SAR (Synthetic Aperture Radar) interference suppression method for intelligent image decomposition

Abstract

The invention discloses an SAR (Synthetic Aperture Radar) interference suppression method based on intelligent image decomposition, which mainly solves the problem of poor SAR imaging interference suppression effect in a complex electromagnetic environment in the existing method. According to the implementation scheme, actually measured data are obtained, and interference signals are generated; respectively imaging measured data, interference signals and mixed data consisting of interference and measured data by using an imaging algorithm, and dividing into a training set and a test set; constructing an image decomposition interference suppression network comprising an image decomposition network and an interference suppression network, and carrying out iterative training on the network by utilizing the training set and adopting back propagation; and inputting the test set into the trained image decomposition interference suppression network to obtain an image after interference suppression. According to the method, interference suppression is carried out on the image level, sufficient interference suppression training can be carried out on a specific scene, the utilization rate of image data and the capability of suppressing interference of the specific scene are improved, and the method can be used for suppressing various interferences caused to radar imaging in a complex electromagnetic environment in target detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar signal processing, and in particular relates to a synthetic aperture radar (SAR) interference suppression method, which can be used in target detection to effectively suppress various interferences to radar imaging caused by complex electromagnetic environments. Background Art

[0002] With the rapid development of electronic technology and communication technology, radar has been widely used in many fields. However, radar is very susceptible to various interferences when working in a complex electromagnetic environment, which seriously affects the accuracy of radar detection of targets. In modern warfare and civilian fields, the electromagnetic environment is becoming increasingly complex, and various intentional or unintentional interference sources continue to emerge, posing a serious threat to the normal operation of radar. For example, in some electromagnetically dense areas, such as urban centers and industrial clusters, the electromagnetic signals generated by civilian equipment may also interfere with radar. In recent years, in order to solve the problem of radar interference, researchers have proposed many methods and achieved some results.

[0003] The patent document with patent application number CN202410783996.1 proposes a "multi-band radar signal analysis and processing method", which uses a multi-band radar receiving antenna to simultaneously collect the returned multi-band radar signal; uses a variational frequency band decomposition model to decompose the multi-band radar signal into several sub-bands; synchronously squeezes the wavelet transform for each sub-band; after nonlinear Fourier transforming each synchronous squeezed wavelet transform result, calculates its synchronous adaptive entropy, and then performs multi-scale adaptive phase adjustment on the synchronous adaptive entropy to obtain a multi-scale adaptive phase adjustment result; based on all multi-scale adaptive phase adjustment results, performs signal reconstruction to extract the target signal from the multi-band radar signal. Although this method can effectively process multi-band signals and improve the accuracy of time-frequency representation, it cannot effectively separate the interference signal from the radar signal when the interference signal band is extremely close to or even overlaps with the radar signal band, so that the interference signal is still mixed in the radar signal band to participate in subsequent processing, resulting in poor interference suppression effect for some interferences with frequencies close to the target signal.

[0004] The patent document with patent application number CN202110056256.4 proposes a "radar slice store-and-forward interference suppression method based on optimized deep model", and its implementation scheme is: obtain radar target echo signal and radar slice store-and-forward interference signal, and pre-process them to obtain training set and verification set; construct a deep U-shaped network model to realize radar interference suppression function, and adjust the hyperparameters of the deep U-shaped network model; use the training set U-shaped network model for training; use the optimized deep model to perform interference suppression processing on radar target echo signal and radar slice store-and-forward interference signal, and complete radar slice store-and-forward interference suppression based on deep model optimization. This method can only achieve interference suppression of radar slice store-and-forward interference, and has limited adaptability when dealing with complex and changeable interference scenarios. Summary of the invention

[0005] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art and propose a SAR interference suppression method based on intelligent image decomposition to effectively deal with complex interference situations, improve the accuracy of radar target detection and imaging, and reduce the time and effort consumed in interference suppression.

[0006] The key technology to achieve the above objectives is: by constructing and training an image decomposition interference suppression network, the interference image is decomposed into pure images and pure interference images, the interference can be directly suppressed at the imaging level, and complex suppression processing of interference signals in the frequency domain and adjustment of specific interference signal types and frequency band parameters can be avoided. The interference images can be processed in parallel using GPUs to improve the efficiency of interference suppression and reduce the time and effort consumed in interference suppression.

[0007] According to the above ideas, the implementation steps of the present invention include the following:

[0008] (1) Create training set and test sample set:

[0009] Acquire synthetic aperture radar measured data from a public website, generate interference signals based on the measured data, signal-to-interference ratio, and interference type; add interference signals to the measured data to obtain mixed data;

[0010] The measured data, interference signals and mixed data are imaged separately, cropped into sub-images of specified sizes, and then divided into training sets and test sets in a ratio of 7:3;

[0011] (2) Constructing image decomposition interference suppression network:

[0012] 2a) constructing an image decomposition network including an image feature extraction module and a feature decomposition module, wherein the image feature extraction module is connected to the two feature decomposition modules in parallel;

[0013] 2b) constructing an interference suppression network including an encoding link, a decoding link and a feature fusion module;

[0014] The encoding chain is an image block linear embedding module, a first feature extraction module, a first downsampling module, a second feature extraction module and a second downsampling module connected in sequence;

[0015] The decoding link is a first upsampling module, a third feature extraction module, a second upsampling module, a fourth feature extraction module and an image merging module connected in sequence;

[0016] The feature fusion module is connected between the first down-sampling module and the second up-sampling module;

[0017] 2c) Connect the two output ports of the image decomposition network to the two interference suppression networks respectively to form an image decomposition interference suppression network, and use the mean square error function as the loss function Loss of the network. MSE ;

[0018] (3) Input the training set into the image decomposition interference suppression network, and iteratively train it using data-driven methods to obtain a trained image decomposition interference suppression network. The Adam algorithm is used as the optimization algorithm.

[0019] (4) Input the test set into the trained image decomposition interference suppression network to obtain the image decomposition interference suppression result.

[0020] Compared with the prior art, the present invention has the following advantages:

[0021] First, since the present invention suppresses interference at the image level, it does not require prior knowledge of specific interference signal bandwidth, frequency, etc., and has stronger robustness and flexibility when facing complex and changeable interference environments, and can work more stably;

[0022] Second, the present invention adds interference signals to the original echo signals and performs imaging to generate SAR images with multiple types and interference levels, thereby being able to perform sufficient interference suppression training for specific scenes and improving the ability to suppress interference for specific scenes;

[0023] Third, since the present invention constructs an image decomposition network including an image decomposition module and an image feature extraction module, it can decompose the interfered SAR image into a pure image and an interference image, thereby improving the utilization of the pure image and the interference image in a data-driven manner and improving the interference suppression capability.

[0024] Fourth, the present invention constructs an interference suppression network including a coding link, a feature fusion module and a decoding link, embeds feature image segmentation into a feature image block matrix through an image segmentation linear embedding module, introduces a multi-head self-attention mechanism through a feature extraction module, extracts image features hierarchically through a downsampling module and an upsampling module, and fuses multi-level image features through a feature fusion module, which not only improves the network's ability to suppress interference, but also enhances the network's ability to extract image features. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a flow chart for realizing the present invention;

[0026] Figure 2 A schematic diagram of the image decomposition interference suppression network structure constructed in the present invention;

[0027] Figure 3 for Figure 2 Schematic diagram of interference suppression network structure in;

[0028] Figure 4 A flowchart of iterative training of an image decomposition interference suppression network in the present invention;

[0029] Figure 5 This is the interference suppression effect diagram of this experiment. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, other embodiments obtained by ordinary technicians in this field without making creative work should all fall within the protection scope of the present invention.

[0031] It should be noted that the step numbers in the specification and claims of the present invention are only for a clear description of the implementation scheme of the present invention to facilitate understanding, and the order of the numbers is not limited.

[0032] Reference Figure 1 The implementation steps of this example include the following:

[0033] Step 1: Set up the training set and test sample set.

[0034] 1.1) Obtain the measured data of the ESR2 satellite from the public website and extract the carrier frequency f of the radar signal from the measured data c and bandwidth B, and convert the measured data into a two-dimensional complex matrix DATA with a size of Na*Nr;

[0035] 1.2) According to the carrier frequency f of the radar signal c and bandwidth B, design a filter with a passband the same as the radar signal frequency band;

[0036] 1.3) Use simulation software to generate a complex Gaussian white noise signal of size Na*Nr, and use the designed filter to filter the generated Gaussian white noise to obtain the noise signal NGWN;

[0037] 1.4) Set the signal-to-interference ratio JSR and calculate the average power P of the measured data signal , and the average power of the noise signal P jammer ;

[0038] 1.5) Calculate the amplitude coefficient J of the interference signal based on the result of step 1.4),

[0039]

[0040] 1.6) Generate the following four different interference signals according to the amplitude coefficient J of the interference signal and the interference type:

[0041] Noise product interference: J1 = J*NGWN;

[0042] Noise amplitude modulation interference: J2 = J*abs(NGWN)

[0043] Noise FM interference: J3 = J*cos(2πf c *t+2π*NGWN)

[0044] Noise phase interference: J4 = J*imag(NGWN)

[0045] Where abs(NFWN) is the amplitude matrix of each point of the noise signal, f c is the carrier frequency of the radar signal, t is the time of the radar signal, and imag(NGWN) is the imaginary part of the noise signal;

[0046] 1.7) Perform matrix addition on the measured data matrix DATA and the four interference signals respectively to obtain the product interference mixed data D1, amplitude modulation interference mixed data D2, frequency modulation interference mixed data D3 and phase modulation interference mixed data D4:

[0047] D1=DATA+J1

[0048] D2=DATA+J2

[0049] D3=DATA+J3

[0050] D4=DATA+J4;

[0051] 1.8) Perform range Doppler algorithm imaging on the measured data, interference signal and mixed data respectively, and crop the measured data image, interference signal image and mixed data image into sub-images of a specified size of 256*256;

[0052] 1.9) Divide these sub-images into training set and test set in a ratio of 7:3.

[0053] Step 2: Construct an image decomposition interference suppression network.

[0054] refer to Figure 2 and Figure 3 This step is implemented as follows:

[0055] 2.1) Construct image decomposition network:

[0056] 2.1.1) Establish an image feature extraction module consisting of the first convolutional layer, the second convolutional layer, the third convolutional layer and the nonlinear activation layer connected in sequence, wherein: the number of convolution kernels in the first convolutional layer is 3, the convolution kernel size is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels in the second layer is 8, the convolution kernel size is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels in the third layer is 16, the convolution kernel size is 3*3, the step size is 1*1, and the padding is 1; the nonlinear activation layer is implemented by the ReLU function;

[0057] 2.1.2) Establish a first image feature decomposition module consisting of a fourth convolutional layer, a fifth convolutional layer, a sixth convolutional layer and a nonlinear activation layer connected in sequence, wherein: the number of convolution kernels in the fourth convolutional layer is 3, the convolution kernel size is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels in the fifth layer is 8, the convolution kernel size is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels in the sixth layer is 16, the convolution kernel size is 3*3, the step size is 1*1, and the padding is 1; the nonlinear activation layer is implemented by the ReLU function;

[0058] 2.1.3) establishing a second image feature decomposition module having the same structure and parameters as the first image feature decomposition module;

[0059] 2.1.4) connecting the image feature extraction module to the first image feature decomposition module and the second image feature decomposition module in parallel to form an image decomposition network;

[0060] 2.2) Construct the encoding link:

[0061] 2.2.1) Establishing an image block linear embedding module including an image segmentation window and an image block embedding layer, wherein the image segmentation window is a sliding window with a size of 16*16 and a moving step of 16, which is used to segment the image into 16*16 image blocks; the image block embedding layer forms an image block matrix according to the positions of the image blocks before segmentation in the image;

[0062] 2.2.2) Establishing a first feature extraction module and a second feature extraction module, wherein both modules include a convolutional layer and a multi-head self-attention layer, wherein the convolutional layer is used to realize feature extraction of feature images, and the multi-head self-attention layer is used to realize feature dependency extraction, wherein the convolutional layer and the multi-head self-attention layer are existing;

[0063] 2.2.3) Establish a first downsampling module consisting of a downsampling first convolutional layer, a downsampling second convolutional layer, a downsampling first nonlinear activation function layer and a downsampling first pooling layer connected in sequence, wherein: the number of convolution kernels of the downsampling first convolutional layer is 8, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels of the downsampling second convolutional layer is 16, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the downsampling first nonlinear activation function layer is implemented by the ReLU function; the downsampling second pooling layer is implemented by the MaxPool function;

[0064] 2.2.4) Establish a second downsampling module consisting of a downsampling third convolutional layer, a downsampling fourth convolutional layer, a downsampling second nonlinear activation function layer and a downsampling second pooling layer connected in sequence, wherein: the second downsampling module is used to reduce the size of the image and increase its dimension; the number of convolution kernels of the downsampling third convolutional layer is 32, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels of the downsampling fourth convolutional layer is 64, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the downsampling second nonlinear activation function layer is implemented by the ReLU function; the downsampling second pooling layer is implemented by the MaxPool function;

[0065] 2.2.5) connecting the image block linear embedding module, the first feature extraction module, the first downsampling module, the second feature extraction module and the second downsampling module in sequence to form a coding chain;

[0066] 2.3) Construct decoding link:

[0067] 2.3.1) Establish a first upsampling module consisting of an upsampling first transposed convolution layer, an upsampling second transposed convolution layer and an upsampling first nonlinear activation function layer connected in sequence, wherein: the number of convolution kernels of the upsampling first transposed convolution layer is 64, the size of the transposed convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels of the upsampling second transposed convolution layer is 32, the size of the transposed convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the upsampling first nonlinear activation function layer is implemented by the ReLU function;

[0068] 2.3.2) Establish a second upsampling module consisting of an upsampling third transposed convolution layer, an upsampling fourth transposed convolution layer and an upsampling second nonlinear activation function layer connected in sequence, wherein: the number of convolution kernels of the upsampling third transposed convolution layer is 8, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels of the upsampling fourth transposed convolution layer is 3, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the upsampling second nonlinear activation function layer is implemented by the ReLU function;

[0069] 2.3.3) Establishing a third feature extraction module and a fourth feature extraction module, wherein both modules include a convolutional layer and a multi-head self-attention layer, wherein the convolutional layer is used to realize feature extraction of feature images, and the multi-head self-attention layer is used to realize feature dependency extraction, wherein the convolutional layer and the multi-head self-attention layer are existing;

[0070] 2.3.4) establishing an image merging module implemented by the mapping layer, which is used to merge the image blocks into a complete image according to the positions of the image blocks in the image block matrix;

[0071] 2.3.5) Connecting the first upsampling module, the third feature extraction module, the second upsampling module, the fourth feature extraction module, and the image merging module in sequence to form a decoding chain;

[0072] 2.4) constructing a feature fusion module for connecting the first down-sampling module and the third feature extraction module, which performs feature fusion on the feature image processed by the first down-sampling module and the feature image processed by the third feature extraction module through a fusion layer, and performs tensor addition on the corresponding channel feature images;

[0073] 2.5) connecting the encoding link with the decoding link, and connecting the feature fusion module between the first down-sampling module of the encoding link and the third feature extraction module of the decoding link to form a first interference suppression network;

[0074] 2.6) establishing a second interference suppression network having the same structure and parameters as the first interference suppression network;

[0075] 2.7) The image decomposition network is connected in parallel with the first interference suppression network and the second interference suppression network to form an image decomposition interference suppression network.

[0076] Step 3 iteratively trains the image decomposition interference suppression network model.

[0077] Reference Figure 4 , this step includes the following:

[0078] 3.1) Set the network parameters, learning rate to 0.0001, batch size to 16, optimization algorithm to Adam, loss function to MSE, number of training rounds to 100, and preset weights of interference loss value to: {β1, β2, ...β i ...β n}, i=1,2...n each β i The value corresponds to the round number;

[0079] 3.2) Inputting the mixed data image in the training set into the image decomposition interference suppression network model, and obtaining a preliminary feature image through the image feature extraction module;

[0080] 3.3) The preliminary feature image is respectively subjected to the first image decomposition module and the second image decomposition module to obtain a decomposed measured data image and a decomposed interference signal image;

[0081] 3.4) Decomposing the measured data image and passing it through the first interference suppression network to obtain the predicted measured data image:

[0082] 3.4.1) The decomposed measured data image is processed by the image block linear embedding module to obtain the decomposed measured data image block matrix;

[0083] 3.4.2) The first feature extraction module performs feature extraction on the decomposed measured data image block matrix to obtain a first feature extracted measured image block matrix;

[0084] 3.4.3) The first downsampling module performs downsampling processing on each image block in the first feature extraction measured image block matrix to obtain a first downsampled measured image block matrix;

[0085] 3.4.4) The second feature extraction module performs feature extraction on the first down-sampled feature measured image block matrix to obtain a second feature extracted measured image block matrix;

[0086] 3.4.5) The second downsampling module performs downsampling processing on each image block in the second feature extraction measured image block matrix to obtain a second downsampling measured image block matrix;

[0087] 3.4.6) The first upsampling module performs upsampling processing on each image block in the second down-sampled measured image block matrix to obtain a first up-sampled measured image block matrix;

[0088] 3.4.7) The third feature extraction module performs feature extraction on the first up-sampled measured image block matrix to obtain a third feature extracted measured image block matrix;

[0089] 3.4.8) The feature fusion module splices and fuses the image blocks at corresponding positions in the first down-sampled measured image block matrix and the third feature extraction measured image block matrix to obtain a measured feature fusion image block matrix;

[0090] 3.4.9) The second upsampling module processes each image block in the feature fusion image block matrix to obtain a second upsampling measured image block matrix;

[0091] 3.4.10) The fourth feature extraction module performs a feature extraction operation on the second up-sampled measured image block matrix to obtain a fourth feature extracted measured image block matrix;

[0092] 3.4.11) The image merging module performs merging processing on the fourth feature extraction measured image block matrix to obtain a predicted measured data image;

[0093] 3.5) Decomposing the interference data image and obtaining the predicted interference data image through the second interference suppression network;

[0094] 3.5.1) The decomposed interference data image is processed by the image block linear embedding module to obtain the decomposed interference data image block matrix;

[0095] 3.5.2) The first feature extraction module performs feature extraction on the decomposed interference data image block matrix to obtain a first feature extracted interference image block matrix;

[0096] 3.5.3) The first down-sampling module performs down-sampling processing on each image block in the first feature extraction interference image block matrix to obtain a first down-sampled interference image block matrix;

[0097] 3.5.4) The second feature extraction module performs feature extraction on the first down-sampled feature interference image block matrix to obtain a second feature extraction interference image block matrix;

[0098] 3.5.5) The second down-sampling module performs down-sampling processing on each image block in the second feature extraction interference image block matrix to obtain a second down-sampled interference image block matrix;

[0099] 3.5.6) The first upsampling module performs upsampling processing on each image block in the second down-sampled interference image block matrix to obtain a first up-sampled interference image block matrix;

[0100] 3.5.7) The third feature extraction module performs feature extraction on the first up-sampled interference image block matrix to obtain a third feature extracted interference image block matrix;

[0101] 3.5.8) The feature fusion module splices and fuses the image blocks at corresponding positions in the first down-sampled interference image block matrix and the third feature extraction interference image block matrix to obtain an interference feature fusion image block matrix;

[0102] 3.5.9) The second upsampling module processes each image block in the interference feature fusion image block matrix to obtain a second upsampling interference image block matrix;

[0103] 3.5.10) The fourth feature extraction module performs a feature extraction operation on the second up-sampled interference image block matrix to obtain a fourth feature extracted interference image block matrix;

[0104] 3.5.11) The image merging module merges the fourth feature extraction interference image block matrix to obtain a predicted interference data image;

[0105] 3.6) Calculate the loss value between the measured data image predicted by the network model output and the measured data image in the training set as the measured loss value loss1, and calculate the loss value between the predicted interference signal image output by the network model and the interference signal image in the training set as the interference loss value loss2; perform weighted calculation on the measured loss value and the interference loss value to obtain the total loss value loss = loss1 + β i loss2;

[0106] 3.7) Back-propagate the total loss value loss and optimize the network parameters;

[0107] 3.8) Repeat steps 3.2) to 3.7) for a total of 100 rounds;

[0108] 3.9) Update the interference loss value weight to β i+1 , repeat steps 3.2) to 3.8) until all preset weight values ​​{β1,β2,...β i ...β n}, and get the trained network model.

[0109] Step 4 obtains the image decomposition interference suppression result.

[0110] 4.1) Input the mixed data image in the test set into the trained image decomposition interference suppression network model, and obtain the preliminary feature image through the image feature extraction module;

[0111] 4.2) The preliminary feature image is processed by the image decomposition module to obtain a decomposed measured data image;

[0112] 4.3) The decomposed measured data image is processed by an image block linear embedding module to obtain a decomposed measured data image block matrix;

[0113] 4.4) The first feature extraction module performs feature extraction on the decomposed measured data image block matrix to obtain a first feature extraction image block matrix;

[0114] 4.5) The first downsampling module performs downsampling processing on each image block in the first feature extraction image block matrix to obtain a first downsampling feature image block matrix;

[0115] 4.6) The second feature extraction module performs feature extraction on the first down-sampled feature image block matrix to obtain a second feature extraction image block matrix;

[0116] 4.7) The second down-sampling module performs down-sampling processing on each image block in the second feature extraction image block matrix to obtain a second down-sampled feature image block matrix;

[0117] 4.8) The first upsampling module performs upsampling processing on each image block in the second down-sampling feature image block matrix to obtain a first down-sampling feature image block matrix;

[0118] 4.9) The third feature extraction module performs feature extraction on the first up-sampled feature image block matrix to obtain a third feature extraction image block matrix;

[0119] 4.10) The feature fusion module splices and fuses the image blocks at corresponding positions in the first down-sampled feature image block matrix and the third feature extraction image block matrix to obtain a feature fusion image block matrix;

[0120] 4.11) The second upsampling module processes each image block in the feature fusion image block matrix to obtain a second upsampling feature image block matrix;

[0121] 4.12) The fourth feature extraction module performs a feature extraction operation on the second up-sampled feature image block matrix to obtain a fourth feature extraction image block matrix;

[0122] 4.13) The image merging module merges the fourth feature extraction image block matrix to obtain the predicted measured data image, that is, the image output result after interference suppression.

[0123] The effect of the present invention can be further illustrated by the following simulation experiment.

[0124] 1. Simulation conditions

[0125] ERS2 satellite data were downloaded from the ASF Data Serch website;

[0126] This experiment uses the MATLAB R2022a software platform to extract measured data, generate interference signals, add interference signals and perform imaging. Figure 5 (a);

[0127] Use the PyTorch framework to build and train the network model.

[0128] 2. Simulation results

[0129] Under the above conditions, the present invention is used to Figure 5 (a) shows the interference suppression of the SAR image containing noise amplitude modulation interference. The result is as follows Figure 5 (b).

[0130] from Figure 5 It can be seen that before interference suppression, it can be observed that the SAR image has obvious interference components. After interference suppression, it can be observed that the interference components in the SAR image are greatly suppressed, and a clear SAR image is obtained.

Claims

1. A SAR interference suppression method based on intelligent image decomposition, characterized in that: The steps include: (1) Create training set and test sample set: Acquire synthetic aperture radar measured data from a public website, generate interference signals based on the measured data, signal-to-interference ratio, and interference type; add interference signals to the measured data to obtain mixed data; The measured data, interference signals and mixed data are imaged separately, cropped into sub-images of specified sizes, and then divided into training sets and test sets in a ratio of 7:3; (2) Constructing image decomposition interference suppression network: 2a) constructing an image decomposition network including an image feature extraction module and a feature decomposition module, wherein the image feature extraction module is connected to the two feature decomposition modules in parallel; 2b) constructing an interference suppression network including an encoding link, a decoding link and a feature fusion module; The encoding chain is an image block linear embedding module, a first feature extraction module, a first downsampling module, a second feature extraction module and a second downsampling module connected in sequence; The decoding link is a first upsampling module, a third feature extraction module, a second upsampling module, a fourth feature extraction module and an image merging module connected in sequence; The feature fusion module is connected between the first down-sampling module and the second up-sampling module; 2c) Connect the two output ports of the image decomposition network to the two interference suppression networks respectively to form an image decomposition interference suppression network, and use the mean square error function as the loss function Loss of the network. MSE ; (3) Input the training set into the image decomposition interference suppression network, and iteratively train it using data-driven methods to obtain a trained image decomposition interference suppression network. The Adam algorithm is used as the optimization algorithm. (4) Input the test set into the trained image decomposition interference suppression network to obtain the image decomposition interference suppression result.

2. The method according to claim 1, characterized in that Step (1) generates an interference signal according to the measured data, the signal-to-interference ratio and the interference type, and the implementation steps include the following: 1a) Use the program provided on the website to extract the carrier frequency f of the radar signal c and bandwidth B, and convert the measured data into a two-dimensional complex matrix DATA with a size of Na*Nr; 1b) Design a filter with a passband that is the same as the radar signal frequency band based on the carrier frequency and bandwidth of the radar signal; 1c) Generate a complex Gaussian white noise signal of size Na*Nr using simulation software, and use the designed filter to filter the generated Gaussian white noise to obtain a noise signal NGWN; 1d ) Calculate the average power P of the measured data signal , and the average power of the noise signal P jammer , calculate the amplitude coefficient J of the interference signal according to the set signal-to-interference ratio JSR, 1e) According to the amplitude coefficient J of the interference signal and the interference type, four interference signals are generated, namely, noise product interference J1, noise amplitude modulation interference J2, noise frequency modulation interference J3 and noise phase modulation interference J4: J1=J*NGWN J2=J*abs(NGWN) J3=J*cos(2πf c *t+2π*NGWN) J4=J*imag(NGWN) abs(NGWN) is the amplitude matrix of each point of the noise signal, f c is the carrier frequency of the radar signal, t is the time of the radar signal, and imag(NGWN) is the imaginary part of the noise signal.

3. The method according to claim 1, characterized in that: Step (1) images the measured data, interference signal and mixed data respectively by adopting the range Doppler imaging algorithm to perform range pulse compression, range migration correction and azimuth pulse compression on the measured data, interference signal and mixed data respectively to obtain the measured data image, interference signal image and mixed data image.

4. The method according to claim 1, characterized in that The two modules in the image decomposition network in step 2a) have the following structures and functions: The image feature extraction module is used to extract image features, and is composed of a first convolutional layer, a second convolutional layer, a third convolutional layer and a nonlinear activation layer connected in sequence, wherein: the number of convolution kernels in the first convolutional layer is 3, the convolution kernel size is 3*3, the step length is 1*1, and the padding is 1; the number of convolution kernels in the second layer is 8, the convolution kernel size is 3*3, the step length is 1*1, and the padding is 1; the number of convolution kernels in the third layer is 16, the convolution kernel size is 3*3, the step length is 1*1, and the padding is 1; the nonlinear activation layer is implemented by the ReLU function; The image feature decomposition module is used to decompose the image, and is composed of a fourth convolutional layer, a fifth convolutional layer, a sixth convolutional layer and a nonlinear activation layer connected in sequence, wherein: the number of convolution kernels in the fourth convolutional layer is 3, the convolution kernel size is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels in the fifth layer is 8, the convolution kernel size is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels in the sixth layer is 16, the convolution kernel size is 3*3, the step size is 1*1, and the padding is 1; the nonlinear activation layer is implemented by a ReLU function.

5. The method according to claim 1, characterized in that The modules that make up the coding link in step 2b) are structured as follows: The image block linear embedding module comprises: an image segmentation window and an image block embedding, wherein: the image segmentation window is a sliding window with a size of 16*16 and a moving step of 16, which is used to segment the image into 16*16 image blocks; the image block embedding is to embed the segmented image blocks into an image block matrix; The first feature extraction module and the second feature extraction module both include a convolution layer and a multi-head self-attention layer, the convolution layer is used to realize feature extraction of the feature image, and the multi-head self-attention layer is used to realize feature dependency extraction; The first downsampling module is composed of a downsampling first convolutional layer, a downsampling second convolutional layer, a downsampling first nonlinear activation function layer and a downsampling first pooling layer connected in sequence, wherein: The number of convolution kernels in the first convolution layer of this downsampling is 8, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; The number of convolution kernels in the second convolution layer of the downsampling is 16, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; The first nonlinear activation function layer of the downsampling is implemented by the ReLU function; The down-sampling second pooling layer is implemented by the MaxPool function; The second downsampling module is composed of a downsampling third convolution layer, a downsampling fourth convolution layer, a downsampling second nonlinear activation function layer and a downsampling second pooling layer connected in sequence, wherein: The second downsampling module is used to reduce the size of the image and increase its dimension; The number of convolution kernels in the third convolution layer of downsampling is 32, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; The number of convolution kernels in the fourth convolution layer of the downsampling is 64, the size of the convolution kernel is 3*3, the step size is 1*1, and the padding is 1; The downsampling second nonlinear activation function layer is implemented by the ReLU function; The downsampling second pooling layer is implemented by the MaxPool function.

6. The method according to claim 1, characterized in that The structure and function of each module of the decoding link in step 2b) are as follows: The third feature extraction module and the fourth feature extraction module both include a convolution layer and a multi-head self-attention layer, wherein the convolution layer realizes feature extraction of the feature image, and the multi-head self-attention layer realizes extraction of feature dependency; The first upsampling module and the second upsampling module both include a transposed convolution layer and a nonlinear activation function layer; The first upsampling module is composed of an upsampling first transposed convolution layer, an upsampling second transposed convolution layer and an upsampling first nonlinear activation function layer connected in sequence; the number of convolution kernels of the upsampling first transposed convolution layer is 64, the size of the transposed convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the number of convolution kernels of the upsampling second transposed convolution layer is 32, the size of the transposed convolution kernel is 3*3, the step size is 1*1, and the padding is 1; the upsampling first nonlinear activation function layer is implemented by a ReLU function; The second upsampling module is composed of an upsampling third transposed convolution layer, an upsampling fourth transposed convolution layer and an upsampling second nonlinear activation function layer connected in sequence; the upsampling third transposed convolution layer has 8 convolution kernels, a convolution kernel size of 3*3, a step size of 1*1, and a padding of 1; the upsampling fourth transposed convolution layer has 3 convolution kernels, a convolution kernel size of 3*3, a step size of 1*1, and a padding of 1; the upsampling second nonlinear activation function layer is implemented by a ReLU function; The image merging module is used to merge the image block matrices into a complete image.

7. The method according to claim 1, characterized in that The feature fusion module in step 2b) realizes feature fusion by concatenating features at the same level in the network, and is used to fuse the features of the shallow and deep layers in the network.

8. The method according to claim 1, characterized in that: Step 2c) The image decomposition interference suppression network loss function Loss MSE , which is expressed as follows: Where (i, j) is the position of the pixel, I target (i, j) is the pixel value at position (i, j) of the measured data image, I generated (i, j) is the pixel value of the predicted measured data image position (i, j), and m and n are the total number of pixels in the row and column respectively.

9. The method according to claim 1, characterized in that: In step (3), the image decomposition interference suppression network is iteratively trained using data-driven methods, and the implementation steps include the following: 3a) Assume that the network training learning rate is 0.0001, the batch size is 16, the optimization algorithm is Adam, and the interference loss value preset weight β is {β1, β2, ...β n }, the number of training rounds for each β value is 100; 3b) The input mixed data image is processed by the network model to obtain a predicted measured data image and a predicted interference signal image, and the loss value between the predicted measured data image and the real measured data image is calculated as the measured loss value loss1, and the loss value between the predicted interference signal image and the real interference signal image is calculated as the interference loss value loss2; the total loss value loss=loss1+β·loss2 is obtained by weighted calculation of the measured loss value and the interference loss value; 3c) Back propagation of the total loss value loss is used to optimize the network parameters; 3d)β i After the corresponding training is completed, the interference loss value weight is updated to β i+1 Repeat 3b) and 3c) until all preset weight values ​​{β1, β2, ...β n }, and get the trained network model.

10. The method according to claim 1, characterized in that In step (4), the test set is input into the trained image decomposition interference suppression network to obtain the image decomposition interference suppression result. The implementation steps include the following: 4a) The mixed image of the test set is decomposed by the image decomposition network to obtain the decomposed measured data feature image; 4b) Decompose the measured data feature image and perform interference suppression processing through the interference suppression network to obtain the final output result of the network model.

Images