A multi-target detection method and device in a complex electromagnetic interference environment for pulse Doppler radar

Abstract

The application discloses a multi-target detection method in a complex electromagnetic interference environment for a pulse Doppler radar, a joint reverse diffusion process is introduced to jointly update interference samples and amplitude samples, and thus the problems of target signals being covered or submerged by interference signals, rising of a detection false alarm rate, and difficulty in effectively separating interference from targets can be overcome; a denoising score matching criterion is introduced to train a neural network of an interference score function, so as to solve the problems of high interference intensity and complex structure in multi-target detection in a complex electromagnetic interference environment; and a sparse Bayesian learning method is introduced to adaptively model a sparse structure of multi-target echo signals. The application further provides a multi-target detection device in a complex electromagnetic interference environment. The method provided by the application can realize modeling and suppression of multiple types of interference, and simultaneously improve the detection probability of multiple low observable targets in the distance and Doppler dimension, the interference suppression capability, and the system robustness in a complex electromagnetic environment.

Description

Technical Field

[0001] This invention belongs to the field of radar signal processing technology, and particularly relates to a method and device for multi-target detection under complex electromagnetic interference environments for pulse Doppler radar. Background Technology

[0002] Pulse Doppler radar, as an important system for high-resolution and high-sensitivity target detection, has broad prospects in applications such as air surveillance, low-altitude defense, and UAV countermeasures. However, in practical applications, radar faces serious challenges from complex electromagnetic interference environments, including structured interference signals from sources such as communication systems, power equipment, and electronic countermeasures platforms, such as communication, frequency modulation, phase modulation, amplitude modulation, and comb spectrum interference. These interference signals have high energy, which can severely overwhelm or mask target echo signals, causing a significant decrease in radar detection performance, manifested as target misses, tracking loss, and a sharp increase in false alarm rates.

[0003] In the presence of interference, traditional linear signal processing methods, namely pulse compression and moving target detection, based on matched filtering and frequency domain filtering, cannot adequately distinguish the characteristics of the target from the interference, making it difficult to accurately model and suppress non-stationary or non-Gaussian interference. Furthermore, most of these traditional methods rely on prior knowledge of interference characteristics, making them ill-suited for complex electromagnetic interference environments and resulting in severe degradation of multi-target detection performance.

[0004] Patent document CN120294733A discloses a multi-target detection method for radar one-dimensional range profiles based on the fusion of CFAR and isolated forest. The method includes: acquiring radar raw data and performing preprocessing to obtain radar one-dimensional range profile data; using CACFAR detection adaptive threshold technology to detect targets in the radar one-dimensional range profile data and normalizing the energy value; then calculating the target score using the isolated forest method; performing weighted fusion of the CACFAR detection and isolated forest results to obtain the fusion result; defining a new detection threshold to confirm the target; setting a new detection threshold and determining whether the target exists.

[0005] Patent document CN119291641A discloses a multi-target detection method and system based on millimeter-wave radar, comprising: acquiring raw echo data collected by millimeter-wave radar, wherein the raw echo data includes target data points of multiple targets; selecting a preset number of data points from the target data points of the multiple targets, and using the selected data points as valid data points; and detecting the multiple targets based on the valid data points. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for multi-target detection in complex electromagnetic interference environments for pulse Doppler radar. This method can model and suppress various types of interference (such as comb spectrum interference, continuous wave interference, linear frequency modulation interference, etc.), while improving the detection probability, interference suppression capability, and system robustness in complex electromagnetic environments for multiple low-observable targets in terms of range and Doppler dimensions.

[0007] To achieve the first objective of this invention, the following technical solution is provided: a method for multi-target detection under complex electromagnetic interference environments for pulse Doppler radar, comprising the following steps:

[0008] Step 1: Obtain interference sample data and construct a fractional function neural model for generating interference variables based on the diffusion model and the denoising score matching criterion.

[0009] Step 2: Acquire the echo signal emitted by the pulse Doppler radar under complex electromagnetic interference environment, and perform digital beamforming on the echo signal to obtain detection information in the fast time dimension and the slow time dimension.

[0010] Based on the detection information and the overcomplete dictionary matrix, construct the complex amplitude vector corresponding to the echo signal;

[0011] Step 3: Construct a joint backdiffusion process and initialize the complex amplitude vector and interference sample data as the initial sampling starting point;

[0012] Step 4: Execute the corrector at specific times:

[0013] For the sample at the current time, update the fractional function of the complex amplitude vector and the fractional function of the disturbance variable, and use the Langevin dynamics criterion to correct the updated sample to obtain the sample after sampling correction.

[0014] Step 5: Update the prior variance of the complex magnitude vector in Step 4 using sparse Bayesian learning, and update the score of the complex magnitude vector.

[0015] Step 6: Execute the predictor:

[0016] Based on the scores obtained in step 5, the gradient of the joint posterior distribution with respect to the complex amplitude vector and the interference variable is calculated, and the gradient is used to generate reconstructed samples according to the backdiffusion stochastic differential equation.

[0017] The prior variance of complex amplitudes in the reconstructed samples is updated using sparse Bayesian learning;

[0018] Step 7: Repeat steps 4-6 until the diffusion terminates to obtain the complex amplitude estimate corresponding to the sample.

[0019] Step 8: Input the complex amplitude estimate into the constant false alarm rate detector to determine whether it contains a valid target, and form a valid target set by combining all samples determined to be valid targets.

[0020] This invention introduces a denoising score matching criterion to train a neural network for the interference score function, addressing the problems of high interference intensity and complex structure in multi-target detection under complex electromagnetic interference environments. Simultaneously, it introduces a joint backdiffusion process to jointly update interference samples and amplitude samples, addressing issues such as target signals being masked or submerged, increased false alarm rates, and difficulty in effectively separating interference from targets in multi-target detection under complex electromagnetic interference environments. Furthermore, it introduces a sparse Bayesian learning method to adaptively model the sparse structure of multi-target echo signals, enhancing the detection capability for weak targets and effectively improving estimation accuracy under interference.

[0021] Specifically, the expression for the denoising score matching criterion is as follows:

[0022] ;

[0023] in, For neural network parameters, Expressing expectations, express It is in the interval Obtained by uniform sampling in the middle, It was obtained from a set of interference samples. This represents the L2 norm.

[0024] Specifically, the detection information is obtained by using two-dimensional data consisting of the fast and slow times of straightening the target.

[0025] Specifically, the overcomplete dictionary matrix is ​​constructed based on Doppler networks and distance networks.

[0026] Specifically, the joint back-diffusion process is described by a stochastic differential equation, the expression of which is as follows:

[0027] ;

[0028] in, Represents the gradient. It is the joint posterior distribution of amplitude and interference. It is the reverse of the standard Wiener process. yes Each target corresponds to a complex amplitude vector of a different Doppler-distance grid.

[0029] Specifically, the expression for the initial sampling starting point is as follows:

[0030] , ;

[0031] in, Indicates the first A complex amplitude vector sample, Indicates the first One interference sample, Let be the transition probability parameter at time 1. The dimension is The identity matrix, The dimension is The identity matrix.

[0032] Specifically, the complex amplitude estimate is obtained by calculating the average of all complex amplitude vectors output at the diffusion termination time.

[0033] To achieve the second objective of this invention, the following technical solution is provided: a multi-target detection device under complex electromagnetic interference environment, used to perform the steps of the above-described multi-target detection method for pulse Doppler radar under complex electromagnetic interference environment.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] (1) Compared with moving target detection methods and traditional pulse compression methods based on matched filtering, this invention introduces a joint back diffusion process to jointly update the interference sample and amplitude sample, which can overcome the problems of target signal being masked or submerged by interference signal, increased false alarm rate, and difficulty in effectively separating interference and target.

[0036] (2) For the problem of high interference intensity and complex structure in multi-target detection under complex electromagnetic interference environment, the present invention can realize feature extraction of interference signal by introducing a denoising score matching criterion to train the neural network of interference score function.

[0037] (3) To address the problem of low detection accuracy in multi-target detection under complex electromagnetic interference environment, this invention introduces a sparse Bayesian learning method to adaptively model the sparse structure of multi-target echo signals, enhance the detection capability of weak targets, and effectively improve the estimation accuracy under interference. Attached Figure Description

[0038] Figure 1 This is a flowchart of a multi-target detection method for pulse Doppler radar under complex electromagnetic interference environment provided in this embodiment;

[0039] Figure 2 This is a schematic diagram of the structure of the multi-target detection device under complex electromagnetic interference environment provided in this embodiment;

[0040] Figure 3This is a schematic diagram of the multi-target detection results under strong electromagnetic interference scenarios provided in this embodiment. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.

[0042] like Figure 1 As shown, this embodiment of the invention provides a method for multi-target detection under complex electromagnetic interference environments for pulse Doppler radar, including the following steps:

[0043] S110: Obtain interference sample data and construct a fractional function neural model for generating interference variables based on the diffusion model and the denoising score matching criterion.

[0044] Collect structured interference sample set Perturbation samples are generated using the forward perturbation process of the diffusion model. ,in This represents the transition probability of the sample. The forward perturbation process is modeled using stochastic differential equations, which can be expressed as:

[0045]

[0046] in, Is Perturbation samples at any given time. It is the drift coefficient. It is the diffusion coefficient. It is a function that controls the amount of noise added at each moment. It is the standard Wiener process used to introduce random noise.

[0047] In this embodiment, the neural network for training the interference score function is trained using a denoising score matching criterion, which can be expressed by the formula:

[0048]

[0049] in, For neural network parameters, Expressing expectations, express It is in the interval Obtained by uniform sampling in the middle, It was obtained from a set of interference samples. This represents the L2 norm.

[0050] S120: Acquire the echo signal emitted by the pulse Doppler radar under complex electromagnetic interference environment, and perform digital beamforming on the echo signal to obtain detection information in the fast time dimension and the slow time dimension.

[0051] Based on the detection information and the overcomplete dictionary matrix, a complex amplitude vector corresponding to the echo signal is constructed.

[0052] After down-conversion demodulation and digital beamforming of the echo signal, the pulse Doppler radar signal under complex electromagnetic interference environment is obtained. Two-dimensional data consisting of fast and slow times for each target. Expressed by the formula:

[0053]

[0054] in, Indicates electromagnetic interference signal; Describe Gaussian noise, and satisfy the following conditions: ,in Indicates a complex Gaussian distribution. Indicates the noise variance; Indicates the amplitude value; Indicates transpose; Indicates distance manifold vectors, Represents the angular frequency of velocity The array manifold vector is expressed by the formula:

[0055] ;

[0056] ;

[0057] ;

[0058] in, Indicates the sampling time. Indicates distance, Represents the speed of light. Indicates a pulse signal. Represents the number of samples in the distance dimension. Indicates the number of pulses. Indicates the carrier frequency. Indicates the pulse repetition period. Indicates the target speed.

[0059] Straighten the two-dimensional data to obtain detection information. Expressed by the formula:

[0060]

[0061] in, ;

[0062] ; ; ; ; ; This indicates that the matrix is ​​being straightened.

[0063] Constructing a Doppler grid and distance grid overcomplete dictionary matrix This can be expressed as a formula:

[0064] ;

[0065] Assuming the target's Doppler and range both fall on the grid, then the detection information... Based on an overcomplete dictionary matrix Modeling, expressed by a formula:

[0066] ;

[0067] in, It is the complex amplitude vector corresponding to the echo signal.

[0068] Will The prior model is zero-mean, and the variance is determined by sparse vectors. The controlled complex Gaussian distribution can be expressed by the formula:

[0069]

[0070] This indicates constructing a diagonal matrix using the elements of a given vector as its diagonal elements. Initialize sparse vectors. for ,in This indicates that the element is 1 and the dimension is 1. The vector.

[0071] S130, construct the joint back diffusion process, and initialize the complex amplitude vector and interference sample data as the initial sampling starting point.

[0072] The joint backdiffusion process is constructed and modeled using stochastic differential equations, which can be expressed as:

[0073] ;

[0074] in, Represents the gradient. It is the joint posterior distribution of amplitude and interference. It is the reverse of the standard Wiener process.

[0075] Initialize the complex amplitude vector sample based on the transition probability of the diffusion process. and interference samples As the sampling starting point of the reverse diffusion chain, it can be expressed by the formula:

[0076] , ;

[0077] in, Indicates the first A complex amplitude vector sample, Indicates the first One interference sample, Let be the transition probability parameter at time 1. The dimension is The identity matrix, The dimension is The identity matrix.

[0078] S140, using time as a node, executes the corrector:

[0079] For the sample at the current time, update the fractional function of the complex amplitude vector and the fractional function of the disturbance variable, and use the Langevin dynamics criterion to correct the updated sample to obtain the sample after sampling correction.

[0080] The fractional function for updating the complex magnitude vector based on the current sample can be expressed by the formula:

[0081]

[0082] in, For the first The score of a complex amplitude sample. This indicates the conjugate operation. express Time Complexity The variance, and .

[0083] The fractional function for updating the disturbance variable based on the current sample can be expressed by the following formula:

[0084]

[0085] in, For the first The score function of each interference sample.

[0086] Using the updated complex magnitude vector and the fractional function of the disturbance variable, the gradient of the log-joint posterior distribution with respect to the complex magnitude and the disturbance can be updated, which can be expressed by the formula:

[0087] ;

[0088] in, This represents the number of samples in the batch. Further, we have:

[0089] ; ;

[0090] , ;

[0091] Based on the gradient of the updated log-joint posterior distribution with respect to the complex amplitude and the disturbance, the Langevin kinetic criterion is applied to update each sample, which can be expressed by the formula:

[0092] ;

[0093] in, for The update step size at any moment, and It is Gaussian noise and satisfies , .

[0094] S150, use sparse Bayesian learning to update the prior variance of the complex magnitude vector in step 4, and update the score of the complex magnitude vector.

[0095] In this embodiment, it is assumed that all complex amplitude vector samples have the same prior distribution, which can be expressed by the formula:

[0096]

[0097] In sparse Bayesian learning The sparsity of complex amplitudes was controlled.

[0098] In this embodiment, the expectation-maximization method is used to update. It can be expressed by the formula:

[0099] ;

[0100] in, This is the updated variance parameter. Further, we have:

[0101] ;

[0102] in, This indicates the conjugate transpose. Represents the determinant of a matrix. Maximize Need to meet the requirements The gradient of is 0, which can be expressed by the formula:

[0103] ;

[0104] Among them, it means Partial derivatives. This leads to the optimal... It can be expressed by the formula:

[0105] ;

[0106] in, express variance express The mean. Further, we have:

[0107]

[0108]

[0109] in, This represents the result of taking the modulus after squaring each element of the vector.

[0110] Based on the updated prior variance parameters of the complex magnitude vector The fractional function for updating the complex magnitude vector can be expressed as:

[0111]

[0112] in, .

[0113] S160, Execute the predictor:

[0114] Based on the scores obtained from S150, the gradient of the joint posterior distribution with respect to the complex amplitude vector and the disturbance variable is calculated, and the gradient is used to generate reconstructed samples according to the backdiffusion stochastic differential equation.

[0115] The prior variance of complex amplitudes in the reconstructed samples is updated using sparse Bayesian learning.

[0116] First, the gradient of the log-joint posterior distribution with respect to the complex amplitude and the disturbance variable is updated using the score obtained from S150. This can be expressed by the formula:

[0117]

[0118] In this embodiment, when further updating the complex amplitude vector samples and interference samples, based on the gradients of the updated log-joint posterior distribution with respect to the complex amplitude and interference, the Euler-Maruyama numerical solution is applied, which can be expressed by the formula:

[0119] ;

[0120] in, For Sampling interval.

[0121] Based on the updated complex amplitude vector samples from the predictor output, sparse Bayesian learning is used to update the prior variance of the complex amplitude. The specific update method is the same as step S140.

[0122] S170, repeat S140, S150 and S160 until the diffusion terminates, to obtain the complex amplitude estimate corresponding to the sample.

[0123] In this embodiment, the diffusion time starts from Advance to Step size is and will Defined as the diffusion termination time. The estimated value of the output complex amplitude at the diffusion termination time is the average of all complex amplitude samples, which can be expressed by the formula:

[0124] ;

[0125] Based on the complex amplitude estimate The range cell and Doppler cell corresponding to the point with the highest amplitude magnitude are obtained. The corresponding range value and Doppler value are used as the estimated results of the effective target range and Doppler, and added to the effective target set.

[0126] S180, the complex amplitude estimate is input to the constant false alarm detector to determine whether it contains a valid target, and all samples determined to be valid targets are combined into a valid target set.

[0127] Remove guard cells near the target's distance and velocity units in the detection information, and calculate the average of the squared magnitudes of all data within the training unit range as an estimate of the noise floor. The threshold for effective targets is determined based on the noise floor level, and the detection volume is calculated. The ratio is compared with a threshold. If the ratio is greater than the threshold, the target is considered to be valid; otherwise, the target is considered to be invalid.

[0128] This embodiment also provides a multi-target detection device 500 under complex electromagnetic interference environment, used to perform the steps of the multi-target detection method for pulse Doppler radar under complex electromagnetic interference environment provided in the above embodiment, specifically including the following:

[0129] Interference score training unit 510, detection information acquisition unit 520, initialization unit 530, corrector unit 540, sparse Bayesian learning unit 550, predictor unit 560, amplitude parameter acquisition unit 570, target output unit 580:

[0130] Among them, the interference score training unit 530 is used to generate samples based on structured interference samples, using the forward perturbation process of the diffusion model, and to train the neural network of the interference score function using the denoising score matching criterion.

[0131] The detection information acquisition unit 520 is used to receive the echo signal emitted by the pulse Doppler radar under complex electromagnetic interference environment, straighten the echo signal to obtain detection information, and construct a complete dictionary matrix to model the detection information.

[0132] The initialization unit 530 is used to construct the joint back diffusion process and initialize the complex amplitude vector sample and interference sample as the sampling starting point of the back diffusion chain.

[0133] The corrector unit 540 is used to execute the corrector step at each time step. First, it updates the complex amplitude vector and the fractional function of the disturbance based on the sample at the current time step. Then, it updates each sample using the Langevin kinetic criterion to complete the sampling correction of the current sample state.

[0134] The sparse Bayesian learning unit 550 is used to update the prior variance of the complex amplitude based on the current sample using sparse Bayesian learning, so that the complex amplitude distribution is adapted to the current observation data, and to update the fractional function of the complex amplitude.

[0135] Predictor unit 560 is used to execute the predictor steps. First, the gradient of the log-joint posterior distribution with respect to the complex amplitude and interference is updated using the updated fractional function. Then, a further update is performed on the complex amplitude samples and interference samples, gradually recovering the original signal structure from the noise along the fractional guidance direction. Finally, the prior variance of the complex amplitude is updated using a sparse Bayesian learning unit based on the updated sample data.

[0136] The amplitude parameter acquisition unit 570 is used by the repetitive corrector unit, the sparse Bayesian learning unit, and the predictor unit to output a complex amplitude estimate until the diffusion termination time.

[0137] The target output unit 580 is used to input the amplitude estimate into the constant false alarm rate detector to determine whether a valid target is included. If a valid target is included, the distance, velocity, and amplitude values ​​corresponding to the valid target are output; otherwise, it is determined that the current detection information does not contain a valid target.

[0138] To better illustrate the technical effects of the solution provided in this embodiment, the following experimental examples are provided.

[0139] The radar parameters were set as follows in the experiment: carrier frequency =3GHz; pulse width Pulse repetition interval ;bandwidth =3MHz; Chirp rate Sampling frequency ; pulse count Then the length of a distance unit is A velocity element has a length of .

[0140] The actual target distances are set to 900, 1800, and 2700 (m), the target speeds are set to -50, 100, and 50 (m / s), and the cumulative signal-to-noise ratio is set to 15 dB. The interference signal is a linear frequency modulated signal with a signal-to-interference ratio of -15.9 dB.

[0141] Its target detection results are as follows Figure 3 As shown, in Figure 3 (a) shows the amplitude detection results using the scheme provided in this embodiment under strong electromagnetic interference scenarios. The data is as follows: the velocity grid is -250~250 m / s with an interval of 50 m / s, and the distance grid is 0~4500 m with an interval of 30 m. The corresponding amplitude estimation results include: Target 1: distance 900 m, velocity -50 m / s; Target 2: distance 2700 m, velocity 50 m / s; Target 3: distance 1800 m, velocity 100 m / s. This proves that the scheme provided in this embodiment can successfully classify interference and target signals and detect all targets.

[0142] like Figure 3 As shown in (b) in the figure, the detection results of the traditional pulse compression method and moving target detection method are as follows under strong electromagnetic interference: the target signal is completely submerged and the traditional method cannot detect the effective target.

[0143] Furthermore, the terms "upper," "lower," "inner," "outer," "front," and "rear" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention.

[0144] Of course, the above description is only a specific embodiment of the present invention and is not intended to limit the scope of the present invention. All equivalent changes or modifications made to the structure, features and principles described in the claims of the present invention should be included in the scope of the claims of the present invention.

[0145] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for multi-target detection under complex electromagnetic interference environments for pulse Doppler radar, characterized in that, Includes the following steps: Step 1: Obtain interference sample data to construct a fractional function neural model for generating interference variables; Step 2: Obtain the echo signal and, using a complete dictionary matrix, construct the complex amplitude vector corresponding to the echo signal. The construction process of the complex amplitude vector is as follows: By performing digital beamforming on the echo signal, detection information in both the fast and slow time dimensions can be obtained. Based on the detection information and the overcomplete dictionary matrix, construct the complex amplitude vector corresponding to the echo signal; The detection information is obtained by using two-dimensional data composed of the fast and slow times of straightening the target; Step 3: Construct a joint backdiffusion process and initialize the complex amplitude vector and interference sample data as the initial sampling starting point. The joint backdiffusion process is described by a stochastic differential equation, the expression of which is as follows: ;in, Represents the gradient. It is the joint posterior distribution of amplitude and interference. It is the reverse of the standard Wiener process. yes Each target corresponds to a complex amplitude vector of a different Doppler-distance grid; the expression for the initial sampling starting point is as follows: , ;in, Indicates the first A complex amplitude vector sample, Indicates the first One interference sample, Let be the transition probability parameter at time 1. The dimension is The identity matrix, The dimension is The identity matrix; Step 4: Using time as the node, execute the corrector: For the sample at the current time, update the complex amplitude vector and the fractional function of the disturbance variable respectively, and use the Langevin dynamics criterion to correct the updated sample to obtain the sample after sampling correction. Step 5: Update the prior variance of the complex magnitude vector in Step 4 using sparse Bayesian learning, and update the score of the complex magnitude vector. Step 6: Execute the predictor: Based on the scores obtained in step 5, the gradient of the joint posterior distribution with respect to the complex amplitude vector and the interference variable is calculated, and the gradient is used to generate reconstructed samples according to the backdiffusion stochastic differential equation. The prior variance of complex amplitudes in the reconstructed samples is updated using sparse Bayesian learning; Step 7: Repeat steps 4-6 until the diffusion terminates to obtain the complex amplitude estimate corresponding to the sample. Step 8: Input the complex amplitude estimate into the constant false alarm rate detector to construct an effective target set.

2. The method for multi-target detection under complex electromagnetic interference environment for pulse Doppler radar according to claim 1, characterized in that, The fractional function neural model was constructed using a diffusion model and a denoised fractional matching criterion.

3. The method for multi-target detection under complex electromagnetic interference environment for pulse Doppler radar according to claim 2, characterized in that, The expression for the denoising score matching criterion is as follows: ;in, For neural network parameters, Expressing expectations, express It is in the interval Obtained by uniform sampling in the middle, It was obtained from a set of interference samples. This represents the L2 norm.

4. The method for multi-target detection under complex electromagnetic interference environment for pulse Doppler radar according to claim 1, characterized in that, The overcomplete dictionary matrix is ​​constructed based on Doppler networks and distance networks.

5. The method for multi-target detection under complex electromagnetic interference environment for pulse Doppler radar according to claim 1, characterized in that, The complex amplitude estimate is obtained by calculating the average of all complex amplitude vectors output at the diffusion termination time.

6. A multi-target detection device under complex electromagnetic interference environment, characterized in that, The steps are for performing the multi-target detection method for pulse Doppler radar under complex electromagnetic interference environment as described in any one of claims 1 to 5.

Images