An on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks

Abstract

This application provides an on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks, including: modeling the aperture brightness temperature of the microwave radiometer using a physical model; modeling the observed brightness temperature of the microwave radiometer using a neural network; and alternately optimizing the parameters of the aperture brightness temperature model and the observed brightness temperature model until the set optimization conditions are met. The advantages of this application are: improving the calibration accuracy and robustness of the model; the combined optimization strategy significantly improves the accuracy and efficiency while maintaining global optimization capabilities; and the alternating optimization strategy decomposes the complex joint optimization problem, which not only reduces the optimization difficulty and improves the computational efficiency, but also makes the convergence more stable.

Description

Technical Field

[0001] The present application belongs to the field of satellite remote sensing data processing technology, and specifically relates to an on-orbit joint calibration method for a satellite-borne microwave radiometer based on the fusion of a physical model and a neural network. Background Art

[0002] A fixed-beam pointing microwave radiometer is a non-scanning passive remote sensing device whose observation antenna is pointed in a fixed direction. This type of radiometer has two main application scenarios: 1) Observing targets in a specific direction; 2) Conducting co-ordinate observations with a radar altimeter, and receiving microwave radiation from the surface, atmosphere, and cold sky background to perform real-time corrections for the wet path delay caused by water vapor and liquid water in clouds when the radar altimeter signal passes through the Earth's atmosphere. The premise for a microwave radiometer (for simplicity, the microwave radiometers mentioned later in this application refer to fixed-beam pointing microwave radiometers) to be able to achieve the above applications is to accurately calibrate it. At present, the calibration of a microwave radiometer can be divided into two stages: antenna aperture brightness temperature T A Calibration and observation brightness temperature T B The antenna aperture brightness temperature T A The calibration is to convert the voltage value measured by the microwave radiometer into the brightness temperature of the antenna aperture; the brightness temperature T B The calibration refers to the antenna aperture brightness temperature T A Converted into the brightness temperature observed within the antenna main lobe field of view, that is, the brightness temperature of the target.

[0003] The observation antenna of a microwave radiometer typically consists of a parabolic reflector and a feed. The reflector primarily reflects the incident microwave signal back to the feed. Since the reflector receives microwave radiation from all targets within the antenna's illumination range, and the contribution of each target's microwave radiation to the aperture brightness temperature is related to the antenna's radiation pattern, microwave radiometers are typically calibrated on-orbit by analyzing the antenna pattern. Antenna pattern correction algorithms represent the observed brightness temperature of the antenna as a linear weighted product of the antenna efficiency and the effective brightness temperature of the corresponding component.

[0004] In the prior art, microwave radiometers are usually calibrated on-orbit in a serial manner, that is, the antenna aperture brightness temperature is calibrated first and then the observed brightness temperature is calibrated. Specifically, the calibration parameters obtained from the ground thermal vacuum test are introduced into the on-orbit observation data to obtain the antenna aperture brightness temperature, and then the observed brightness temperature is obtained through the antenna pattern correction algorithm. However, this serial method fails to fully utilize the complementarity of the two. In addition, due to the multi-solution nature of the aperture calibration equation, the accuracy of the antenna pattern correction algorithm modeling, the changes in the dynamic environment during the on-orbit operation of the microwave radiometer, and the lack of an effective global optimization strategy in the optimization process, the on-orbit calibration accuracy of the microwave radiometer still has a lot of room for improvement. Summary of the Invention

[0005] The purpose of this application is to address the above-mentioned shortcomings and propose an on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks, including:

[0006] Step 1: Use the physical model to model the brightness temperature of the microwave radiometer aperture;

[0007] Step 2: Use neural network to model the brightness temperature observed by microwave radiometer;

[0008] Step 3: Alternately optimize the parameters of the observation aperture model and the observation brightness temperature model until the set optimization conditions are met.

[0009] As an improvement to the above method, step 1 includes:

[0010] Establish the microwave radiometer aperture brightness temperature model:

[0011]

[0012]

[0013] in, and They represent the brightness temperature values of the aperture predicted by the observation branch and the low-temperature calibration branch respectively; a 1 、 a 2 、 a 3 、 c 1 、 c 2 、 c 3 and u are the physical model parameters to be optimized; T aw and T cw represent the physical temperatures of the waveguide in the observation branch and the low-temperature calibration branch respectively; T as and T cs Respectively represent the physical temperature of the switches in the observation branch and the low-temperature calibration branch; Q A Represents the nonlinear error of the aperture brightness temperature; V A 、 V H and V C Respectively represent the output voltage of the receiver on the observation branch and the high and low temperature calibration branches; represents the equivalent input brightness temperature of the high-temperature calibration reference source; represents the equivalent input brightness temperature of the low-temperature calibration reference source; represents the equivalent input brightness temperature of the observation branch; Q C Indicates the nonlinear error of the brightness temperature of the aperture in the low temperature calibration branch;

[0014] A ground thermal vacuum test was conducted, and multiple working temperature test points of the microwave radiometer were set. At each working temperature test point, when the low-temperature calibration variable temperature source was stable at the set temperature, the observation variable temperature source was heated or cooled with a step size S1 in the range of T1 K to T2 K, and the data set groundDataTa was obtained. When the observation variable temperature source was stable at the set temperature, the low-temperature calibration variable temperature source was heated or cooled with a step size S2 in the range of T1 K to T2 K, and the data set groundDataTc was obtained. At the same time, the temperature measurement data of the observation variable temperature source, the high and low temperature variable temperature sources, the corresponding output voltages, and the attenuation components on each transmission branch were collected. When V A =V H 、V A =V C and V H =V C The corresponding datasets groundDataVAVH, groundDataVAVC and groundDataVHVC are obtained respectively.

[0015] As an improvement to the above method, step 2 includes:

[0016] The on-orbit observation data of the microwave radiometer is used as the original data set, and the original data set is quality controlled and normalized to obtain the test data;

[0017] Divide the experimental dataset into training set and test set;

[0018] Choose between feedforward neural networks, convolutional neural networks, or recurrent neural networks, as well as their variants, to model the observed brightness temperature.

[0019] As an improvement to the above method, step 3 includes:

[0020] Step 3-1: Initialize the parameters of the aperture brightness temperature model and the observation brightness temperature model and set the upper and lower limits of each parameter;

[0021] Step 3-2: Alternately optimize the parameters of the surface brightness temperature model and the observation brightness temperature model, including:

[0022] Step 3-2-1: Fix the parameters of the observation brightness temperature model and optimize the parameters of the aperture brightness temperature model;

[0023] Step 3-2-2: Fix the parameters of the aperture brightness temperature model and optimize the parameters of the observation brightness temperature model;

[0024] Step 3-2-3: Calculate whether the total residual of the aperture brightness temperature model and the observation brightness temperature model meets the set stop optimization condition. If not, repeat steps 3-2-1 and 3-2-2 until the set stop optimization condition is met.

[0025] As an improvement to the above method, step 3-2-1 includes:

[0026] Define the objective function groundStageObjective, combine the ground thermal vacuum test data and the on-orbit observation data, and adjust the aperture brightness temperature model parameters to minimize the joint residual groundStageLoss of the ground and on-orbit data.

[0027] The objective function groundStageObjective is expressed as:

[0028]

[0029] in, m It represents the data volume of the dataset groundDataTa obtained when the low-temperature calibration variable temperature source is stable at the set temperature and the temperature of the observation variable temperature source is increased or decreased in the range of T1 K to T2 K with a step size S1; n It represents the data volume of the dataset groundDataTc obtained when the low-temperature calibration variable temperature source is heated or cooled in the range of T1 K to T2 K with a step size S2 when the observed variable temperature source is stable at the set temperature; Indicates the i The true value of the brightness temperature of the mouth of each observation branch; Indicates the i The brightness temperature of the mouth predicted by each observation branch; Indicates the j The true value of the brightness temperature of the aperture of the low-temperature calibration branch; Indicates the j The brightness temperature of the aperture predicted by the low-temperature calibration branch;

[0030] The joint residual groundStageLoss is expressed as:

[0031]

[0032] in, Loss groundDataTa and Loss groundDataTc They represent T during the ground thermal vacuum test. A Curve and T C The loss function of the curve; Loss orbitTrain represents the loss function of the on-orbit training set data;

[0033] The global optimization algorithm is used to globally optimize the parameters of the aperture brightness temperature model;

[0034] An optimization algorithm with strong constraints is used to locally optimize the parameters of the surface brightness temperature model. For the groundDataVAVH, groundDataVAVC, and groundDataVHVC datasets, the following strong constraints need to be met:

[0035] a 1 × T A + a 2 × T aw ≈ (1 - a 3 )× T as

[0036] a 1 × T A + a 2 × T aw + a 3 × T as ≈ c 1 × T C + c 2 × T cw + c 3 × T cs

[0037] c 1 × T C + c 2 × T cw ≈ (1 - c 3 )× T cs

[0038] in, T AIndicates oral surface brightness temperature; T C Indicates the temperature of the low-temperature calibration source.

[0039] As an improvement to the above method, step 3-2-2 includes:

[0040] Define the objective function nnStageObjective and use the latest aperture brightness temperature model parameters to generate the aperture brightness temperature prediction value of the observation antenna , combining the observed brightness temperature model parameters and on-orbit observation data, calculate the joint residual nnStageLoss of ground and on-orbit data;

[0041] The global optimization algorithm is used to train the observation brightness temperature model parameters so that the observed brightness temperature prediction value is close to the true value; the local optimization algorithm is used to locally optimize the observation brightness temperature model parameters;

[0042] The objective function nnStageObjective is expressed as:

[0043]

[0044] in, L represents the amount of observation data in the on-orbit training set; represents the true value of brightness temperature observed on orbit; Represents the brightness temperature prediction value of the radiometer on-orbit observation.

[0045] As an improvement to the above method, step 3-2-3 includes:

[0046] The total residual of the aperture brightness temperature model and the observation brightness temperature model is calculated, and the optimization ends when the set number of cycles is reached or the convergence threshold of the total residual reaches the set threshold.

[0047] As an improvement to the above method, the method further includes:

[0048] Step 4: Use the optimized parameters to construct the final apertural brightness temperature model and the observation brightness temperature model, test them on an independent test set, and evaluate the accuracy and generalization ability of the model.

[0049] Compared with the prior art, the advantages of this application are:

[0050] 1. A physical calibration model is used for the surface brightness temperature, while a neural network model is used for the observed brightness temperature. The physical model provides prior knowledge and physical constraints, while the neural network model uses a data-driven approach to learn the complex nonlinear relationships during on-orbit radiometer operation. This fusion approach provides physical constraints for the data-driven model, maintaining model interpretability while improving calibration accuracy and robustness.

[0051] 2. During the calibration of the aperture brightness temperature, the introduction of constraint conditions ensures the rationality of the physical parameters. This constraint mechanism not only improves the stability of the model, but also further enhances the optimization accuracy.

[0052] 3. Global optimization algorithms can effectively explore the solution space and avoid local optimality. Optimization with strong constraints and local optimization algorithms can perform detailed searches near the optimal solution found by the global optimization algorithm, further improving the accuracy of the solution. This combined optimization strategy significantly improves optimization accuracy and efficiency while maintaining global optimization capabilities.

[0053] 4. The alternating optimization strategy decomposes the complex joint optimization problem, which not only reduces the optimization difficulty and improves the computational efficiency, but also makes the convergence more stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 The figure shows the front-end structure of the microwave radiometer receiver.

[0055] Figure 2 The figure shows the data processing flow chart of microwave radiometer on-orbit calibration;

[0056] Figure 3 Shown is the probability density map of the MWR 18.7 GHz on-orbit calibration results in the test set (green: pure physics model, red: physics + neural network model). DETAILED DESCRIPTION

[0057] The technical solution of this application is described in detail below with reference to the accompanying drawings.

[0058] The present invention aims to provide an on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks. The aperture brightness temperature is modeled using a physical model, while the observed brightness temperature is modeled using a neural network. By alternately optimizing the physical model parameters and the neural network model parameters, combined with the strong constraints of ground thermal vacuum test data and the soft constraints of on-orbit observation data, the global optimization of the model parameters is achieved to calibrate the radiometer on-orbit, and its aperture brightness temperature T is obtained. A and the observed brightness temperature T B .

[0059] like Figure 2 As shown, the present invention provides an on-orbit joint calibration method for a spaceborne microwave radiometer based on the fusion of a physical model and a neural network. The method is as follows:

[0060] Step 1: Modeling the brightness temperature of the microwave radiometer aperture, including:

[0061] like Figure 1As shown, the microwave radiometer uses a periodic two-point calibration method for calibration. By periodically switching the internal switch of the receiver, the observation signal and the high and low temperature calibration source signals enter the receiver through different transmission paths. The transmission path usually includes microwave components such as the feed, connecting waveguide, and switch. For the observation branch and the low temperature calibration branch, respectively, there are:

[0062] (1)

[0063] (2)

[0064] in, and They represent the aperture calibration equations of the microwave radiometer observation branch and the cryogenic calibration branch based on the physical model respectively; and Respectively represent the input vectors of the calibration equations in the observation branch and the low-temperature calibration branch (including the physical temperature and output voltage of each microwave component, etc.); and They represent the brightness temperature values of the aperture predicted by the observation branch and the low-temperature calibration branch respectively.

[0065] During the ground thermal vacuum test, the microwave radiometer's observation variable temperature source and low-temperature calibration source (with known radiation characteristics, controllable and variable temperature, and a temperature range of Tl K to T2 K) are used to simulate on-orbit observation targets and low-temperature calibration reference sources, respectively. During the test, specific temperature test points are usually set for the working temperature of the microwave radiometer. At each temperature test point, when the low-temperature calibration variable temperature source is stable at the preset temperature, the observation variable temperature source is heated or cooled in the range of Tl K to T2 K with a step size of S1 to obtain the data set groundDataTa; when the observation variable temperature source is stable to the preset temperature, the low-temperature calibration variable temperature source is heated or cooled in the range of Tl K to T2 K with a step size of S2 to obtain the data set groundDataTc. At the same time, the data acquisition unit of the microwave radiometer will collect temperature measurement data of the observation variable temperature source and high and low temperature variable temperature sources, the corresponding output voltages, and each attenuation component on each transmission branch. In addition, during the cycle, by setting the voltage equalization point: V A =V H 、V A =V C and V H =V C The corresponding data sets groundDataVAVH, groundDataVAVC and groundDataVHVC are obtained respectively. A 、V H and V C Respectively represent the output voltage of the receiver on the observation branch and the high and low temperature calibration branches.

[0066] For the datasets groundDataTa and groundDataTc, equations (1) and (2) must be satisfied. For the datasets groundDataVAVH, groundDataVAVC, and groundDataVHVC, the nonlinear error in the observation branch or the cryogenic calibration branch of the microwave radiometer is minimal (nearly zero). At this point, the microwave radiometer system is nearly linear, and no correction for its nonlinear error is required. At this point, the predicted values of the aperture brightness temperature of the observation branch and the cryogenic calibration branch are as follows:

[0067] (3)

[0068] (4)

[0069] in, and They represent the aperture calibration equations without nonlinear correction established based on the physical model for the microwave radiometer observation branch and the low-temperature calibration branch respectively.

[0070] During ground-based thermal vacuum tests, the microwave radiometer's observations and cryogenic feeds are each aimed at a corresponding variable-temperature source. Therefore, the observations and brightness temperatures of the cryogenic calibration sources, both of which have known radiation characteristics, can be directly used as the "true" values of their respective aperture brightness temperatures.

[0071] Step 2: Modeling brightness temperature from microwave radiometer observations, including:

[0072] Step 2-1: Calculate the brightness temperature observed along track: The simulated brightness temperature along track of the microwave radiometer generated based on the radiation transfer model (or the observed brightness temperature at the cross-matching point with the benchmark payload based on the star cross comparison method) is used as the "true value" of the brightness temperature observed on track by the radiometer to be calibrated.

[0073] Step 2-2: Data Quality Control: Collect raw data and perform data cleaning (e.g., missing value processing, outlier detection, and duplicate data processing). Furthermore, since the observational accuracy of the radiative transfer model (or benchmark payload) is easily affected by factors such as surface type (land, sea, or ice), cloud thickness, and precipitation, it is necessary to further select "stable" data for the dataset.

[0074] Step 2-3: Dataset Partitioning: After quality control, randomly divide the data into a training set (orbitTrain), a validation set (orbitValidation), and a test set (orbitTest) in the ratio m:n:k (or a fixed ratio, etc.). The training set is used to calibrate the model parameters, the validation set is used to adjust the model parameters, and the test set is used to evaluate the model.

[0075] Step 2-4: Construct model input and output: Normally, the aperture brightness temperature of a microwave radiometer is a linear weighted sum of the antenna efficiency and the effective brightness temperature of the corresponding part. Unlike explicitly calculating the contribution of each part to the aperture brightness temperature, the neural network mainly establishes the relationship between the observed brightness temperature and the aperture brightness temperature through the main features of the sample. Therefore, the input of the observation brightness temperature model is data related to the effective brightness temperature of each part, such as the aperture brightness temperature of the antenna, the latitude of the satellite, and the physical temperature of the parabolic reflector. By calculating the correlation between each feature and the observed brightness temperature, the feature with higher correlation is selected as the input. The final selected input can be T A 、T A 2 , Tref (physical temperature of the reflecting surface), latitude lat, satellite cabin temperature, sea surface temperature, etc. The output of the model is the predicted value of the brightness temperature observed by the microwave radiometer.

[0076] Steps 2-5: Normalization: Input data may have different dimensions or ranges. To avoid "exploding gradients" caused by excessively large input values, it is usually necessary to normalize the network's input and output data so that all indicators are on the same dimension. Common normalization methods include Z-score normalization and Min-Max normalization.

[0077] Step 2-6: Select a network model: For microwave radiometer observation data, a feedforward neural network (or convolutional neural network, recurrent neural network, or its variants, such as LSTM and GRU) can be used to fit the data. At the same time, the corresponding network structure parameters need to be set (such as the number of hidden layers, number of neurons, and activation function in the feedforward neural network; the number, size, stride, and pooling method of the convolutional neural network).

[0078] Step 3: Alternately optimize and jointly solve the physical model parameters and the neural network parameters, including:

[0079] Step 3-1: Initialize the physical model parameters and neural network parameters and set upper and lower limits for each parameter. By establishing a brightness temperature radiation transmission model based on the physical parameters of the radiometer's microwave components (such as insertion loss and input and output voltage reflection coefficients), the upper and lower limits of each parameter in the physical model are calculated to ensure the rationality of each parameter during the optimization process. Neural network parameters primarily refer to the parameters that need to be set within the model's network structure.

[0080] Step 3-2: Execute the alternating optimization strategy: Alternately optimize the physical model parameters and the neural network model parameters, including the following steps:

[0081] Step 3-2-1: Fix the neural network parameters and optimize the physical model parameters:

[0082] By defining the objective function groundStageObjective, combining the ground thermal vacuum test data and on-orbit observation data, and adjusting the physical model parameters, the joint residual groundStageLoss of the ground and on-orbit data is minimized.

[0083] The objective function groundStageObjective is expressed as:

[0084] (5)

[0085] in, m It represents the data volume of the dataset groundDataTa obtained when the low-temperature calibration variable temperature source is stable at the set temperature and the temperature of the observation variable temperature source is increased or decreased in the range of T1 K to T2 K with a step size S1; n It represents the data volume of the dataset groundDataTc obtained when the low-temperature calibration variable temperature source is heated or cooled in the range of T1 K to T2 K with a step size S2 when the observed variable temperature source is stable at the set temperature; Indicates the i The true value of the brightness temperature of the observation branch; Indicates the i The brightness temperature of the mouth predicted by each observation branch; Indicates the j The true value of the brightness temperature of the aperture of the low-temperature calibration branch; Indicates the j The brightness temperature of the aperture predicted by the low-temperature calibration branch. (6)

[0086] in, Loss groundDataTa and Loss groundDataTc They represent T during the ground thermal vacuum test. A Curve and T C The loss function of the curve; Loss orbitTrain Represents the loss function of the on-orbit training set data. The loss function can be the root mean square error or the mean error.

[0087] Use global optimization algorithms (such as whale optimization algorithm, genetic algorithm, etc.) to globally optimize the physical model parameters.

[0088] The physical model parameters are optimized using an optimization algorithm with strong constraints: for the groundDataVAVH, groundDataVAVC, and groundDataVHVC datasets, the approximate equations need to be satisfied respectively. 、 and Strong constraints:

[0089] : a 1 × T A + a 2 × T aw ≈ (1 - a 3 )× T as (7)

[0090] : a 1 × T A + a 2 × T aw + a 3 × T as ≈ c 1 × T C + c 2 × T cw + c 3 × T cs (8)

[0091] : c 1 × T C + c 2 × T cw ≈ (1 - c 3 )× T cs (9)

[0092] in, T A Indicates oral surface brightness temperature; T C Indicates the temperature of the low-temperature calibration source. The approximate meaning of the above formula is that the difference between the values at both ends of the formula is within the set range.

[0093] Step 3-2-2: Fix the physical model parameters and optimize the neural network parameters:

[0094] By defining the objective function nnStageObjective, the latest physical model parameters are used to generate the predicted brightness temperature of the observation antenna. , combining the network parameters and on-orbit observation data, and calculating the joint residual nnStageLoss of ground and on-orbit data.

[0095] The objective function nnStageObjective is expressed as:

[0096] (10)

[0097] in, L represents the amount of observation data in the on-orbit training set; represents the true value of brightness temperature observed on orbit; Represents the brightness temperature prediction value of the radiometer on-orbit observation.

[0098] Use global optimization algorithms (such as whale optimization algorithm, genetic algorithm, ant colony algorithm, etc.) to train the neural network parameters so that the observed brightness temperature prediction value is close to the "true value".

[0099] Use local optimization algorithms (such as gradient descent algorithm, Levenberg-Marquardt algorithm LM, etc.) to locally optimize the neural network parameters.

[0100] Step 3-2-3: Calculate the total residual totalLoss (groundStageLoss + nnStageLoss) and check whether it converges. If the convergence condition is met, stop the optimization; otherwise, continue to the next round of iteration and continue to execute steps 3-2-1) to 3-2-3) until convergence.

[0101] Step 4: Model output verification and evaluation. This involves using the optimized parameters to construct the final model and obtain the radiometer's aperture and observed brightness temperatures. Model parameters can be continuously adjusted on the validation set, and model performance can be evaluated on the test set.

[0102] The on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks provided in this application has the following advantages:

[0103] 1. The aperture brightness temperature of the microwave radiometer is modeled using a physical model, and the observed brightness temperature is modeled using a neural network model. This breaks through the bottleneck of the calibration accuracy of the traditional pure physical model and overcomes the problem of poor interpretability of the pure neural network model. The fusion method of the present invention achieves a good balance between accuracy and interpretability.

[0104] 2. Bidirectional Alternating Optimization Framework: This paper proposes an alternating iterative mechanism featuring strong constraints on ground data, soft constraints on on-orbit data, and joint residual feedback. By decomposing the complex joint optimization problem, it not only reduces the optimization difficulty and improves computational efficiency, but also ensures more stable convergence.

[0105] 4. Hybrid Optimization Algorithm: Ground data is optimized using a combination of global optimization and a nonlinear algorithm with strong constraints. Neural network models are optimized using a combination of global and local optimization algorithms. The global optimization algorithm effectively explores the solution space and avoids local optima. The nonlinear optimization algorithm with strong constraints and the local optimization algorithm enable a detailed search near the optimal solution found by the global optimization algorithm, further improving solution accuracy. This combined optimization strategy significantly improves optimization accuracy and efficiency while maintaining global optimization capabilities.

[0106] Simulation experiment:

[0107] Assume that the triple-frequency fixed-pointing microwave radiometer MWR is operating at an altitude of 970 km and an inclination of 66°. The observation frequencies of MWR are 18.7 GHz, 23.8 GHz and 37 GHz. The data products of MWR are released in a cycle of 10 days. The observation frequency of MWR L2 data is 1 second. Figure 2 It can be seen that the main steps of the on-orbit calibration of the microwave radiometer include: aperture brightness temperature modeling based on the physical model, preprocessing / quality control / dataset division / normalization of on-orbit data, observation brightness temperature modeling based on neural network, and alternating optimization solution of calibration parameters.

[0108] Modeling of the brightness temperature of the mouth surface based on the physical model:

[0109] (1) Ideally, the receiver of a microwave radiometer is linear, and the relationship between the input brightness temperature and the output voltage of the receiver is as follows:

[0110] (11)

[0111] in, V N Indicates the output voltage of the receiver; a and b is the calibration factor of the radiometer; T N ' represents the equivalent input brightness temperature of the receiver; N =( C 、 A 、 H ) represents the different brightness temperature signals received by the receiver, where C Indicates the cold source, A Indicates observation,H Represents a heat source. Use the high and low temperature calibration reference source’s equivalent input brightness temperature T H ' and T C ' and the corresponding output voltage V H and V C The calibration coefficient of the radiometer can be obtained a and b , and then the equivalent input brightness temperature of the observation branch is obtained T A ' :

[0112] (12)

[0113] However, in actual operation, the nonlinearity of the power response of the detector diode in the radiometer will cause nonlinear errors in the system. For this nonlinear error, the U coefficient method proposed by the Jet Propulsion Laboratory (JPL) in the United States can be used to correct it. The calculation formula is as follows:

[0114] (13)

[0115] in, Q A Represents the nonlinear error of the aperture brightness temperature; u Used to represent the nonlinearity of the radiometer system; G Represents the system gain. For the observation branch, when the brightness temperature is T A After the input signal enters the feed and passes through microwave components such as waveguides, isolators and switches, the equivalent input brightness temperature at the receiver input port is:

[0116] (14)

[0117] In summary, the predicted brightness temperature of the observation antenna is for:

[0118] (15)

[0119] Similarly: For the low temperature calibration branch, the predicted value of the brightness temperature of the aperture is for:

[0120] (16)

[0121] in, a 1 、 a 2 、a 3 、 c 1 、 c 2 、 c 3 and u are the physical model parameters to be optimized; T aw and T cw represent the physical temperatures of the waveguide in the observation branch and the low-temperature calibration branch, respectively. T as and T cs Respectively represent the physical temperature of the switches in the observation branch and the low-temperature calibration branch; Q C It represents the nonlinear error of the brightness temperature of the aperture in the low temperature calibration branch.

[0122] (2) In the ground thermal vacuum test, the temperature of the observation variable temperature source and the low temperature calibration source with controllable temperature are cycled (for example, when the low temperature calibration source is stabilized at 90 K, the observation variable temperature source is changed from 90 K to 330 K; or when the observation variable temperature source is stabilized at 90 K, the low temperature calibration source is changed from 90 K to 330 K), and the data set groundDataTa (T A curve) and groundDataTc (Tc curve). In addition, during the temperature cycle, there will be a voltage equalization point (V A = V H 、V C = V C or V H = V C ), you can get the voltage equal point data sets groundDataVAVH, groundDataVAVC and groundDataVHVC.

[0123] (3) The observation and cryogenic feed sources of the radiometer are respectively aligned with the corresponding variable temperature sources. Therefore, the brightness temperature values of the observation and cryogenic calibration sources with known radiation characteristics can be directly used as the "true values" of their respective aperture brightness temperatures.

[0124] Preprocessing / quality control / dataset partitioning / normalization of on-orbit data:

[0125] Data preprocessing: The on-orbit observation data of the radiometer cycle 003, cycle 013, cycle 022 and cycle 031 are used as the original data set, and missing value processing, outlier detection and duplicate data processing are performed on the data.

[0126] Quality control: To obtain “stable” on-orbit observation data, the data quality control indicators are as follows: 1) Use land and sea markers to remove observation points located on land; 2) Use sea ice markers to remove observation points in sea ice-covered areas; 3) Limit the latitude of observation points to between 60°N and 60°S; 4) Eliminate clouds with liquid water content greater than 0.1 kg / m 2 5) Eliminate data points where the deviation between the radiometer aperture brightness temperature and the simulated brightness temperature is greater than 20 K.

[0127] Dataset Partitioning: After quality control, the data is randomly divided into a training set and a test set in a ratio of 2:1. The training set data and the thermal vacuum test data are used to solve the calibration model parameters. The test set is used to verify the solved calibration model parameters.

[0128] Normalization: The Min-Max method is used to normalize the input and output data.

[0129] Observational brightness temperature modeling based on neural network:

[0130] Model input and output: The network input contains 4 nodes: T A 、T A 2 , Tref (physical temperature of the reflecting surface), and latitude lat. The network output is the predicted brightness temperature observed by the radiometer. This application uses the fifth-generation reanalysis data product ERA-5 provided by ECMWF to simulate the along-track brightness temperature of the radiometer, which is used as the "true value" of the MWR brightness temperature observation. The atmospheric millimeter wave radiation transfer model MPM93 is used to calculate atmospheric coefficients, and the fast microwave emissivity model FASTEM-5 is used to calculate the ocean surface emissivity.

[0131] Model parameter settings: This application uses a feedforward neural network to fit the MWR observation data. The relevant parameters are set as follows: a 4-layer neural network (1 input layer, 2 hidden layers, and 1 output layer); the number of nodes in each layer is 4, 6, 3, and 1 respectively; the activation functions are tansig (hyperbolic tangent), tansig, and linear respectively; and the network's initial weights and biases are random numbers between 0 and 1.

[0132] Alternating optimization solution of calibration parameters:

[0133] Parameter initialization and upper and lower limit setting: establish a brightness temperature radiation transmission model for the physical parameters (insertion loss, input and output voltage reflection coefficient) on the MWR observation branch and the low temperature calibration branch, and calculate the physical model parameters ( a 1 、 a 2 、 a 3、 c 1 、 c 2 、 c 3 、 u The number of alternating optimizations is set to 20, and the convergence threshold is set to 1e-4.

[0134] Phase 1: Fix the neural network model parameters and optimize the physical model parameters using the whale optimization algorithm and then a constrained nonlinear optimization algorithm (the corresponding approximate equality constraints that the voltage equalization point dataset must satisfy) to minimize the loss function of the ground stage objective function, groundStageObjective.

[0135] (17)

[0136] in, i represents the data points collected when the low-temperature calibration source is kept at 90 K and the variable temperature source is observed to change from 90 K to 330 K (T A curve data points); j represents the data points collected when the low-temperature calibration source changes from 90 K to 330 K while the observation variable temperature source is kept at 90 K (T C curve data points).

[0137] Phase 2: Fixed physical model parameters and generated using the latest physical model parameters , the whale optimization algorithm and LM local optimization algorithm are used to train the parameters (weights and biases) of the neural network, so that the observed brightness temperature values predicted by the radiometer are close to the "true value".

[0138] Stage 3: Calculate the total residual from stages 1 and 2 and check the convergence conditions.

[0139] Repeat stage 1 to stage 3 until the total residual converges or the maximum number of iterations is reached.

[0140] Model validation: The final output model (outputting the MWR's surface brightness temperature and observed brightness temperature) was constructed using the optimized parameters and tested on an independent test set to evaluate the model's accuracy and generalization ability.

[0141] from Figure 3 The figure shows the probability density distribution curves of the deviations (OMB) from the true observed brightness temperature (simulated brightness temperature) for the MWR 18.7 GHz channel in the test set, obtained using a purely physical model (TbPhy) (green) and a hybrid model (physics + neural network) (TbMix) (red). The figure shows that the average deviation of OMB is smaller after calibration using the hybrid model.

[0142] In order to quantitatively evaluate the brightness temperature of the MWR after on-orbit calibration, the statistical results of the OMB deviation of the MWR 18.7 GHz channel in the training set and the test set are listed in Table 1. Among them, the mean absolute deviation MAD, standard deviation STD, root mean square error RMSE and goodness of fit R 2 They are used to measure the brightness temperature prediction value of MWR observation and the brightness temperature T along the track simulation. B,simu The average difference, dispersion of the difference, accuracy, and goodness of fit between the two groups are shown in Table 1. As can be seen from Table 1, the OMB bias of the MWR 18.7 GHz channel is reduced after hybrid model calibration, indicating that the systematic bias between the observed and simulated brightness temperatures of the MWR can be eliminated after on-orbit hybrid model calibration. After on-orbit hybrid model calibration, the MAD, STD, and RMSE values of the OMB in the 18.7 GHz channel test set are 1.16 K, 1.49 K, and 1.48 K, respectively. Compared with the MAD, STD, and RMSE values of the OMB bias between the observed and simulated brightness temperatures of the MWR after calibration using a pure physical model, these values decrease by 31.4%, 3.2%, and 27.8%, respectively. These results demonstrate that this on-orbit calibration algorithm can improve the calibration accuracy and stability of the MWR on-orbit observed brightness temperature.

[0143] Table 1: OMB statistics (K) of the 18.7 GHz channel after on-orbit MWR calibration for the training and test sets (TbPhy represents the observed brightness temperature after calibration using a pure physical model; TbMix represents the observed brightness temperature after calibration using a hybrid physical and neural network model)

[0144]

[0145] Finally, it should be noted that the above embodiments are intended only to illustrate the technical solutions of this application and are not intended to limit the scope of the present invention. Although this application has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions to the technical solutions of this application do not depart from the spirit and scope of the technical solutions of this application and should be encompassed by the claims of this application.

Claims

1. An on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks, comprising: Step 1: Modeling the aperture brightness temperature of the microwave radiometer using a physical model; the physical model includes the aperture calibration equations of the microwave radiometer observation branch and the low-temperature calibration branch; Step 2: Use a neural network to model the brightness temperature observed by the microwave radiometer. The input of the neural network includes the brightness temperature of the aperture, the physical temperature of the reflecting surface, and the satellite latitude. The output is the predicted value of the brightness temperature observed by the microwave radiometer. Step 3: Alternately optimize the parameters of the aperture brightness temperature model and the observation brightness temperature model until the set optimization conditions are met; The step 3 includes: Step 3-1: Initialize the parameters of the aperture brightness temperature model and the observation brightness temperature model and set the upper and lower limits of each parameter; Step 3-2: Alternately optimize the parameters of the surface brightness temperature model and the observation brightness temperature model, including: Step 3-2-1: Fix the parameters of the observation brightness temperature model and optimize the parameters of the aperture brightness temperature model; Step 3-2-2: Fix the parameters of the aperture brightness temperature model and optimize the parameters of the observation brightness temperature model; Step 3-2-3: Calculate whether the total residual of the aperture brightness temperature model and the observation brightness temperature model meets the set stop optimization condition. If not, repeat steps 3-2-1 and 3-2-2 until the set stop optimization condition is met. The step 3-2-1 includes: defining an objective function groundStageObjective, combining ground thermal vacuum test data and on-orbit observation data, adjusting the aperture brightness temperature model parameters so that the joint residual groundStageLoss of the ground and on-orbit data is minimized; The joint residual groundStageLoss is expressed as: ; Among them, LossgroundDataTa and LossgroundDataTc represent the ground thermal vacuum test process T A Curve and T C The loss function of the curve; LossorbitTrain represents the loss function of the on-track training set data; The step 3-2-2 includes: Define the objective function nnStageObjective and use the latest aperture brightness temperature model parameters to generate the aperture brightness temperature prediction value of the observation antenna , combining the observed brightness temperature model parameters and on-orbit observation data, calculate the joint residual nnStageLoss of ground and on-orbit data; The total residual in step 3-2-3 is the sum of groundStageLoss and nnStageLoss.

2. The on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks according to claim 1 is characterized in that: The step 1 comprises: Establish the microwave radiometer aperture brightness temperature model: ; ; in, and They represent the brightness temperature values of the aperture predicted by the observation branch and the low-temperature calibration branch respectively; a 1 、 a 2 、 a 3 、 c 1 、 c 2 、 c 3 and u are the physical model parameters to be optimized; T aw and T cw represent the physical temperatures of the waveguide in the observation branch and the low-temperature calibration branch respectively; T as and T cs Respectively represent the physical temperature of the switches in the observation branch and the low-temperature calibration branch; Q A Represents the nonlinear error of the aperture brightness temperature; V A 、 V H and V C Respectively represent the output voltage of the receiver on the observation branch and the high and low temperature calibration branches; represents the equivalent input brightness temperature of the high-temperature calibration reference source; represents the equivalent input brightness temperature of the low-temperature calibration reference source; represents the equivalent input brightness temperature of the observation branch; Q C Indicates the nonlinear error of the brightness temperature of the aperture in the low temperature calibration branch; A ground thermal vacuum test was conducted, and multiple working temperature test points of the microwave radiometer were set. At each working temperature test point, when the low-temperature calibration variable temperature source was stable at the set temperature, the observation variable temperature source was heated or cooled with a step size S1 in the range of T1 K to T2 K, and the data set groundDataTa was obtained. When the observation variable temperature source was stable at the set temperature, the low-temperature calibration variable temperature source was heated or cooled with a step size S2 in the range of T1 K to T2 K, and the data set groundDataTc was obtained. At the same time, the temperature measurement data of the observation variable temperature source, the high and low temperature variable temperature sources, the corresponding output voltages, and each attenuation component on each transmission branch were collected. When V A =V H 、V A =V C and V H =V C The corresponding datasets groundDataVAVH, groundDataVAVC and groundDataVHVC are obtained respectively.

3. The on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks according to claim 1 is characterized in that: The step 3-2-1 further includes: The objective function groundStageObjective is expressed as: ; in, m It represents the data volume of the dataset groundDataTa obtained when the low-temperature calibration variable temperature source is stable at the set temperature and the temperature of the observation variable temperature source is increased or decreased in the range of T1 K to T2 K with a step size S1; n It represents the data volume of the dataset groundDataTc obtained when the low-temperature calibration variable temperature source is heated or cooled in the range of T1 K to T2 K with a step size S2 when the observed variable temperature source is stable at the set temperature; Indicates the i The true value of the brightness temperature of the observation branch; Indicates the i The brightness temperature of the mouth predicted by each observation branch; Indicates the j The true value of the brightness temperature of the aperture of the low-temperature calibration branch; Indicates the j The brightness temperature of the aperture predicted by the low-temperature calibration branch; The global optimization algorithm is used to globally optimize the parameters of the aperture brightness temperature model; An optimization algorithm with strong constraints is used to locally optimize the parameters of the surface brightness temperature model. For the groundDataVAVH, groundDataVAVC, and groundDataVHVC datasets, the following strong constraints need to be met: a 1 × T A + a 2 × T aw ≈ (1 - a 3 )× T as a 1 × T A + a 2 × T aw + a 3 × T as ≈ c 1 × T C + c 2 × T cw + c 3 × T cs c 1 × T C + c 2 × T cw ≈ (1 - c 3 )× T cs in, T A Indicates oral surface brightness temperature; T C Indicates the temperature of the low-temperature calibration source.

4. The on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks according to claim 1 is characterized in that: The step 3-2-2 further includes: The global optimization algorithm is used to train the observation brightness temperature model parameters so that the observed brightness temperature prediction value is close to the true value; the local optimization algorithm is used to locally optimize the observation brightness temperature model parameters; The objective function nnStageObjective is expressed as: ; in, L represents the amount of observation data in the on-orbit training set; T Bk represents the true value of brightness temperature observed on orbit; Represents the brightness temperature prediction value of the radiometer on-orbit observation.

5. The on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks according to claim 1 is characterized in that: The step 3-2-3 includes: The total residual of the aperture brightness temperature model and the observation brightness temperature model is calculated, and the optimization ends when the set number of cycles is reached or the convergence threshold of the total residual reaches the set threshold.

6. The on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks according to claim 1 is characterized in that: The step 2 includes: The on-orbit observation data of the microwave radiometer is used as the original data set, and the original data set is quality controlled and normalized to obtain the test data; Divide the experimental dataset into training set and test set; Choose between feedforward neural networks, convolutional neural networks, or recurrent neural networks, as well as their variants, to model the observed brightness temperature.

7. The on-orbit joint calibration method for spaceborne microwave radiometers based on the fusion of physical models and neural networks according to claim 1 is characterized in that: Also includes: Step 4: Use the optimized parameters to construct the final apertural brightness temperature model and the observation brightness temperature model, test them on an independent test set, and evaluate the accuracy and generalization ability of the model.

Images