L-band high-orbit SAR (Synthetic Aperture Radar) full-link radiometric calibration method based on satellite-ground cooperation

Abstract

The invention discloses an L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground cooperation, and the method comprises the following steps: S1, obtaining a hardware gain function changing along with a space angle based on echo data and RCS information between a micro-nano satellite carrying an active transponder and a high-orbit SAR satellite; s2, a pre-trained deep learning model is utilized to obtain software gains changing along with working parameters; and S3, performing high-precision radiometric calibration on the input to-be-calibrated high-orbit SAR image by using a full-link gain model obtained based on a hardware gain function and a software gain, and inverting the absolute radar cross-sectional area sigma of the target. According to the method, the problems of high-sea far-field calibration difficulty, long calibration period, uncalibrated ground processing software gain, obvious atmospheric effect influence and the like of a high-orbit SAR satellite can be solved, the separation calibration of the gain of a whole link from a space antenna to the ground processing can be realized, and the difficulty of software and hardware gain coupling in a traditional method is solved.

Description

Technical Field

[0001] This invention relates to the field of satellite radar calibration technology, and in particular to a full-link radiometric calibration method for L-band high-orbit SAR based on satellite-ground coordination. Background Technology

[0002] High-orbit L-band synthetic aperture radar, with its high-orbit characteristics, possesses significant advantages such as wide-area coverage, high revisit capability, and all-weather, day-and-night imaging, demonstrating irreplaceable application value in fields such as maritime surveillance, large-scale disaster assessment, agricultural surveys, and strategic remote sensing. However, its imaging quality and the level of quantitative application of data fundamentally depend on high-precision system radiometric calibration, among which on-orbit measurement and calibration of antenna patterns is a crucial and highly challenging link in the entire calibration chain.

[0003] The purpose of radiometric calibration is to establish a precise correspondence between the grayscale value of each pixel in a synthetic aperture radar image and the true backscattering coefficient of the target, so that the image grayscale can accurately reflect the physical characteristics of the ground objects and achieve quantitative remote sensing. Calibration is generally divided into internal calibration and external calibration. Internal calibration monitors changes in parameters such as transmitter and receiver gain through the calibration loop within the system, but it cannot measure passive components such as antenna patterns. External calibration, on the other hand, involves direct end-to-end measurements of the system by deploying calibrators with known scattering characteristics (such as corner reflectors) on the ground or by utilizing large-area uniformly distributed targets (such as tropical rainforests).

[0004] Currently, radiation calibration technology has mainly formed a technical system with ground-based calibration fields as the core and natural scene calibration as a supplement, and has achieved significant results on low-orbit synthetic aperture radar satellites. However, when serving high-orbit synthetic aperture radar, the existing technical system has revealed the following fundamental limitations that are difficult to overcome:

[0005] First, the geographical location of the ground-based calibration field is fixed, resulting in insufficient spatial coverage. While high-orbit synthetic aperture radar (HEAP) has an extremely wide coverage area, its calibration relies on satellites passing over fixed ground calibration fields. This leaves gaps in radiometric calibration data for vast, remote areas such as the open ocean and polar regions, making it impossible to guarantee imaging accuracy in these critical areas and limiting the global application capabilities of HEAP.

[0006] Secondly, the calibration cycle is limited by the satellite revisit cycle. Calibration relies on the satellite periodically visiting ground calibration fields, and its frequency is constrained by the satellite orbit and the distribution of ground stations. For high-orbit synthetic aperture radar, the revisit cycle can be as long as several days or even longer, making it difficult to meet the requirements of rapid, high-frequency calibration for routine, operational use. Slow drifts in system performance are also difficult to monitor and correct in a timely manner.

[0007] Furthermore, existing calibration methods cannot effectively overcome the effects of lower atmospheric (troposphere, ionosphere) on microwave signal attenuation, refraction, and delay. This effect is particularly significant in long-distance space-to-ground transmission links and changes dynamically with time and space, leading to systematic errors and uncertainties in ground measurement results, making them difficult to directly use for calibrating antenna hardware characteristics in space.

[0008] Finally, existing calibration methods primarily focus on measuring and correcting gain variations in radar hardware systems (including antennas, transmitters, and receivers), generally neglecting the gain uncertainties introduced by different imaging modes, algorithm selections, and signal processing parameters in ground data processing software. This software-processed gain is coupled with the hardware gain, and in traditional end-to-end calibration, it is uniformly calibrated as a single overall gain coefficient. When the software algorithm is updated or the processing parameters change, the entire calibration relationship becomes invalid, requiring a complete recalibration in the field, lacking flexibility and adaptability.

[0009] In recent years, low-Earth orbit (LEO) micro-nano satellite technology has offered new possibilities for space radiometric calibration due to its advantages of low cost, rapid deployment, and flexibility. By designing appropriate orbits and using LEO satellites carrying calibrators as dynamic reference targets, it is theoretically possible to achieve space-based on-orbit measurements of the radiation patterns of high-orbit synthetic aperture radar (SAR) antennas. This overcomes geographical limitations, shortens the calibration cycle, and avoids most atmospheric transmission effects. However, relying solely on space-based measurements cannot solve the calibration problem of ground-based processing software gains. The coupling problem between hardware and software gains remains, affecting the final accuracy and practicality of end-to-end calibration. Summary of the Invention

[0010] To address the problems existing in the prior art, the present invention aims to provide a satellite-ground coordinated L-band high-orbit SAR full-link radiometric calibration method, which overcomes the problems of high-orbit synthetic aperture radar satellites, such as difficulty in calibration in the far sea and distant areas, long calibration cycle, uncalibrated ground processing software gain, and significant atmospheric effects.

[0011] To achieve the above-mentioned objectives, this invention provides a full-link radiometric calibration method for L-band high-orbit SAR based on satellite-ground coordination, comprising the following steps:

[0012] Step S1: Based on the echo data and RCS information between the micro / nano satellite equipped with an active transponder and the high-orbit SAR satellite, obtain the hardware gain function that varies with the spatial angle. ;

[0013] Step S2: Using a pre-trained deep learning model, obtain the software gain that varies with the operating parameters. ;

[0014] Step S3: Utilize the hardware gain function and software gain The obtained end-link gain model High-precision radiometric calibration is performed on the input high-orbit SAR image to be calibrated, and the absolute radar cross-section σ of the target is retrieved.

[0015] According to one technical solution of the present invention, step S1 specifically includes:

[0016] Step S101: Deploy at least one microsatellite operating in low Earth orbit, wherein the microsatellite is equipped with an active transponder with dynamic RCS value adjustment function;

[0017] Step S102: Based on high-precision spatiotemporal reference and precise orbit and attitude determination technology, adjust the attitude of the high-orbit SAR satellite and the micro-nano satellite so that the beam center of the active transponder is aligned with the beam center of the high-orbit SAR satellite, and establish and maintain a stable inter-satellite calibration signal link.

[0018] Step S103: After the inter-satellite calibration signal link is established, control the active transponder to forward at least three calibration signals with known and progressive RCS information at multiple spatial sampling positions within the main lobe coverage area of ​​the high-orbit SAR satellite in a time-division manner.

[0019] Step S104: The high-orbit SAR satellite receives and records the echo data of the calibration signal, and simultaneously records its own orbit and attitude data;

[0020] Step S105: Download the echo data, orbit and attitude data, and RCS information to the ground station;

[0021] Step S106: The ground station processes echo data, orbit and attitude data, and RCS information. Linear regression analysis is performed on multiple sets of RCS-received power data for each spatial sampling point to calculate the discrete hardware gain value. And a continuous hardware gain function is obtained through curve fitting. .

[0022] According to one technical solution of the present invention, in step S101, the micro-nano satellite operates at an orbital altitude of 500km to 1000km, and adopts a polar orbit or a sun-synchronous orbit; the dynamic RCS value of the active transponder is adjusted within the range of 60dBsm to 90dBsm.

[0023] According to one technical solution of the present invention, in step S103, the RCS values ​​corresponding to the at least three known and progressively calibrated signals are increased in a logarithmic scale.

[0024] In step S106, the model used for curve fitting is established based on the physical characteristics of the antenna pattern.

[0025] According to one technical solution of the present invention, in step S2, the training process of the deep learning model includes:

[0026] Step S201: Construct a system simulation model for a high-orbit SAR satellite. The input to the system simulation model is a combination of operating parameters. The combination of operating parameters Including at least one of the following: imaging mode, signal bandwidth, pulse duration, and weighting function;

[0027] Step S202: Generate multiple different combinations of working parameters within the parameter space. For each set of working parameter combinations Simulate the echo signal of a point target with a known RCS value. ;

[0028] Step S203: The simulated echo signal Input into the ground processing software, according to the combination of working parameters. The data is processed, and the image power value of the target point is measured in the output simulation image. ;

[0029] Step S204, according to the formula Calculate the corresponding combination of working parameters The true value of software gain This generates a combination of multiple sets of working parameters. , The training dataset consists of}

[0030] Step S205: Using the training dataset, train a deep neural network regression model, with the input being a combination of working parameters. Output bit software gain .

[0031] According to one technical solution of the present invention, the pre-trained deep learning model is deployed for use in performing actual imaging tasks by high-orbit SAR satellites. Real-time prediction of corresponding software processing gain .

[0032] According to a technical solution of the present invention, in step S201, the combination of working parameters It also includes point target quality evaluation parameters, which include at least one of integral sidelobe ratio, peak sidelobe ratio and 3dB resolution;

[0033] The point target quality evaluation parameters are extracted from the output of the simulation image and used as input features of the deep learning model.

[0034] According to one technical solution of the present invention, in step S205, the deep learning model includes an input layer, at least one hidden layer, and an output layer;

[0035] The number of nodes and the combination of working parameters in the input layer The dimensions correspond;

[0036] The output layer includes a node for outputting the predicted software gain value.

[0037] According to one technical solution of the present invention, step S3 specifically includes:

[0038] Step S301, according to the formula Construct a full-link gain model for the system;

[0039] Step S302: Obtain the actual combination of working parameters used for the high-orbit SAR image to be calibrated. And obtain the off-axis angle corresponding to the target pixel in the image. ;

[0040] Step S303: Combine according to actual working parameters and off-axis angle From the full-link gain model Determine the corresponding system gain value. ;

[0041] Step S304: Measure the pixel power value of the target in the high-orbit SAR image to be calibrated. ;

[0042] Step S305: Calculate the absolute radar cross-section of the target. , represented as:

[0043] .

[0044] According to one technical solution of the present invention, the full-link radiometric calibration of high-orbit SAR satellites achieved by the method described above has a calibration error of less than 1 dB.

[0045] Compared with existing technologies, the L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground cooperation provided by this invention has the following significant technical advantages:

[0046] The present invention provides a full-link radiometric calibration method for L-band high-orbit SAR based on space-ground collaboration. It utilizes low-orbit auxiliary satellites as dynamic calibration reference targets and can directly measure the hardware gain of high-orbit synthetic aperture radar antennas in space. This eliminates the dependence on fixed ground calibration fields and makes system calibration possible in traditional calibration blind areas such as the open sea and polar regions, achieving true global coverage calibration capability.

[0047] This invention decomposes the total system gain into spatial hardware gain. and ground software gain The system consists of two parts, which are calibrated independently through space-based measurements and ground-based modeling, respectively. This solves the problem of hardware and software gain coupling in traditional end-to-end calibration. After software algorithm updates or processing parameter adjustments, there is no need to perform expensive and time-consuming field calibration again. Only the software gain model needs to be updated, which greatly enhances the adaptability and flexibility of the calibration system.

[0048] This invention, through the dynamic establishment and measurement of inter-satellite links, can complete a spatial sampling measurement of the main lobe pattern of a high-orbit synthetic aperture radar antenna within tens of minutes, reducing the calibration time, which traditionally relies on revisit cycles, from several days or even months to minutes. This meets the high-frequency, routine calibration requirements of high-orbit synthetic aperture radar and is beneficial for real-time monitoring of system performance changes.

[0049] The space-based inter-satellite measurement link of this invention is basically located outside the atmosphere, and the influence of the atmosphere in the signal transmission path is negligible. This avoids the adverse effects of the lower atmosphere on measurement accuracy, and can more directly and purely reflect the true characteristics of the antenna hardware, thereby improving the accuracy and reliability of hardware gain measurement.

[0050] This invention employs sophisticated techniques such as active transponders, multi-RCS linear regression, curve fitting based on physical characteristics, and deep learning prediction, and separates hardware and software error sources. This enables the overall error of full-link radiometric calibration of high-orbit synthetic aperture radar to be controlled at a low level, improving the calibration accuracy to 1dB, and providing a solid data foundation for high-quality quantitative remote sensing applications. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the 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 creative effort.

[0052] Figure 1 The schematic diagram illustrates a process flow of an L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to an embodiment of the present invention.

[0053] Figure 2 The diagram illustrates a detailed flowchart of a satellite-ground coordinated L-band high-orbit SAR full-link radiometric calibration method according to an embodiment of the present invention.

[0054] Figure 3This diagram illustrates the overall architecture of end-to-end calibration according to one embodiment of the present invention.

[0055] Figure 4 A schematic diagram illustrating the attitude control of a low-Earth orbit satellite according to an embodiment of the present invention;

[0056] Figure 5 This diagram illustrates a space hardware calibration according to one embodiment of the present invention.

[0057] Figure 6 This diagram illustrates the received power and the RCS of the relay signal according to one embodiment of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of this application can be combined with each other. The following embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the patent application. It should be pointed out that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the protection scope of this application.

[0060] This invention provides a full-link radiometric calibration method for L-band high-orbit SAR based on satellite-ground coordination, such as... Figure 3 As shown, the space-based hardware calibration of this invention includes a low-Earth orbit (LEO) micro-nano calibration satellite, a high-Earth orbit (HEO) L-band SAR satellite, and a GNSS spatiotemporal synchronization and orbit determination / attitude determination system. The LEO micro-nano calibration satellite is deployed at an orbital altitude of 500 to 1000 km, using polar or sun-synchronous orbits with an inclination of 85° to 90°. The HEO L-band SAR satellite operates in geosynchronous orbit and performs radar imaging via the L-band. The ground control station is responsible for coordinating and controlling the calibration process and processing data. The GNSS system provides a time synchronization reference for the entire calibration process, with a time synchronization accuracy of less than 1 μs, ensuring that the beam center of the active transponder is aligned with the beam center of the HEO SAR antenna during the orbit determination / attitude determination process.

[0061] The low-Earth orbit (LEO) micro-nano calibration satellite carries an active transponder, a communication module, an attitude control system, and a GNSS time synchronization unit. The active transponder has an adjustable RCS output capability, with a dynamic adjustment range of 60 dBsm to 90 dBsm, an accuracy of 0.1 dB, and a response time of less than 1 ms. The communication module uses a Ka-band communication link, operating in the 26 to 40 GHz frequency band, supporting direct communication between the LEO satellite and ground stations. The attitude control system is equipped with an attitude sensor to ensure that the transponder beam is always aligned with the center of the main lobe beam of the high-Earth orbit radar satellite.

[0062] like Figures 1 to 6 As shown, the present invention provides a full-link radiometric calibration method for L-band high-orbit SAR based on satellite-ground cooperation, comprising the following steps:

[0063] Step S1: Based on the echo data and RCS information between the micro / nano satellite equipped with an active transponder and the high-orbit SAR satellite, obtain the hardware gain function that varies with the spatial angle. ;

[0064] Step S1 is the process of acquiring space-based hardware gain. By deploying low-Earth orbit micro / nano satellites equipped with active transponders, and while flying over the high-Earth orbit SAR field of view, an inter-satellite calibration link is established using high-precision spatiotemporal references and precise orbit and attitude determination technology; such as... Figure 5 As shown, a known RCS signal is forwarded at multiple spatial locations within the main lobe beam range by an active transponder, and spatial sampling of the main lobe pattern is obtained by combining this with the high-speed motion of the low-orbit satellite. Based on inter-satellite measurement data, linear regression and curve fitting methods are used to directly obtain the hardware gain of the main lobe pattern of the high-orbit SAR antenna. Complete on-orbit calibration of space hardware;

[0065] The high-orbit L-band SAR satellite operates in geosynchronous orbit and performs radar imaging via the L-band.

[0066] The low-Earth orbit (LEO) micro / nano calibration satellite employs a polar or sun-synchronous orbit (inclination 85°–90°) and carries an active transponder, a communication module, an attitude control system, and a GNSS time synchronization unit. The active transponder has adjustable RCS output capability, relaying three signals with known and progressively increasing radar cross-sections (RCS). The communication module uses a Ka-band communication link, supporting direct communication between the LEO satellite and ground stations. The attitude control system is equipped with an attitude sensor to ensure the transponder beam is always aligned with the center of the main lobe beam of the high-Earth orbit radar satellite. The GNSS system provides a time synchronization reference for the entire calibration process, with a time synchronization accuracy of less than 1 μs.

[0067] The inter-satellite measurement data (including RCS settings, received power, and spatiotemporal reference information) is transmitted to the ground station via a Ka-band communication link. The ground station performs linear regression analysis on multiple sets of RCS-power data from each space sampling point to calculate discrete hardware gain values. By curve fitting based on the physical characteristics of the antenna pattern, the hardware gain function of the main lobe pattern of a continuous high-orbit SAR antenna is obtained. .

[0068] Step S2: Using a pre-trained deep learning model, obtain the software gain that varies with the operating parameters. ;

[0069] Step S2 is the process of acquiring ground software gain. It quantifies and predicts the signal processing gain, i.e., the software gain, introduced by the high-orbit SAR ground processing system due to different operating parameter settings. .

[0070] A simulation dataset covering combinations of operating parameters for high-orbit SAR satellites was constructed. For each set of parameters, the echo of an ideal point target was simulated and processed using ground-based processing software. By comparing the output power with the input ideal RCS value, the true value of the software gain under the corresponding combination of operating parameters was calculated. A large number of such datasets were then processed. , The data pairs constitute the training dataset; a deep neural network is used to train a regression model, establishing a mapping relationship from operating parameters to software gain, thereby realizing software gain based on actual operating conditions. Real-time prediction to complete gain calibration for ground processing;

[0071] The combination of high-orbit SAR working parameters The simulation dataset includes parameters such as imaging mode, signal bandwidth, pulse duration, and weighting function. For each set of parameters... Simulation of known RCS ( Ideal point target echo The output power is obtained by inputting it into the ground processing software. Calculate the true value of software gain ;

[0072] The neural network consists of one input layer, three hidden layers, and one output layer. The input layer has d nodes, corresponding to... The components are arranged in three fully connected layers with 128, 64, and 32 nodes respectively, using ReLU activation. The output layer has one node. The loss function used is the mean squared error loss function.

[0073] Step S3: Utilize the hardware gain function and software gain The obtained end-link gain model High-precision radiometric calibration is performed on the input high-orbit SAR image to be calibrated, and the absolute radar cross-section σ of the target is retrieved.

[0074] Based on the space-based hardware gain obtained in step S1 and the ground-based software gain obtained in step S2, high-precision radiometric calibration is performed on the actually acquired high-orbit SAR images. The space-based hardware gain... Software gain compared to ground prediction Multiply by each product to construct an end-link gain model that covers the total system gain across the entire signal link. When it is necessary to calibrate a high-orbit SAR image, the actual working conditions of the high-orbit SAR should be considered. The system gain value is determined by the off-axis angle θ of each target pixel in the image relative to the beam center, and the true radar cross section of the target is inverted using the radiation calibration equation. Complete the full-link radiation calibration and application.

[0075] In some embodiments of the present invention, such as Figure 2 As shown, step S1 specifically includes:

[0076] Step S101: Deploy at least one microsatellite operating in low Earth orbit, wherein the microsatellite is equipped with an active transponder with dynamic RCS value adjustment function;

[0077] In step S101, the microsatellite operates at an orbital altitude of 500km to 1000km, using a polar orbit or a sun-synchronous orbit; the dynamic RCS value of the active transponder is adjusted within the range of 60dBsm to 90dBsm.

[0078] Employing a low-Earth orbit micro-nano satellite platform offers advantages such as low cost, flexible launch, and mass deployment. The onboard active transponder generates a known and precisely controllable RCS signal, providing a foundation for subsequent quantitative measurements and linear regression analysis. Compared to passive reflectors, it offers a higher signal-to-noise ratio and more reliable measurements.

[0079] Step S102: When the micro-nano satellite enters the potential line of sight of the high-orbit SAR satellite, the calibration process is initiated. Based on high-precision spatiotemporal reference and precise orbit and attitude determination technology, the attitudes of the high-orbit SAR satellite and the micro-nano satellite are adjusted so that the beam center of the active transponder is aligned with the beam center of the high-orbit SAR satellite, and a stable inter-satellite calibration signal link is established and maintained.

[0080] High-precision attitude control and beam alignment ensured the effective transmission and reception of calibration signals, maximizing the utilization of the main lobe gain of the high-orbit SAR antenna and improving the strength and quality of the measurement signal.

[0081] Step S103: After the inter-satellite calibration signal link is established, the active transponder is controlled to forward at least three calibration signals with known and progressive RCS information at multiple spatial sampling positions within the main lobe coverage area of ​​the high-orbit SAR satellite in a time-division manner.

[0082] In step S103, the RCS values ​​corresponding to the at least three known and progressively scaled calibration signals are incremented on a logarithmic scale (e.g., 70dBsm, 80dBsm, 90dBsm).

[0083] The high-speed motion of low-Earth orbit satellites naturally enables spatial sampling of the radiation pattern of high-Earth orbit SAR antennas in the angular dimension (θ). Multiple incremental RCS values ​​are measured at each sampling point to perform linear regression analysis on the measurement data. By verifying whether a good linear relationship is maintained between the received power and the known RCS, the reliability of the measurement data at that point can be verified. Furthermore, the slope of the regression line is used to more accurately calculate the hardware gain value at that angle, improving the robustness and accuracy of single-point measurements.

[0084] Step S104: The high-orbit SAR satellite receives and records the echo data of the calibration signal, and simultaneously records its own orbit and attitude data;

[0085] Step S105: Download the echo data, orbit and attitude data, and RCS information to the ground station;

[0086] Step S106: The ground station processes echo data, orbit and attitude data, and RCS information. Linear regression analysis is performed on multiple sets of RCS-received power data for each spatial sampling point to calculate the discrete hardware gain value. And a continuous hardware gain function is obtained through curve fitting. .

[0087] In step S106, the model used for curve fitting is established based on the physical characteristics of the antenna pattern.

[0088] Step S1 consists of two main parts: attitude adjustment and forwarding calibration signals.

[0089] When a low-Earth orbit (LEO) micro / nano satellite enters the field of view window of a high-Earth orbit (HEO) SAR satellite, the LEO micro / nano satellite adjusts its attitude based on GNSS positioning information, pointing its active transponder antenna towards the HEO SAR satellite. Simultaneously, the HEO SAR satellite adjusts its antenna beam pointing to ensure the LEO micro / nano satellite is within the center of its antenna main lobe. After attitude alignment is complete, the active transponder receives the HEO SAR signal and estimates the received power, then forwards the signal with a linearly varying RCS value. For example... Figure 6As shown, for signals received at different times, the power of the received signal is first estimated, and then the RCS value of the forwarded signal is dynamically set. The solid bars represent the estimated received power value, and the dashed bars represent the RCS value of the forwarded signal.

[0090] The receiving antenna receives L-band signals from the high-orbit SAR. A power detector performs real-time power estimation on the filtered signal, achieving a measurement accuracy of 0.1 dB. The digital signal processor runs an RCS control algorithm, dynamically adjusting the RCS through a variable gain amplifier to generate three logarithmically increasing RCS values. This ensures that the SNR of the relayed calibration signal is at least 20 dB higher than the background clutter, while avoiding SAR receiver saturation. The dynamically adjusted RCS signal is then relayed via a time-division multiplexing method using the transmitting antenna.

[0091] Linear regression processing eliminates random errors from single measurements and improves the reliability of gain values ​​at each angle sampling point. Curve fitting based on the physical model not only yields a continuous function that facilitates subsequent calculations but also suppresses measurement noise to some extent and ensures that the fitted radiation pattern shape conforms to the basic physical laws of the antenna, making the results more reasonable and accurate.

[0092] In some embodiments of the present invention, such as Figure 2 As shown, in step S2, the training process of the deep learning model includes:

[0093] Step S201: Construct a system simulation model for a high-orbit SAR satellite. The input to the system simulation model is a combination of operating parameters. The combination of operating parameters Including at least one of the following: imaging mode, signal bandwidth, pulse duration, and weighting function;

[0094] First, it is necessary to define the key operating parameters that affect the gain of the ground processing software and combine them into a parameter vector to obtain the operating parameter combination. Then, a system simulation model of a high-orbit SAR satellite is constructed, and the input of this model is the combination of these operating parameters. .

[0095] For example, define a working parameter vector:

[0096]

[0097] Where BW represents bandwidth. For pulse width, For carrier frequency, For repetition frequency, and These are the azimuth / range weighting coefficients. For scanning angle, This is an imaging method. Hypercubic sampling is used to generate N sets of parameter combinations. .

[0098] Step S202: Generate multiple different combinations of working parameters within the parameter space. For each set of working parameter combinations Simulate a system with a known RCS value (set value, such as...) Simulated echo signal of point target ;

[0099] Step S203: The simulated echo signal Input into the ground processing software, according to the combination of working parameters. The data is processed, and the image power value of the target point is measured in the output simulation image. ;

[0100] Experimental design methods are used to generate a large number (e.g., 10,000) of different combinations of working parameters. For each combination of working parameters... Using the simulation model established in step S201, the echo signal of an ideal point target is simulated and generated. The original echo data from this simulation The data is input into the actual high-orbit SAR ground processing software and processed strictly according to the working parameter combinations. The settings are processed to obtain a simulated SAR image. Simulated point targets are located, and their image power values ​​are accurately measured after imaging. .

[0101] Step S204, according to the formula Calculate the corresponding combination of working parameters The true value of software gain This generates a combination of multiple sets of working parameters. , The training dataset consists of}

[0102] Step S205: Using the training dataset, train a deep neural network regression model. Where d is the parameter dimension, and the input is the combination of working parameters. Output bit software gain .

[0103] Construct a deep neural network regression model, with the following combination of input layer nodes and operating parameters: The dimension d is matched. The network contains at least one hidden layer, for example, a structure with three fully connected layers, with the number of neurons being 128, 64, and 32 respectively. The activation function used is ReLU (rectified linear function) to introduce non-linear fitting capability. The output layer is a single neuron that directly outputs the predicted software gain. The training dataset generated in step S204 is input into the network, and the mean squared error (MSE) is used as the loss function. The network is trained using a backpropagation algorithm (such as the Adam optimizer) until the model converges. The training objective is to teach the network a complex mapping function F from the operating parameters Ω to the software gain G2.

[0104] The loss function uses mean squared error:

[0105]

[0106] Where N is the number of training samples, It is a network prediction value. It is the actual value.

[0107] In some embodiments of the present invention, see again Figure 2 After the trained deep learning model is deployed, when radiometric calibration of an actual high-orbit SAR image is required, the deep learning model will use the methods employed by the high-orbit SAR satellite when performing the actual imaging task. Real-time prediction of corresponding software processing gain .

[0108] In some embodiments of the present invention, in step S201, the combination of operating parameters... It also includes point target quality evaluation parameters, which include at least one of integral sidelobe ratio, peak sidelobe ratio and 3dB resolution;

[0109] The point target quality evaluation parameters are extracted from the output of the simulation image and used as input features of the deep learning model.

[0110] To improve the accuracy and robustness of the prediction model, the working parameter combination Ω can also include point target quality assessment parameters extracted from the simulation output image, such as integral sidelobe ratio (ISLR), peak sidelobe ratio (PSLR), and 3dB resolution. These parameters reflect the quality of the imaging processing and are also indirectly related to the actual gain effect of the software. Providing them as additional input features to the deep learning model helps the model to more comprehensively understand the impact of processing parameters on the final result, thereby making more accurate predictions.

[0111] In some embodiments of the present invention, step S3 specifically includes:

[0112] Step S301, according to the formula Construct the system's end-to-end gain model, where Calibrate the gain for space-based hardware. Calibrate the gain for ground processing software;

[0113] Step S302: Obtain the actual combination of working parameters used for the high-orbit SAR image to be calibrated. and obtain the target pixels in the image. corresponding off-axis angle ;

[0114] Step S303: Combine according to actual working parameters and off-axis angle From the full-link gain model Determine the corresponding system gain value. ;

[0115] Step S304: Measure the pixel power value of the target in the high-orbit SAR image to be calibrated. ;

[0116] Step S305: Calculate the absolute radar cross-section of the target. , represented as:

[0117] .

[0118] According to one technical solution of the present invention, the full-link radiometric calibration of high-orbit SAR satellites achieved by the method described above has a calibration error of less than 1 dB.

[0119] 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 satellite-ground coordinated L-band high-orbit SAR full-link radiometric calibration method as described in any of the above technical solutions.

[0120] 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.

[0121] 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.

[0122] 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 satellite-ground coordinated L-band high-orbit SAR full-link radiometric calibration method as described in any of the above technical solutions.

[0123] 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.

[0124] This invention provides a satellite-ground collaborative L-band high-orbit SAR full-link radiometric calibration method. It involves deploying a low-orbit micro / nano satellite equipped with an active transponder to measure the hardware gain of the high-orbit SAR antenna's main lobe pattern in orbit, and then processing the data on the ground to obtain its hardware gain function. A simulation dataset covering high-orbit SAR operating conditions was constructed. A deep neural network was used to obtain the mapping relationship between operating parameters and software processing gain, and the software processing gain was predicted in real time based on the operating conditions. Multiplying the hardware gain by the software gain yields the total link gain. Substitute into the radiation calibration equation to invert the true radar cross section of the target. This invention achieves full-link radiometric calibration of the main lobe pattern. Compared with traditional ground-based calibration methods, this invention overcomes the bottlenecks of difficult calibration in distant sea areas, significant environmental interference, and long calibration cycles. It enables minute-level calibration of the main lobe pattern of high-orbit SAR, decouples errors between space hardware and ground processing software, and improves calibration accuracy to 1 dB. This provides low-cost, high-efficiency, and high-precision calibration support for high-quality distant imaging and quantitative remote sensing applications of high-orbit L-band SAR.

[0125] The above description is merely one embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A full-link radiometric calibration method for L-band high-orbit SAR based on satellite-ground coordination, characterized in that, Includes the following steps: Step S1: Based on the echo data and RCS information between the micro / nano satellite equipped with an active transponder and the high-orbit SAR satellite, obtain the hardware gain function that varies with the spatial angle. ; Step S2: Using a pre-trained deep learning model, obtain the software gain that varies with the operating parameters. ; Step S3: Utilize the hardware gain function and software gain The obtained end-link gain model High-precision radiometric calibration is performed on the input high-orbit SAR image to be calibrated, and the absolute radar cross-section σ of the target is retrieved.

2. The L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to claim 1, characterized in that, Step S1 specifically includes: Step S101: Deploy at least one microsatellite operating in low Earth orbit, wherein the microsatellite is equipped with an active transponder with dynamic RCS value adjustment function; Step S102: Based on high-precision spatiotemporal reference and precise orbit and attitude determination technology, adjust the attitude of the high-orbit SAR satellite and the micro-nano satellite so that the beam center of the active transponder is aligned with the beam center of the high-orbit SAR satellite, and establish and maintain a stable inter-satellite calibration signal link. Step S103: After the inter-satellite calibration signal link is established, control the active transponder to forward at least three calibration signals with known and progressive RCS information at multiple spatial sampling positions within the main lobe coverage area of ​​the high-orbit SAR satellite in a time-division manner. Step S104: The high-orbit SAR satellite receives and records the echo data of the calibration signal, and simultaneously records its own orbit and attitude data; Step S105: Download the echo data, orbit and attitude data, and RCS information to the ground station; Step S106: The ground station processes echo data, orbit and attitude data, and RCS information. Linear regression analysis is performed on multiple sets of RCS-received power data for each spatial sampling point to calculate the discrete hardware gain value. And a continuous hardware gain function is obtained through curve fitting. .

3. The L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to claim 2, characterized in that, In step S101, the microsatellite operates at an orbital altitude of 500km to 1000km, using a polar orbit or a sun-synchronous orbit; the dynamic RCS value of the active transponder is adjusted within the range of 60dBsm to 90dBsm.

4. The L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to claim 2, characterized in that, In step S103, the RCS values ​​corresponding to the at least three known and progressively scaled calibration signals are increased using a logarithmic scale. In step S106, the model used for curve fitting is established based on the physical characteristics of the antenna pattern.

5. The L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to claim 1, characterized in that, In step S2, the training process of the deep learning model includes: Step S201: Construct a system simulation model for a high-orbit SAR satellite. The input to the system simulation model is a combination of operating parameters. The combination of operating parameters Including at least one of the following: imaging mode, signal bandwidth, pulse duration, and weighting function; Step S202: Generate multiple different combinations of working parameters within the parameter space. For each set of working parameter combinations Simulate the echo signal of a point target with a known RCS value. ; Step S203: The simulated echo signal Input into the ground processing software, according to the combination of working parameters. The data is processed, and the image power value of the target point is measured in the output simulation image. ; Step S204, according to the formula Calculate the corresponding combination of working parameters The true value of software gain This generates a combination of multiple sets of working parameters. , The training dataset consists of} Step S205: Using the training dataset, train a deep neural network regression model, with the input being a combination of working parameters. The output is software gain. .

6. The L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to claim 5, characterized in that, The pre-trained deep learning model is deployed for use in performing actual imaging tasks based on the data from high-orbit SAR satellites. Real-time prediction of corresponding software processing gain .

7. The L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to claim 5, characterized in that, In step S201, the combination of operating parameters It also includes point target quality evaluation parameters, which include at least one of integral sidelobe ratio, peak sidelobe ratio and 3dB resolution; The point target quality evaluation parameters are extracted from the output of the simulation image and used as input features of the deep learning model.

8. The L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to claim 5, characterized in that, In step S205, the deep learning model includes an input layer, at least one hidden layer, and an output layer; The number of nodes and the combination of working parameters in the input layer The dimensions correspond; The output layer includes a node for outputting the predicted software gain value.

9. The L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to claim 1, characterized in that, Step S3 specifically includes: Step S301, according to the formula Construct a full-link gain model for the system; Step S302: Obtain the actual combination of working parameters used for the high-orbit SAR image to be calibrated. And obtain the off-axis angle corresponding to the target pixel in the image. ; Step S303: Combine according to actual working parameters and off-axis angle From the full-link gain model Determine the corresponding system gain value. ; Step S304: Measure the pixel power value of the target in the high-orbit SAR image to be calibrated. ; Step S305: Calculate the absolute radar cross-section of the target. , is represented as: 。 10. The L-band high-orbit SAR full-link radiometric calibration method based on satellite-ground coordination according to claim 1, characterized in that, The calibration error is less than 1 dB.

Images