Method and device for establishing backscattering coefficient correction model

Abstract

The invention provides a method and device for establishing a backscattering coefficient correction model, and the method comprises the steps: obtaining multi-source data comprising an observation value corresponding to a backscattering coefficient; performing space-time matching on the multi-source data, generating a data unit corresponding to each satellite observation pixel, and filtering the data units through a preset rule to obtain a reserved data set; outputting a simulation value through a physical model based on the reserved data set not including the observation value, constructing a training set based on the simulation value and the reserved data set, and using the observation value as a target variable of the training set; and training a machine learning model by using the training set, and performing grid search optimization on preset hyper-parameters in the machine learning model to obtain a final backscattering coefficient correction model. A random forest algorithm is adopted to effectively learn and fit a complex and high-order nonlinear relationship between a physical model simulation value and a satellite observation value.

Description

Technical Field

[0001] This invention relates to the field of parameter correction technology, and in particular to a method and apparatus for establishing a backscattering coefficient correction model. Background Technology

[0002] Spaceborne microwave scatterometers are key remote sensing sensors for acquiring parameters such as global sea surface wind fields.

[0003] Among related technologies, the polarimetric bidirectional reflectance distribution function (pBRDF) model, as a physical model, can be used to simulate the microwave scattering process of the sea surface and simulate the NRCS observed by active sensors; however, there is a significant deviation between the simulation results of this model and the actual satellite observations. Traditional deviation correction methods often rely on empirical linear regression or lookup tables. These methods struggle to effectively handle complex, nonlinear deviations caused by the combined effects of model defects, instrument calibration, and atmospheric transmission, resulting in limited correction accuracy and generalization ability. While CN119556247A constructs a calibration model dataset with a uniformly distributed azimuth angle and uses the least squares method to fit the system's nonlinear deviations for correction, the actual ocean wind field exhibits inherently non-uniform wind direction distribution. Therefore, the assumption of a uniform azimuth angle distribution easily introduces systematic errors. Furthermore, the least squares method is essentially a linear or quasi-linear fitting tool, which is insufficient for capturing and correcting complex, high-order nonlinear deviations caused by the coupled effects of multiple factors such as theoretical simplification of physical models, inaccurate instrument calibration, electromagnetic wave atmospheric transmission attenuation, and low-wind-speed model failure. Based on the above description, this method's correction accuracy and generalization ability are fundamentally limited when addressing deviations with complex causes.

[0004] Based on the above analysis of the development status of this technology field, the existing technologies lack a solution that combines physical models with data-driven methods to systematically correct the bias of elements in the polarization bidirectional reflection distribution function (pBRDF) based on the random forest machine learning algorithm. Summary of the Invention

[0005] The purpose of this invention is to provide a method and apparatus for establishing a backscattering coefficient correction model, aiming to solve the above-mentioned problems in the prior art.

[0006] According to a first aspect of the present invention, a method for establishing a backscattering coefficient correction model is provided, comprising: Acquire multi-source data including observations of backscattering coefficients; Spatiotemporal matching of multi-source data is performed to generate data units corresponding to each satellite observation pixel. Data units are then filtered using preset rules to obtain a retained dataset. Based on the retained dataset excluding observations, simulated values ​​are output by the physical model. A training set is constructed based on the simulated values ​​and the retained dataset, and the observed values ​​are used as the target variable of the training set. The machine learning model is trained using the training set, and the preset hyperparameters in the machine learning model are optimized by grid search to obtain the final backscattering coefficient correction model.

[0007] According to a second aspect of the present invention, an apparatus for establishing a backscattering coefficient correction model is provided, comprising: The multi-source loading module is used to acquire multi-source data, including observations of backscattering coefficients. The spatiotemporal matching module is used to perform spatiotemporal matching on multi-source data, generate data units corresponding to each satellite observation pixel, and filter the data units through preset rules to obtain the retained dataset. The physics simulation module is used to output simulated values ​​from a reserved dataset that does not include observations, construct a training set based on the simulated values ​​and the reserved dataset, and use the observations as the target variable of the training set. The model training module is used to train a machine learning model using a training set and to perform grid search optimization on the preset hyperparameters in the machine learning model to obtain the final backscattering coefficient correction model.

[0008] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method for establishing a backscattering coefficient correction model as provided in the first aspect of the present disclosure.

[0009] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which an information transmission implementation program is stored, which, when executed by a processor, implements the steps of the method for establishing a backscattering coefficient correction model provided in the first aspect of the present disclosure.

[0010] The technical solution provided by the embodiments of the present invention has the following beneficial effects: the random forest algorithm is used to effectively learn and fit the complex, high-order nonlinear relationship between the simulated values ​​of the physical model and the satellite observation values, breaking through the bottleneck of the traditional linear regression or lookup table method in dealing with composite biases, thereby achieving higher accuracy in systematic bias correction.

[0011] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of the method for establishing the backscattering coefficient correction model according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the implementation framework of an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the corrective effect of an embodiment of the present invention; Figure 4 This is a schematic diagram of the apparatus for establishing the backscattering coefficient correction model according to an embodiment of the present invention; Figure 5 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0015] Method Implementation Examples According to an embodiment of the present invention, a method for establishing a backscattering coefficient correction model is provided. Figure 1 This is a flowchart of the method for establishing the backscattering coefficient correction model according to an embodiment of the present invention, as follows: Figure 1 As shown, the method for establishing the backscattering coefficient correction model according to an embodiment of the present invention specifically includes: In step S110, multi-source data including the observed values ​​corresponding to the backscattering coefficients are acquired, specifically including: This process is implemented through the system's data input interface module, which is responsible for reading raw data from two sources from the storage medium; Read the L1 level data file (usually in HDF5 or NetCDF format) from the spaceborne scatterometer (such as ASCAT), and analyze the normalized radar backscattering cross section value as the observation value, denoted as . The angle of incidence is denoted as Azimuth is denoted as Latitude and longitude are recorded as and timestamp as The parameters of the spaceborne scatterometer were obtained. From reanalysis data (such as Fifth Generation of ECMWF Atmospheric Reanalyses for Climate, ERA5) and ocean model data (such as Hybrid Coordinate Ocean Model, HYCOM), obtain the following driving parameters that are spatiotemporally matched with satellite observations; the sea surface wind speed is denoted as... WS Sea wind direction is recorded as WD Sea surface temperature is denoted as SST and sea surface salinity is recorded as SSS To obtain the driving parameters; The parameters of the onboard scatterometer and the driving parameters are used together as multi-source data.

[0016] In step S120, spatiotemporal matching of multi-source data is performed to generate data units corresponding to each satellite observation pixel. The data units are then filtered using preset rules to obtain a retained dataset. Specifically, this includes: This process is implemented through an algorithm in the data preprocessing and matching module, which aims to align the two types of data in step S110 in time and space.

[0017] For current satellite observation pixels Based on its latitude, longitude, and timestamp, the parameter values ​​corresponding to the current location and time are calculated using bilinear interpolation, or further derivation is performed by finding its grid point in auxiliary data, which is also within the scope of protection of this application. Formula 1 is used to express the calculation of relative wind direction. : Formula 1; in, Indicates the wind direction at sea. Indicates the azimuth angle of satellite observation; Generate data units corresponding to each satellite observation pixel. That is, a preliminary data unit, in which, Represents the observed value. Indicates the angle of incidence. Indicates wind speed over the sea surface. Indicates sea surface temperature. Indicates sea surface salinity.

[0018] This process is used to remove invalid or low-quality data to ensure the reliability of subsequent training data. Iterate through each satellite observation pixel and perform the following judgment: if any condition in the preset rules is met, filter out the data cell. The preset rules include: Sea ice concentration is above the sea ice concentration threshold. , This represents the sea ice concentration threshold, which is set to 0 in this embodiment of the invention. Rainfall rate higher than rainfall rate threshold , The rainfall rate threshold is set to 0.2 mm / hr in this embodiment of the invention. Sea surface wind speed is less than the sea surface wind speed threshold. , The threshold value for sea surface wind speed is 3 m / s in this embodiment of the invention.

[0019] In the third preset rule mentioned above, the sea surface wind speed can be obtained directly from the data unit, while other parameters are obtained from their respective files.

[0020] In step S130, simulated values ​​are output from the physical model based on the retained dataset excluding observations. A training set is constructed based on the simulated values ​​and the retained dataset, using the observations as the target variable of the training set. Specifically, this includes: The preserved dataset, excluding observations, is used to obtain simulated values ​​of the physical model for NRCS under specific conditions through a pre-deployed pBRDF model. Simulated values ​​corresponding to the backscattering coefficients based on the physical model are then obtained using the pBRDF model according to its internal theoretical formula, namely the two-scale rough surface scattering model. ; The core output expression of the physics model: ; At this point, each data unit has been expanded to: .

[0021] The feature engineering module constructs the optimal input feature set, i.e., the training set, for the machine learning model; Define the target variable: satellite observations Defined as the target variable that the machine learning model needs to predict. ; Input feature selection: Through physical analysis and feature importance evaluation, the features most relevant to the bias are selected from all parameters to construct the input feature vector. ; Construct training set samples as ,in, , Sea surface temperature (SST) and salinity (SSS) were discarded during the construction of the training set, and all samples were discarded. The sets are combined to form a training set for machine learning.

[0022] In step S140, a machine learning model is trained using the training set, and the preset hyperparameters in the machine learning model are optimized using a grid search to obtain the final backscattering coefficient correction model, specifically including: The core of this invention's embodiments is bias correction through machine learning; The training set is used to train the random forest model as a machine learning model. Preferably, a predetermined number of samples are randomly selected from the preprocessed full data to form the training dataset of the machine learning model. The minimum number of leaf nodes is used as a preset hyperparameter for grid search optimization. The training set is divided into sampled samples and out-of-bag samples. Out-of-bag samples are samples that were not selected by the bootstrap sampling set. During the grid search optimization process, a random forest is constructed using the sampled samples. That is, a set of candidate values ​​for the minimum number of leaf nodes is preset as the grid. The best solution is selected by trying them one by one until the optimal solution is found. Finally, the candidate value that minimizes the out-of-bag error under the preset number of trees is selected as the optimal minimum number of leaf nodes.

[0023] The prediction error was calculated using formula 2 based on out-of-bag samples. : Formula 2; in, Indicates the number of samples outside the bag. Represents the target variable. Indicates the predicted value of samples outside the bag; After determining the optimal values ​​of the preset hyperparameters, training is performed using the entire training set data to execute the training process of the random forest algorithm. After training, a backscattering coefficient correction model is obtained, which encapsulates the complex nonlinear mapping relationship from input features to target correction values.

[0024] The method further includes: In step S150, the correction result is output and saved using the backscattering coefficient correction model, specifically including: The trained random forest model is applied to the data to be corrected to achieve batch and automated bias correction and obtain more accurate backscattering coefficients. For any NRCS data requiring calibration, including test and validation sets or the full dataset, the output of the calibration physical model is first generated by generating its corresponding feature vector. Then, the feature vector The input is fed into the trained random forest model; the random forest model performs forward computation through its internally integrated multiple decision trees, and averages the outputs of all decision trees to finally generate the corrected NRCS value. ).

[0025] The calculation process can be represented as follows: ; in, This represents the total number of decision trees in the model. Indicates the first A decision tree for input The predicted output.

[0026] This is achieved through the results output module, which is responsible for saving and displaying the final results; and for displaying the correction results. Together with the corresponding metadata, such as time, geographic location, original observations, and original simulations, it is written into a standard NetCDF data file to complete the entire calibration process.

[0027] The above technical solutions of the embodiments of the present invention will be illustrated with reference to the following accompanying drawings.

[0028] Figure 2 This is a schematic diagram of the implementation framework of an embodiment of the present invention, as shown below. Figure 2 As shown, the overall implementation framework is illustrated, including the training and application process. The main processes include observation calculation, spatiotemporal matching, quality control, simulated value calculation, training and application of the random forest model. Figure 3 This is a schematic diagram illustrating the corrective effect of an embodiment of the present invention, such as... Figure 3 As shown, the calibration effect is illustrated. The left figure compares the uncalibrated model simulation values ​​with the ASCAT scatterometer observation values, where the horizontal axis represents the ASCAT observation values ​​and the vertical axis represents the pBRDF simulation values. The right figure compares the calibrated model simulation values ​​with the ASCAT scatterometer observation values, where the horizontal axis represents the ASCAT observation values ​​and the vertical axis represents the calibrated simulation values. Bias represents the deviation value.

[0029] In summary, addressing the existing problems, this invention proposes a method for establishing a backscattering coefficient correction model. Employing the random forest algorithm, it effectively learns and fits the complex, high-order nonlinear relationship between simulated physical model values ​​and satellite observations, overcoming the limitations of traditional linear regression or lookup table methods in handling composite biases. This achieves higher-precision systematic bias correction. It does not rely on idealized assumptions such as uniform azimuth distribution, but learns intrinsic patterns from real, non-uniformly distributed data. The trained model exhibits stable correction performance for unknown data and different sea states, with generalization capabilities far superior to existing methods relying on specific distribution assumptions. The entire correction process is standardized and automated, enabling the model to efficiently and in batches process massive amounts of remote sensing data, significantly improving operational processing efficiency and result reproducibility. Furthermore, it can calibrate composite systematic biases introduced by multiple coupling factors based on constructed features.

[0030] Device Examples According to an embodiment of the present invention, an apparatus for establishing a backscattering coefficient correction model is provided. Figure 4 This is a schematic diagram of the apparatus for establishing the backscattering coefficient correction model according to an embodiment of the present invention, as shown below. Figure 4 As shown, the apparatus for establishing the backscattering coefficient correction model according to an embodiment of the present invention specifically includes: Multi-source loading module 40 is used to acquire multi-source data including observations corresponding to backscattering coefficients, specifically for: The normalized radar backscattering cross section value was analyzed as the observation value, and the incident angle, azimuth angle, latitude and longitude and timestamp were analyzed to obtain the parameters of the spaceborne scatterometer. By analyzing sea surface wind speed, sea surface wind direction, sea surface temperature, and sea surface salinity, the driving parameters can be obtained. The parameters of the onboard scatterometer and the driving parameters are used together as multi-source data.

[0031] The spatiotemporal matching module 42 is used to perform spatiotemporal matching on multi-source data, generate data units corresponding to each satellite observation pixel, and filter the data units through preset rules to obtain a retained dataset. Specifically, it is used for: For current satellite observation pixels Based on its latitude, longitude, and timestamp, the parameter values ​​corresponding to the current location and time are calculated using bilinear interpolation. Formula 1 is used to express the calculation of relative wind direction. : Formula 1; in, Indicates the wind direction at sea. Indicates the azimuth angle of satellite observation; Generate data units corresponding to each satellite observation pixel. ,in, Represents the observed value. Indicates the angle of incidence. Indicates wind speed over the sea surface. Indicates sea surface temperature. Indicates sea surface salinity.

[0032] The system iterates through all satellite observation pixels and filters out any data cell that meets any of the preset rules. The preset rules include: sea ice concentration is higher than the sea ice concentration threshold, rainfall rate is higher than the rainfall rate threshold, and sea surface wind speed is lower than the sea surface wind speed threshold.

[0033] The physics simulation module 44 is used to output simulated values ​​from a retained dataset that does not include observations, construct a training set based on the simulated values ​​and the retained dataset, and use the observations as the target variable of the training set. Specifically, it is used for: The retained dataset, excluding observations, is simulated using a pre-deployed pBRDF model based on its internal two-scale rough surface scattering model, yielding simulated values ​​for the backscattering coefficients corresponding to the physical model. ; Construct training set samples as ,in, , .

[0034] Model training module 46 is used to train a machine learning model using the training set and to perform grid search optimization on the preset hyperparameters in the machine learning model to obtain the final backscattering coefficient correction model. Specifically, it is used for: The random forest model was trained using the training set and used as the machine learning model. The minimum number of leaf nodes is used as a preset hyperparameter for grid search optimization. The training set is divided into sampled samples and out-of-bag samples. During the grid search optimization process, a random forest is constructed using the sampled samples, and the prediction error is calculated using Equation 2 with the out-of-bag samples. : Formula 2; in, Indicates the number of samples outside the bag. Represents the target variable. Indicates the predicted value of samples outside the bag; After determining the optimal values ​​of the preset hyperparameters, the model is trained to obtain the backscattering coefficient correction model.

[0035] The device further includes: The application module is used to output and save the correction results using the backscattering coefficient correction model.

[0036] In summary, addressing the existing problems, this invention presents a backscattering coefficient correction model establishment device. Employing the random forest algorithm, it effectively learns and fits the complex, high-order nonlinear relationship between simulated physical model values ​​and satellite observations, overcoming the limitations of traditional linear regression or lookup table methods in handling composite biases. This achieves higher-precision systematic bias correction. It does not rely on idealized assumptions such as uniform azimuth distribution, but learns intrinsic patterns from real, non-uniformly distributed data. The trained model exhibits stable correction performance on unknown data and different sea states, with generalization capabilities far superior to existing methods relying on specific distribution assumptions. The entire correction process is standardized and automated, enabling the model to efficiently and in batches process massive amounts of remote sensing data, significantly improving operational processing efficiency and result reproducibility. Furthermore, it can calibrate composite systematic biases introduced by multiple coupling factors based on constructed features.

[0037] Electronic device examples Figure 5This is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device 500 may include at least one processor 510 and a memory 520. The processor 510 can execute instructions stored in the memory 520. The processor 510 is communicatively connected to the memory 520 via a data bus. In addition to the memory 520, the processor 510 can also be communicatively connected to an input device 530, an output device 540, and a communication device 550 via the data bus.

[0038] Processor 510 can be any conventional processor, such as a commercially available CPU. Processors may also include graphics processing units (GPUs), field-programmable gate arrays (FPGAs), systems-on-chips (SoCs), application-specific integrated circuits (ASICs), or combinations thereof.

[0039] The memory 520 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0040] In this embodiment of the present disclosure, the memory 520 stores executable instructions, and the processor 510 can read the executable instructions from the memory 520 and execute the instructions to implement all or part of the steps of the method for establishing a backscattering coefficient correction model in any of the exemplary embodiments described above.

[0041] Computer-readable storage medium embodiments In addition to the methods and apparatus described above, exemplary embodiments of this disclosure may also be a computer program product or a computer-readable storage medium storing the computer program product, the computer product including computer program instructions that can be executed by a processor to implement all or part of the steps described in the method for establishing any of the backscattering coefficient correction models in the exemplary embodiments described above.

[0042] Computer program products can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. Programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages, and scripting languages ​​(e.g., Python). The program code can be executed entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0043] Computer-readable storage media may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media include: static random access memory (SRAM) having one or more electrically connected wires; electrically erasable programmable read-only memory (EEPROM); erasable programmable read-only memory (EPROM); programmable read-only memory (PROM); read-only memory (ROM); magnetic storage; flash memory; magnetic disk or optical disk; or any suitable combination thereof.

[0044] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for establishing a backscattering coefficient correction model, characterized in that, include: Acquire multi-source data including observations of backscattering coefficients; Spatiotemporal matching is performed on the multi-source data to generate data units corresponding to each satellite observation pixel. The data units are then filtered using preset rules to obtain a retained dataset. Based on a retained dataset excluding observations, simulated values ​​are output through a physical model. A training set is constructed based on the simulated values ​​and the retained dataset, and the observed values ​​are used as the target variable of the training set. The training set is used to train a machine learning model, and the preset hyperparameters in the machine learning model are optimized by grid search to obtain the final backscattering coefficient correction model.

2. The method according to claim 1, characterized in that, The method further includes: The correction results are output and saved using the backscattering coefficient correction model.

3. The method according to claim 1, characterized in that, The acquisition of multi-source data, including the observed values ​​corresponding to the backscattering coefficients, specifically includes: The normalized radar backscattering cross section value was analyzed as the observation value, and the incident angle, azimuth angle, latitude and longitude and timestamp were analyzed to obtain the parameters of the spaceborne scatterometer. By analyzing sea surface wind speed, sea surface wind direction, sea surface temperature, and sea surface salinity, the driving parameters can be obtained. The spaceborne scatterometer parameters and the driving parameters are used together as multi-source data.

4. The method according to claim 1, characterized in that, The step of performing spatiotemporal matching on the multi-source data to generate data units corresponding to each satellite observation pixel specifically includes: For current satellite observation pixels Based on its latitude, longitude, and timestamp, the parameter values ​​corresponding to the current location and time are calculated using bilinear interpolation. Formula 1 is used to express the calculation of relative wind direction. : Official 1; in, Indicates the wind direction at sea. Indicates the azimuth angle of satellite observation; Generate data units corresponding to each satellite observation pixel. ,in, Represents the observed value. Indicates the angle of incidence. Indicates wind speed over the sea surface. Indicates sea surface temperature. Indicates sea surface salinity.

5. The method according to claim 1, characterized in that, The step of filtering the data units using preset rules to obtain the retained dataset specifically includes: The data unit is filtered out if any of the preset rules are met, and the preset rules include: sea ice concentration is higher than the sea ice concentration threshold, rainfall rate is higher than the rainfall rate threshold, and sea surface wind speed is lower than the sea surface wind speed threshold.

6. The method according to claim 4, characterized in that, The construction of a training set based on the retained dataset (excluding observations) using a physical model to output simulated values, and then constructing the training set based on the simulated values ​​and the retained dataset, specifically includes: The retained dataset, excluding observations, is used in a pre-deployed pBRDF model. The pBRDF model, based on its internal two-scale rough surface scattering model, simulates the backscattering coefficients according to the physical model. ; Construct training set samples as ,in, , .

7. The method according to claim 1, characterized in that, The process of training a machine learning model using the training set and optimizing the preset hyperparameters in the machine learning model using a grid search to obtain the final backscattering coefficient correction model specifically includes: The training set is used to train a random forest model as a machine learning model. The minimum number of leaf nodes is used as a preset hyperparameter for grid search optimization. The training set is divided into sampled samples and out-of-bag samples. During the grid search optimization process, a random forest is constructed using the sampled samples, and the prediction error is calculated using Equation 2 with the out-of-bag samples. : Official 2; in, Indicates the number of samples outside the bag. Represents the target variable. Indicates the predicted value of samples outside the bag; After determining the optimal values ​​of the preset hyperparameters, the model is trained to obtain the backscattering coefficient correction model.

8. A device for establishing a backscattering coefficient correction model, characterized in that, include: The multi-source loading module is used to acquire multi-source data, including observations of backscattering coefficients. The spatiotemporal matching module is used to perform spatiotemporal matching on the multi-source data, generate data units corresponding to each satellite observation pixel, and filter the data units through preset rules to obtain a retained dataset. The physics simulation module is used to output simulated values ​​through a physics model based on a reserved dataset that does not include observations, construct a training set based on the simulated values ​​and the reserved dataset, and use the observations as the target variable of the training set. The model training module is used to train a machine learning model using the training set and to perform grid search optimization on the preset hyperparameters in the machine learning model to obtain the final backscattering coefficient correction model.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method for establishing the backscattering coefficient correction model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an information transmission implementation program, which, when executed by a processor, implements the steps of the method for establishing the backscattering coefficient correction model as described in any one of claims 1 to 7.

Images