Synthetic aperture radiometer error correction method, system, terminal device and medium

Abstract

The application provides a kind of comprehensive aperture radiometer error correction method, system, terminal equipment and medium, including obtaining the original brightness temperature image of target near-field imaging scene, obtaining the initial near-field visibility function corresponding;Initial near-field visibility function is input into the near-field error compensation network model constructed in advance, extract frequency domain feature information, compensate near-field error, obtain new near-field visibility function;Calculate the error of new near-field visibility function and target far-field visibility function, if the error is less than or equal to the preset error threshold, then Fourier transform inversion is carried out on new near-field visibility function, to obtain the corrected reconstructed brightness temperature image;Otherwise, according to the error, construct loss function, carry out back propagation on near-field error compensation network model, obtain trained near-field error compensation network model, and based on the trained near-field error compensation network model, obtain the corrected reconstructed brightness temperature image.The application can correct near-field error, improve the quality of reconstructed brightness temperature image.

Description

Technical Field

[0001] This invention belongs to the field of microwave remote sensing and detection technology, specifically relating to a method, system, terminal equipment, and medium for error correction of an integrated aperture radiometer. Background Technology

[0002] In recent years, applications such as security imaging of concealed dangerous goods, close-range target imaging under adverse visual conditions, and detection of buried objects have created an urgent need for high-resolution passive microwave near-field imaging. Therefore, research on near-field imaging technology is of great significance. Under near-field conditions, the target signal received by the synthetic aperture radiometer appears as a spherical wave, not a plane wave. This difference means that the visibility function of the system and the target brightness temperature distribution cannot be described by a simple Fourier transform relationship, thus causing overlapping and blurring problems in the near-field reconstructed brightness temperature image. Summary of the Invention

[0003] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a comprehensive aperture radiometer error correction method, system, terminal equipment, and medium. Its purpose is to correct near-field errors and improve the quality of reconstructed brightness temperature images. This solves the technical problem in existing technologies where near-field errors lead to overlapping and blurring of reconstructed brightness temperature images.

[0004] In a first aspect, the present invention provides a method for correcting errors in a synthetic aperture radiometer, the method comprising the following steps:

[0005] Acquire the original brightness temperature image of the target near-field imaging scene, and obtain the initial near-field visibility function corresponding to the target near-field imaging scene based on the original brightness temperature image;

[0006] The initial near-field visibility function is input into a pre-constructed near-field error compensation network model to obtain a new near-field visibility function. The near-field error compensation network model is used to extract the frequency domain feature information of the initial near-field visibility function and to compensate for the near-field error of the initial near-field visibility function based on the frequency domain feature information. The near-field error includes phase error and amplitude error.

[0007] The error between the new near-field visibility function and the preset target far-field visibility function is calculated. If the error is less than or equal to the preset error threshold, a Fourier transform inversion is performed on the new near-field visibility function to obtain the reconstructed brightness temperature image after the target near-field imaging scene correction. Otherwise, a loss function is constructed based on the error, and the near-field error compensation network model is backpropagated using the loss function until the error is less than or equal to the preset error threshold to obtain the trained near-field error compensation network model. The initial near-field visibility function is then input into the trained near-field error compensation network model. After performing a Fourier transform inversion on the new near-field visibility function output by the trained near-field error compensation network model, the corrected reconstructed brightness temperature image is obtained.

[0008] Optionally, the near-field error compensation network model includes an information extraction module, a first feature extraction module, a second feature extraction module, an anti-overfitting module, a feature recombination module, and a feature aggregation module. The information extraction module is used to extract the amplitude and phase information of the visibility function. The first feature extraction module is used to extract the shallow frequency domain features of the visibility function based on the amplitude and phase information. The shallow frequency domain features characterize the overall structure and distribution of the frequency domain data, mainly reflecting the changing trends of the near-field and far-field visibility functions in terms of global amplitude and phase. The second feature extraction module is used to extract the deep frequency domain features of the visibility function from the shallow frequency domain features. The deep frequency domain features characterize the local detail differences of the frequency domain data and depict the fine structure deviations and local distortions caused by the near-field effect. The anti-overfitting module is used to randomly deactivate some neurons in the deep frequency domain features to prevent model overfitting. The feature recombination module is used to reconstruct the deep frequency domain features output by the anti-overfitting module to obtain a multi-channel feature map. The feature aggregation module is used to perform average pooling on the multi-channel feature map to achieve near-field error compensation.

[0009] Optionally, both the first feature extraction module and the second feature extraction module consist of five sequentially connected residual convolutional blocks;

[0010] The shallow frequency domain features are represented as follows:

[0011] F1 = X5(F0) + conv(F0)

[0012] X i (F0)=ReLU(BN(conv2(ReLU(conv1(F0)))))+shortcut,i=1

[0013] X i (F0)=ReLU(BN(conv2(ReLU(conv1(X i-1 (F0))))))+shortcut,i=2,3,4,5

[0014] Where F1 represents shallow frequency domain features, F0 represents the amplitude and phase information of the visibility function output by the information extraction module, F0 = Normalize(stack(Flatten(Re(X)),Flatten(Im(X)))), Re(X) represents the amplitude information extracted from the real part of the visibility function X, Im(X) represents the phase information extracted from the imaginary part of the visibility function X, Flatten(·) represents the flattening operation, stack(·) represents the concatenation operation, Normalize(·) represents the normalization operation, and conv(F0) represents the convolution operation on F0 to extract the original features of F0. iconv1 represents the output of the i-th residual convolutional block, conv1 represents the first convolutional layer in the residual convolutional block, ReLU(·) represents the activation function, conv2 represents the second convolutional layer in the residual convolutional block, BN(·) represents batch normalization, and shortcut represents skip connections.

[0015] The deep frequency domain features are represented as follows:

[0016] F2 = X5(F1) + conv(F1)

[0017] X i (F1)=ReLU(BN(conv2(ReLU(conv1(F1)))))+shortcut,i=1

[0018] X i (F1)=ReLU(BN(conv2(ReLU(conv1(X i-1 (F1))))))+shortcut,i=2,3,4,5

[0019] Here, F2 represents the deep frequency domain characteristics.

[0020] Optionally, the deep frequency domain features output by the anti-overfitting module are reconstructed to obtain a multi-channel feature map, including:

[0021] Through calculation formula

[0022] F4=conv4(LeakyReLU(conv3(F3)))

[0023] F3 = dropout(F2)

[0024] The multi-channel feature map F4 is obtained; conv3 represents the third convolutional layer in the feature reorganization module, LeakyReLU(·) represents the activation function, which is used to retain some gradient responses to negative values ​​and improve the network's ability to capture subtle features, conv4 represents the fourth convolutional layer in the feature reorganization module, and dropout(F2) represents the result of the anti-overfitting module randomly deactivating some neurons in F2.

[0025] Optionally, the error is the mean absolute error.

[0026] Optionally, the expression for the error loss function is:

[0027]

[0028] Where Loss represents the error loss value, n = 1, ..., N, N represents the total number of tiny pixels in the original brightness temperature image, and V label V represents the preset target far-field visibility function.out This represents the near-field visibility function.

[0029] Optionally, the raw brightness temperature image includes multiple tiny facets.

[0030] Secondly, the present invention provides an integrated aperture radiometer error correction system, comprising:

[0031] The data acquisition module is used to acquire the original brightness temperature image of the target near-field imaging scene, and to obtain the initial near-field visibility function corresponding to the target near-field imaging scene based on the original brightness temperature image;

[0032] The error compensation module is used to input the initial near-field visibility function into a pre-built near-field error compensation network model to obtain a new near-field visibility function. The near-field error compensation network model is used to extract the frequency domain feature information of the initial near-field visibility function and compensate for the near-field error of the initial near-field visibility function based on the frequency domain feature information. The near-field error includes phase error and amplitude error.

[0033] The reconstruction module calculates the error between the new near-field visibility function and the preset target far-field visibility function. If the error is less than or equal to the preset error threshold, a Fourier transform inversion is performed on the new near-field visibility function to obtain the reconstructed brightness temperature image after the target near-field imaging scene correction. Otherwise, a loss function is constructed based on the error, and the near-field error compensation network model is backpropagated using the loss function until the error is less than or equal to the preset error threshold, resulting in a trained near-field error compensation network model. The initial near-field visibility function is then input into the trained near-field error compensation network model, and a Fourier transform inversion is performed on the new near-field visibility function output by the trained near-field error compensation network model to obtain the corrected reconstructed brightness temperature image.

[0034] Thirdly, the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method.

[0035] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0036] The beneficial effects of this invention are:

[0037] The integrated aperture radiometer error correction method provided by this invention extracts the frequency domain feature information of the initial near-field visibility function and compensates for the near-field error of the initial near-field visibility function based on the frequency domain feature information, thereby correcting the near-field error and improving the imaging quality. A loss function is constructed based on the error, and the near-field error compensation network model is backpropagated using the loss function until the error is less than or equal to a preset error threshold. This not only achieves the mapping from the near-field visibility function to the target far-field visibility function in the frequency domain, but also allows for flexible adjustment of the compensation intensity based on the target far-field visibility function, further improving the imaging quality. Attached Figure Description

[0038] Figure 1 This is a schematic diagram illustrating the imaging principle of a synthetic aperture radiometer.

[0039] Figure 2 This is a flowchart of a method for correcting errors in an integrated aperture radiometer according to one embodiment of this application;

[0040] Figure 3 This is a model structure diagram of the near-field error compensation network model in one embodiment of this application;

[0041] Figure 4a This is a reconstructed image before near-field error correction in one embodiment of this application;

[0042] Figure 4b This is a reconstructed image corresponding to a preset target far-field visibility function in one embodiment of this application;

[0043] Figure 4c This is a reconstructed image corrected using a synthetic aperture radiometer error correction method in one embodiment of this application;

[0044] Figure 5a This is a real part spectrum image of the brightness temperature image in the u direction corresponding to different visibility functions in one embodiment of this application;

[0045] Figure 5b This is a real part spectrum image of the brightness temperature image in the v direction corresponding to different visibility functions in one embodiment of this application;

[0046] Figure 5c This is a spectral image of the imaginary part of the brightness temperature image in the u direction corresponding to different visibility functions in one embodiment of this application;

[0047] Figure 5d This is a spectral image of the imaginary part in the v direction of the brightness temperature image corresponding to different visibility functions in one embodiment of this application;

[0048] Figure 6aThis is an IFFT reconstructed image after correction using the integrated aperture radiometer error correction method in one embodiment of this application;

[0049] Figure 6b This is an IFFT reconstructed image after conventional method correction in one embodiment of this application;

[0050] Figure 7 This is a schematic diagram of the integrated aperture radiometer error correction system in one embodiment of this application;

[0051] Figure 8 This is a schematic diagram of the structure of a terminal device in one embodiment of this application. Detailed Implementation

[0052] To address the technical problem of near-field errors causing overlapping and blurring of near-field reconstructed brightness and temperature images in existing technologies, this invention discloses a method, system, terminal device, and medium for correcting errors in integrated aperture radiometers. This method extracts the frequency domain features of the initial near-field visibility function and compensates for near-field errors based on these features, thereby correcting near-field errors and improving imaging quality. A loss function is constructed based on the error, and this function is used to backpropagate the near-field error compensation network model until the error is less than or equal to a preset error threshold. This not only achieves the mapping from the near-field visibility function to the target far-field visibility function in the frequency domain but also allows for flexible adjustment of the compensation intensity based on the target far-field visibility function, further improving imaging quality.

[0053] To facilitate understanding, the imaging principle of the integrated aperture radiometer will be explained first.

[0054] Specifically, such as Figure 1 As shown, there is a radiation source S in the target scene plane, which is divided into multiple tiny surface elements ds. OXY is the antenna plane, located in the plane where z = 0. P1 and P2 are two receiving antennas, and D represents the distance between the two antennas P1 and P2. The distance from the target scene plane to the antenna plane is d, and the distances from the radiating surface element to the two antenna elements are R1 and R2, respectively. The distances from the antennas to the origin are d1 and d2, respectively, and the distance from the radiating surface element to the origin is R. S The zenith angle of the radiating element is θ, and the azimuth angle is...

[0055] Assume the coordinates of a small surface element ds in the Cartesian coordinate system are (x, y, z), and the two antennas are (x1, y1, 0) and (x2, y2, 0). The radiation field intensities received by the two antenna elements are represented as E1(R1, t) and E2(R2, t), respectively. The visibility function obtained after complex correlation processing of the two signals is:

[0056]

[0057] In the formula, T is the angle formed by the radiating surface element to each antenna element. r For system noise, For the Fringewashing function, D i F is the directivity coefficient of the antenna. n This is the field strength pattern of the antenna. The exponent represents the phase difference caused by the different distances of the incident electromagnetic wave to the two antenna elements forming the baseline. T B This represents the brightness temperature image of a scene in the spatial domain, and * represents the conjugate complex number.

[0058] For the entire integrated aperture measurement system, the visibility function expression obtained from the correlation output of the entire observation field plane is:

[0059]

[0060] The distance R from the antenna element to the point source i for:

[0061]

[0062] Under far-field conditions, i.e. R i Approximating it using a first-order Taylor series, i.e.:

[0063]

[0064] in again Therefore, the above formula can be expressed as:

[0065]

[0066] therefore

[0067]

[0068] Where D x =x j -x i D y =y j -y i Under far-field conditions make and Defined as the normalized difference between the positions and wavelengths of two antenna elements, and let Representing the direction cosine, substituting the above expression into the formula yields...

[0069]

[0070] Define the corrected brightness temperature as

[0071]

[0072] Ignoring the Fringe washing function and assuming all antenna elements are identical, ideally, the relationship between the visibility function and brightness temperature data can be expressed as:

[0073]

[0074] It can be observed that under far-field conditions, the visibility function and brightness temperature data have a Fourier transform relationship.

[0075] However, in near-field imaging scenarios (such as security imaging of concealed dangerous goods on the human body), the radiation waves from radiation sources located in the near-field region cannot be treated as plane waves when they reach the antenna, but rather as spherical waves. Therefore, the distance difference between the target source and different antennas cannot be ignored. Thus, the visibility function with near-field error is expressed as:

[0076]

[0077] The biggest difference between the near-field and far-field visibility functions is that the near-field visibility function has more features. This near-field error term. The existence of the near-field error term causes aberrations such as blurring and defocusing in the imaging results when directly using traditional synthetic aperture imaging methods. Moreover, the closer the target is to the system, the more severe the degradation effect.

[0078] To address the aforementioned problems, this invention provides a method for correcting errors in a comprehensive aperture radiometer, such as... Figure 2 As shown, the method includes the following steps:

[0079] Step 21: Obtain the original brightness temperature image of the target near-field imaging scene, and obtain the initial near-field visibility function corresponding to the target near-field imaging scene based on the original brightness temperature image.

[0080] It should be understood that a synthetic aperture radiometer transmits signals via electromagnetic waves of a range of frequencies (typically in the microwave or millimeter-wave band) and receives echoes from the target area. These signals include radiation reflected from the target object and radiation emitted from the target surface. By measuring the intensity and spectral characteristics of this radiation, the radiometer can deduce the temperature distribution of the target surface. The brightness temperature data acquired and processed by the radiometer can be presented using graphical methods. These images are typically displayed as pseudo-color or grayscale images, with each pixel in the brightness temperature image representing the surface temperature at a specific location.

[0081] The initial near-field visibility function corresponding to the near-field imaging scene of the target, obtained from the original brightness temperature image, has been described in detail in the previous section on imaging principles, and will not be repeated here.

[0082] Step 22: Input the initial near-field visibility function into the pre-built near-field error compensation network model to obtain the new near-field visibility function.

[0083] In embodiments of the present invention, a near-field error compensation network model is used to extract frequency domain feature information of the initial near-field visibility function, and to compensate for the near-field error of the initial near-field visibility function based on the frequency domain feature information. The aforementioned near-field error includes phase error and amplitude error.

[0084] The structure of the near-field error compensation network model is explained below.

[0085] Specifically, such as Figure 3 As shown, the near-field error compensation network model includes an information extraction module 31, a first feature extraction module 32, a second feature extraction module 33, an anti-overfitting module 34, a feature recombination module 35, and a feature aggregation module 36 connected in sequence.

[0086] Among them, the information extraction module 31 is used to extract the amplitude information and phase information of the visibility function.

[0087] Specifically, in the embodiments of the present invention, considering that the visibility function includes a real part and an imaginary part, with the real part containing amplitude information and the imaginary part containing phase information, directly inputting it into the near-field error compensation network model cannot effectively decompose the intrinsic characteristics of the visibility function. Therefore, before inputting the data into the network, flattening and normalization operations are performed on the data to obtain a 1*165 matrix. Then, the real and imaginary parts of the visibility function are extracted separately, and the data are combined into a 2*165 two-channel data using the stack function. This not only effectively decomposes the intrinsic characteristics of the visibility function but also optimizes the learning process of the network. This process is specifically represented as F0 = Normalize(stack(Flatten(Re(X)), Flatten(Im(X)))), where Re(X) represents the amplitude information extracted from the real part of the visibility function X, Im(X) represents the phase information extracted from the imaginary part of the visibility function X, Flatten(·) represents the flattening operation, stack(·) represents the stacking operation, and Normalize(·) represents the normalization operation.

[0088] The first feature extraction module 32 is used to extract shallow frequency domain features of the visibility function based on amplitude and phase information.

[0089] Specifically, in the embodiments of the present invention, both the first feature extraction module 32 and the second feature extraction module 33 consist of five sequentially connected residual convolutional blocks (residual convolutional block 3A). Figure 3 3A in the middle), residual convolution block 3B ( Figure 3 3B in the middle), residual convolution block 3C ( Figure 3 3C in the middle), residual convolution block 3D ( Figure 3 3D in the middle), residual convolution block 3E ( Figure 3 The system is composed of 3E layers, with each residual convolutional block consisting of two one-dimensional convolutions (conv1, conv2), residual connections, batch normalization (BN), and ReLU activation operations. Furthermore, two skip connections are used between modules. The first skip connection maps the amplitude and phase information of the visibility function output by the information extraction module 31 into multi-channel data, adding it to the output of the first feature extraction module 32, preserving the feature details of the input data and enhancing feature extraction. The second skip connection maps the amplitude and phase information of the visibility function output by the information extraction module 31 into more channels, adding it to the deep frequency domain features output by the second feature extraction module 33, achieving effective fusion of shallow and deep frequency domain features and improving feature representation capabilities.

[0090] Specifically, the shallow frequency domain features extracted by the first feature extraction module 32 are represented as follows:

[0091] F1 = X5(F0) + conv(F0)

[0092] X i (F0)=ReLU(BN(conv2(ReLU(conv1(F0)))))+shortcut,i=1

[0093] X i (F0)=ReLU(BN(conv2(ReLU(conv1(X i-1 (F0))))))+shortcut,i=2,3,4,5

[0094] Where F1 represents shallow frequency domain features, F0 represents the amplitude and phase information of the visibility function output by information extraction module 31, conv(F0) represents the convolution operation on F0 to extract the original features of F0, and X i conv1 represents the output of the i-th residual convolutional block, conv1 represents the first convolutional layer in the residual convolutional block, ReLU(·) represents the activation function, conv2 represents the second convolutional layer in the residual convolutional block, BN(·) represents batch normalization, and shortcut represents skip connections.

[0095] The second feature extraction module 33 is used to extract deep frequency domain features of the visibility function from shallow frequency domain features. The deep frequency domain features extracted by the second feature extraction module 33 are represented as follows:

[0096] F2 = X5(F1) + conv(F1)

[0097] Xi (F1)=ReLU(BN(conv2(ReLU(conv1(F1)))))+shortcut,i=1

[0098] X i (F1)=ReLU(BN(conv2(ReLU(conv1(X i-1 (F1))))))+shortcut,i=2,3,4,5

[0099] Here, F2 represents the deep frequency domain characteristics.

[0100] Anti-overfitting module 34 is used to randomly deactivate some neurons in deep frequency domain features to prevent model overfitting.

[0101] Specifically, the process of randomly deactivating some neurons in deep frequency domain features can be represented as F3 = dropout(F2). This module can improve the network's generalization ability.

[0102] The feature recombination module 35 is used to recombine the deep frequency domain features output by the anti-overfitting module 34 to obtain a multi-channel feature map.

[0103] Specifically, through the calculation formula

[0104] F4=conv4(LeakyReLU9conv3(F3)))

[0105] F3 = dropout(F2)

[0106] The multi-channel feature map F4 is obtained; conv3 represents the third convolutional layer in the feature reorganization module 35, which is used to reduce the number of channels, filter out key features in the multi-channel and reduce redundancy. LeakyReLU(·) represents the activation function, which is used to retain some gradient responses for negative values ​​and improve the network's ability to capture subtle features. conv4 represents the fourth convolutional layer in the feature reorganization module 35, which is used to recombine the dimensionality-reduced features. This combination can integrate the multi-scale features extracted by the residual module and provide rich feature information for subsequent global pooling and output estimation.

[0107] The feature aggregation module 36 is used to perform average pooling on the multi-channel feature map to achieve near-field error compensation.

[0108] Step 23: Calculate the error between the new near-field visibility function and the preset target far-field visibility function. If the error is less than or equal to the preset error threshold, perform a Fourier transform inversion on the new near-field visibility function to obtain the reconstructed brightness temperature image after the target near-field imaging scene correction. Otherwise, construct a loss function based on the error, and use the loss function to backpropagate the near-field error compensation network model until the error is less than or equal to the preset error threshold to obtain the trained near-field error compensation network model. Input the initial near-field visibility function into the trained near-field error compensation network model, and perform a Fourier transform inversion on the new near-field visibility function output by the trained near-field error compensation network model to obtain the corrected reconstructed brightness temperature image.

[0109] Specifically, in the embodiments of the present invention, the above-mentioned error is the mean absolute error.

[0110] The expression for the loss function constructed based on the error is:

[0111]

[0112] Where Loss represents the error loss value, n = 1, ..., N, N represents the total number of tiny pixels in the original brightness temperature image, and V label V represents the preset target far-field visibility function. out This represents the near-field visibility function.

[0113] In an embodiment of the present invention, the process of inputting the initial near-field visibility function into the trained near-field error compensation network model to obtain the corrected reconstructed brightness temperature image includes:

[0114] The initial near-field visibility function is input into the trained near-field error compensation network model to obtain the corrected new near-field visibility function;

[0115] The corrected near-field visibility function is inverted by Fourier transform to obtain the reconstructed brightness temperature image of the target near-field imaging scene after correction.

[0116] To verify the effectiveness of the present invention, the following comparative experiment was conducted.

[0117] Example 1

[0118] In this embodiment, the original brightness temperature images are selected from the NWPU-RESISC45 remote sensing image scene classification dataset. The NWPU-RESISC45 dataset is a large-scale public dataset for remote sensing image scene classification, which includes 45 scene categories such as airports, basketball courts, deserts, industrial areas, forests, and rivers. Each scene has 700 remote sensing images, totaling 31,500 images. The images are rich in information and cover a wide range of content.

[0119] After obtaining the original scene brightness temperature image dataset, the visibility function is obtained through an ideal uniform sampling synthetic aperture system simulation program. Specifically, each antenna pair receives the radiation signal from the original scene radiation source, and after complex correlation calculations and redundancy removal operations, visibility functions under near-field and far-field conditions are generated respectively. The generated simulation data is divided into training, validation, and test training sets according to a certain ratio. The antenna array used in the ideal uniform sampling synthetic aperture system simulation program is a matrix antenna array consisting of 24 antenna elements.

[0120] After determining the parameters of the simulation program, the steps to generate the simulation dataset are as follows:

[0121] 1) Select any original scene brightness temperature image (P) from the original scene brightness temperature image dataset and input P into the ideal integrated aperture radiometer simulation program;

[0122] 2) Set the distance between the scene plane and the antenna plane to 0.8m and 5m respectively, and generate the simulated synthetic aperture visibility function V(u,v) under ideal conditions;

[0123] 3) After processing each original scene image through steps 1 and 2 respectively, all the obtained samples are combined into a simulation dataset (D);

[0124] 4) Divide the simulation dataset D into training, validation, and test datasets according to a certain ratio.

[0125] The 0.8m visibility function serves as the network's input dataset, while the 5m visibility function serves as the network's label dataset. To remove irrelevant DC components and allow the network to focus more on meaningful frequency domain information, the values ​​at zero frequencies in the datasets were set to zero. After completing the test and obtaining the corrected visibility function, the values ​​at zero frequencies from the input datasets were extracted and filled into the zero-frequency region of the corrected visibility function.

[0126] Experimental results are as follows Figure 4a , Figure 4b , Figure 4c As shown, where, Figure 4a The image is the reconstructed image before near-field error correction. Figure 4b The reconstructed image corresponding to the preset target far-field visibility function, Figure 4c The image is a reconstructed image after correction using the integrated aperture radiometer error correction method provided in this application. Figures 4a to 4c It can be seen that the brightness temperature image obtained after near-field error correction ( Figure 4c Compared to the image before correction ( Figure 4aThe image obtained by directly inverting near-field frequency domain data is much clearer. The image obtained directly from near-field frequency domain data shows significant blurring and distortion. This is because, under near-field conditions, electromagnetic waves are spherical waves, and the intensity of the received electromagnetic waves varies at different locations. Furthermore, due to the close distance, interference and diffraction phenomena may occur, leading to distortion and discrepancies in the near-field brightness temperature image. Intuitively, the brightness temperature image corresponding to the corrected visibility function (…) Figure 4c ) and the image corresponding to the 5m visibility function used as the training label in the experiment ( Figure 4b The effect is similar to that of the previous method, and it even reduces background error to some extent. Therefore, the integrated aperture radiometer error correction method proposed in this paper has a better correction effect and yields a high-quality corresponding spatial domain image.

[0127] After qualitatively comparing the images obtained from the visibility function inversion before and after correction, in another embodiment of the present invention, RMSE, PSNR, and SSIM are used to quantitatively compare the quality of different images. Both RMSE and PSNR are calculated based on the original scene brightness temperature image corresponding to the data. RMSE is a comprehensive measure of the error between the inverted brightness temperature image and the original scene brightness temperature image; the smaller the RMSE, the closer the image is to the original scene. PSNR is also a comprehensive measure of the quality of the inverted brightness temperature image; the larger the value, the less distortion in the reconstructed brightness temperature image.

[0128] The formulas for calculating RMSE, PSNR, and SSIM are as follows:

[0129]

[0130]

[0131] Where M and N are the number of pixels in the reconstructed image in two directions, respectively, and T i These are the brightness temperature images obtained by inverting the visibility function in the near field, far field, and near field correction, respectively. T is the original scene brightness temperature image, and S is the dynamic range of the brightness temperature values ​​in the original scene image. and μ T They are T i The average value of T; and They are T i The variance of T; It is T i The covariance between T and T; c1 and c2 are constants used for the stabilizing formula.

[0132] The brightness temperature images obtained by inverting all near-field visibility functions, far-field visibility functions, and corrected visibility functions were compared with their corresponding original target scenes, and the results are shown in Table 1. The mean RMSE values ​​were 35.73K, 29.07K, and 28.20K, respectively; the mean PSNR values ​​were 12.34dB, 15.92dB, and 15.81dB, respectively; and the mean SSIM values ​​were 0.1628, 0.2331, and 0.2287, respectively.

[0133] Table 1

[0134]

[0135] As shown in Table 1, the brightness temperature image after near-field error correction exhibits a significant advantage in RMSE, indicating higher image quality and a substantial reduction in error. In terms of PSNR and SSIM, the corrected image is significantly superior to the uncorrected near-field image and closely approximates the far-field image. The improved PSNR indicates that the corrected image more closely resembles the real target scene in terms of overall brightness temperature distribution, while the improved SSIM suggests a certain improvement in structural similarity, with some restoration of edge and texture details. Furthermore, the fact that the corrected image's PSNR and SSIM are very close to those of the far-field image further demonstrates that the integrated aperture radiometer error correction method provided in this application can accurately learn and correct the differences between the near-field and far-field frequency domains.

[0136] To more intuitively evaluate the frequency domain correction effect of the MSR-NFECNet method (the integrated aperture radiometer error correction method provided in this application), in another embodiment of this application, the near-field, far-field, and corrected visibility functions are normalized to obtain a normalized spectrum image. Since the differences in the two-dimensional spectrum images are not significant, the maximum points of the two-dimensional spectrum images are extracted, and one-dimensional curves are truncated along the u and v directions. A locally magnified spectrum is then plotted in the region near zero frequency. By comparing the RMSE values ​​of the real and imaginary parts of these curves, the frequency domain error correction capability of the integrated aperture radiometer error correction method provided in this application can be further verified. Specifically, as follows... Figures 5a to 5d As shown. By Figures 5a to 5d It can be seen that there are significant differences between the near-field, far-field, and corrected spectra and the original image spectrum. In the imaginary part, the corrected spectrum is significantly closer to the far-field image and the original scene image spectrum in both the u and v directions. However, the difference may not be very obvious in the real part. In another embodiment of this application, RMSE was used to quantitatively compare the near-field, far-field, and corrected frequency domain spectra with the original scene spectrum. The average results are shown in Table 2.

[0137] Table 2

[0138] Real(u) 0.0689 0.0605 0.1132 Real(v) 0.0681 0.0621 0.1232 Imag(u) 0.2145 0.1894 0.1965 Imag(v) 0.2162 0.2091 0.2125

[0139] As shown in Table 2, the corrected spectrum exhibits significant optimization in the imaginary part. Compared to the near-field spectrum, the corrected imaginary RMSE decreases to 0.1965 and 0.2125 in the u and v directions, respectively, more closely resembling the trend of the far-field spectrum. This indicates that the correction method has a clear advantage in correcting errors in the imaginary spectrum. However, the corrected result does not outperform the near-field spectrum in terms of the real part RMSE; the real part RMSE in the u and v directions is larger than that of the near-field spectrum. Although the real part correction effect is not as expected, the overall imaging quality after correction is significantly better than before. The sharpness and detail recovery of targets in the reconstructed image are improved, indicating that the correction method successfully optimizes error propagation during the imaging process while performing spectral correction. The improvement in the imaginary spectrum may have a more significant impact on imaging quality, which to some extent compensates for the shortcomings of real part correction.

[0140] To comprehensively evaluate the model's generalization ability and performance stability, in another embodiment of this application, near-field data at different distances are used for testing and combination for training. By verifying the model's performance under different distance conditions, it is determined whether the synthetic aperture radiometer error correction method can adapt to previously unseen near-field error characteristics.

[0141] Specifically, different distance data were input for testing. Test set data with target scene and array distances of 0.5m, 1.5m, 2m, 3m and 4m were generated respectively and input into a network with a visibility function of 0.8m as the training set. The output results were then inverted through inverse Fourier transform. The inverted image was compared with the original scene image, and the average RMSE was calculated as shown in Table 3.

[0142] Table 3

[0143] RMSE 37.28 38.23 40.62 41.13 41.18 PSNR 12.7727 13.5469 13.2861 13.3480 13.3990 SSIM 0.1222 0.1447 0.1349 0.1452 0.1518

[0144] Table 3 shows that the closer the data is to 0.8m, the higher the quality of the inverted image. However, there is still room for improvement in the overall performance. Therefore, to further verify the network's adaptability to different distance visibility functions, we attempted to generate a training set using combinations of data at different distances.

[0145] Visibility functions at different distances were combined for training. A training set was generated by combining visibility functions at 0.8m and 1.5m for model training. During testing, data at 0.8m, 1.5m, and the combined training set (0.8-1.5m) were input respectively, and the results are shown in Table 4.

[0146] Table 4

[0147] RMSE 28.8881 28.8974 28.8928 PSNR 15.8061 15.8183 15.8122 SSIM 0.2293 0.2296 0.2295

[0148] As shown in Table 4, the test results were significantly improved after training with a combination of visibility functions at two different distances. This indicates that the integrated aperture radiometer error correction method provided in this application can learn a wider range and more diverse error characteristics from multi-distance data, thereby improving the ability to correct near-field errors. The data combination method effectively enhances the model's generalization ability, making it more robust to different scenarios.

[0149] To verify the contribution of each module in the near-field error compensation network model of this application to the performance improvement, an ablation experiment was also conducted in another embodiment of this application. The specific experimental results are shown in Table 5. In this ablation experiment, the second feature extraction module, skip links, anti-overfitting module, and feature aggregation module in the near-field error compensation network model were removed respectively, and the RMSE values ​​of the model on the test set under different combinations were recorded.

[0150] Table 5

[0151] Res-CNN × √ √ √ 28.67 Res-CNN √ × √ √ 28.82 Res-CNN √ √ × √ 29.21 Res-CNN √ √ √ × 29.01 Res-CNN √ √ √ √ 28.2

[0152] As shown in Table 5, when the second feature extraction module was removed, the RMSE value of the model increased to 28.82, indicating that the residual block helps improve the prediction accuracy of the model. When the skip links were removed, the RMSE value increased slightly to 28.67, indicating that the skip links are helpful for efficient information transfer and stable gradient updates. After removing the anti-overfitting module, the RMSE value further increased to 29.21, indicating that the anti-overfitting module plays an important role in reducing overfitting. When the feature aggregation module was removed, the RMSE value increased to 29.01, indicating that the feature aggregation module has a positive effect on extracting effective features and reducing computational complexity.

[0153] In summary, the complete near-field error compensation network model exhibits the lowest RMSE value of 28.2. This indicates that the synergistic effect of the network components significantly improves the model's performance, validating the rationality and effectiveness of the network structure design.

[0154] To verify the robustness of the proposed method and further evaluate the near-field error correction effect of the integrated aperture radiometer error correction method provided in this application under different error intensities, noise was added to the visibility functions in both the training and test sets based on an ideal simulation dataset. The added noise consisted of Gaussian noise with a mean of 0 and standard deviations of 0.05, 0.1, 0.15, 0.2, 0.25, and 0.3.

[0155] In a training scenario using noise-free data and testing with noisy data, Gaussian noise of varying intensities (ranging from 0.05 to 0.3, with mean 0 and standard deviations of 0.05, 0.1, 0.15, 0.2, 0.25, and 0.3) was added to the 0.8m visibility function in the test set. Under different noise intensities, the integrated aperture radiometer error correction method provided in this application was used to correct the noise-added data, and then the corresponding spatial domain image was obtained through inversion. Simultaneously, near-field noise data was also directly inverted, and finally, the images obtained by the two methods were compared with the original scene image. Two samples were randomly selected from the test dataset. Under different noise intensities, all data in the test set were processed using two different methods, and the average RMSE, PSNR, and SSIM of the reconstructed image compared to the original scene image were calculated. The results are shown in Tables 6(a) and 6(b). Table 6(a) shows the values ​​calculated after correcting the visibility function using the integrated aperture radiometer error correction method provided in this application, while Table 6(b) shows the corresponding index values ​​of the image obtained by directly inverting the uncorrected visibility function using inverse Fourier transform.

[0156] Table 6(a)

[0157] RMSE 28.5107 28.7079 29.0049 29.4123 29.8541 30.3695 PSNR 15.6370 15.5428 15.4097 15.2283 15.0700 14.8789 SSIM 0.2252 0.2182 0.2103 0.2012 0.1939 0.1864

[0158] Table 6(b)

[0159] RMSE 35.7352 35.7566 35.7920 35.8398 35.9028 35.9780 PSNR 12.3414 12.3463 12.3569 12.3723 12.3871 12.4047 SSIM 0.1624 0.1616 0.1604 0.1590 0.1576 0.1560

[0160] As shown in Tables 6(a) and 6(b), the image quality deteriorates with increasing noise intensity after noise is added to the test dataset. However, the integrated aperture radiometer error correction method provided in this application exhibits good noise resistance. Compared with traditional methods, the integrated aperture radiometer error correction method provided in this application consistently achieves higher PSNR, SSIM, and lower RMSE for all noise intensities. The RMSE of the brightness temperature image reconstructed by the visibility function after correction using the integrated aperture radiometer error correction method provided in this application is consistently below 31K, and the PSNR is consistently above 14dB. In contrast, the RMSE of the brightness temperature image obtained by the traditional method is consistently above 35.5K, and the PSNR is consistently below 12dB. Furthermore, the SSIM value corresponding to the integrated aperture radiometer error correction method provided in this application is consistently higher than that of the traditional method.

[0161] In the case of training and testing with noisy data, a training set with a noise intensity of 0.3 was used for training, and various datasets with different noise intensities were also used to test the integrated aperture radiometer error correction method provided in this application. The experimental results are shown in Table 7.

[0162] Table 7

[0163] RMSE 27.9293 27.9548 28.0167 28.1029 28.2331 28.4182 30.3695 PSNR 15.0293 15.0221 14.9980 14.9579 14.9024 14.8266 14.8789 SSIM 0.2250 0.2236 0.2212 0.2178 0.2132 0.2074 0.1864

[0164] Based on Table 7, comparing the test results in Table 6 with a noise intensity of 0.3 reveals that training with noisy data improves both the RMSE and PSNR of the near-field corrected image, especially the RMSE, which drops from 30.37K to 28.42K. Testing with data of other noise intensities shows that lower noise levels yield better results in all aspects, almost surpassing the performance of training with noise-free data. This is likely because radiometer measurements in actual operation are essentially a mixture of target signals and various noises. Training with noisy data more realistically simulates the characteristics of radiometer measurements, making the model more closely resemble the actual data distribution. Furthermore, adding noisy training data enhances the model's adaptability and robustness to noise, prompting it to focus on more universal and significant features in the data, thus effectively avoiding overfitting.

[0165] In another embodiment of this application, to verify the near-field error correction capability of the integrated aperture radiometer error correction method provided in this application for the system's measured scene, multiple observation scenes were selected as source scenes for experiments. Based on the trained MSR-NFECNet network, different methods were used to correct the visibility function obtained by the system, and finally, the brightness temperature image reconstruction results of the source scene under different near-field error correction methods were obtained as follows: Figures 6a-6b As shown. Figure 6a The image of the reconstructed integrated aperture radiometer after near-field error correction using the integrated aperture radiometer error correction method provided in this application. Figure 6b This is an image reconstructed using the synthetic aperture IFFT transform after near-field error correction using traditional methods. Combined with... Figures 6a to 6b It is known that after correcting the system's measured visibility using traditional near-field error correction methods, the background error in the image still remains relatively large when using the IFFT transform method for brightness temperature reconstruction. However, after correcting the system's measured visibility using the integrated aperture radiometer error correction method provided in this application, the background is cleaner and the Gibbs oscillation is smaller when using the IFFT transform method for brightness temperature reconstruction. The image quality is better than that after correction using the traditional near-field error correction method, verifying the ability of the integrated aperture radiometer error correction method provided in this application to correct the measured visibility of the system in an extended source scene.

[0166] Since the true brightness temperature distribution of the exposed scene is unknown, the relative variance (RSD) is used to measure the correction effect of different near-field error correction methods on the visibility function of the system measurement of the exposed scene, as well as the noise characteristics of the reconstructed brightness temperature image. The formula is as follows:

[0167]

[0168] For 13 sets of measured data in the exposed source scenario, the relative variances of the visibility function after correction using the traditional near-field error correction method and the integrated aperture radiometer error correction method provided in this application, respectively, after brightness temperature image reconstruction using the IFFT method, were 0.5681 and 0.3932, respectively. A comparison of the two methods shows that the integrated aperture radiometer error correction method provided in this application can reduce the relative variance of the inverted image by 30.79%. The results indicate that for the exposed source scenario, the integrated aperture radiometer error correction method provided in this application can improve the visibility function correction effect of the ASR-Y system under near-field measurement conditions, thereby improving the quality of the reconstructed brightness temperature image.

[0169] In summary, the integrated aperture radiometer error correction method provided by this invention extracts the frequency domain features of the initial near-field visibility function and compensates for the near-field error of the initial near-field visibility function based on these features, thereby correcting near-field errors and improving imaging quality. Furthermore, by constructing a loss function based on the error and using this loss function to backpropagate the near-field error compensation network model until the error is less than or equal to a preset error threshold, this method not only achieves the mapping from the near-field visibility function to the target far-field visibility function in the frequency domain but also allows for flexible adjustment of the compensation intensity based on the target far-field visibility function, further improving imaging quality. This provides a data foundation for applications such as security imaging, concealed object detection, medical imaging, and near-field remote sensing.

[0170] The following describes the integrated aperture radiometer error correction system provided by the present invention, such as... Figure 7 As shown, the integrated aperture radiometer error correction system 700 includes:

[0171] The data acquisition module 701 is used to acquire the original brightness temperature image of the target near-field imaging scene, and to acquire the initial near-field visibility function corresponding to the target near-field imaging scene based on the original brightness temperature image;

[0172] The error compensation module 702 inputs the initial near-field visibility function into a pre-constructed near-field error compensation network model to obtain a new near-field visibility function. The near-field error compensation network model is used to extract the frequency domain feature information of the initial near-field visibility function and to compensate for the near-field error of the initial near-field visibility function based on the frequency domain feature information. The near-field error includes phase error and amplitude error.

[0173] The reconstruction module 703 is used to calculate the error between the new near-field visibility function and the preset target far-field visibility function. If the error is less than or equal to the preset error threshold, the new near-field visibility function is subjected to Fourier transform inversion to obtain the reconstructed brightness temperature image after the target near-field imaging scene is corrected. Otherwise, a loss function is constructed based on the error, and the near-field error compensation network model is backpropagated using the loss function until the error is less than or equal to the preset error threshold to obtain the trained near-field error compensation network model. The initial near-field visibility function is input into the trained near-field error compensation network model, and the new near-field visibility function output by the trained near-field error compensation network model is subjected to Fourier transform inversion to obtain the corrected reconstructed brightness temperature image.

[0174] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. Their specific functions and technical effects can be found in the method embodiments section, and will not be repeated here. Those skilled in the art will understand that, for the sake of convenience and brevity, the division of the above-mentioned functional units and modules is only used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0175] like Figure 8 As shown, embodiments of the present invention provide a terminal device, such as... Figure 8 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 8 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100, wherein the processor D100 executes the computer program D102 to implement the steps in any of the above method embodiments.

[0176] Specifically, when the processor D100 executes the computer program D102, it acquires the original brightness temperature image of the target near-field imaging scene and obtains the initial near-field visibility function corresponding to the target near-field imaging scene based on the original brightness temperature image; it inputs the initial near-field visibility function into a pre-constructed near-field error compensation network model to obtain a new near-field visibility function; it calculates the error between the new near-field visibility function and the preset target far-field visibility function. If the error is less than or equal to a preset error threshold, it performs a Fourier transform inversion on the new near-field visibility function to obtain the corrected reconstructed brightness temperature image of the target near-field imaging scene; otherwise, it constructs a loss function based on the error and uses the loss function to backpropagate the near-field error compensation network model until the error is less than or equal to the preset error threshold to obtain a trained near-field error compensation network model. It then inputs the initial near-field visibility function into the trained near-field error compensation network model and performs a Fourier transform inversion on the new near-field visibility function output by the trained near-field error compensation network model to obtain the corrected reconstructed brightness temperature image. Specifically, by extracting the frequency domain feature information of the initial near-field visibility function and compensating for the near-field error of the initial near-field visibility function based on the frequency domain feature information, the near-field error can be corrected and the imaging quality improved. A loss function is constructed based on the error, and the near-field error compensation network model is backpropagated using the loss function until the error is less than or equal to a preset error threshold. This not only enables the mapping from the near-field visibility function to the target far-field visibility function in the frequency domain, but also allows for flexible adjustment of the compensation intensity based on the target far-field visibility function, further improving the imaging quality.

[0177] The processor D100 can be a central processing unit (CPU), or it can be 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.

[0178] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.

[0179] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0180] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.

[0181] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of protection of this application is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this application as described above, which are not provided in detail for the sake of brevity.

[0182] One or more embodiments in this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments in this application should be included within the protection scope of this application.

Claims

1. A method for correcting errors in a synthetic aperture radiometer, characterized in that, include: Acquire the original brightness temperature image of the target near-field imaging scene, and obtain the initial near-field visibility function corresponding to the target near-field imaging scene based on the original brightness temperature image; The initial near-field visibility function is input into a pre-constructed near-field error compensation network model to obtain a new near-field visibility function. The near-field error compensation network model is used to extract the frequency domain feature information of the initial near-field visibility function and to compensate for the near-field error of the initial near-field visibility function based on the frequency domain feature information. The near-field error includes phase error and amplitude error. The error between the new near-field visibility function and the preset target far-field visibility function is calculated. If the error is less than or equal to a preset error threshold, a Fourier transform inversion is performed on the new near-field visibility function to obtain the reconstructed brightness temperature image of the target near-field imaging scene after correction. Otherwise, a loss function is constructed based on the error, and the loss function is used to backpropagate the near-field error compensation network model until the error is less than or equal to the preset error threshold to obtain a trained near-field error compensation network model. The initial near-field visibility function is then input into the trained near-field error compensation network model. After performing a Fourier transform inversion on the new near-field visibility function output by the trained near-field error compensation network model, the corrected reconstructed brightness temperature image is obtained.

2. The method for correcting errors in a synthetic aperture radiometer according to claim 1, characterized in that, The near-field error compensation network model includes an information extraction module, a first feature extraction module, a second feature extraction module, an anti-overfitting module, a feature recombination module, and a feature aggregation module. The information extraction module extracts the amplitude and phase information of the visibility function. The first feature extraction module extracts shallow frequency domain features of the visibility function based on the amplitude and phase information. These shallow frequency domain features characterize the overall structure and distribution of the frequency domain data, reflecting the changing trends of the near-field and far-field visibility functions in terms of global amplitude and phase. The second feature extraction module extracts deep frequency domain features of the visibility function from the shallow frequency domain features. These deep frequency domain features characterize local detail differences in the frequency domain data, depicting fine structural deviations and local distortions caused by near-field effects. The anti-overfitting module randomly deactivates some neurons in the deep frequency domain features to prevent model overfitting. The feature recombination module recombines the deep frequency domain features output by the anti-overfitting module to obtain a multi-channel feature map. The feature aggregation module performs average pooling on the multi-channel feature map to achieve near-field error compensation.

3. The method for correcting errors in a synthetic aperture radiometer according to claim 2, characterized in that, Both the first feature extraction module and the second feature extraction module consist of five sequentially connected residual convolutional blocks; The shallow frequency domain features are represented as follows: in, This represents the shallow frequency domain features. This represents the amplitude and phase information of the visibility function output by the information extraction module. , Represents the visibility function The amplitude information extracted from the real part, Represents the visibility function Phase information extracted from the imaginary part, Indicates the flattening operation. This indicates a splicing operation. This indicates a normalization operation. Indicates to Perform convolution operations to extract... The original characteristics, Indicates the first The output of each residual convolutional block This represents the first convolutional layer in the residual convolutional block. This represents the activation function. This represents the second convolutional layer in the residual convolutional block. Indicates batch normalization, Indicates a skip connection; The deep frequency domain features are represented as follows: in, This represents the deep frequency domain characteristics.

4. The radiometer error correction method according to claim 3, characterized in that, The process of reconstructing the deep frequency domain features output by the anti-overfitting module to obtain a multi-channel feature map includes: Through calculation formula The multi-channel feature map is obtained. ; This represents the third convolutional layer in the feature recombination module. This represents the activation function, used to retain some gradient responses to negative values, thereby improving the network's ability to capture subtle features. This represents the fourth convolutional layer in the feature recombination module. This indicates that the anti-overfit module is... The result of randomly inactivating some neurons in the brain.

5. The method for correcting errors in a synthetic aperture radiometer according to claim 4, characterized in that, The error mentioned is the mean absolute error.

6. The method for correcting errors in a synthetic aperture radiometer according to claim 5, characterized in that, The expression for the error loss function is: in, This represents the error loss value. , This represents the total number of tiny facets in the original brightness temperature image. This represents the preset target far-field visibility function. This represents the new near-field visibility function.

7. The method for correcting errors in a synthetic aperture radiometer according to claim 6, characterized in that, The original brightness temperature image comprises multiple micro-areas.

8. A composite aperture radiometer error correction system, characterized in that, include: The data acquisition module is used to acquire the original brightness temperature image of the target near-field imaging scene, and to acquire the initial near-field visibility function corresponding to the target near-field imaging scene based on the original brightness temperature image; An error compensation module is used to input the initial near-field visibility function into a pre-constructed near-field error compensation network model to obtain a new near-field visibility function. The near-field error compensation network model is used to extract the frequency domain feature information of the initial near-field visibility function and, based on the frequency domain feature information, compensate for the near-field error of the initial near-field visibility function. The near-field error includes phase error and amplitude error. The reconstruction module is used to calculate the error between the new near-field visibility function and the preset target far-field visibility function. If the error is less than or equal to a preset error threshold, the new near-field visibility function is subjected to a Fourier transform inversion to obtain the reconstructed brightness temperature image of the target near-field imaging scene after correction. Otherwise, a loss function is constructed based on the error, and the near-field error compensation network model is backpropagated using the loss function until the error is less than or equal to the preset error threshold to obtain a trained near-field error compensation network model. The initial near-field visibility function is input into the trained near-field error compensation network model, and the new near-field visibility function output by the trained near-field error compensation network model is subjected to a Fourier transform inversion to obtain the corrected reconstructed brightness temperature image.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images