Training methods, radiometric calibration methods and equipment for radiometric calibration neural network models

Abstract

This application provides a training method, a radiometric calibration method, an apparatus, and a computer storage medium for a neural network model used in the radiometric calibration of infrared remote sensing instruments. The training method includes: acquiring at least one sample data point of a preset neural network model, wherein the preset neural network model is established based on the correlation between state parameters affecting the radiometric response performance of the infrared remote sensing instrument during operation and calibration coefficients; the state parameters affecting the radiometric response performance of the infrared remote sensing instrument during operation include the temperature of the instrument's components; and training the preset neural network model based on the at least one sample data point to obtain a radiometric response model for the infrared remote sensing instrument. Using this radiometric response model for radiometric calibration of infrared remote sensing instruments enables instruments lacking calibration resources or with low calibration accuracy to achieve high-precision and high-efficiency radiometric calibration.

Description

Technical Field

[0001] This application relates to the field of infrared remote sensing technology, and in particular to a training method, radiometric calibration method, device, and computer storage medium for a radiometric calibration neural network model of an infrared remote sensing instrument. Background Technology

[0002] Infrared remote sensing technology is based on the principle of infrared radiation, observing objects by receiving infrared light. It provides crucial data for acquiring atmospheric temperature and humidity information, monitoring meteorological information, and determining surface and sea surface parameters. Therefore, infrared remote sensing instruments are widely used in the aviation and aerospace fields. Infrared remote sensing instruments are often divided into imaging instruments and non-imaging instruments, with imaging instruments (such as infrared imagers) being more widely used. In a common application scenario, infrared remote sensing instruments are mounted on satellites to observe the Earth (the target object). These instruments receive infrared light radiated by the Earth and generate output values ​​based on the energy of the received infrared light. The output value of an infrared imager is a dimensionless quantity without actual physical meaning; calibration is required to convert it into physically meaningful quantities such as brightness, surface reflectivity, and surface temperature. Currently, infrared remote sensing instruments require calibration resources for calibration. However, some infrared remote sensing instruments lack calibration resources (for example, some satellites do not carry radiation reference sources), and some infrared remote sensing instruments, although they have calibration resources, have unsatisfactory calibration resources (for example, blackbodies with an actual emissivity of less than 1), resulting in low calibration accuracy, long calibration operation time, and low efficiency. Summary of the Invention

[0003] In view of this, embodiments of this application provide a radiometric calibration method, apparatus, electronic device, and computer storage medium for an infrared remote sensing instrument to solve some or all of the above-mentioned problems.

[0004] According to a first aspect of this application, a method for training a neural network model for radiometric calibration of an infrared remote sensing instrument is provided. The method includes: acquiring at least one sample data of a preset neural network model, wherein the preset neural network model is established based on the correlation between state parameters affecting the radiometric response performance of the infrared remote sensing instrument during operation and calibration coefficients, and the state parameters affecting the radiometric response performance of the infrared remote sensing instrument during operation include the temperature of the components of the infrared remote sensing instrument. The method further includes: training the preset neural network model based on the at least one sample data to obtain a radiometric response model of the infrared remote sensing instrument.

[0005] According to a second aspect of the embodiments of this application, a radiometric calibration method for an infrared remote sensing instrument is provided. The method includes: acquiring the output value and state parameters of the infrared remote sensing instrument under an actual operating state, wherein the state parameters are numerical values ​​of state parameters affecting the radiometric response performance of the infrared remote sensing instrument; inputting the state parameters of the infrared remote sensing instrument under the actual operating state into a preset radiometric response model to obtain calibration coefficients of the infrared remote sensing instrument under the aforementioned actual operating state, wherein the preset neural network model is established based on the correlation between the state parameters affecting the radiometric response performance of the infrared remote sensing instrument during operation and the calibration coefficients, and the state parameters affecting the radiometric response performance of the infrared remote sensing instrument during operation include the temperature of the components of the infrared remote sensing instrument; and obtaining the entrance pupil radiation corresponding to the output value of the infrared remote sensing instrument based on the output value and the obtained calibration coefficients of the infrared remote sensing instrument under the corresponding operating state, and based on a functional relationship, the functional relationship is:

[0006]

[0007] Where x represents the output value of the infrared remote sensing instrument under actual working conditions, y represents the entrance pupil radiance corresponding to the output value of the infrared remote sensing instrument, and a i This represents the i-th scaling coefficient in a set of scaling coefficients under the corresponding working state, where N is the number of scaling coefficients in the set, N is an integer greater than 0, and i is an integer in [0, N].

[0008] According to a third aspect of the present application, an electronic device is provided, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; the memory is used to store at least one executable instruction, which causes the processor to perform an operation corresponding to the radiometric calibration method of the infrared remote sensing instrument of the first aspect.

[0009] According to a fourth aspect of the embodiments of this application, a computer storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the radiometric calibration method of an infrared remote sensing instrument as described in the first aspect.

[0010] According to one or more embodiments of this application, the temperature of the components of the infrared remote sensing instrument includes at least one of the temperatures of the mechanical support components, optical path components, detector assembly, and radiation cooler of the infrared remote sensing instrument.

[0011] According to one or more embodiments of this application, the status parameters of the infrared remote sensing instrument also include at least one of the scanning mirror rotation angle of the infrared remote sensing instrument and an identifier of the time the infrared remote sensing instrument has been operating.

[0012] According to one or more embodiments of this application, obtaining at least one sample data of a preset neural network model includes: using a radiation reference source to obtain the calibration coefficients of an infrared remote sensing instrument in one or more operating states and the state parameters in the one or more operating states, to obtain at least one sample data, each sample data including the state parameters of the infrared remote sensing instrument in one operating state and the calibration coefficients in that operating state.

[0013] According to one or more embodiments of this application, training the preset neural network model based on the at least one sample data includes: inputting state parameters contained in the sample data into the preset neural network model to obtain predicted calibration coefficients; calculating a loss function value based on the predicted calibration coefficients and the corresponding calibration coefficients contained in the sample data; and adjusting the parameters contained in the preset neural network model based on the loss function value to reduce the loss function value, wherein the closer the predicted calibration coefficients and the corresponding calibration coefficients contained in the sample data are, the smaller the loss function value.

[0014] According to the embodiments of this application, a training method for a neural network model for radiometric calibration of infrared remote sensing instruments, a radiometric calibration method for infrared remote sensing instruments, electronic equipment, and computer storage media are provided. A preset neural network model is established based on the correlation between state parameters affecting the radiometric response performance of infrared remote sensing instruments and calibration coefficients. The state parameters affecting the radiometric response performance of the infrared remote sensing instrument during operation include the temperature of the instrument's components. The radiometric response model of the infrared remote sensing instrument can be obtained by training the preset neural network model using sample data. Therefore, only the state parameters affecting the radiometric response performance of the infrared remote sensing instrument need to be obtained. By inputting the state parameters into the obtained radiometric response model, the calibration coefficients of the infrared remote sensing instrument can be predicted, thus achieving radiometric calibration of the infrared remote sensing instrument. Therefore, the embodiments of this application utilize a specific neural network model relating state parameters and calibration coefficients that affect the radiometric response performance of infrared remote sensing instruments, enabling radiometric calibration of infrared remote sensing instruments that lack calibration resources or have low calibration accuracy due to their own calibration resources. That is, the infrared remote sensing instrument does not need to be equipped with calibration resources, which simplifies the structure and design of the infrared remote sensing instrument. Meanwhile, using a neural network model that integrates the state parameters and calibration coefficients of infrared remote sensing instruments, which include the influence of temperature on the components of the infrared remote sensing instrument, results in higher calibration accuracy, shorter calibration time, and higher efficiency. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 A flowchart illustrating the operation of a training method for a neural network model used for radiometric calibration of infrared remote sensing instruments, provided in an embodiment of this application.

[0017] Figure 2 A flowchart illustrating a radiometric calibration method for an infrared remote sensing instrument, provided as an embodiment of this application;

[0018] Figure 3 A schematic diagram of a radiometric calibration method for an infrared remote sensing instrument provided in an embodiment of this application;

[0019] Figure 4 A structural block diagram of a calibration device for an infrared remote sensing instrument provided in this application embodiment;

[0020] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions in the embodiments of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art should fall within the protection scope of this application. It should be noted that the "one" working state mentioned in this application refers to a state in which the state parameters affecting the infrared radiation response performance (such as the temperature of the components of the infrared remote sensing instrument) have not changed substantially, and the calibration coefficient has not changed substantially.

[0022] The specific implementation of the embodiments of this application will be further described below with reference to the accompanying drawings.

[0023] Reference Figure 1 This document illustrates a flowchart of an operation method for training a neural network model for radiometric calibration of infrared remote sensing instruments, as provided in an embodiment of this application. The training method for this neural network model includes the following operations:

[0024] Operation S10: Obtain at least one sample data of a preset neural network model. The preset neural network model is established based on the correlation between state parameters affecting the radiation response performance of the infrared remote sensing instrument during operation and calibration coefficients. The state parameters affecting the radiation response performance of the infrared remote sensing instrument during operation include the temperature of the components of the infrared remote sensing instrument.

[0025] The preset neural network model is established based on the correlation between state parameters affecting the radiation response performance of an infrared remote sensing instrument and calibration coefficients during operation. Here, state parameters represent the operating state of the infrared remote sensing instrument. The state parameters affecting the radiation response performance of an infrared remote sensing instrument mainly include the temperature of its components. These include, for example, one or more of the temperatures of the mechanical support components (e.g., the temperature of the scanning mirror bracket), the optical path components (mainly including the temperature of the optical lenses), the detector assembly, and the radiation cooler. The preset neural network model can include a backpropagation (BP) neural network model. Since the number of state parameters affecting the remote sensing instrument is relatively small, using a BP neural network model is simple and can achieve the required calibration accuracy. However, it should be noted that other neural network models can also be used in the implementation of this application.

[0026] According to one or more embodiments of this application, the sample data of the preset neural network model can be obtained by using a radiation reference source (e.g., a blackbody reference source) to acquire the state parameters of the infrared remote sensing instrument in one or more operating states and the corresponding calibration coefficients in each operating state, and to obtain at least one set of sample data. Each set of sample data includes the state parameters of the infrared remote sensing instrument in one operating state and the calibration coefficients in the corresponding operating state. For example, the infrared remote sensing instrument can be made to detect the infrared radiation of a blackbody reference source and generate an output value in one operating state. The values ​​of the state parameters (i.e., state parameters) that affect the radiation response performance of the infrared remote sensing instrument in that operating state are obtainable. Various existing methods, such as sensors, can be used to acquire the state parameters of the infrared remote sensing instrument in each operating state. For example, a temperature sensor can be used to detect the temperature of the mechanical support components, the optical path components, the detector assembly, and the radiation cooler of the infrared remote sensing instrument.

[0027] Meanwhile, the entrance pupil radiance of the blackbody reference source is known. The following pre-defined functional relationship exists between the calibration coefficients, the corresponding output value of the infrared remote sensing instrument, and the entrance pupil radiance:

[0028]

[0029] Where x represents the output value of the infrared remote sensing instrument (x i That is, the output value raised to the power of i), y represents the entrance pupil radiance of the infrared remote sensing instrument (i.e., the radiance value of the infrared light energy received by the infrared remote sensing instrument), a i This represents the i-th calibration coefficient in a set of calibration coefficients, where N is the number of calibration coefficients in the set, N is a positive integer, and i is an integer in the range [0, N]. The number of calibration coefficients can be determined based on experience, the properties of the infrared remote sensing instrument being calibrated (e.g., the physical properties of the photoelectric conversion electronic devices in an infrared imager), or the required calibration accuracy. For example, N can be 1, 2, or 3.

[0030] By substituting the output value of the infrared remote sensing instrument and the corresponding entrance pupil radiation into the above functional relationship, the corresponding calibration coefficients can be obtained.

[0031] The state parameters obtained in the above manner and the corresponding scaling coefficients under the state constitute a sample data.

[0032] Optionally, a set of sample data may include a set of state parameters and a set of calibration coefficients corresponding to the set of state parameters. Here, the set of state parameters may be multiple state parameters of an infrared remote sensing instrument in a working state, such as the values ​​of the optical path temperature and the mechanical scanning mirror rotation angle of the infrared remote sensing instrument.

[0033] The following example illustrates how to obtain sample data for a pre-defined neural network model, including: acquiring historical data obtained during the historical calibration process for at least one infrared remote sensing instrument. This historical data includes at least one set of calibration coefficients for the infrared remote sensing instrument and at least one set of corresponding state parameters. Using one set of calibration coefficients and one set of state parameters as a single sample data set, at least one sample data set is extracted from the historical data. It should be noted that, in one implementation, the historical data obtained during the historical calibration process may include state parameters and calibration coefficients obtained from calibration of an infrared remote sensing instrument under different operating conditions or using different calibration methods. Different operating conditions may refer to state parameters, i.e., state parameters, having different values ​​affecting the infrared response performance of the infrared remote sensing instrument. Optionally, the state parameters include optical path temperature. Obtaining at least one sample data set for the pre-defined neural network model includes: acquiring the calibration coefficients of the infrared remote sensing instrument under different optical path temperatures and obtaining at least one sample data set, where the sample data includes the optical path temperature and the corresponding calibration coefficients. Calibration methods may include blackbody calibration, cross-calibration, ground-based substitution calibration, pre-launch calibration, etc. By extracting sample data from historical data of an infrared remote sensing instrument, a preset neural network model corresponding to that instrument can be trained. The resulting preset neural network model can reflect the correlation between the state parameters and calibration coefficients of the infrared remote sensing instrument. When the infrared remote sensing instrument needs to be calibrated, the preset neural network model and the current state parameters can be used for calibration, making the infrared remote sensing instrument detection more accurate.

[0034] Of course, this is merely an illustrative example. Optionally, in another implementation, the calibration sample data may include the state parameters and their corresponding calibration coefficients obtained during the calibration of the infrared remote sensing instrument using a blackbody reference source within a preset time period. Since the state of an infrared remote sensing instrument is usually continuously changing during operation, the calibration coefficients for a specific operating state obtained through the radiation response model can be used for a period of time (i.e., the error within that period is small enough not to affect the calibration results, or the impact on the calibration results is negligible). Therefore, the "corresponding calibration coefficients" here do not need to be strictly in the same operating state as the state parameters. As long as the calibration coefficients do not undergo substantial changes, the calibration coefficients in the "operating state" that do not affect the required accuracy of the calibration results can be used as the "corresponding calibration coefficients."

[0035] The training method for the neural network model used for radiometric calibration of infrared remote sensing instruments further includes: operating S20, training the preset neural network model based on the at least one sample data to obtain the radiometric response model of the infrared remote sensing instrument.

[0036] By inputting one or more sets of sample data, consisting of the calibration coefficients obtained in operation S10 and the state parameters under the corresponding state, into a preset neural network model such as BP, the radiation response model of the infrared remote sensing instrument can be obtained.

[0037] When multiple scaling coefficients are included, there are also multiple preset neural network models. Each scaling coefficient corresponds to a preset neural network model, and each preset neural network model needs to be trained. The preset neural network models for each scaling coefficient can be the same or different. In addition, the same sample data or different sample data can be used to train each preset neural network model. It can also be considered that the state parameters corresponding to different scaling coefficients can be the same or different, depending on which state parameters affect the size of the scaling coefficient. That is, for scaling coefficient A, the neural network model corresponding to scaling coefficient A is trained using sample data composed of state parameters B that affect the size of scaling coefficient A. There can be one or more state parameters affecting a scaling coefficient, and the embodiments of this application do not limit this.

[0038] Besides the temperature of the instrument's components, the following state parameters can also affect the radiation response performance of infrared remote sensing instruments: For infrared remote sensing instruments with scanning mirrors, if the scanning mirror rotates during calibration, the scanning mirror angle can be included as a state parameter affecting the instrument's radiation response performance, because changes in the scanning mirror's deflection angle alter the imaging optical path. Additionally, if the calibrated infrared remote sensing instrument has been used for a considerable period (e.g., more than one year), the accumulated usage time will affect the calibration coefficient. Therefore, the state parameters affecting the radiation response performance of the infrared remote sensing instrument can include an indication of the instrument's operational time. This indication can be absolute or relative, as long as it accurately represents the amount of time the instrument has been used.

[0039] It should also be noted that the state parameter B corresponding to the calibration coefficient A can include all or some of the state parameters that affect the magnitude of the calibration coefficient A. It does not need to include all state parameters that affect the calibration coefficient A. Some state parameters with little impact on the calibration coefficient A can be discarded to simplify the calibration model and improve calibration efficiency. For example, when calibrating an infrared remote sensing instrument with a relatively short operating time, it is unnecessary to consider the state parameter indicating the instrument's operating time, as this state parameter has almost no impact on the calibration coefficient of an infrared remote sensing instrument with a relatively short operating time.

[0040] This section provides a specific example illustrating how to train a pre-defined neural network model. In one example of this application, training the pre-defined neural network model based on at least one set of sample data includes: inputting state parameters contained in the sample data into the pre-defined neural network model to obtain predicted scaling coefficients; calculating a loss function value based on the predicted scaling coefficients and the corresponding scaling coefficients contained in the sample data; and adjusting the state parameters contained in the pre-defined neural network model according to the loss function value to reduce the loss function value. The closer the predicted scaling coefficients and the corresponding scaling coefficients contained in the sample data are, the smaller the loss function value. Optionally, the loss function can be set as follows:

[0041]

[0042] Where L represents the loss function value, M represents the number of sample data, and P k Q represents the scaling coefficient of the prediction obtained by inputting the k-th sample data into the neural network model. k Let M represent the calibration coefficients contained in the k-th sample data, where M is an integer greater than 0 and k is an integer in the range [1, M]. Since the loss function represents the difference between the predicted calibration coefficients and the corresponding calibration coefficients contained in the sample data, training based on reducing the loss function can make the calibration coefficients predicted by the preset neural network model increasingly closer to the calibration coefficients contained in the sample data, thus ensuring that the calibration coefficients predicted by the preset neural network model conform to the actual working state of infrared remote sensing instruments.

[0043] If there are multiple calibration coefficients, corresponding to multiple different neural network models, the state parameters can be input into multiple neural network models to obtain multiple calibration coefficients. These multiple calibration coefficients constitute a set of calibration coefficients, which can determine the functional relationship between a calibration coefficient and the output value of the corresponding infrared remote sensing instrument and the entrance pupil radiation.

[0044] It should be noted that, since the operating state of infrared remote sensing instruments is usually continuously changing, the calibration coefficients obtained through the radiation response model can be used for a period of time (i.e., the error within this period is small enough not to affect the calibration results, or the impact on the calibration results is negligible). Therefore, the "corresponding operating state" here does not need to be the same "single" operating state where the state parameters do not change at all. Multiple operating states where the state parameters change slightly, the calibration coefficients do not change substantially, and the calibration results do not affect the required accuracy can all be used as the "corresponding operating state" here.

[0045] The method for training a neural network model for radiometric calibration of infrared remote sensing instruments provided in this application establishes a preset neural network model based on the correlation between the temperature state parameters of the components of the infrared remote sensing instrument and the calibration coefficients. Training this preset neural network model yields a radiometric response model. This radiometric response model enables accurate and rapid radiometric calibration of infrared remote sensing instruments that lack calibration resources or whose calibration accuracy using their own calibration resources is low, without requiring the infrared remote sensing instrument to carry calibration resources or possess high-precision calibration resources.

[0046] Another aspect of this application provides a radiometric calibration method for an infrared remote sensing instrument, which can be applied to a calibration device for an infrared remote sensing instrument, i.e., a device for performing the radiometric calibration method for an infrared remote sensing instrument. This application calibrates an infrared remote sensing instrument based on the principle of infrared radiation. Infrared light (i.e., infrared radiation) is a type of electromagnetic wave. The generation of infrared light is closely related to temperature. Objects in nature, when their temperature is above absolute zero (i.e., -273.15℃), will radiate infrared light outward, which is the principle of infrared radiation. The energy of the radiated infrared light is determined by the surface temperature of the object. When observing an object (i.e., a target object) using an infrared remote sensing instrument, the infrared remote sensing instrument receives the infrared light radiated by the target object. Based on the energy of the infrared light, the output value is mainly generated using photoelectric conversion devices (such as the image sensor of an infrared imager). The greater the energy of the infrared light received by the infrared remote sensing instrument, the larger the output value; the smaller the energy of the infrared light received by the infrared remote sensing instrument, the smaller the output value.

[0047] like Figure 2 As shown, the radiometric calibration method for the aforementioned infrared remote sensing instrument includes the following operations.

[0048] Operation 101: Obtain the output value and status parameters of the infrared remote sensing instrument under actual working conditions.

[0049] The state parameters here refer to the values ​​of state parameters that affect the radiation response performance of the infrared remote sensing instrument. State parameters affecting the radiation response performance of the infrared remote sensing instrument include the temperature of its components. Furthermore, for infrared remote sensing instruments with scanning mirrors, if the scanning mirror rotates during calibration, the state parameters affecting the radiation response performance include the scanning mirror rotation angle, because changes in the scanning mirror deflection angle alter the imaging optical path of the infrared remote sensing instrument. For example, if the calibrated infrared remote sensing instrument has been used for a long time (e.g., more than one year), the time elapsed will affect the calibration coefficient; therefore, the state parameters affecting the radiation response performance of the infrared remote sensing instrument can include an indication of the time the instrument has been used. This indication can be an absolute or relative time indication, as long as it represents the elapsed time of the infrared remote sensing instrument. Sensors can be used to acquire the state parameters of the infrared remote sensing instrument in various operating states. For example, temperature sensors can be used to detect the temperature of the mechanical support components, optical path components, detector assembly, and radiation cooler of the infrared remote sensing instrument in an actual operating state.

[0050] During infrared remote sensing observations, the instrument receives the radiant energy of infrared light, generating an output value corresponding to this energy. This output value can be directly read from the instrument. The instrument's calibration device can also obtain the output value under the same actual operating conditions from the instrument via circuitry or communication.

[0051] Operation 102: Input the state parameters of the infrared remote sensing instrument under the above-mentioned actual working state into the radiation response model obtained according to the training method described above, and obtain the calibration coefficients of the infrared remote sensing instrument under the corresponding working state.

[0052] The radiation response model obtained according to the training method described above can be pre-assembled into the calibration device of the infrared remote sensing instrument. Then, the state parameters of the infrared remote sensing instrument under actual working conditions can be input into the radiation response model to obtain the calibration coefficients of the infrared remote sensing instrument under the aforementioned actual working conditions.

[0053] Operation 103: Based on the output value and the corresponding calibration coefficients of the infrared remote sensing instrument, obtain the entrance pupil radiance corresponding to the output value of the infrared remote sensing instrument based on the functional relationship. The above functional relationship is:

[0054]

[0055] Where x represents the output value of the infrared remote sensing instrument in an actual working state, y represents the entrance pupil radiation corresponding to the output value of the infrared remote sensing instrument, and ai This represents the i-th scaling coefficient in a set of scaling coefficients under the corresponding working state, where N is the number of scaling coefficients in the set, N is an integer greater than 0, and i is an integer in [0, N].

[0056] During the observation process of infrared remote sensing instruments, the instruments receive the energy of infrared light and obtain an output value. Substituting this output value into the aforementioned preset function, the entrance pupil radiance corresponding to that output value can be calculated; that is, the radiant value of the infrared light energy received by the instrument. This entrance pupil radiance reflects physically significant quantities such as brightness, surface reflectivity, and surface temperature. Therefore, staff use this calculated entrance pupil radiance to predict weather and conduct meteorological and environmental monitoring.

[0057] Based on the radiometric calibration method for infrared remote sensing instruments described in operations 101-103 above, a specific application scenario is given here to illustrate the method, such as... Figure 3 As shown, Figure 3 This is a schematic diagram of a scenario for a radiometric calibration method for an infrared remote sensing instrument provided in an embodiment of this application. Figure 3 The image shows Earth 20 and satellite 21, with the satellite carrying an infrared remote sensing instrument 211, a temperature sensor 212, and a processor 213.

[0058] Infrared remote sensing instruments can monitor or measure the Earth to obtain detection data, and satellites can transmit the monitoring data obtained by infrared remote sensing instruments back to Earth. However, the accuracy of infrared remote sensing instruments can decrease during the monitoring process due to environmental influences or instrument degradation and aging. Therefore, calibration of infrared remote sensing instruments is necessary. Optionally, periodic calibration can be performed, for example, a cycle can be 5 minutes, 30 minutes, or 1 hour.

[0059] It should be noted that before satellite launch, the infrared remote sensing instruments aboard the satellite need to be calibrated using different methods to obtain state parameters and calibration coefficients. In this example, state parameters may include optical path temperature. For instance, ground-based calibration or pre-launch calibration can be used to obtain calibration coefficients at different optical path temperatures. Multiple sets of state parameters and calibration coefficients are then used to train a pre-defined neural network model, which can be used for calibration after satellite launch. Alternatively, after satellite launch, blackbody calibration or cross-calibration can be used to obtain multiple sets of state parameters and calibration coefficients, which are then used to train a pre-defined neural network model using a processor.

[0060] Based on the above Figure 2The corresponding embodiments describe a radiometric calibration method for infrared remote sensing instruments. This application also provides a calibration apparatus for an infrared remote sensing instrument, used to execute the radiometric calibration method provided in this application. (Refer to...) Figure 4 This diagram illustrates a structural block diagram of a calibration device for an infrared remote sensing instrument according to an embodiment of this application. The calibration device 40 for the infrared remote sensing instrument includes:

[0061] The sample module 401 is used to acquire at least one sample data of a preset neural network model. The preset neural network model is established based on the correlation between the state parameters affecting the radiation response performance of the infrared remote sensing instrument and the calibration coefficients under the working state of the infrared remote sensing instrument. The state parameters include the temperature of the components of the infrared remote sensing instrument.

[0062] Training module 402 is used to train a preset neural network model based on at least one sample data;

[0063] The state parameter module 403 is used to acquire the state parameters of the infrared remote sensing instrument in a working state, wherein the state parameters are the values ​​of the state parameters that affect the radiation response performance of the infrared remote sensing instrument.

[0064] The prediction module 404 is used to input the state parameters into the preset neural network model to obtain the calibration coefficients of the infrared remote sensing instrument in the corresponding working state.

[0065] Optionally, in one example, the state parameters include the optical path temperature, and the sample module 401 is used to obtain the calibration coefficients of the infrared remote sensing instrument under different optical path temperatures and to obtain at least one sample data, which includes the optical path temperature and the corresponding calibration coefficients.

[0066] Optionally, in one example, the optical path temperature includes at least one of the temperatures of the mechanical structure of the infrared remote sensing instrument, the temperatures of the optical path components, and the temperatures of the detector assembly.

[0067] Optionally, in one example, the sample module 401 is used to acquire historical data obtained for at least one infrared remote sensing instrument during the historical calibration process. The historical data includes at least one set of calibration coefficients for the infrared remote sensing instrument and at least one set of corresponding state parameters. The set of calibration coefficients and the set of state parameters are used as sample data to extract at least one sample data from the historical data.

[0068] The calibration coefficients, the corresponding output values ​​of the infrared remote sensing instrument, and the entrance pupil irradiance have the following functional relationship:

[0069]

[0070] Where x represents the output value of the infrared remote sensing instrument, y represents the entrance pupil radiance of the infrared remote sensing instrument (i.e., the radiance value of the infrared light energy received by the infrared remote sensing instrument), and a i This represents the i-th calibration coefficient in a set of calibration coefficients, where N is the number of calibration coefficients in the set, where N is a positive integer and i is an integer in the range [0, N]. The number of calibration coefficients can be determined based on experience, calibration accuracy, and the properties of the infrared remote sensing instrument (especially the properties of the photoelectric conversion component), for example, N can be 1, 2, or 3.

[0071] Optionally, in one example, the training module 402 is used to input the state parameters contained in the sample data into a preset neural network model to obtain the predicted scaling coefficients; calculate the loss function value based on the predicted scaling coefficients and the corresponding scaling coefficients contained in the sample data; and adjust the parameters contained in the preset neural network model according to the loss function value to reduce the loss function value. The closer the predicted scaling coefficients and the corresponding scaling coefficients contained in the sample data are, the smaller the loss function value is.

[0072] Optionally, in one example, the training module 402 is used to train the slope parameter model and the offset parameter model respectively using sample data.

[0073] The calibration device for infrared remote sensing instruments provided in this application establishes a preset neural network model based on the correlation between state parameters and calibration coefficients, and trains it to obtain a radiometric response model. During calibration, only the state parameters of the infrared remote sensing instrument need to be acquired and input into the radiometric response model to obtain the corresponding calibration coefficients. Through this neural network model, remote sensing instruments lacking calibration resources, or those with low calibration accuracy using their own calibration resources, can achieve more accurate and faster radiometric calibration.

[0074] Based on the above Figure 2 In a corresponding embodiment, this application provides an electronic device 50 for performing the above-described... Figure 2 The radiometric calibration method for infrared remote sensing instruments described in the corresponding embodiments is shown in Figure 5. Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The specific embodiments of this application do not limit the specific implementation of the electronic device. For example, the electronic device 50 may be a device that includes an infrared remote sensing instrument, or it may be a device that is independent of the infrared remote sensing instrument.

[0075] like Figure 5As shown, the electronic device 50 may include: a processor 502, a communications interface 504, a memory 506, and a communications bus 508.

[0076] The processor 502, communication interface 504, and memory 506 communicate with each other via communication bus 508. Communication interface 504 is used to communicate with other electronic devices, such as terminal devices or servers. Processor 502 executes program 510, specifically performing the relevant operations in the above-described embodiments of the radiometric calibration method for infrared remote sensing instruments. Specifically, program 510 may include program code, which includes computer operation instructions. Processor 502 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The electronic device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs. Memory 506 stores program 510. Memory 506 may include high-speed RAM and may also include non-volatile memory, such as at least one disk storage device. Program 510 can specifically be used to cause processor 502 to execute the radiometric calibration method for any of the infrared remote sensing instruments described in Embodiment 1. The specific implementation of each operation in program 510 can be found in the corresponding operations and units described in the above-mentioned embodiment of the radiometric calibration method for infrared remote sensing instruments, and will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the devices and modules described above can be referred to the corresponding process descriptions in the aforementioned method embodiments, and will not be repeated here.

[0077] The electronic device provided in this application establishes a preset neural network model based on the correlation between state parameters and calibration coefficients, and trains it to obtain a radiation response model. During the calibration process, it only needs to obtain the state parameters of the infrared remote sensing instrument and input the state parameters into the radiation response model to obtain the corresponding calibration coefficients. Through the neural network model based on the relationship between state parameters and calibration coefficients, infrared remote sensing instruments that lack calibration resources or whose calibration accuracy is low using their own calibration resources can achieve more accurate calibration.

[0078] According to the radiometric calibration method, apparatus, electronic device, and computer storage medium for infrared remote sensing instruments provided in this application, a preset neural network model is established based on the correlation between state parameters and calibration coefficients affecting the radiometric response performance of the infrared remote sensing instrument. The radiometric response model is obtained by training the preset neural network model with sample data. Radiometric calibration of the infrared remote sensing instrument can be achieved by inputting the state parameters and output values ​​of the infrared remote sensing instrument in its actual working state into the radiometric response model. Therefore, the embodiments of this application utilize the neural network model between state parameters and calibration coefficients affecting the radiometric response performance of the infrared remote sensing instrument, enabling radiometric calibration of infrared remote sensing instruments that lack calibration resources or have low calibration accuracy due to their own calibration resources. That is, the infrared remote sensing instrument does not need to be equipped with calibration resources, which simplifies the structure and design of the infrared remote sensing instrument. Simultaneously, using the neural network model between state parameters and calibration coefficients affecting the radiometric response performance of the infrared remote sensing instrument results in higher calibration accuracy, shorter calibration time, and higher efficiency.

[0079] It should be noted that, depending on the implementation needs, the various components / operations described in the embodiments of this application can be split into more components / operations, or two or more components / operations or parts of components / operations can be combined into new components / operations to achieve the purpose of the embodiments of this application.

[0080] The methods described above according to the embodiments of this application can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or a non-transitory machine-readable medium and subsequently stored on a local recording medium, downloaded via a network. Thus, the methods described herein can be stored on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA) for such software processing. It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components (e.g., RAM, ROM, flash memory, etc.) capable of storing or receiving software or computer code that, when accessed and executed by the computer, processor, or hardware, implements the radiometric calibration method for the infrared remote sensing instrument described herein. Furthermore, when a general-purpose computer accesses code for implementing the radiometric calibration method for the infrared remote sensing instrument shown herein, the execution of the code transforms the general-purpose computer into a dedicated computer for executing the radiometric calibration method for the infrared remote sensing instrument shown herein.

[0081] Those skilled in the art will recognize that the unit and method operations of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the embodiments of this application.

[0082] The above embodiments are only used to illustrate the embodiments of this application, and are not intended to limit the embodiments of this application. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the embodiments of this application. Therefore, all equivalent technical solutions also fall within the scope of the embodiments of this application, and the patent protection scope of the embodiments of this application should be defined by the claims.

Claims

1. A training method for a neural network model used for radiometric calibration of infrared remote sensing instruments, characterized in that, include: Obtain at least one sample data of a preset neural network model. The preset neural network model is established based on the correlation between state parameters affecting the radiation response performance of the infrared remote sensing instrument during operation and calibration coefficients. The state parameters affecting the radiation response performance of the infrared remote sensing instrument during operation are parameters representing the working state of the infrared remote sensing instrument. The state parameters include the temperature of the components of the infrared remote sensing instrument. and The preset neural network model is trained based on the at least one sample data to obtain the radiation response model of the infrared remote sensing instrument. The process of obtaining at least one sample data of a preset neural network model includes: obtaining the state parameters of an infrared remote sensor in one or more operating states and determining the corresponding radiometric calibration coefficients using a radiation reference source in each operating state, and obtaining the at least one sample data.

2. The method according to claim 1, characterized in that, The temperature of the components of the infrared remote sensing instrument includes at least one of the following: the temperature of the mechanical support components, the temperature of the optical path components, the temperature of the detector assembly, and the temperature of the radiation cooler.

3. The method according to claim 1, characterized in that, The status parameters also include at least one of the scanning mirror rotation angle of the infrared remote sensing instrument and an indicator of the time the infrared remote sensing instrument has been operating.

4. The method according to claim 1, characterized in that, The acquisition of at least one sample data of a preset neural network model includes: The calibration coefficients and state parameters of the infrared remote sensing instrument under one or more operating conditions are obtained using a radiation reference source to obtain at least one sample data. Each sample data includes the state parameters of the infrared remote sensing instrument under one operating condition and the calibration coefficients under that operating condition.

5. The method according to claim 1, characterized in that, The step of training the preset neural network model based on the at least one sample data includes: The state parameters contained in the sample data are input into the preset neural network model to obtain the predicted calibration coefficients; Based on the predicted scaling coefficients and the corresponding scaling coefficients contained in the sample data, a loss function value is calculated. The parameters contained in the preset neural network model are adjusted according to the loss function value to reduce the loss function value. The closer the predicted scaling coefficients and the corresponding scaling coefficients contained in the sample data are, the smaller the loss function value.

6. A radiometric calibration method for an infrared remote sensing instrument, characterized in that, include: The output value and state parameters of the infrared remote sensing instrument under an actual working state are obtained, wherein the state parameters are the values ​​of state parameters that affect the radiation response performance of the infrared remote sensing instrument. By inputting the state parameters of the infrared remote sensing instrument under the actual working state into the radiation response model obtained by the method according to any one of claims 1-5, the calibration coefficients of the infrared remote sensing instrument under the actual working state are obtained. and Based on the output value and the calibration coefficients corresponding to the infrared remote sensing instrument, the entrance pupil radiation corresponding to the output value of the infrared remote sensing instrument is obtained based on a functional relationship, which is: Where x represents the output value of the infrared remote sensing instrument in an actual working state, y represents the entrance pupil radiation corresponding to the output value of the infrared remote sensing instrument, and a i This represents the i-th scaling coefficient in a set of scaling coefficients under the corresponding working state. N is the number of scaling coefficients in the set of scaling coefficients. N is an integer greater than 0, and i is an integer in [0, N].

7. The method according to claim 6, characterized in that, The temperature of the components of the infrared remote sensing instrument includes at least one of the following: the temperature of the mechanical support components, the temperature of the optical path components, the temperature of the detector assembly, and the temperature of the radiation cooler.

8. The method according to claim 6, characterized in that, The status parameters also include at least one of the scanning mirror rotation angle of the infrared remote sensing instrument and an indicator of the time the infrared remote sensing instrument has been operating.

9. An electronic device, comprising: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation corresponding to the radiometric calibration method of the infrared remote sensing instrument as described in any one of claims 6-8.

10. A computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the radiometric calibration method of an infrared remote sensing instrument as described in any one of claims 6-8.

Images