, and short-wave infrared bands; the radar images include Sentinel-1 VV polarization images and / or VH polarization images.
[0015] Acquiring satellite remote sensing image data specifically includes: Acquire initial satellite remote sensing image data; perform resampling, projection unification, and cropping preprocessing on the initial satellite remote sensing image data to obtain satellite remote sensing image data.
[0016] Step 200: Determine the auxiliary dataset based on satellite remote sensing imagery data. The auxiliary dataset is used as the spectral physical constraint for the HVCE algorithm.
[0017] The auxiliary dataset determined based on satellite remote sensing imagery data specifically includes: The spectral index is determined based on the optical image, and the correlation between the spectral index and the reflectivity of each optical band and the radar characteristic quantity corresponding to the radar image is determined, and the corresponding correlation coefficient is obtained.
[0018] The candidate spectral index that maximizes the evaluation function is selected as the optimal spectral index, and the auxiliary dataset is determined based on the optimal spectral index; wherein, the evaluation function is constructed with the objective of maximizing the average absolute value of the correlation coefficient.
[0019] The mathematical expression corresponding to the auxiliary dataset is: .
[0020] in, For auxiliary datasets; This is the correlation coefficient between the spectral index and the reflectance of various optical bands; This is the correlation coefficient between the spectral index and the backscattering coefficient corresponding to the radar image.
[0021] Step 300: The satellite remote sensing image data is fused using a constrained improved fusion algorithm to obtain a fused image. The constrained improved fusion algorithm includes spectral physical constraints based on the HVCE algorithm, and constrained improvements to the Singular Value Decomposition (SVD) algorithm and Discrete Wavelet Transform (DWT) algorithm, forming new fusion algorithms H-SVD and H-DWT.
[0022] The mathematical expression for the fused image is: .
[0023] in, To merge images; , All are weights; , All are design matrices; , All data are satellite remote sensing images; This is the matrix transpose.
[0024] Step 400: Determine the soil moisture inversion results based on the fused imagery using the soil moisture inversion model. The soil moisture inversion model is a machine learning regression model that performs correlation analysis between the fused imagery of known in-situ soil moisture and the corresponding satellite remote sensing imagery data, and uses the coefficient of determination and normalized root mean square error to quantitatively analyze the inversion accuracy.
[0025] This application aims to address the current problems of optical-SAR heterogeneous image fusion methods lacking constraints for soil moisture inversion tasks, easily generating spectral distortion and spatial information loss, resulting in insufficient correlation between the fusion results and soil moisture, and limited inversion accuracy and stability.
[0026] This application constructs a fusion algorithm for soil moisture retrieval tasks by introducing an adaptive mechanism with auxiliary data constraints. By improving the constraint of SVD / DWT using the HVCE framework, it effectively suppresses spectral distortion and spatial information loss caused by heterogeneous image fusion while enhancing moisture-sensitive information, thereby improving the correlation between fused images and soil moisture and significantly improving the accuracy and stability of soil moisture retrieval. At the same time, this application implements the above process in a systematic module, which is convenient for promotion and application in different regions and different vegetation growth stages.
[0027] In practical applications, the operation steps of the method mentioned in this application can also be as follows: Step 1: Download satellite remote sensing images and preprocess data. Obtain radar and optical images of the same time phase, and perform preprocessing such as resampling, projection unification and cropping on the radar and optical images. Also, organize the in-situ soil moisture data and remove outliers.
[0028] Preprocessing also includes unifying the spatial resolution of optical and radar images and cropping the study area.
[0029] Specifically, SAR satellite data (radar imagery) download: Based on the requirements of this application, GRD (Ground Range Detected) imagery data from Level-1 products in IW (Interferometric Wide) imaging mode was selected. The satellite imagery data was obtained from the European Space Agency's Copernicus Sentinel Science Centre.
[0030] Optical satellite data (optical imagery) download: Level-2A products were selected based on the requirements of this application. Level-2A products primarily contain atmospherically corrected lower atmospheric reflectivity data. The satellite imagery data was obtained from the European Space Agency's Copernicus Sentinel Science Centre.
[0031] The imaging time difference between Sentinel-1 and Sentinel-2 image data should be less than 7 days, and the imaging time should be basically synchronized with the ground sampling time.
[0032] Sentinel-1 data preprocessing: SNAP software was used to preprocess Sentinel-1 image data. The processing steps for GRDH (Ground Range Detected High Resolution) data included: orbit correction, thermal noise removal, radiometric calibration, speckle filtering, and terrain correction.
[0033] Sentinel-2 data preprocessing: Sentinel-2 L2A level data has already undergone atmospheric correction. Since the resolution varies between Sentinel-2 bands, it needs to be unified to the same resolution. The Sne2Res plugin in SNAP is used to perform super-resolution synthesis of the Sentinel-2 image data across each band, reconstructing the Sentinel-2 image data to a 10m resolution and ensuring the consistency of spatial resolution between Sentinel-1 and Sentinel-2 satellite imagery.
[0034] Step 2: Auxiliary dataset selection and adaptive spectral constraint construction. Spectral indices are calculated based on optical images. The correlation between the spectral indices and the reflectivity of each optical band and radar features is calculated. The optimal auxiliary dataset is adaptively determined according to the preset selection criteria. Spectral physical constraints for the HVCE (Helmert Variance Component Estimation) algorithm are constructed to achieve error constraint optimization for soil moisture inversion tasks.
[0035] The Sentinel-2 multispectral image data is used to calculate the corresponding spectral indices, and the optimal spectral indices are selected based on preset screening criteria. A schematic diagram of the optimal spectral index calculation results is shown below. Figure 2 As shown. Figure 2 In this context, VV represents the backscattering coefficient, B represents the multispectral blue band, and I... VI The values are calculated for spectral constraints. Preset selection criteria include: calculating the correlation coefficients between candidate spectral indices and the reflectivity of each optical band, as well as their correlation coefficients with radar characteristic quantities; constructing an evaluation function with the objective of maximizing the average absolute value of the correlation coefficients; selecting the candidate spectral index that maximizes the evaluation function as the optimal spectral index; using the optimal spectral index to determine the optimal auxiliary dataset and generate adaptive spectral physical constraints.
[0036] Spectral index calculation: Import Sentinel-2 images matching the sampling date into ENVI (Environment for Visualizing Images) software to calculate the 18 spectral indices shown in Table 1.
[0037] Table 1. Formulas for Calculating Spectral Indices
[0038] The meanings of the letters appearing in the table are as follows: VSDI stands for Visible and Shortwave Infrared Drought Index.
[0039] Blue (B); Green (G); Red (R); Near-infrared band (NIR); Red Edge 1 (RE1), Sentinel-2 standard; Red Edge 2 (RE2); Red Edge 3 (RE3); Short-Wave Infrared 1 (S1); Short-Wave Infrared 2 (S2).
[0040] Optimal auxiliary dataset selection: To facilitate the construction of the HVCE framework, the optimal auxiliary dataset is mainly calculated using formula (1) as the spectral constraint information for the HVCE framework. (1) The optimal spectral index auxiliary data is calculated using formula (1) to assist in the construction of the HVCE error equation.
[0041] Step 3: Construction of fusion algorithm and image fusion under HVCE constraints. Under the spectral physical constraints, the Singular Value Decomposition (SVD) algorithm and the Discrete Wavelet Transform (DWT) algorithm are constrained and improved to form new fusion algorithms H-SVD and H-DWT. The radar image and optical image are fused based on the H-SVD algorithm or H-DWT algorithm to output the fused image.
[0042] Entropy weights and HVCE spectral physical constraints are introduced during the injection of feature components obtained from SVD decomposition and the fusion of multi-scale detail components and approximate components obtained from DWT. These constraints limit spectral shift and distortion in the fused image and enhance the moisture-sensitive information of the radar image, thus forming the H-SVD and H-DWT algorithms. In other words, the SVD and DWT algorithms are constrained and improved under spectral physical constraints, resulting in the H-SVD and H-DWT algorithms respectively. This reduces spectral distortion and spatial information loss caused by differences in optical and radar imaging mechanisms, and outputs a fused image.
[0043] Specifically, the HVCE algorithm improved based on optimal spectral index auxiliary data: To verify the effectiveness of the HVCE framework in the field of multi-source remote sensing image data fusion, the HVCE framework improved based on optimal spectral index auxiliary data was mainly used to fuse multispectral satellite image data without prior error information. (i represents the multispectral band number), SAR satellite image data This study is the first to verify the feasibility of the HVCE framework and its advantages in preserving image information.
[0044] The satellite image data input to HVCE is the low-frequency approximate component or high-frequency detail component of the satellite image X after decomposition by SVD or DWT.
[0045] Assuming components (z=1 indicates) The component, when z=2, represents (components), component express The components are given by w, where w = 1, 2, ..., m × n. m is the number of horizontal pixels in the input image; n is the number of vertical pixels in the input image.
[0046] To meet the application requirements of multi-source image data fusion, the following approach is adopted. Constructing error equations for auxiliary data: The key to selecting auxiliary data lies in its correlation with the dataset to be fused. Therefore, Each band and The fusion of data and the selection of auxiliary data all adopt the absolute value average maximization criterion.
[0047] The improved HVCE framework is detailed below: First, an error equation is established based on the selected optimal spectral image data to determine the coefficient matrix. (z=1 indicates) The coefficient matrix of the components, when z=2 represents The coefficient matrix of the components), the error equation can be expressed as: (2) and These are the coefficients of the stochastic model. Coefficient matrix: .
[0048] It is worth noting that, unlike traditional methods, this application innovatively uses optimized spectral indices to construct a stochastic model and establish an error equation, which will effectively improve the HVCE framework's ability to retain soil moisture information.
[0049] Suppose two independent image data and The weight matrices are respectively and : .
[0050] Indicates the weight.
[0051] initialization weight Let's set it to 1. The corresponding residual equation is: (3) (4) (5) (6) in, For the corresponding residuals; Estimate the parameters; This is the coefficient matrix of the overall normal equation; This is the coefficient matrix of the overall normal equation; The normal equation coefficient matrix is formed independently from the first set of observation data; The normal equation coefficient matrix is formed independently from the second set of observation data; The vector of constant terms in the normal equations, independently constructed from the first set of observation data; This is the vector of constant terms in the normal equations, independently constructed from the second set of observation data.
[0052] Using the sum of squared residuals of various corrections after each adjustment and Perform iterative estimation of the unit weight variance of the two types of data. and Establish the relationship between the sum of squared residuals and the unit weighted variance: (7) (8) The unit weighted variance is estimated using the following procedure: (9) (10) By updating weights , The iteration continues until equation (11) is satisfied.
[0053] (11) Output weights after stopping iteration when the condition is met. , .
[0054] According to the least squares solution of the linear system, we can obtain the result using formula (12). : (12) This is the corresponding design matrix.
[0055] The above algorithm framework is an improved HVCE framework with spectral indices as error constraints. The above steps are used to... , Image data is fused to obtain .
[0056] The H-SVD and H-DWT algorithms are developed: An adaptive fusion framework based on HVCE is used to uniformly adjust the error distribution and information injection intensity in the fusion process. Specifically, the weighted average fusion rule in traditional multi-resolution singular value decomposition (SVD) or discrete wavelet decomposition (DWT) algorithms is replaced with the HVCE framework. In the SVD path, the input radar and optical images are decomposed at multiple resolutions to obtain corresponding feature components. In the component injection and reconstruction stage, the variance component weights estimated by HVCE are introduced to dynamically adjust the information injection ratio and weight allocation of the feature components, thereby enhancing soil moisture sensitivity information while suppressing spectral loss, resulting in the improved H-SVD algorithm. In the DWT path, the input image is decomposed at multiple wavelet scales to obtain approximate components and multi-scale detail components. In the component fusion and inverse transform reconstruction stage, the fusion weights of the components are also adaptively estimated by HVCE, and the injection of detail information at different scales and directions is controlled under spectral constraints to reduce spectral distortion caused by heterogeneous fusion and preserve spatial details, resulting in the improved H-DWT algorithm. In this way, the algorithm can be upgraded without changing the basic decomposition-fusion-reconstruction framework of SVD / DWT, thus forming an H-SVD and H-DWT fusion algorithm for soil moisture inversion tasks.
[0057] Output fused image: Multispectral satellite imagery and SAR satellite imagery are decomposed using H-SVD and H-DWT algorithms respectively, and then fused using the inverse operation of H-SVD and H-DWT algorithms to obtain the fused imagery. and fused image or .
[0058] Figure 3 A visual representation of RGB images before and after image fusion in the QLB-NET experimental site.
[0059] Step 4: Evaluation of fusion effect and soil moisture inversion modeling. Qualitative and quantitative evaluation of the fused images is carried out, and the correlation between the image features before and after fusion and the in-situ soil moisture is compared and analyzed. Then, a soil moisture inversion model is constructed and the soil moisture inversion results are output. Figure 4 This is a schematic diagram illustrating the changes in the correlation between the image before and after fusion and the in-situ soil moisture.
[0060] When performing qualitative and quantitative evaluations on fused images, the quantitative evaluation includes one or more of ERGAS, PSNR, RMSE, and FSIM. The improvement in soil moisture characterization capability of the fusion is assessed by comparing the changes in the correlation between image features before and after fusion and in-situ soil moisture. Furthermore, the soil moisture retrieval model is constructed using a machine learning regression model, specifically an XGBoost model, with the coefficient of determination R0... 2 The normalized root mean square error (nRMSE) is used as an accuracy evaluation index.
[0061] Specifically, the fusion effect evaluation and soil moisture inversion modeling are performed, and the soil moisture inversion results are output: The quality of the fused image output in step 3 is assessed. This quality assessment includes at least a qualitative evaluation of the fused image and a quantitative evaluation based on ERGAS (Erreur Relative Globale Adimensionnelle de Synthèse, dimensionless global relative synthesis error), PSNR (Peak Signal-to-Noise Ratio), RMSE (Root Mean Square Error), and FSIM (Feature Similarity Index). Furthermore, the enhancement effect of the fused image on soil moisture-sensitive information is determined by comparing the changes in the correlation between image features before and after fusion and in-situ soil moisture. Based on this, a soil moisture inversion model is constructed, using the band reflectance and spectral index of the fused image as input and in-situ soil moisture as output. This inversion model is an XGBoost machine learning regression model, and the coefficient of determination R is used. 2 The inversion accuracy was evaluated using the normalized root mean square error (nRMSE), and the final output was the soil moisture inversion results for the study area. Figure 5 This is a schematic diagram illustrating the accuracy of the soil moisture inversion model.
[0062] Table 2. Quantitative evaluation table of fused image quality (bands: B, G, R) In one exemplary embodiment, such as Figure 6 As shown, a soil moisture retrieval device based on multi-source remote sensing information fusion is provided, comprising: The data acquisition module is used to acquire satellite remote sensing image data; the satellite remote sensing image data includes radar images and optical images of the same time phase.
[0063] An auxiliary dataset determination module is used to determine an auxiliary dataset based on the satellite remote sensing image data; the auxiliary dataset is used as a spectral physical constraint for the HVCE algorithm.
[0064] The fusion processing module is used to perform fusion processing on the satellite remote sensing image data using a constrained improved fusion algorithm to obtain a fused image. The constrained improved fusion algorithm includes spectral physical constraints based on the HVCE algorithm, and constrained improvements to the Singular Value Decomposition (SVD) algorithm and Discrete Wavelet Transform (DWT) algorithm to form new fusion algorithms H-SVD algorithm and H-DWT algorithm.
[0065] The inversion module is used to determine the soil moisture inversion results based on the fused image according to the soil moisture inversion model. The soil moisture inversion model is a machine learning regression model that performs correlation comparison analysis between the fused image of known in-situ soil moisture and the corresponding satellite remote sensing image data, and uses the coefficient of determination and normalized root mean square error to quantitatively analyze the inversion accuracy.
[0066] In practical applications, the device may also include the following modules: The data acquisition and preprocessing module is used to acquire radar images and optical images of the same time phase and perform resampling, projection unification and cropping preprocessing.
[0067] The adaptive spectral constraint generation module is used to calculate spectral indices and analyze their correlation with optical bands and radar characteristics, and adaptively determine the optimal auxiliary dataset based on preset screening criteria.
[0068] The HVCE constrained fusion module is used to constrain and improve the SVD and DWT algorithms under the constraints of the spectral index to form the H-SVD and H-DWT algorithms, and output the fused image based on the H-SVD or H-DWT algorithm.
[0069] The inversion and evaluation module is used to perform qualitative and quantitative evaluation of fused images, conduct correlation comparison analysis before and after fusion, and establish a soil moisture inversion model to output soil moisture inversion results.
[0070] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 7As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores soil moisture inversion data based on multi-source remote sensing information fusion. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a soil moisture inversion method based on multi-source remote sensing information fusion.
[0071] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0072] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0073] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0074] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).
[0075] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logic devices, etc., and are not limited to these.
[0076] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0077] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for soil moisture retrieval based on multi-source remote sensing information fusion, characterized in that, include: Acquire satellite remote sensing image data; the satellite remote sensing image data includes radar images and optical images of the same time phase; An auxiliary dataset is determined based on the satellite remote sensing image data; the auxiliary dataset is used as the spectral physical constraint for the HVCE algorithm. The satellite remote sensing image data is fused using a constrained improved fusion algorithm to obtain a fused image. The constrained improved fusion algorithm includes spectral physical constraints based on the HVCE algorithm, and constrained improvements to the Singular Value Decomposition (SVD) algorithm and Discrete Wavelet Transform (DWT) algorithm to form new fusion algorithms H-SVD algorithm and H-DWT algorithm. Based on the soil moisture inversion model, the soil moisture inversion results are determined according to the fused images. The soil moisture retrieval model is a machine learning regression model that compares the correlation between the fused image of known in-situ soil moisture and the corresponding satellite remote sensing image data, and uses the coefficient of determination and normalized root mean square error to quantitatively analyze the retrieval accuracy.
2. The soil moisture retrieval method based on multi-source remote sensing information fusion according to claim 1, characterized in that, The auxiliary dataset is determined based on the aforementioned satellite remote sensing image data, specifically including: Based on the optical image, the spectral index is determined, and the correlation between the spectral index and the reflectivity of each optical band and the radar characteristic quantity corresponding to the radar image is determined, so as to obtain the corresponding correlation coefficient. The candidate spectral index that maximizes the evaluation function is selected as the optimal spectral index, and the auxiliary dataset is determined based on the optimal spectral index; wherein the evaluation function is constructed with the objective of maximizing the average absolute value of the correlation coefficient.
3. The soil moisture retrieval method based on multi-source remote sensing information fusion according to claim 2, characterized in that, The mathematical expression corresponding to the auxiliary dataset is: ； in, For auxiliary datasets; This is the correlation coefficient between the spectral index and the reflectance of various optical bands; This is the correlation coefficient between the spectral index and the backscattering coefficient corresponding to the radar image.
4. The soil moisture retrieval method based on multi-source remote sensing information fusion according to claim 1, characterized in that, The optical images include Sentinel-2 multispectral images; the Sentinel-2 multispectral images include at least a portion of the blue, green, red, red-edge, near-infrared, and short-wave infrared bands; the radar images include Sentinel-1 VV polarization images and / or VH polarization images.
5. The soil moisture retrieval method based on multi-source remote sensing information fusion according to claim 1, characterized in that, Acquiring satellite remote sensing image data specifically includes: Acquire initial satellite remote sensing image data; The initial satellite remote sensing image data is preprocessed by resampling, projection unification, and cropping to obtain the satellite remote sensing image data.
6. The soil moisture retrieval method based on multi-source remote sensing information fusion according to claim 1, characterized in that, The mathematical expression corresponding to the fused image is: ； in, To merge images; , All are weights; , All are design matrices; , All data are satellite remote sensing images; This is the matrix transpose.
7. A soil moisture retrieval device based on multi-source remote sensing information fusion, characterized in that, include: The data acquisition module is used to acquire satellite remote sensing image data; the satellite remote sensing image data includes radar images and optical images of the same time phase. An auxiliary dataset determination module is used to determine an auxiliary dataset based on the satellite remote sensing image data; the auxiliary dataset is used as a spectral physical constraint for the HVCE algorithm. The fusion processing module is used to perform fusion processing on the satellite remote sensing image data using a constrained improved fusion algorithm to obtain a fused image. The constrained improved fusion algorithm includes spectral physical constraints based on the HVCE algorithm, and constrained improvements to the Singular Value Decomposition (SVD) algorithm and Discrete Wavelet Transform (DWT) algorithm to form new fusion algorithms H-SVD algorithm and H-DWT algorithm. The inversion module is used to determine the soil moisture inversion results based on the fused image according to the soil moisture inversion model; The soil moisture retrieval model is a machine learning regression model that compares the correlation between the fused image of known in-situ soil moisture and the corresponding satellite remote sensing image data, and uses the coefficient of determination and normalized root mean square error to quantitatively analyze the retrieval accuracy.
8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the soil moisture inversion method based on multi-source remote sensing information fusion as described in any one of claims 1-6.