Artificial intelligence automatic mining microwave rainforest radiation field radiation modeling method and device

Abstract

The invention discloses an artificial intelligence automatic mining microwave rainforest radiation field radiation modeling method and device, and belongs to the technical field of electronic information, and the method comprises the steps: selecting a series of time points, and obtaining the measurement values of an internal environment temperature field and a humidity field of a rainforest; obtaining an observation geometric angle when the external microwave radiation reference source observes the rainforest and an observation result of the external microwave radiation reference source; taking measured values of an internal environment temperature field and a humidity field of the rainforest and an observation geometric angle when the rainforest is observed at the microwave radiation reference source as input of a neural network, taking an observation result of an external microwave radiation reference source as output of the neural network, and training the neural network; inputting a rainforest temperature field measurement value, a rainforest humidity field measurement value, an observation zenith angle and an azimuth angle into the trained model, and obtaining a calibration result of the remote sensing instrument at any moment through network output. The method can provide theoretical radiation values of the remote sensing instrument under any working condition and different observation geometric conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic information, and in particular to a method and device for automatically mining microwave rainforest radiation field radiation modeling by artificial intelligence. Background Art

[0002] As a surface radiation calibration target for microwave space remote sensing instruments, rainforests have the advantages of stable brightness temperature and relatively high degree of isotropic radiation isotropy. The Amazon rainforest is a commonly used radiation calibration field for microwave remote sensing instruments. The rainforest in Simao, Yunnan, China is a ground-based radiation calibration field that is widely used by domestic satellites. Calibration based on radiation correction fields often adopts the same-pupil observation method. When remote sensing satellites observe radiation correction fields, they use near-simultaneous observations of reference satellites and ground observation equipment to form radiation correction field observation data with radiation references to calibrate the remote sensing instruments to be calibrated.

[0003] However, in fact, for microwave remote sensing, when a specific microwave remote sensing instrument observes a rainforest, the microwave penetration determines the observed radiation (or brightness temperature), which is a complex mixture of multiple factors such as surface emissivity, rainforest temperature environment, humidity environment, canopy emissivity, etc. according to the weights formed by different observation conditions. This complex process leads to a relatively large observation uncertainty of a single microwave remote sensing instrument on the rainforest calibration field. Although the rainforest is a relatively good ground-based radiation calibration field for microwave remote sensing instruments, there is still a certain gap for high-precision calibration requirements.

[0004] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention and should not be regarded as an acknowledgment or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art. Summary of the invention

[0005] The purpose of the present invention is to provide a microwave rainforest radiation field radiation modeling method and device automatically mined by artificial intelligence to solve some or all of the above problems.

[0006] To achieve the above object, the present invention provides a microwave rainforest radiation field radiation modeling method automatically mined by artificial intelligence, comprising the following steps:

[0007] S1: Select a series of time points to obtain the measured values ​​of the temperature field and humidity field of the rainforest environment at the series of time points; and obtain the observation geometry angle when an external microwave radiation reference source observes the rainforest and the observation results of the external microwave radiation reference source;

[0008] S2: Take the measured values ​​of the temperature field and humidity field of the rainforest environment at a certain time point and the observed geometric angle when the rainforest is observed by the microwave radiation reference source as the input of the neural network, and take the observation result of the external microwave radiation reference source corresponding to the time point as the output of the neural network to train the neural network;

[0009] S3: Input the rainforest temperature field measurement, rainforest humidity field measurement, observation zenith angle and azimuth angle into the trained model, and obtain the calibration results of the remote sensing instrument at any time through the network output.

[0010] In one embodiment of the present invention, the observation geometric angles of the external microwave radiation reference source when observing the rainforest include the zenith angle and the azimuth angle.

[0011] In one embodiment of the present invention, step S2 specifically includes the following steps:

[0012] S201: The measured values ​​of the temperature field in the rainforest environment [T1, T2, ..., T M ], rainforest humidity field measurement values ​​[P1, P2, …, P N ], observation zenith angle β i and azimuth α i The observation result R of the external microwave radiation reference source corresponding to the time point is taken as input. i The output of the neural network is the label;

[0013] S202: Normalize all inputs and labels;

[0014] S203: creating a fully connected type including an input layer, at least one hidden layer and an output layer as a neural network model to be trained;

[0015] S204: Train the neural network model using the normalized input and labels.

[0016] In one embodiment of the present invention, step S304 specifically includes the following steps: During the training process, the forward propagation calculates the predicted value y output by the neural network pred , according to the predicted value y pred and the true label y to calculate the loss value of the current model; according to the loss value in each round of training, calculate and record the gradient, then clear the gradient and execute the back propagation algorithm according to the loss value to calculate the new gradient value of the model parameters, and adjust the parameters of the model according to the calculated new gradient value; repeat the above operation until the loss value is less than the set threshold and stop the training process.

[0017] In one embodiment of the present invention, in step S304:

[0018] The loss function is In the form of pred represents the predicted value of the neural network output, and y represents the cross-calibration result value obtained in step S2.

[0019] The present invention also provides a microwave rainforest radiation field radiation modeling device automatically mined by artificial intelligence, including: a data processing module, a training module and a data acquisition module;

[0020] The data processing module is used to select a series of time points to obtain the measured values ​​of the temperature field and humidity field of the rainforest environment at the series of time points; and to obtain the observation geometric angle when the external microwave radiation reference source observes the rainforest and the observation results of the external microwave radiation reference source;

[0021] The training module is used to use the measured values ​​of the temperature field and humidity field of the rainforest environment at a certain time point and the observed geometric angle when the rainforest is observed by the microwave radiation reference source as the input of the neural network, and the observation result of the external microwave radiation reference source corresponding to the time point as the output of the neural network to train the neural network;

[0022] The data acquisition module is used to input the rainforest temperature field measurement values, rainforest humidity field measurement values, observation zenith angles and azimuth angles into the trained model, and obtain the calibration results of the remote sensing instrument at any time through network output.

[0023] In one embodiment of the present invention, the observation geometric angles of the external microwave radiation reference source when observing the rainforest include the zenith angle and the azimuth angle.

[0024] In one embodiment of the present invention, the training module includes: an input submodule, a processing submodule, a model creation submodule and a training submodule.

[0025] The input submodule is used to input the temperature field measurement values ​​[T1, T2, ..., T M ], rainforest humidity field measurement values ​​[P1, P2, …, P N ], observation zenith angle β i and azimuth α i The observation result R of the external microwave radiation reference source corresponding to the time point is taken as input. i The output of the neural network is the label;

[0026] The processing submodule is used to normalize all inputs and labels;

[0027] The model creation submodule is used to create a fully connected type neural network model to be trained, which includes an input layer, at least one hidden layer and an output layer;

[0028] The training submodule is used to train the neural network model using the normalized input and labels.

[0029] In one embodiment of the present invention, the training submodule is specifically used for: during the training process, forward propagation calculates the predicted value y_pred output by the neural network, and calculates the loss value of the current model according to the predicted value y_pred and the true label y; according to the loss value in each round of training, the gradient is calculated and recorded, and then the gradient is cleared and the back propagation algorithm is executed according to the loss value to calculate the new gradient value of the model parameter, and the parameters of the model are adjusted according to the calculated new gradient value; the above operation is repeated until the loss value is less than the set threshold, and the training process is stopped.

[0030] In one embodiment of the present invention, the loss function is In the form of pred represents the predicted value of the neural network output, and y represents the cross-calibration result value obtained in step S2.

[0031] Compared with the prior art, the microwave rainforest radiation field radiation modeling method and device automatically mined by artificial intelligence according to the present invention greatly improves the efficiency and accuracy of modeling by utilizing the powerful learning ability of neural networks and automatically mining the complex relationship between the external radiation performance of the rainforest and the indicator parameters through artificial intelligence. Artificial intelligence can process a large amount of data and quickly find the laws and relationships therein, reducing human intervention and errors, and ensuring that the theoretical radiation value of the remote sensing instrument under different observation geometric conditions can be obtained under any working condition (i.e., the state represented by the rainforest temperature field and humidity field). BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a flow chart of a microwave rainforest radiation field radiation modeling method automatically mined by artificial intelligence according to an embodiment of the present invention;

[0033] Figure 2 It is a structural schematic diagram of a microwave rainforest radiation field radiation modeling device automatically mined by artificial intelligence according to one embodiment of the present invention. DETAILED DESCRIPTION

[0034] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings, but it should be understood that the protection scope of the present invention is not limited by the specific implementation modes.

[0035] Unless explicitly stated otherwise, throughout the specification and claims, the term “comprise” or variations such as “include” or “comprising”, etc., will be understood to include the stated elements or components but not to exclude other elements or components.

[0036] like Figure 1As shown, according to the preferred embodiment of the present invention, the microwave rainforest radiation field radiation modeling method automatically mined by artificial intelligence is applied to the calibration device of the microwave rainforest radiation correction field radiation modeling automatically mined by artificial intelligence, that is, the device that executes the microwave rainforest radiation correction field modeling method automatically mined by artificial intelligence.

[0037] The present invention is based on the basic principle of microwave remote sensing, and establishes a microwave rainforest radiation correction field radiation modeling method automatically mined by artificial intelligence. In the present invention, the internal environmental variables of the rainforest (including temperature field, humidity field) and time variables (characterizing the growth characteristics of the rainforest at different times) are used as the indicator parameters of the external performance of the rainforest radiation, and then the observed radiation (or brightness temperature) of the rainforest is used as the benchmark of the external radiation performance of the rainforest by an external microwave radiation reference source (including a microwave remote sensing satellite with high precision calibration, a microwave radiometer on a manned aircraft / unmanned aircraft, or a microwave radiometer on a vehicle). The complex relationship between the external radiation performance of the rainforest and the indicator parameters is automatically mined by artificial intelligence. Considering the radiation differences of the rainforest targets in all directions, the observation of the external microwave radiation reference source is described by the observation geometry angle, and characterized by the zenith angle and the azimuth angle.

[0038] Since the expression process of the relationship between external radiation manifestations and indicator parameters is relatively complex, it is difficult to find an accurate formula to describe this objective physical process, especially to find accurate formula parameter values. Therefore, the present invention uses artificial intelligence technology to automatically mine this relationship and obtains the relationship by training a neural network with a large amount of sample data.

[0039] Specifically, the present invention comprises the following steps:

[0040] S1: Select a series of time points to obtain the measured values ​​of the temperature field and humidity field of the rainforest environment at the series of time points; and obtain the observation geometry angle when an external microwave radiation reference source observes the rainforest and the observation results of the external microwave radiation reference source.

[0041] In this step, a series of time points are selected as anchor points for description, here [t1, t2, …, t K ] represents, where K represents the number of time points at which the rainforest radiation correction field is observed with an external microwave radiation reference source.

[0042] The internal environmental temperature field of the rainforest at these time points is expressed as [T 1K , T 2K ,…,T MK ], and the humidity field is represented by [P 1K , P 2K ,…,P NK ] represents, where M represents the number of rainforest temperature fields that can be measured, and N represents the number of rainforest humidity fields that can be measured.

[0043] The observation geometry angles of the external microwave radiation reference source when observing the rain forest include the zenith angle {β i |i=1,2,…,K} and azimuth {α i |i=1,2,…,K},the observation results of the external microwave radiation reference source are expressed as {R i |i=1,2,…,K}.

[0044] S2: Take the measured values ​​of the internal environmental temperature field and humidity field of the rainforest at a certain point in time and the observed geometric angle when observing the rainforest with a microwave radiation reference source as the input of the neural network, and take the observation results of the external microwave radiation reference source corresponding to the time point as the output of the neural network to train the neural network.

[0045] In this step, the temperature field measurement values ​​of the rainforest internal environment at a specific time point [T1, T2, ..., T M ], rainforest humidity field measurement values ​​[P1, P2, …, P N ], observation zenith angle β i and azimuth α i The combined M+N+2-dimensional vector is used as the input of the neural network, and then the observation result R of the external microwave radiation reference source corresponding to the time point is used. i As the output of the neural network. At all K time points that meet the requirements of the method, (M+N+2)×K-dimensional input and K-dimensional output are formed, and then the neural network is trained. The present invention names it as Rainforest Radiometric Neural Network (RRNN). Regarding the number of samples, generally, the sample size is sufficient to carry out the work of this application.

[0046] The neural network may be selected from (but not limited to) a fully connected neural network, and the optimizer (Adam) may be used (but not limited to).

[0047] When training a neural network, the following steps are performed:

[0048] S201: The measured values ​​of the temperature field in the rainforest environment [T1, T2, ..., T M ], rainforest humidity field measurement values ​​[P1, P2, …, P N ], observation zenith angle β i and azimuth α i The observation result R of the external microwave radiation reference source corresponding to the time point is taken as input. i The output of the neural network is the label.

[0049] S202: Normalize all inputs and labels:

[0050]

[0051] Among them, d i represents the i-th input, d i-max and d i-min is the maximum and minimum value of the input vector, d i ′ is the value after normalization. The label normalization process adopts the same method as above.

[0052] S203: Create a fully connected neural network including an input layer, at least one hidden layer and an output layer as a model to be trained.

[0053] The fully connected neural network in the present invention obtains the output (calibration result) by processing the input data (temperature field measurement values ​​of the rainforest internal environment, rainforest humidity field measurement values, observation zenith angle and azimuth) through a series of linear transformations and activation functions.

[0054] During the training phase, known inputs and corresponding labels (normalized results) are used to adjust the model's parameters so that the model can predict the calibration results of new input data as accurately as possible.

[0055] In the present invention, the activation function used is the relu function. Other activation functions can also be used if they meet the accuracy requirements after testing.

[0056] The microwave rainforest radiation field radiation modeling network (RRNN) automatically mined by artificial intelligence adopts the following code (using Python code as an example, but not limited to Python code):

[0057]

[0058]

[0059] S204: Train the model using the normalized input and labels.

[0060] During the training process, the RRNN forward propagation is used to calculate the predicted value y of the neural network output pred , according to the predicted value y pred And the true label y calculates the loss value y of the current model loss ; According to the loss value y in each round of training loss , calculate and record the gradient, then clear the gradient and calculate the loss value y lossExecute the back propagation algorithm to calculate the new gradient value of the model parameters, and adjust the parameters of the model according to the calculated new gradient value; repeat the above operation until the loss value is less than the set threshold value, and stop the training process. In the present invention, the threshold value is set to 0.1, which is used to adjust the required radiation calibration accuracy. The smaller the value, the higher the required radiation calibration accuracy. When the required radiation calibration accuracy is reduced, the value can be increased.

[0061] During the training process of the model, an optimizer is also provided to update the parameters of the model according to the calculated gradient, so that the model is adjusted in the direction of reducing the loss. The parameters of the model are the network and training parameters in the training process.

[0062] Among them, the loss function adopts In the form of pred represents the predicted value of the neural network output, y represents the label, that is, the observation result of the external microwave radiation reference source obtained in step S1. This formula only shows the form of the loss function.

[0063] During the training process, through multiple cycles, the model parameters are continuously adjusted to minimize the loss function, so that the model can better fit the given training data.

[0064] The logical process of training is in the form of the following code (using Python code as an example, but not limited to Python code):

[0065]

[0066]

[0067] By repeating forward propagation, calculating loss, back propagation and parameter updating, the model parameters are gradually adjusted so that the model can better fit the training data. During the training process, the training effect of the model can be monitored by observing the change of loss value, using the validation set for evaluation, etc., and the hyperparameters (such as learning rate, number of layers, number of neurons, etc.) can be adjusted as needed to improve the performance of the model.

[0068] S3: Input the rainforest temperature field measurement, rainforest humidity field measurement, observation zenith angle and azimuth angle into the trained model, and obtain the calibration results of the remote sensing instrument at any time through the network output.

[0069] The calibration result is:

[0070] R=RRNN.out(T1,T2,…,T M , P1, P2, …, P M , β, α).

[0071] When the external microwave radiation reference source is a microwave remote sensing satellite, the calibration result R in the above formula is the extra-atmospheric observation result of the rainforest radiation correction field. When the external radiation reference source is airborne / vehicle-mounted, the calibration result R in the above formula is the radiation value of the upper layer of the rainforest, not the extra-atmospheric observation result. It needs to be further calculated through radiation transmission to the theoretical observation value outside the atmosphere.

[0072] The present invention is based on the high-precision fitting ability of neural networks to functions, and takes the physical factors that affect the calibration accuracy (i.e., the rainforest temperature field and humidity field) as part of the input, ensuring that the method can obtain the theoretical radiation value of the remote sensing instrument under different observation geometric conditions under any working condition (i.e., the state represented by the rainforest temperature field and humidity field).

[0073] like Figure 2 As shown, the microwave rainforest radiation field radiation modeling device automatically mined by artificial intelligence according to the preferred embodiment of the present invention includes: a data processing module 1, a training module 2 and a data acquisition module 3.

[0074] The data processing module 1 is used to select a series of time points to obtain the measured values ​​of the temperature field and humidity field of the rainforest environment at the series of time points; and to obtain the observation geometric angle when an external microwave radiation reference source observes the rainforest and the observation results of the external microwave radiation reference source.

[0075] Specifically, a series of time points are selected as anchor points for description, here [t1, t2, …, t K ] represents, where K represents the number of time points at which the rainforest radiation correction field is observed with an external microwave radiation reference source.

[0076] The internal environmental temperature field of the rainforest at these time points is expressed as [T 1K , T 2K ,…,T MK ], and the humidity field is represented by [P 1K , P 2K ,…,P NK ] represents, where M represents the number of rainforest temperature fields that can be measured, and N represents the number of rainforest humidity fields that can be measured.

[0077] The observation geometry angles of the external microwave radiation reference source when observing the rain forest include the zenith angle {β i |i=1,2,…,K} and azimuth {α i |i=1,2,…,K}, the observation results of the external microwave radiation reference source are expressed as {R i | i=1,2,…,K}.

[0078] Training module 2 is used to use the measured values ​​of the internal environmental temperature field and humidity field of the rainforest at a certain point in time and the observed geometric angle when the rainforest is observed by a microwave radiation reference source as the input of the neural network, and the observation results of the external microwave radiation reference source corresponding to the time point as the output of the neural network to train the neural network.

[0079] In training module 2, the temperature field measurements of the rainforest interior environment at a specific time point [T1, T2, …, T M ], rainforest humidity field measurement values ​​[P1, P2, …, P N ], observation zenith angle β i and azimuth α i The combined M+N+2-dimensional vector is used as the input of the neural network, and then the observation result R of the external microwave radiation reference source corresponding to the time point is used. i As the output of the neural network. At all K time points that meet the requirements of the method, (M+N+2)×K-dimensional input and K-dimensional output are formed, and then the neural network is trained. The present invention names it as Rainforest Radiometric Neural Network (RRNN). Regarding the number of samples, generally, the sample size is sufficient to carry out the work of this application.

[0080] The neural network may be selected from (but not limited to) a fully connected neural network, and the optimizer (Adam) may be used (but not limited to).

[0081] When training the neural network, the training module 2 specifically includes:

[0082] The input submodule 201 is used to input the temperature field measurement values ​​[T1, T2, ..., T M ], rainforest humidity field measurement values ​​[P1, P2, …, P N ], observation zenith angle β i and azimuth α i The observation result R of the external microwave radiation reference source corresponding to the time point is taken as input. i The output of the neural network is the label;

[0083] The processing submodule 202 is used to normalize all inputs and labels:

[0084]

[0085] Among them, d i represents the i-th input, d i-max and d i-min is the maximum and minimum value of the input vector, d i′ is the value after normalization. The label normalization process adopts the same method as above.

[0086] The model creation submodule 203 is used to create a fully connected neural network including an input layer, at least one hidden layer and an output layer as a model to be trained.

[0087] The fully connected neural network in the present invention obtains the output (calibration result) by subjecting the input data (the output value of the remote sensing instrument to be calibrated and the temperature measurement value of the optical path) to a series of linear transformations and activation functions. In the training phase, the parameters of the model are adjusted using the known input and the corresponding labels (normalized calibration results) so that the model can predict the calibration results of new input data as accurately as possible.

[0088] In the present invention, the activation function used is the relu function. Other activation functions can also be used if they meet the accuracy requirements after testing.

[0089] The microwave rainforest radiation field radiation modeling network (RRNN) automatically mined by artificial intelligence adopts the following code (using Python code as an example, but not limited to Python code):

[0090]

[0091]

[0092] The training submodule 204 is used to train the model using the normalized input and labels.

[0093] During the training process, the training submodule 204 is used to calculate the predicted value y of the neural network output using the RRNN forward propagation. pred , according to the predicted value y pred And the true label y calculates the loss value y of the current model loss ; According to the loss value y in each round of training loss , calculate and record the gradient, then clear the gradient and calculate the loss value y loss Execute the back propagation algorithm to calculate the new gradient value of the model parameters, and adjust the parameters of the model according to the calculated new gradient value; repeat the above operation until the loss value is less than the set threshold value, and stop the training process. In the present invention, the threshold value is set to 0.1, which is used to adjust the required radiation calibration accuracy. The smaller the value, the higher the required radiation calibration accuracy. When the required radiation calibration accuracy is reduced, the value can be increased.

[0094] During the training process of the model, an optimizer is also provided to update the parameters of the model according to the calculated gradient, so that the model is adjusted in the direction of reducing the loss. The parameters of the model are the network and training parameters in the training process.

[0095] Among them, the loss function adopts In the form of pred Represents the predicted value of the neural network output, y represents the label, that is, the observation result of the external microwave radiation reference source. This formula only shows the form of the loss function.

[0096] During the training process, through multiple cycles, the model parameters are continuously adjusted to minimize the loss function, so that the model can better fit the given training data.

[0097] The logical process of training is in the form of the following code (using Python code as an example, but not limited to Python code):

[0098]

[0099] By repeating forward propagation, calculating loss, back propagation and parameter updating, the model parameters are gradually adjusted so that the model can better fit the training data. During the training process, the training effect of the model can be monitored by observing the change of loss value, using the validation set for evaluation, etc., and the hyperparameters (such as learning rate, number of layers, number of neurons, etc.) can be adjusted as needed to improve the performance of the model.

[0100] The data acquisition module 3 is used to input the rainforest temperature field measurement values, rainforest humidity field measurement values, observation zenith angle and azimuth angle into the trained model, and obtain the calibration results of the remote sensing instrument at any time through network output.

[0101] The calibration result is:

[0102] R=RRNN.out(T1,T2,…,T M , P1, P2, …, P M , β, α).

[0103] When the external microwave radiation reference source is a microwave remote sensing satellite, the calibration result R in the above formula is the extra-atmospheric observation result of the rainforest radiation correction field. When the external radiation reference source is airborne / vehicle-mounted, the calibration result R in the above formula is the radiation value of the upper layer of the rainforest, not the extra-atmospheric observation result. It needs to be further calculated through radiation transmission to the theoretical observation value outside the atmosphere.

[0104] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made based on the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of the present invention is intended to be limited by the claims and their equivalents.

Claims

1. A microwave rainforest radiation field modeling method automatically mined by artificial intelligence, characterized in that: The following steps are involved: S1: Select a series of time points to obtain the measured values ​​of the temperature field and humidity field of the rainforest environment at the series of time points; and obtain the observation geometry angle when an external microwave radiation reference source observes the rainforest and the observation results of the external microwave radiation reference source; S2: Take the measured values ​​of the temperature field and humidity field of the rainforest environment at a certain time point and the observed geometric angle when the rainforest is observed by the microwave radiation reference source as the input of the neural network, and take the observation result of the external microwave radiation reference source corresponding to the time point as the output of the neural network to train the neural network; S3: Input the rainforest temperature field measurement, rainforest humidity field measurement, observation zenith angle and azimuth angle into the trained model, and obtain the calibration results of the remote sensing instrument at any time through the network output.

2. The microwave rainforest radiation field radiation modeling method automatically mined by artificial intelligence according to claim 1, characterized in that: The observation geometric angles of the external microwave radiation reference source when observing the rainforest include the zenith angle and the azimuth angle.

3. The microwave rainforest radiation field radiation modeling method automatically mined by artificial intelligence as claimed in claim 1 is characterized in that: Step S2 specifically includes the following steps: S201: The measured values ​​of the temperature field in the rainforest environment [T1, T2, …, T M ], rainforest humidity field measurement values ​​[P1, P2, …, P N ], observation zenith angle β i and azimuth α i The observation result R of the external microwave radiation reference source corresponding to the time point is taken as input. i The output of the neural network is the label; S202: normalize all inputs and labels; S203: creating a fully connected type including an input layer, at least one hidden layer and an output layer as a neural network model to be trained; S204: Train the neural network model using the normalized input and labels.

4. The microwave rainforest radiation field radiation modeling method automatically mined by artificial intelligence as claimed in claim 3 is characterized in that: Step S304 specifically includes the following steps: During the training process, forward propagation calculates the predicted value y output by the neural network pred , according to the predicted value y pred and the true label y to calculate the loss value of the current model; according to the loss value in each round of training, calculate and record the gradient, then clear the gradient and execute the back propagation algorithm according to the loss value to calculate the new gradient value of the model parameters, and adjust the parameters of the model according to the calculated new gradient value; repeat the above operation until the loss value is less than the set threshold and stop the training process.

5. The microwave rainforest radiation field radiation modeling method automatically mined by artificial intelligence as claimed in claim 4 is characterized in that: In step S304: The loss function is In the form of pred represents the predicted value of the neural network output, and y represents the cross-calibration result value obtained in step S2.

6. A microwave rainforest radiation field modeling device automatically mined by artificial intelligence, characterized in that: include: Data processing module, training module and data acquisition module; The data processing module is used to select a series of time points to obtain the measured values ​​of the temperature field and humidity field of the rainforest environment at the series of time points; and to obtain the observation geometric angle when the external microwave radiation reference source observes the rainforest and the observation results of the external microwave radiation reference source; The training module is used to use the measured values ​​of the temperature field and humidity field of the rainforest environment at a certain time point and the observed geometric angle when the rainforest is observed by the microwave radiation reference source as the input of the neural network, and the observation result of the external microwave radiation reference source corresponding to the time point as the output of the neural network to train the neural network; The data acquisition module is used to input the rainforest temperature field measurement values, rainforest humidity field measurement values, observation zenith angles and azimuth angles into the trained model, and obtain the calibration results of the remote sensing instrument at any time through network output.

7. The microwave rainforest radiation field radiation modeling device automatically mined by artificial intelligence as claimed in claim 6, characterized in that: The observation geometric angles of the external microwave radiation reference source when observing the rainforest include the zenith angle and the azimuth angle.

8. The microwave rainforest radiation field radiation modeling device automatically mined by artificial intelligence as claimed in claim 6, characterized in that: The training module includes: an input submodule, a processing submodule, a model creation submodule and a training submodule The input submodule is used to convert the temperature field measurement values ​​[T1, T2, ..., T M ], rainforest humidity field measurement values ​​[P1, P2, …, P N ], observation zenith angle β i and azimuth α i The observation result R of the external microwave radiation reference source corresponding to the time point is taken as input. i The output of the neural network is the label; The processing submodule is used to normalize all inputs and labels; The model creation submodule is used to create a fully connected type neural network model to be trained, which includes an input layer, at least one hidden layer and an output layer; The training submodule is used to train the neural network model using the normalized input and labels.

9. The microwave rainforest radiation field radiation modeling device automatically mined by artificial intelligence as claimed in claim 8, characterized in that: The training submodule is specifically used for: in the training process, forward propagation is used to calculate the predicted value y_pred output by the neural network, and the loss value of the current model is calculated according to the predicted value y_pred and the true label y; according to the loss value in each round of training, the gradient is calculated and recorded, and then the gradient is cleared and the back propagation algorithm is executed according to the loss value to calculate the new gradient value of the model parameter, and the parameters of the model are adjusted according to the calculated new gradient value; the above operation is repeated until the loss value is less than the set threshold, and the training process is stopped.

10. The microwave rainforest radiation field radiation modeling device automatically mined by artificial intelligence as claimed in claim 9, characterized in that: The loss function is In the form of pred represents the predicted value of the neural network output, and y represents the cross-calibration result value obtained in step S2.

Images