Radar multi-target distinguishing method based on sparse reconstruction and waveform optimization

Abstract

The invention discloses a radar multi-target distinguishing method based on sparse reconstruction and waveform optimization. The radar multi-target distinguishing method comprises the following steps: determining waveform characteristics of a transmitted signal and a receiving mode of an echo signal; transmitting a signal at the transmitting end based on the determined waveform characteristics; loading a radar multi-target resolution model; the radar multi-target resolution model is used for simulating spatial positions and motion directions of multiple targets, carrying out target spatial distribution modeling and realizing simulation of the spatial positions and the motion directions of the multiple targets; receiving the target echo signal at the receiving end, constructing a space structure of a target space by combining a radar multi-target resolution model, and performing target resolution; and according to the target resolution result, adjusting the bandwidth and the modulation mode of the next frame of transmitted waveform in combination with the adjustment strategy to determine the waveform, and iteratively executing signal transmission. According to the technical scheme, detection, tracking and accurate interception of multiple targets of an incoming attack cluster can be effectively realized, and the defense combat effectiveness of our army in a complex battlefield environment is enhanced.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and more specifically, to a radar multi-target resolution method based on sparse reconstruction and waveform optimization. Background Technology

[0002] In multi-target attack scenarios, the small distances between targets, complex motion trajectories, and noise interference can cause radar echo signals to overlap in the time, frequency, or spatial domains. This can lead the radar seeker to misidentify multiple targets as a single target, making it impossible to accurately separate the motion trajectories of each target. Furthermore, existing multi-target resolution methods for radar (such as time-frequency analysis and beamforming) are limited by the Rayleigh limit, resulting in limited resolution capabilities when targets are too close together. Secondly, traditional Fourier transform-based algorithms require high sampling rates, leading to high computational complexity and difficulty in real-time processing. Finally, in multi-target environments, sidelobe interference from strong scattering targets can increase the false alarm rate.

[0003] Therefore, in dense target environments, radar multi-target resolution not only needs to solve the problem of multi-target echo coupling, but also needs to be optimized in terms of computational complexity and resolution accuracy. Summary of the Invention

[0004] To achieve the above objectives, this application provides a radar multi-target resolution method based on sparse reconstruction and waveform optimization, comprising the following steps:

[0005] Determine the waveform characteristics of the transmitted signal and the method of receiving the echo signal;

[0006] Based on the determined waveform characteristics, a signal is transmitted at the transmitting end;

[0007] Load a radar multi-target resolution model; the radar multi-target resolution model is used to simulate the spatial position and motion direction of multiple targets; wherein, constructing the radar multi-target resolution model includes: constructing a dynamic dictionary matrix based on target motion priors, modeling the echo signal y as a sparse linear combination; updating the dynamic dictionary matrix; performing target position analysis based on the dynamic dictionary matrix to suppress false alarms; performing target spatial distribution modeling to realize the simulation of the spatial position and motion direction of multiple targets;

[0008] At the receiving end, target echo signals are received, and the spatial structure of the target space is constructed by combining the radar multi-target resolution model to perform target resolution.

[0009] Based on the target resolution result, the bandwidth and modulation method of the next frame's transmitted waveform are adjusted in conjunction with the adjustment strategy to determine the waveform, and the transmission signal at the transmitting end is iteratively executed.

[0010] The dynamic dictionary matrix based on the prior knowledge of the target motion is represented as: y = φs + n.

[0011] Where y is the echo signal, φ is the dynamic dictionary matrix, s is the sparse coefficient vector, and n is the noise;

[0012] Each column of the dynamic dictionary matrix φ corresponds to the echo model of a potential target, and the position and amplitude of the non-zero elements in s correspond to the presence and intensity of the target, respectively.

[0013] Updating the dynamic dictionary matrix refers to iteratively analyzing the unmatched signal components through the residual signal after sparse reconstruction, where the residual signal is r = y - φs; the update includes:

[0014] Calculate the correlation C between the current residual and all atoms. t (i); Select the atom index i that has the highest correlation between the current residual and all atoms. t =argmax|C t (i)|, add to index set Λ t =Λ t-1 ∪{i t};

[0015] Selected atoms are determined based on the atom index with the highest relevance. Optimize the distance-Doppler parameters and update the parameters of the corresponding atoms in the dynamic dictionary matrix φ;

[0016] The coefficients are updated by solving for the sparse coefficients corresponding to the current index set using the least squares method, as follows:

[0017]

[0018] Performing a residual update is represented as: residual r t =y-φ(Λ t )s t .

[0019] During the iterative process of updating the dynamic dictionary matrix, when ||r t The iteration terminates when ||2 < σ or t = K; where r t Let K be the residual, K be the sparsity, and σ be the noise level;

[0020] At this point, the dynamic dictionary matrix φ has been updated.

[0021] Target location analysis based on a dynamic dictionary matrix includes:

[0022] Multi-target resolution modeling is performed, represented as: Where z is the introduced auxiliary variable, s is the sparse coefficient vector, and Π c The velocity-distance continuity constraint function;

[0023] The optimization based on the multi-objective resolution model includes: decoupling by alternately optimizing the sparse coefficient vectors s and z; solving the problem by least squares after fixing z and s to achieve sparsity recovery; achieving sparsity by soft thresholding and shrinking s and z to achieve constrained projection; and adjusting the Lagrange multipliers to balance sparsity and constraints to achieve multiplier update.

[0024] A Markov random field model is introduced to model the spatial distribution of the target.

[0025] Furthermore, the spatial structure of the target space constructed by combining the radar multi-target resolution model includes:

[0026] The receiving end performs low-noise amplification, down-conversion, and sampling on the target echo signal. After pulse compression and clutter suppression processing on the sampled digital signal, it calls the radar multi-target resolution model to generate the spatial structure of the target space. The generation of the spatial structure of the target space includes: dynamic dictionary matrix initialization, iterative execution of sparse reconstruction and dictionary update to generate a spatial structure of the target space that matches the echo signal.

[0027] Among them, dynamic dictionary matrix initialization includes parameter space discretization, atom generation, and motion model expansion;

[0028] The parameter space discretization refers to discretizing the range-Doppler parameter space of the target information into a range grid and a velocity network according to the resolution requirements of the radar system; the range grid is represented as: The velocity grid is represented as: Where B is the signal bandwidth, T c λ is the coherent processing time, λ is the wavelength, and c is the speed of light.

[0029] When the atoms are generated, each dictionary atom φ i,j Corresponding to a hypothetical objective in (R) i v j The echo model at point () is expressed as: in, For time delay, The Doppler frequency shift is represented by rect(·), and the pulse envelope function is represented by rect(·).

[0030] If the target has acceleration a k Then a three-parameter grid (R) is introduced. i v j a k ), perform dictionary atomic expansion, represented as:

[0031] The results of target discrimination include: the distance difference between the two nearest targets, the maximum velocity difference of the target group, the dispersion of the target velocity distribution, the maximum acceleration of the target group, the concentration of the target group's motion direction, the proportion of ground / meteorological interference in the echo, and the frequency band location of external electromagnetic interference.

[0032] Further adjustments to the strategy include:

[0033] The minimum target spacing d estimated based on the current frame min Adjust the signal bandwidth B of the next frame. new =α·c / (2d) min ), where α is the oversampling factor;

[0034] If a large Doppler spread is detected in the target, the number of OFDM subcarriers is increased to improve velocity resolution.

[0035] Furthermore, the transmitting end adopts a phased array radar array and is equipped with an FPGA to realize real-time waveform modulation; the transmitted signal is a multi-carrier orthogonal frequency division multiplexing waveform, and the waveform characteristics include bandwidth, waveform, pulse repetition frequency and spectrum;

[0036] The receiving end uses a multi-channel ADC for signal sampling.

[0037] The radar multi-target discrimination method provided in this invention overcomes the Rayleigh limit through sparse reconstruction and dynamic dictionary learning, achieving effective discrimination even when target spacing is small. Secondly, dynamic waveform optimization suppresses time-frequency aliasing of multi-target echoes, reduces false alarm interference, and improves the measurement and tracking accuracy of the radar seeker, particularly enhancing detection accuracy in dense target environments. This invention can effectively detect, track, and accurately intercept multiple incoming targets in a cluster, enhancing the PLA's defensive combat effectiveness in complex battlefield environments. It is applicable to fields such as intelligent acceleration and military reconnaissance. Attached Figure Description

[0038] Figure 1 This is a step diagram of a radar multi-target resolution method provided according to an embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of the signal processing process of the radar multi-target resolution method provided in the embodiment of the present invention. Detailed Implementation

[0040] This invention provides a radar multi-target resolution method that combines adaptive waveform optimization and dynamic sparse dictionary learning. By using sparse reconstruction and dynamic dictionary learning, it breaks through the Rayleigh limit and achieves target super-resolution. During the resolution process, dynamic waveform optimization is used to suppress time-frequency aliasing of multi-target echoes and reduce false alarm interference, thereby improving the detection accuracy in dense target environments.

[0041] The specific implementation of the present invention will now be described in detail with reference to the accompanying drawings.

[0042] The steps of the radar multi-target resolution method provided by this invention are as follows: Figure 1 As shown, it includes:

[0043] Step S100: Determine the waveform characteristics of the transmitted signal and the reception method of the echo signal:

[0044] First, after determining the transmission rules of the radar signal, the waveform characteristics of the transmitted signal are determined:

[0045] Specifically, the transmitting end uses a phased array radar array and is equipped with an FPGA to realize real-time waveform adjustment; the transmitted signal is a multi-carrier orthogonal frequency division multiplexing waveform, and the waveform characteristics of the transmitted signal can be adjusted according to requirements. The waveform characteristics include information such as bandwidth, waveform, pulse repetition frequency and spectrum.

[0046] Secondly, it is necessary to determine the reception method of the echo signal corresponding to the transmission rules: the receiver uses a multi-channel ADC to sample the signal, receives the target echo signal y, and performs adaptive filtering to suppress noise.

[0047] Step S110: Based on the determined waveform characteristics of the transmitted signal, transmit the signal at the transmitting end;

[0048] For example, the transmitted signal is an OFDM waveform, and the subcarrier spacing is adaptively adjusted according to the target density.

[0049] Step S120: Load the radar multi-target resolution model; the radar multi-target resolution model is used to simulate the spatial position and direction of motion of multiple targets;

[0050] Before loading the radar multi-target resolution model, the radar multi-target resolution model is constructed, including the following steps:

[0051] Step S121: Construct a dynamic dictionary matrix based on the prior knowledge of target motion, and combine it with compressed sensing theory to model the echo signal y as a sparse linear combination; expressed as: y=φs+n(1)

[0052] Where φ is the dynamic dictionary matrix, s is the sparse coefficient vector, and n is noise.

[0053] Each column (called an "atom") of the dynamic dictionary matrix φ corresponds to an echo model of a potential target, and the position and amplitude of the non-zero elements in s correspond to the presence and intensity of the target, respectively.

[0054] Step S122: The dynamic dictionary matrix φ is updated iteratively using an improved OMP (Orthogonal Matching Pursuit) algorithm, so that the dynamic dictionary matrix can adaptively match the target scattering characteristics.

[0055] In each iteration, the unmatched signal components are analyzed using the residual signal r = y - φs after sparse reconstruction; the analysis process includes:

[0056] 1) Atom selection: Calculate the correlation C between the current residual and all atoms. t :

[0057] C t =φ H r t -1 (2)

[0058] Where t is the iteration number and φ is the dynamic dictionary.

[0059] The atom index i with the highest correlation between the pre-selection residual and all atoms is selected. t =argmax|C t (i)|, and add to the index set Λ t =Λ t-1 ∪{i t}

[0060] 2) Parameter fine-tuning: Selecting atoms based on the atom index with the highest correlation. The distance-Doppler parameters are optimized using Newton's method, and the parameters of the corresponding atoms in the dictionary φ are updated accordingly:

[0061]

[0062] Where y is the observed signal.

[0063] 3) Coefficient Update: The sparse coefficients corresponding to the current index set are solved using the least squares method, expressed as:

[0064]

[0065] 4) Residual update, represented as: r t =y-φ(Λ t )s t (5)

[0066] During the iteration process, when ||r t When ||2<σ(noise level) or t=K(K is sparsity), the iteration terminates and the dynamic dictionary matrix φ is updated.

[0067] Step S123: Combine the multi-target parameters analyzed from the echo signal and perform target position analysis based on the dynamic dictionary matrix φ:

[0068] The information obtained after analyzing the echo signal constitutes multi-target parameters, which may include the position, velocity, acceleration of each target, interference signal, and the frequency band location of the interference.

[0069] 1) The target's position, velocity, and scattering intensity parameters are jointly optimized using the Alternating Direction Multiplier Method (ADMM);

[0070] In this step, multi-object resolution modeling is performed to achieve a joint optimization problem of sparse recovery and spatial constraints, expressed as:

[0071] Where z is an introduced auxiliary variable, Π c The velocity-distance continuity constraint function;

[0072] The optimization process based on the multi-objective resolution model includes: decoupling by alternately optimizing the sparse coefficient vectors s and z; then fixing z and s and solving them using the least squares method to achieve sparsity recovery; achieving sparsity by fixing s and z and shrinking them through soft thresholding, i.e., constrained projection; and finally adjusting the Lagrange multipliers to balance sparsity and constraints to achieve multiplier update.

[0073] 2) Introduce a Markov random field (MRF) model to model the spatial distribution of targets, assign higher joint probabilities to neighboring targets, and punish isolated false alarms to suppress false alarms;

[0074] In practical applications, especially at low SNR, false alarms are prone to occur. To address this issue, Markov random fields are used to model the radar observation scene as a spatially correlated grid, with each network cell establishing a correspondence with the target. The MRF algorithm is used to analyze neighborhood information to enhance the correlation of weak targets, determine whether the state of each grid cell is affected by its neighboring cells, and determine whether noise or interference is mistakenly identified as an isolated target.

[0075] Specifically, a Markov random field (MRF) defines the prior probability of the target distribution as follows:

[0076]

[0077] Where i to j represent adjacent resolution units, and β is the spatial smoothing factor.

[0078] This completes the modeling of the spatial distribution of the target and enables the simulation of the spatial position and motion direction of multiple targets.

[0079] Step S130: Receive the target echo signal y at the receiving end, construct the spatial structure of the target space in combination with the radar multi-target resolution model, and perform target resolution;

[0080] like Figure 2 As shown in step S200, the receiver performs low-noise amplification, down-conversion and sampling on the target echo signal, and further performs pulse compression and clutter suppression on the sampled digital signal. Then, it calls the radar multi-target resolution model and executes step S210 to generate the spatial structure of the target space.

[0081] Specifically, the following steps are included:

[0082] 1) Target prior dynamic dictionary matrix initialization: At this point, an initial set of atoms is generated by combining the prior information of the target echo signal and the radar multi-target resolution model, and the echo signal is modeled as a sparse linear combination; the implementation process includes: parameter space discretization, atom generation and motion model expansion;

[0083] Parameter space discretization means: obtaining multi-target parameters according to the resolution requirements of the radar system. The resolution requirements include conditions such as maximum detection range and velocity range. The target parameters (such as range-Doppler parameters) obtained after parsing the target information are spatially discretized into range grids and velocity networks.

[0084] The distance grid is represented as:

[0085] The velocity grid is represented as:

[0086] Where B is the signal bandwidth, T c λ is the coherent processing time, λ is the wavelength, and c is the speed of light.

[0087] During atom generation, each dictionary atom φ i,j Corresponding to a hypothetical objective in (R) i v j The echo model at point () is expressed as:

[0088] in, For time delay, denoted as Doppler frequency shift, and rect(·) is the pulse envelope function.

[0089] If the target parameter contains acceleration 'a' k Then a three-parameter grid (R) is introduced. i v j a k ), perform dictionary atomic expansion, represented as:

[0090] Furthermore, the echo signal is modeled based on sparse linear combination, i.e.: y=φs+n(12)

[0091] Where φ is the dynamic dictionary matrix, s is the sparse coefficient vector, and n is noise.

[0092] 2) Iteratively perform sparse reconstruction and dictionary update, execute the improved OMP (Orthogonal Matching Pursuit) algorithm, and combine the dynamic dictionary to reconstruct the signal and match the target scattering characteristics.

[0093] In this step, the atomic selection, parameter fine-tuning, coefficient update and residual update in step S122 are performed to generate the spatial structure of the target space that matches the echo signal; the target position analysis based on step S123 is performed to solve the problem of multi-target coupling in the spatial structure of the target space, so as to realize target discrimination through the spatial structure of the target space.

[0094] The results of target resolution include: the distance difference between the two nearest targets, the maximum velocity difference of the target group, the dispersion of the target velocity distribution, the maximum acceleration of the target group, the concentration of the target group's motion direction, the proportion of ground / meteorological interference in the echo, and the frequency band location of external electromagnetic interference.

[0095] like Figure 2 As shown in step S220, waveform adjustments can be made based on waveform optimization feedback: if the distance difference between the two nearest targets is too small, the bandwidth of the transmitted signal needs to be adjusted; if the velocity / acceleration of the target group is large, waveform modulation optimization is required; if the dispersion is large, it will increase velocity ambiguity, so the pulse repetition frequency (PRF) needs to be adjusted; if there is interference frequency band, the spectrum resources need to be reallocated.

[0096] Step S140: Based on the target resolution result, adjust the bandwidth and modulation mode of the next frame transmitted waveform, execute step S110, and transmit the waveform at the transmitting end.

[0097] The adjustment strategy is as follows:

[0098] 1) Resolution feedback: The minimum target spacing d estimated based on the current frame. min Adjust the signal bandwidth B of the next frame. new =α·c / (2d) min ), where α is the oversampling factor;

[0099] 2) Modulation and coding optimization: If a large Doppler spread of the target is detected, the number of OFDM subcarriers is increased to improve the speed resolution.

[0100] The radar multi-target resolution method provided by this invention utilizes dynamic dictionary learning to avoid the "basis mismatch" problem caused by fixed networks and improve parameter estimation accuracy. Secondly, through waveform adaptive adjustment, the system resolution dynamically changes with the environment, achieving closed-loop optimization. Finally, the OMP algorithm reduces invalid atom searches and lowers complexity through dynamic updates.

[0101] The above-disclosed embodiments are merely a few specific examples of the present invention. However, the present invention is not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A radar multi-target resolution method based on sparse reconstruction and waveform optimization, characterized in that, Includes the following steps: Determine the waveform characteristics of the transmitted signal and the method of receiving the echo signal; Based on the determined waveform characteristics, a signal is transmitted at the transmitting end; Load the radar multi-target resolution model; The radar multi-target resolution model is used to simulate the spatial position and motion direction of multiple targets. The construction of the radar multi-target resolution model includes: constructing a dynamic dictionary matrix based on prior knowledge of target motion, modeling the echo signal y as a sparse linear combination; updating the dynamic dictionary matrix; performing target position analysis based on the dynamic dictionary matrix to suppress false alarms; and modeling the spatial distribution of targets to simulate the spatial position and motion direction of multiple targets. The receiving end receives the target echo signal, and combines it with the radar multi-target resolution model to construct the spatial structure of the target space for target resolution. Based on the target resolution result, the bandwidth and modulation method of the next frame's transmitted waveform are adjusted in conjunction with the adjustment strategy to determine the waveform, and the transmission of the signal at the transmitting end is iteratively executed.

2. The radar multi-target resolution method according to claim 1, characterized in that, The constructed dynamic dictionary matrix based on the prior knowledge of the target motion is represented as: y = φs + n. Where y is the echo signal, φ is the dynamic dictionary matrix, s is the sparse coefficient vector, and n is the noise; Each column of the dynamic dictionary matrix φ corresponds to the echo model of a potential target, and the position and amplitude of the non-zero elements in s correspond to the presence and intensity of the target, respectively.

3. The radar multi-target resolution method according to claim 2, characterized in that, The updating of the dynamic dictionary matrix refers to iteratively analyzing the unmatched signal components through the residual signal after sparse reconstruction, wherein the residual signal is r = y - φs; The update includes: Calculate the correlation C between the current residual and all atoms. t (i); Select the atom index i that has the highest correlation between the current residual and all atoms. t =argmax|C t (i)|, add to index set Λ t =Λ t-1 ∪{i t }; The selected atom φ is determined based on the atom index with the highest relevance. it Optimize the distance-Doppler parameters and update the parameters of the corresponding atoms in the dynamic dictionary matrix φ; The coefficients are updated by solving for the sparse coefficients corresponding to the current index set using the least squares method, as follows: Performing a residual update is represented as: r t =y-φ(Λ t )s t .

4. The radar multi-target resolution method according to claim 3, characterized in that, During the iterative process of updating the dynamic dictionary matrix, when ||r t The iteration terminates when ||2 < σ or t = K; where r t Let K be the residual, K be the sparsity, and σ be the noise level; At this point, the dynamic dictionary matrix φ has been updated.

5. The radar multi-target resolution method according to claim 3, characterized in that, The target location analysis based on the dynamic dictionary matrix includes: Multi-target resolution modeling is performed, represented as: Where z is the introduced auxiliary variable, s is the sparse coefficient vector, and Π c The velocity-distance continuity constraint function; The optimization based on the multi-objective resolution model includes: decoupling by alternately optimizing the sparse coefficient vectors s and z; solving the problem by least squares after fixing z and s to achieve sparsity recovery; achieving sparsity by soft thresholding and shrinking s and z to achieve constrained projection; and adjusting the Lagrange multipliers to balance sparsity and constraints to achieve multiplier update. A Markov random field model is introduced to model the spatial distribution of the target.

6. The radar multi-target resolution method according to claim 1, characterized in that, The spatial structure of the target space constructed by combining the radar multi-target resolution model includes: The receiving end performs low-noise amplification, down-conversion, and sampling on the target echo signal. After pulse compression and clutter suppression processing on the sampled digital signal, it calls the radar multi-target resolution model to generate the spatial structure of the target space. The generation of the spatial structure of the target space includes: dynamic dictionary matrix initialization, iterative execution of sparse reconstruction and dictionary update to generate a spatial structure of the target space that matches the echo signal.

7. The radar multi-target resolution method according to claim 6, characterized in that, The dynamic dictionary matrix initialization includes parameter space discretization, atom generation, and motion model expansion. The parameter space discretization refers to discretizing the range-Doppler parameter space of the target information into a range grid and a velocity network according to the resolution requirements of the radar system; the range grid is represented as: The velocity grid is represented as: Where B is the signal bandwidth, T c λ is the coherent processing time, λ is the wavelength, and c is the speed of light. When the atoms are generated, each dictionary atom φ i,j Corresponding to a hypothetical objective in (R) i v j The echo model at point () is expressed as: in, For time delay, The Doppler frequency shift is represented by rect(·), and the pulse envelope function is represented by rect(·). If the target has acceleration a k Then a three-parameter grid (R) is introduced. i v j a k ), perform dictionary atomic expansion, represented as:

8. The radar multi-target resolution method according to claim 1, characterized in that, The results of target discrimination include: the distance difference between the two nearest targets, the maximum velocity difference of the target group, the dispersion of the target velocity distribution, the maximum acceleration of the target group, the concentration of the target group's motion direction, the proportion of ground / meteorological interference in the echo, and the frequency band location of external electromagnetic interference.

9. The radar multi-target resolution method according to claim 1, characterized in that, The adjustment strategy includes: The minimum target spacing d estimated based on the current frame min Adjust the signal bandwidth B of the next frame. new =α·c / (2d) min ), where α is the oversampling factor; If a large Doppler spread of the target is detected, the number of OFDM subcarriers is increased to improve velocity resolution.

10. The radar multi-target resolution method according to claim 1, characterized in that, The transmitting end adopts a phased array radar array and is equipped with an FPGA to realize real-time waveform modulation; the transmitted signal is a multi-carrier orthogonal frequency division multiplexing waveform, and the waveform characteristics include bandwidth, waveform, pulse repetition frequency and spectrum; The receiving end uses a multi-channel ADC for signal sampling.

Images