Satellite-borne SAR (Synthetic Aperture Radar) radiometric calibration compensation method under non-uniform rainfall condition

Abstract

The invention relates to a satellite-borne SAR (Synthetic Aperture Radar) radiometric calibration compensation method under a non-uniform rainfall condition. The satellite-borne SAR radiometric calibration compensation method comprises the following steps: constructing a'satellite-borne SAR observation system parameter-three-dimensional rainfall spatial distribution-radiometric calibration deviation 'database; taking the satellite-borne SAR observation system parameters in the database and the three-dimensional rainfall spatial distribution as input characteristics, taking the corresponding radiometric calibration deviation as an output label, training the deep learning model, and obtaining a radiometric calibration deviation prediction model; and acquiring a to-be-compensated satellite-borne SAR image, predicting a radiometric calibration deviation value of each pixel point or region in the to-be-compensated SAR image by using the radiometric calibration deviation prediction model according to the corresponding satellite-borne SAR observation system parameters and the time-space three-dimensional rainfall spatial distribution, and performing radiation compensation on the to-be-compensated SAR image based on the radiometric calibration deviation value. According to the method, the radiation error compensation capability with higher precision is realized, the radiation error space difference among the pixel points of the SAR image can be reduced, and the quantitative remote sensing performance of the satellite-borne SAR in land and sea observation can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radar remote sensing, and particularly relates to a satellite-borne SAR radiation calibration compensation method under non-uniform rainfall conditions. BACKGROUND

[0002] As an active microwave remote sensing system, Synthetic Aperture Radar (SAR) has the ability of all-weather and all-time observation with high resolution, and has been widely used in quantitative remote sensing fields such as ground deformation monitoring, resource exploration, disaster assessment and ocean observation. The core premise of SAR quantitative application is accurate radiation calibration, that is, converting the gray value or power value of SAR image into normalized backscatter coefficient (σ0) with clear physical meaning. Traditional satellite-borne SAR radiation calibration mainly relies on passive calibrators (such as corner reflectors) with known radar cross section (RCS) laid on the ground, or uses large-area uniformly distributed natural scenes (such as Amazon tropical rainforest, ocean, desert, etc.) as reference targets, to realize the radiation deviation correction of the whole image by establishing the function relationship between image digital value (DN) and backscatter coefficient.

[0003] However, the above-mentioned traditional calibration method has significant deficiencies in processing SAR data affected by atmosphere, especially high-frequency band satellite-borne SAR in rainfall conditions. Raindrops in the rain can cause absorption and scattering of microwave signals, resulting in signal attenuation on the propagation path. This attenuation is closely related to rainfall intensity, raindrop spectrum distribution, signal frequency and polarization mode, etc. Because the rain usually presents strong spatial non-uniformity within the coverage range of tens to hundreds of kilometers of SAR image, it causes the signal to pass through different rainfall intensity areas, so that different pixels in the SAR image receive different degrees of attenuation, and then produce radiation errors with obvious spatial differences. The traditional average calibration method based on the whole image or limited point targets cannot accurately depict and compensate the pixel-level radiation deviation caused by the non-uniform spatial distribution of rainfall, which seriously restricts the quantitative application efficiency of SAR in rainy weather.

[0004] At present, some studies try to use meteorological models or measured meteorological data to roughly correct the atmospheric influence, but most of them are based on the assumption of uniform distribution of rainfall, or use low-resolution meteorological data for regional average correction, which cannot fundamentally solve the problem of spatial difference of radiation error caused by spatial heterogeneity of rainfall. In addition, the diversity of SAR system working modes (such as strip, spotlight, scan) and observation parameters (such as frequency, polarization, incidence angle, azimuth angle) further increases the complexity of the influence of rainfall attenuation, making it extremely difficult to establish a general analytical correction model.

[0005] Therefore, how to accurately obtain the spatial distribution information of rainfall, and establish the quantitative relationship between the radiation deviation and the multi-mode, multi-parameter SAR observation condition, and then realize the intelligent and fine radiation compensation of the SAR image, becomes a key technical challenge to improve the quantitative application ability of the spaceborne SAR under complex weather conditions. SUMMARY

[0006] In view of the above technical problems, the present application provides a spaceborne SAR radiation calibration compensation method under non-uniform rainfall conditions, which fuses high-resolution three-dimensional rainfall spatial distribution data and SAR system observation parameters to construct a radiation deviation prediction model, and realizes high-precision intelligent radiation compensation of SAR images affected by rainfall at the pixel level or regional level.

[0007] The technical solution for solving the technical problems of the present application is: a spaceborne SAR radiation calibration compensation method under non-uniform rainfall conditions, comprising the following steps:

[0008] Step S1, a "spaceborne SAR observation system parameter-three-dimensional rainfall spatial distribution-radiation calibration deviation" database is constructed; wherein the spaceborne SAR observation system parameters include the working mode and parameters of the SAR sensor; the three-dimensional rainfall spatial distribution is high-resolution gridded rainfall data that is simultaneous in space and time with SAR image observation;

[0009] Step S2, the spaceborne SAR observation system parameters and three-dimensional rainfall spatial distribution in the database are taken as input features, and the corresponding radiation calibration deviation is taken as output label, a deep learning model is trained, and a radiation calibration deviation prediction model is obtained;

[0010] Step S3, a spaceborne SAR image to be compensated is obtained, according to the corresponding spaceborne SAR observation system parameters and the three-dimensional rainfall spatial distribution that is simultaneous in space and time, the radiation calibration deviation prediction model is used to predict the radiation calibration deviation value of each pixel or region in the SAR image to be compensated, and the SAR image to be compensated is compensated based on the radiation calibration deviation value.

[0011] According to one technical solution of the present application, the step S1 specifically comprises:

[0012] Step S11, a plurality of spaceborne SAR images covering at least three passive calibrators with known radar cross sections are obtained;

[0013] Step S12, for each spaceborne SAR image, its corresponding spaceborne SAR observation system parameters, three-dimensional rainfall spatial distribution data that is simultaneous in space and time, and radiation calibration deviation data calculated based on the passive calibrators are obtained;

[0014] Step S13: Associate the spaceborne SAR observation system parameters, three-dimensional precipitation spatial distribution data and radiometric calibration deviation data obtained in step S12 to form a set of data pairs. The database is composed of multiple sets of the data pairs.

[0015] According to one technical solution of the present invention, step S12, based on the radiation calibration deviation data calculated by the passive calibrator, includes the following steps:

[0016] Acquire a reference SAR image that is observed from the same viewpoint and in the same mode as the SAR image under clear weather conditions, and calculate the received power of the corresponding pixel of the passive calibrator as the reference power.

[0017] Obtain the received power of the corresponding pixel of the passive calibrator in the SAR image to be tested under rainfall conditions;

[0018] The difference between the received power under the rainfall conditions and the reference power is calculated to obtain the radiometric calibration deviation of the pixel at the passive calibrator.

[0019] According to one technical solution of the present invention, the operating modes in the parameters of the spaceborne SAR observation system include strip mode, spotlight mode and scanning mode;

[0020] The parameters include frequency, polarization mode, azimuth angle, and incident angle;

[0021] The three-dimensional rainfall spatial distribution data is a three-dimensional grid data with a horizontal spatial resolution of 1~4km and a vertical resolution of 200~500m, and the rainfall measurement accuracy is better than 2mm / h.

[0022] According to one technical solution of the present invention, in step S2, the deep learning model is a convolutional neural network;

[0023] The input features also include an estimated total rainfall impact index calculated based on the geometric relationship between the SAR signal transmission and reception path and the three-dimensional rainfall grid.

[0024] According to one technical solution of the present invention, the total rainfall impact index The theoretical expression is: The rainfall intensity passing through the precipitation layer along the signal path LOS is used to represent this.

[0025] ,

[0026] in, The intensity of rainfall is determined by latitude. ,longitude and height The function, the integration path This represents the two-way transmission and reception path of the SAR signal within the three-dimensional spatial distribution of rainfall.

[0027] In actual calculations, the total rainfall impact index Its valuation is calculated using gridded data. The calculation method is as follows: for the mesh The rainfall intensity within the grid is The signal travels through the grid for a length of time. The total rainfall impact index estimate The following is obtained by summing the rainfall intensity within each three-dimensional rainfall grid traversed by the SAR signal path:

[0028] ,

[0029] Where 2 represents a two-way path.

[0030] According to a technical solution of the present invention, in step S3, radiometric compensation is performed on the SAR image to be compensated based on the radiometric calibration deviation value, specifically including:

[0031] Based on the predicted radiometric calibration deviation value of the pixel or region The backscattering coefficient value of the corresponding pixel or region in the SAR image to be compensated. Correction is performed to obtain the corrected backscattering coefficient value. , is represented as:

[0032] .

[0033] According to one aspect of the present invention, a spaceborne SAR radiometric calibration compensation system under non-uniform rainfall conditions is proposed, comprising:

[0034] The database construction module is used to construct a database of "spaceborne SAR observation system parameters - three-dimensional precipitation spatial distribution - radiometric calibration deviation"; wherein, the spaceborne SAR observation system parameters include the working mode and parameters of the SAR sensor; the three-dimensional precipitation spatial distribution is high-resolution gridded precipitation data that is simultaneously observed with the SAR image;

[0035] The model training module is used to train a deep learning model with the parameters of the spaceborne SAR observation system and the three-dimensional spatial distribution of rainfall in the database as input features and the corresponding radiometric calibration deviation as output label, so as to obtain a radiometric calibration deviation prediction model.

[0036] The radiation compensation module is used to acquire the spaceborne SAR image to be compensated, and based on the corresponding spaceborne SAR observation system parameters and the three-dimensional spatial distribution of rainfall in time and space, it uses the radiation calibration deviation prediction model to predict the radiation calibration deviation value of each pixel or region in the SAR image to be compensated, and performs radiation compensation on the SAR image to be compensated based on the radiation calibration deviation value.

[0037] According to one technical solution of the present invention, the database construction module includes:

[0038] The data acquisition unit is used to acquire multiple SAR images covering the passive calibrator, the corresponding parameters of the spaceborne SAR observation system, and three-dimensional precipitation data in space and time.

[0039] The deviation calculation unit is used to calculate the radiation calibration deviation based on the difference in echo power of the passive calibrator under sunny and rainy conditions.

[0040] The data association storage unit is used to associate and store the parameters of the spaceborne SAR observation system, three-dimensional rainfall data, and radiometric calibration deviation to form the database.

[0041] According to one technical solution of the present invention, the radiation compensation module includes:

[0042] The prediction unit is used to input the observation conditions of the SAR image to be compensated and the three-dimensional rainfall data into the prediction model to obtain the radiometric calibration deviation prediction map;

[0043] The correction unit is used to perform pixel-by-pixel correction on the SAR image to be compensated based on the radiometric calibration deviation prediction map.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] This invention provides a method for radiometric calibration compensation of spaceborne SAR under non-uniform rainfall conditions. It proposes and constructs a correlation database of "SAR observation conditions - three-dimensional rainfall distribution - radiometric deviation", which accurately correlates the fine distribution of rainfall in the horizontal and vertical directions with the SAR signal propagation path. This method breaks through the limitations of traditional methods that assume uniform rainfall influence or use coarse resolution data, and can more accurately characterize the spatial differences in rainfall attenuation, thus solving the radiometric error problem caused by the spatial non-uniformity of rainfall.

[0046] This invention, by introducing a deep learning model, can automatically learn and establish a complex nonlinear mapping relationship from multidimensional input (SAR parameters + three-dimensional rainfall data) to radiation bias. This model not only considers the influence of different rainfall intensities, but also integrates the differences in the sensitivity of rainfall attenuation to factors such as different SAR working modes, frequencies, polarization, and observation angles. This significantly improves the adaptability and accuracy of radiation compensation under different observation conditions, and realizes high-precision and intelligent radiation bias prediction.

[0047] This invention utilizes a trained prediction model to predict and compensate for radiation errors in SAR images with a wide coverage area, either pixel-by-pixel or region-by-region. This effectively addresses the problem of significant differences in scattering characteristics in different regions of an image due to uneven rainfall distribution, thereby improving the consistency and spatial uniformity of the radiation accuracy of the entire SAR image. Attached Figure Description

[0048] Figure 1 This is a flowchart illustrating the overall process of the spaceborne SAR radiometric calibration and compensation method under non-uniform rainfall conditions provided in this embodiment of the invention.

[0049] Figure 2 This is a schematic diagram of a typical scenario for spaceborne SAR radiometric calibration under heavy rainfall conditions in an embodiment of the present invention;

[0050] Figure 3 This is a schematic diagram of the passive calibration method for SAR images under no rainfall conditions in an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram illustrating the method for calculating the radiation deviation of the pixel where the corner reflector is located under the influence of rainfall in an embodiment of the present invention;

[0052] Figure 5 This is a schematic cross-sectional view of the SAR signal transceiver link passing through a non-uniformly distributed rain layer in an embodiment of the present invention;

[0053] Figure 6 This is a schematic cross-sectional view of the SAR signal transceiver link passing through a three-dimensional rainfall tomography grid in an embodiment of the present invention;

[0054] Figure 7 This is a schematic diagram illustrating the construction of the intelligent prediction model for radiation deviation in an embodiment of the present invention. Detailed Implementation

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the embodiments will be briefly described below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0056] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The embodiments cannot be described in detail here, but the embodiments of the present invention are not limited to the following embodiments.

[0057] like Figures 1 to 7 As shown, the present invention provides a method for satellite-borne SAR radiometric calibration compensation under non-uniform rainfall conditions, which mainly includes three core steps: database construction, model training, and intelligent compensation using the trained model. Specifically, the method includes the following steps:

[0058] Step S1: Acquire and analyze multiple SAR data images to construct a database of "spaceborne SAR observation system parameters - three-dimensional precipitation spatial distribution - radiometric calibration bias"; wherein, the spaceborne SAR observation system parameters include the operating mode and parameters of the SAR sensor; the three-dimensional precipitation spatial distribution is high-resolution gridded precipitation data from the same time and space as the SAR image observation, specifically including:

[0059] Step S11: Acquire multiple spaceborne SAR images covering at least three passive calibrators with known radar cross sections;

[0060] Step S12: For each spaceborne SAR image, acquire its corresponding spaceborne SAR observation system parameters, three-dimensional spatial distribution data of rainfall in space and time, and radiometric calibration deviation data calculated based on the passive calibrator.

[0061] Step S13: Associate the spaceborne SAR observation system parameters, three-dimensional precipitation spatial distribution data and radiometric calibration deviation data obtained in step S12 to form a set of data pairs. The database is composed of multiple sets of the data pairs.

[0062] Step S1 is the data foundation for the high-precision calibration of this invention. The specific implementation is as follows:

[0063] First, a large amount of sample data needs to be acquired. This data comes from geometrically corrected spaceborne SAR (C-band or X-band) images, and at least three ground corner reflectors with known RCS (radiative cross-section) must be fixedly deployed as passive calibrators within the observation area corresponding to each image. Simultaneously, high-resolution three-dimensional rainfall spatial distribution data must be acquired, synchronously with the observation time of each SAR image and covering the same spatial area. This rainfall data should be a gridded product, with a recommended horizontal resolution of 1–4 km and a recommended vertical resolution of 200–500 m, and a rainfall intensity measurement accuracy better than 2 mm / h, to accurately describe the distribution of rainfall in three-dimensional space.

[0064] Secondly, the true radiometric calibration deviation of the corner reflector positions in each sample SAR image is calculated. Specific methods (e.g.) Figure 3 ,Figure 4 (As shown): Under clear, cloudless, and rainless weather conditions, multiple SAR observations were conducted on the same area, from the same viewpoint, and using the same mode. A baseline calibration curve was established by fitting the theoretical RCS value of the corner reflector with its echo power (averaged multiple times) in clear images, representing the relationship between image power and RCS values ​​under this observation mode without rainfall. Then, for SAR images acquired under rainfall conditions, the actual echo power of the corresponding pixel of the corner reflector was read and compared with the theoretical echo power calculated based on the baseline calibration curve and the RCS value of that corner reflector. The difference represents the radiometric calibration deviation caused by rainfall at that corner reflector location.

[0065] Finally, data association and database entry are performed. For each SAR image under rainfall conditions, the observation conditions (including imaging mode, center frequency, polarization, average azimuth, average incident angle, etc.), the synchronous three-dimensional rainfall distribution data at the time of image acquisition, and the calculated radiometric calibration deviation values ​​for each corner reflector position are precisely matched and associated in time and space to form a complete data record. By accumulating numerous data records covering different seasons, climate zones, rainfall types and intensities, and SAR observation modes, a large-scale and information-rich database of relationships between "spaceborne SAR observation system parameters—three-dimensional rainfall spatial distribution—radiative deviation" can be constructed to support model training, prediction, and subsequent data expansion.

[0066] Reference Figure 2In a typical scenario of spaceborne SAR radiometric calibration under heavy rainfall conditions, a SAR satellite flies along the azimuth, transmitting microwave signals to target areas on the Earth's surface via its antenna and receiving echoes to generate SAR images covering land or sea. SAR signals are transmitted and received at specific frequencies (e.g., C-band or X-band), polarizations (V / H polarization, circular polarization, etc.), and specific incident and azimuth angles. Different frequencies and polarizations cause varying degrees of scattering and absorption loss when the signal passes through the rain layer. Different incident and azimuth angles result in different propagation path lengths and traversed areas within the rain layer, directly affecting the power and scattering characteristics of the signal upon reaching the Earth's surface and returning to the satellite. During satellite signal propagation, the satellite passes through a non-uniformly distributed rain layer. Due to the uneven spatial distribution of rainfall, different areas within the SAR image coverage experience varying rainfall intensities during signal propagation, leading to different attenuation and scattering effects along the signal propagation path, resulting in different radiation deviations in the echo signal across different areas. Multiple corner reflectors are deployed on the ground to acquire the true scattering intensity of SAR observation echoes under different rainfall conditions. By comparing theoretical scattering values ​​with actual observation values, the radiation deviation of SAR signals caused by rainfall under different observation conditions is quantified. The data measured by these corner reflectors will be used to construct the "Spaceborne SAR Observation System Parameters - Three-Dimensional Rainfall Spatial Distribution - Radiometric Calibration Deviation" database of this invention, providing reliable training samples for subsequent deep learning modeling and SAR image radiometric compensation.

[0067] Step S2: Using the parameters of the spaceborne SAR observation system and the three-dimensional spatial distribution of rainfall in the database as input features, and the corresponding radiometric calibration deviation as output label, train a deep learning model to obtain a radiometric calibration deviation prediction model.

[0068] Step S2 is the core of the invention's intelligence, its function being to learn patterns from data. The specific implementation is as follows:

[0069] The database constructed in step S1 is used as the training sample set. For each sample data point, an input feature vector for the model needs to be constructed. The input features mainly consist of two parts: the first part is the SAR observation condition parameters, represented in numerical or coded form; the second part is the three-dimensional rainfall distribution characteristics, which, in order to be effectively input into the model, need to be combined with the SAR observation geometry (incident angle, azimuth angle). Specifically, based on the satellite position, imaging geometry, and digital elevation model (DEM), the two-way propagation path (LOS) of the SAR signal from the satellite to each corner reflector (i.e., ground target point) can be retrieved. This path is then overlaid with the three-dimensional rainfall grid (e.g., ...). Figure 5 , Figure 6 (As shown), calculate the rainfall intensity within each rainfall grid cell traversed by the signal path. Length of the path through the grid The sum of the products, i.e., the estimate of the total rainfall impact index. (Factor 2 represents two-way travel). This exponent R, as a highly condensed composite feature consistent with the physical process, together with SAR observation parameters (frequency f, polarization Pol, incident angle θ, etc.) constitutes the input feature vector.

[0070] The model's output label is the radiation calibration deviation value calculated for that sample.

[0071] A suitable deep learning model, such as a multilayer perceptron (MLP) or convolutional neural network (CNN), is selected and supervised training is performed with the input feature vector X and the radiometric bias Y. The training objective is to minimize the error (e.g., mean squared error, MSE) between the model's predicted bias and the actual bias. Through training with a large number of samples, the model can learn the complex nonlinear laws governing the impact of rainfall of different intensities and spatial structures on signal attenuation under different SAR system configurations, ultimately resulting in a stable and reliable radiometric calibration bias prediction model.

[0072] Step S3: Obtain the spaceborne SAR image to be compensated. Based on the corresponding spaceborne SAR observation system parameters and the three-dimensional spatial distribution of rainfall in space and time, use the radiometric calibration deviation prediction model to predict the radiometric calibration deviation value of each pixel or region in the SAR image to be compensated, and perform radiometric compensation on the SAR image to be compensated based on the radiometric calibration deviation value.

[0073] Step S3 involves applying the model trained in step S2 in a practical manner. The specific implementation is as follows:

[0074] When radiometric calibration is required for a new spaceborne SAR image affected by rainfall, the SAR observation condition parameters of the image and high-resolution three-dimensional rainfall distribution data at the same time and in the same space as the image are first obtained.

[0075] Then, the image to be processed is considered as being composed of numerous virtual "target points" (i.e., each pixel or a cell block). For each point (pixel) that needs correction, its corresponding signal propagation path and total rainfall impact index estimate are calculated based on its geographical location and SAR imaging geometry. Estimate the total rainfall impact index at this point. Combined with SAR observation condition parameters, this forms the input feature vector for that point.

[0076] Inputting this feature vector into the radiometric calibration bias prediction model trained in step S2 will allow the model to output the radiometric bias at that point. The predicted value.

[0077] Finally, based on the predicted radiometric calibration deviation values ​​of pixels or regions... The backscattering coefficient value of the corresponding pixel or region in the SAR image to be compensated. Correction is performed to obtain the corrected backscattering coefficient value. , is represented as:

[0078] .

[0079] By constructing a dedicated multi-source correlation database, complex physical processes that are difficult to model directly are transformed into data-driven learning problems. Utilizing the powerful nonlinear fitting capabilities of deep learning models, a mapping from specific observation conditions and rainfall scenarios to precise radiometric biases is achieved. This method differs from traditional approaches such as "one calibration coefficient per image" or "coarse regional average correction," achieving spatially adaptive, pixel-level high-precision radiometric compensation for spaceborne SAR images under rainfall conditions. Practical applications demonstrate that within a rainfall intensity range of 2-10 mm / h, this method effectively optimizes radiometric calibration errors caused by rainfall, significantly improving the consistency of radiometric accuracy across different regions of the image. This provides crucial technical support for quantitative SAR applications during rainfall (such as soil moisture retrieval and sea surface wind field extraction).

[0080] Considering that the observation range of a SAR image is tens to hundreds of kilometers, the rainfall in different areas of the image usually varies greatly. The attenuation of C / X band radar signals by rainfall leads to significant differences in the scattering characteristics of different areas of an image. In addition, the impact of rainfall on SAR observations varies with different modes and parameters. Therefore, the non-uniformity of rainfall in the plane and vertical direction must be considered when radiometrically calibrating the image. The radiometric calibration of each pixel of the image needs to be accurately corrected by the SAR working mode and parameters as well as the three-dimensional distribution of rainfall.

[0081] In some embodiments of this invention, step S12, based on the radiation calibration deviation data calculated by the passive calibrator, includes the following steps:

[0082] Acquire a reference SAR image that is observed from the same viewpoint and in the same mode as the SAR image under clear weather conditions, and calculate the received power of the corresponding pixel of the passive calibrator as the reference power.

[0083] Obtain the received power of the corresponding pixel of the passive calibrator in the SAR image to be tested under rainfall conditions;

[0084] The difference between the received power under the rainfall conditions and the reference power is calculated to obtain the radiometric calibration deviation of the pixel at the passive calibrator.

[0085] The key step in this process is to calculate the radiation calibration deviation using a corner reflector.

[0086] For a fixed ground corner reflector array (containing at least 3 corner reflectors with different RCS values), under prolonged clear and rainless conditions (e.g., two consecutive weeks without rainfall), using the same spaceborne SAR sensor, with the same imaging mode (e.g., strip mode), and a similar range of incident angles, record the echo power of the corresponding pixels of the corner reflectors. Assume the true RCS of corner reflector i is... The echo power is P ri The average value is calculated from observation data collected on multiple (usually more than 10) sunny days. As a reference value for the echo power of the corner reflector in clear weather Scatter plots of received power and RCS were drawn and fitted to establish a curve showing the relationship between received power and true RCS under conditions of no rainfall. .

[0087] Reference Figure 3 As shown, under clear weather conditions with no rainfall, the echo power returned by SAR satellites under different RCS values ​​was obtained by deploying multiple corner reflectors in the observation area. In the figure, the small gray dots represent the observed echo power values ​​at the corresponding locations of corner reflectors in SAR images obtained from multiple observations under no-rainfall conditions; the black triangles represent the average echo power calculated based on multiple observation results. For multiple groups... By performing linear fitting on the scatter plots, the functional relationship between the echo power of the corner reflector and the RCS under the condition of no rainfall is obtained (i.e., the gray dashed line in the figure).

[0088] Reference Figure 4 When rainfall occurs, the SAR satellite observes the area again, acquiring images from the same viewpoint and using the same mode as the SAR observations under the same rainy weather conditions. During propagation, the SAR signal is scattered and absorbed by the rain layer, causing echo power attenuation. Therefore, the actual received echo power will be lower than the reference echo power under the no-rainy conditions, resulting in a difference from the reference value under no-rainy conditions. This represents the observed echo power values ​​at the corresponding locations of each corner reflector under rainfall conditions.

[0089] Calculate the difference between the reference value of the echo power of each corner reflector under no-rain conditions and the measured value under rain conditions. This refers to the actual radiation deviation value caused by rainfall for corner reflectors with different RCS.

[0090] To ensure database quality, a signal-to-noise ratio (SNR) threshold can be set. Only when the SNR of the corner reflector in both clear and rainy images exceeds a certain threshold (e.g., 20 dB) will the calculated bias data be included in the database. Simultaneously, the specific SAR image row and column numbers corresponding to the calculated bias are recorded to facilitate precise spatial correlation with the 3D rainfall data.

[0091] In some embodiments of the present invention, the operating modes in the parameters of the spaceborne SAR observation system include strip mode, spotlight mode, and scan mode;

[0092] The parameters include frequency, polarization mode, azimuth angle, and incident angle;

[0093] High-resolution three-dimensional precipitation spatial distribution refers to precipitation data that is gridded in the horizontal direction and has a tortuous structure in the vertical direction. The horizontal spatial resolution is 1~4km, the vertical resolution is 200~500m, the precipitation measurement accuracy is better than 2mm / h, and the overall precipitation presents a three-dimensional grid shape. The precipitation is the precipitation in the same area at the same time as the SAR image observation.

[0094] In some embodiments of the present invention, in step S2, the deep learning model is a convolutional neural network;

[0095] The input features also include an estimated total rainfall impact index calculated based on the geometric relationship between the SAR signal transmission and reception path and the three-dimensional rainfall grid.

[0096] In this embodiment, such as Figure 7 As shown, a specific construction scheme for a radiation deviation prediction model based on a convolutional neural network is proposed, wherein the input parameters include:

[0097] (1) SAR mode and parameter vector (frequency, polarization, incident angle, azimuth angle, imaging mode coding);

[0098] (2) The three-dimensional distribution characteristics of rainfall in the corresponding observation area at the observation time (can be input after dimensionality reduction encoding);

[0099] (3) Distribution of SAR image pixels in terms of geographic location, azimuth and incident angle.

[0100] The output is the radiation deviation value for each corner reflector.

[0101] The model improves the accuracy of predicting radiation deviation in SAR images under heavy rainfall conditions by minimizing the mean square error (MSE) or weighted loss function between the prediction deviation and the measured deviation in the database through supervised learning.

[0102] The training process can involve randomly dividing the constructed database into a training set (e.g., 70%), a validation set (e.g., 15%), and a test set (e.g., 15%). The network weights are iteratively optimized using the training set data, while the training process is monitored using the validation set to prevent overfitting. Training is stopped early when the validation set performance no longer improves. Finally, the generalization performance of the model is evaluated using the test set. Evaluation metrics include the root mean square error (RMSE) of the predicted bias versus the true bias, the mean absolute error (MAE), and the correlation coefficient (R²). A well-trained model should have an RMSE significantly smaller than the typical bias range caused by rainfall.

[0103] Reference Figure 5 When a SAR satellite is in orbit, it transmits and receives microwave signals to the Earth's surface through its antenna, and the imaging swath is S. SAR During transmission and reception, the signal passes through the rain layer located between the satellite and the Earth's surface at an oblique angle. The different thicknesses of the dashed lines represent the intensity of the rainfall; thicker lines indicate heavier rainfall, while thinner lines indicate lighter rainfall. The intervals between the dashed lines... The planar resolution of the rainfall is given. The height of the rainfall layer varies significantly at different ground distances, and the spatial distribution of rainfall intensity is also non-uniform. Within the imaging swath, the length of the SAR signal propagation path through the rainfall layer, the rainfall intensity it passes through, and the height of the rainfall it passes through are all different at different ground distances. For a certain ground distance, the SAR signal will pass through the corresponding rainfall area on its propagation path. The length of the path and the rainfall intensity of the area it passes through will jointly determine the signal propagation loss and scattering effect at that location.

[0104] In some embodiments of the present invention, the total rainfall impact index The theoretical expression is: The rainfall intensity passing through the precipitation layer along the signal path LOS is used to represent this.

[0105] ,

[0106] in, The intensity of rainfall is determined by latitude. ,longitude and height The function, the integration path This represents the two-way transmission and reception path of the SAR signal within the three-dimensional spatial distribution of rainfall.

[0107] Where 2 represents a two-way path;

[0108] In practical applications, refer to Figure 6 To quantitatively describe the non-uniformity of rainfall spatial distribution, this invention employs a three-dimensional rainfall tomography mesh model to depict the spatial distribution of rainfall.

[0109] The rainfall layer is discretized into a three-dimensional grid in both the vertical and horizontal directions;

[0110] The horizontal resolution is 2-3 km, and the vertical resolution is 200-300 m;

[0111] In actual calculations, the total rainfall impact index Its valuation is calculated using gridded data. The calculation method is as follows: for the mesh The rainfall intensity within the grid is The signal travels through the grid for a length of time. .

[0112] like Figure 6 As shown, the dark gray grid areas in the figure represent the precipitation layer grid cells that the SAR signal passes through on its transmission and reception path. As the incident angle of the signal varies at different positions in the ground distance direction, the length of the grid path it traverses and the intensity of the precipitation passing through the grids also differ. This directly leads to spatial non-uniformity in the radiation deviation exhibited by the pixel or region in the SAR image. The three-dimensional grid (x-axis) through which the signal passes is determined by the SAR azimuth and incident angle. i ,y i ,h i The total rainfall impact index mentioned above will then be calculated. It can be replaced with the total rainfall impact index estimate. It is obtained by accumulating the rainfall intensity within each three-dimensional rainfall grid traversed by the SAR signal path, and is expressed as:

[0113] ,

[0114] Where 2 represents a two-way path.

[0115] According to one aspect of the present invention, a spaceborne SAR radiometric calibration compensation system under non-uniform rainfall conditions is proposed, comprising:

[0116] The database construction module is used to construct a database of "spaceborne SAR observation system parameters - three-dimensional precipitation spatial distribution - radiometric calibration deviation"; wherein, the spaceborne SAR observation system parameters include the working mode and parameters of the SAR sensor; the three-dimensional precipitation spatial distribution is high-resolution gridded precipitation data that is simultaneously observed with the SAR image;

[0117] The model training module is used to train a deep learning model with the parameters of the spaceborne SAR observation system and the three-dimensional spatial distribution of rainfall in the database as input features and the corresponding radiometric calibration deviation as output label, so as to obtain a radiometric calibration deviation prediction model.

[0118] The radiation compensation module is used to acquire the spaceborne SAR image to be compensated, and based on the corresponding spaceborne SAR observation system parameters and the three-dimensional spatial distribution of rainfall in time and space, it uses the radiation calibration deviation prediction model to predict the radiation calibration deviation value of each pixel or region in the SAR image to be compensated, and performs radiation compensation on the SAR image to be compensated based on the radiation calibration deviation value.

[0119] According to one technical solution of the present invention, the database construction module includes:

[0120] The data acquisition unit is used to acquire multiple SAR images covering the passive calibrator, the corresponding parameters of the spaceborne SAR observation system, and three-dimensional precipitation data in space and time.

[0121] The deviation calculation unit is used to calculate the radiation calibration deviation based on the difference in echo power of the passive calibrator under sunny and rainy conditions.

[0122] The data association storage unit is used to associate and store the parameters of the spaceborne SAR observation system, three-dimensional rainfall data, and radiometric calibration deviation to form the database.

[0123] According to one technical solution of the present invention, the radiation compensation module includes:

[0124] The prediction unit is used to input the observation conditions of the SAR image to be compensated and the three-dimensional rainfall data into the prediction model to obtain the radiometric calibration deviation prediction map;

[0125] The correction unit is used to perform pixel-by-pixel correction on the SAR image to be compensated based on the radiometric calibration deviation prediction map.

[0126] According to one aspect of the present invention, an electronic device is provided, comprising: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are stored in the memory; when the electronic device is running, the processor executes the one or more computer programs stored in the memory to cause the electronic device to perform a spaceborne SAR radiometric calibration compensation method under non-uniform rainfall conditions as described in any of the above technical solutions.

[0127] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0128] The memory can be an internal storage unit of the terminal device, such as a hard drive or RAM. Alternatively, it can be an external storage device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory can include both internal and external storage units. The memory is used to store the computer program and other programs and data required by the terminal device. It can also be used to temporarily store data that has been output or will be output.

[0129] According to one aspect of the present invention, a computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement a spaceborne SAR radiometric calibration compensation method under non-uniform rainfall conditions as described in any of the above technical solutions.

[0130] For example, computer-readable storage media can be read-only memory (ROM), random access memory (RAM), read-only optical disc (CD-ROM), magnetic tape, floppy disk, and optical data storage devices. They can be implemented using computer-executable program code, thus allowing them to be stored in a storage device for execution by a computing device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Therefore, this invention is not limited to any particular hardware and software combination.

[0131] In summary, this invention proposes a satellite-borne SAR radiometric calibration compensation method under non-uniform rainfall conditions, which solves the problem of insufficient correction of radiometric deviation caused by the assumption of uniform rainfall influence in SAR images in traditional radiometric calibration methods, and significantly improves calibration accuracy.

[0132] This invention uses deep learning to establish a nonlinear prediction relationship between pixel brightness value and RCS, replacing the traditional empirical formula, and improving the compensation accuracy under different rainfall intensities, types and observation conditions;

[0133] This invention, through regional intelligent compensation, greatly enhances the quantitative monitoring capability of SAR under extreme weather conditions such as heavy precipitation, typhoons, and rainstorms, thereby improving the application efficiency of quantitative remote sensing.

[0134] This invention has the ability to perform radiometric calibration compensation under rainfall intensities of 2~10 mm / h, which can effectively improve the radiometric quality of SAR images under severe weather conditions such as heavy rainfall, typhoons, and rainstorms, and provide a reliable data foundation for quantitative remote sensing applications with high data accuracy requirements, such as hydrological monitoring, disaster emergency response, and marine dynamic environment inversion.

[0135] Furthermore, it should be noted that the embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0136] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0137] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0138] It should also be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0139] Finally, it should be noted that the above description represents a preferred embodiment of the present invention. It should be pointed out that although preferred embodiments have been described, those skilled in the art, once they understand the basic inventive concept of the present invention, can make various improvements and modifications without departing from the principles described herein. These improvements and modifications should also be considered within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A method for radiometric calibration and compensation of spaceborne SAR under non-uniform rainfall conditions, characterized in that, Includes the following steps: Step S1: Construct a database of "spaceborne SAR observation system parameters - three-dimensional precipitation spatial distribution - radiometric calibration deviation"; wherein, the spaceborne SAR observation system parameters include the working mode and parameters of the SAR sensor; the three-dimensional precipitation spatial distribution is high-resolution gridded precipitation data at the same time and space as the SAR image observation; Step S2: Using the parameters of the spaceborne SAR observation system and the three-dimensional spatial distribution of rainfall in the database as input features, and the corresponding radiometric calibration deviation as output label, train a deep learning model to obtain a radiometric calibration deviation prediction model. Step S3: Obtain the spaceborne SAR image to be compensated. Based on the corresponding spaceborne SAR observation system parameters and the three-dimensional spatial distribution of rainfall in space and time, use the radiometric calibration deviation prediction model to predict the radiometric calibration deviation value of each pixel or region in the SAR image to be compensated, and perform radiometric compensation on the SAR image to be compensated based on the radiometric calibration deviation value.

2. The method according to claim 1, characterized in that, Step S1 specifically includes: Step S11: Acquire multiple spaceborne SAR images covering at least three passive calibrators with known radar cross sections; Step S12: For each spaceborne SAR image, acquire its corresponding spaceborne SAR observation system parameters, three-dimensional spatial distribution data of rainfall in space and time, and radiometric calibration deviation data calculated based on the passive calibrator. Step S13: Associate the spaceborne SAR observation system parameters, three-dimensional precipitation spatial distribution data and radiometric calibration deviation data obtained in step S12 to form a set of data pairs. The database is composed of multiple sets of the data pairs.

3. The method according to claim 2, characterized in that, In step S12, the radiation calibration deviation data calculated based on the passive calibrator includes the following steps: Acquire a reference SAR image that is observed from the same viewpoint and in the same mode as the SAR image under clear weather conditions, and calculate the received power of the corresponding pixel of the passive calibrator as the reference power. Obtain the received power of the corresponding pixel of the passive calibrator in the SAR image to be tested under rainfall conditions; The difference between the received power under the rainfall conditions and the reference power is calculated to obtain the radiometric calibration deviation of the pixel at the passive calibrator.

4. The method according to claim 2 or 3, characterized in that, The operating modes in the parameters of the spaceborne SAR observation system include strip mode, spotlight mode, and scan mode; The parameters include frequency, polarization mode, azimuth angle, and incident angle; The three-dimensional rainfall spatial distribution data is a three-dimensional grid data with a horizontal spatial resolution of 1~4km and a vertical resolution of 200~500m, and the rainfall measurement accuracy is better than 2mm / h.

5. The method according to claim 1, characterized in that, In step S2, the deep learning model is a convolutional neural network; The input features also include an estimated total rainfall impact index calculated based on the geometric relationship between the SAR signal transmission and reception path and the three-dimensional rainfall grid.

6. The method according to claim 5, characterized in that, Total Rainfall Impact Index The theoretical expression is: The rainfall intensity passing through the precipitation layer along the signal path LOS is used to represent this. ， in, Rainfall intensity is measured in latitude. ,longitude and height The function, the integration path This represents the two-way transmission and reception path of the SAR signal within the three-dimensional spatial distribution of rainfall. In actual calculations, the total rainfall impact index Its valuation is calculated using gridded data. The calculation method is as follows: for the mesh The rainfall intensity within the grid is The signal travels through the grid for a length of time. The total rainfall impact index estimate The following is obtained by summing the rainfall intensity within each three-dimensional rainfall grid traversed by the SAR signal path: ， Where 2 represents a two-way path.

7. The method according to claim 6, characterized in that, In step S3, radiometric compensation is performed on the SAR image to be compensated based on the radiometric calibration deviation value, specifically including: Based on the predicted radiometric calibration deviation value of the pixel or region The backscattering coefficient value of the corresponding pixel or region in the SAR image to be compensated. The corrected backscattering coefficient value is obtained by performing correction. , is represented as: 。 8. A spaceborne SAR radiometric calibration compensation system under non-uniform rainfall conditions, characterized in that, include: The database construction module is used to construct a database of "spaceborne SAR observation system parameters - three-dimensional precipitation spatial distribution - radiometric calibration deviation"; wherein, the spaceborne SAR observation system parameters include the working mode and parameters of the SAR sensor; the three-dimensional precipitation spatial distribution is high-resolution gridded precipitation data that is simultaneously and spatially with the SAR image observation; The model training module is used to train a deep learning model with the parameters of the spaceborne SAR observation system and the three-dimensional spatial distribution of rainfall in the database as input features and the corresponding radiometric calibration deviation as output label, so as to obtain a radiometric calibration deviation prediction model. The radiation compensation module is used to acquire the spaceborne SAR image to be compensated, and based on the corresponding spaceborne SAR observation system parameters and the three-dimensional spatial distribution of rainfall in time and space, it uses the radiation calibration deviation prediction model to predict the radiation calibration deviation value of each pixel or region in the SAR image to be compensated, and performs radiation compensation on the SAR image to be compensated based on the radiation calibration deviation value.

9. The system according to claim 8, characterized in that, The database construction module includes: The data acquisition unit is used to acquire multiple SAR images covering the passive calibrator, the corresponding parameters of the spaceborne SAR observation system, and three-dimensional precipitation data in space and time. The deviation calculation unit is used to calculate the radiation calibration deviation based on the difference in echo power of the passive calibrator under sunny and rainy conditions. The data association storage unit is used to associate and store the parameters of the spaceborne SAR observation system, three-dimensional rainfall data, and radiometric calibration deviation to form the database.

10. The system according to claim 8, characterized in that, The radiation compensation module includes: The prediction unit is used to input the observation conditions of the SAR image to be compensated and the three-dimensional rainfall data into the prediction model to obtain the radiometric calibration deviation prediction map; The correction unit is used to perform pixel-by-pixel correction on the radiometric values ​​of the SAR image to be compensated based on the radiometric calibration deviation prediction map.

Images