A signal acquisition and processing method and system based on a multifunction radar

Abstract

The application discloses a signal collection processing method and system based on a multifunctional radar, relates to the technical field of signal collection processing, and comprises the following steps: optimizing radar emission waveform parameters by adopting a quantum genetic algorithm, wherein the parameters include a center frequency, a bandwidth and a frequency modulation slope, and a nonlinear frequency modulation waveform is generated by FPGA hardware; receiving target echo signals by adopting a non-uniform sparse array, synchronously implementing time domain random interval sampling and frequency domain pseudo-random frequency point selection, and forming time-space-frequency three-dimensional compressed sensing observation data; performing three-linear tensor joint sparse reconstruction on the time-space-frequency three-dimensional compressed sensing observation data, decomposing polarization scattering matrix eigenvalues from the reconstructed signals, calculating eigenvalue entropy and micro-Doppler frequency coupling characteristic quantities; inputting the coupling characteristic quantities into a deep reinforcement learning model, and dynamically outputting a constant false alarm detection threshold, a moving target display filter order and resource allocation weights; and performing resource allocation based on the constant false alarm detection threshold.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of signal acquisition and processing, in particular to a signal acquisition and processing method and system based on a multifunctional radar. BACKGROUND

[0002] The technical field of signal acquisition and processing involves a series of methods and technologies for efficient acquisition, compression, transmission, reconstruction and analysis of raw signals from sensors using advanced mathematical algorithms and hardware platforms. Therefore, how to use advanced technical means to improve the intelligent level and security of signal acquisition and processing has become one of the problems to be solved at present.

[0003] In the field of signal acquisition and processing, the traditional radar waveform is mostly fixed or preset, which is difficult to dynamically adjust according to the target characteristics and environmental changes. The optimization of waveform parameters relies on manual trial and error or local search algorithm, which has the problems of slow convergence and easy to fall into local optimum, resulting in insufficient resolution and anti-interference ability in complex scenes. In the strong ground object, sea clutter or dense interference environment, the suppression ability of the traditional moving target display filter is limited, and the fixed parameters are difficult to adapt to the changing scene, resulting in high false alarm rate and serious weak target missing detection. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a signal acquisition and processing method based on a multifunctional radar to solve the problem that the traditional radar waveform is mostly fixed or preset, which is difficult to dynamically adjust according to the target characteristics and environmental changes. The optimization of waveform parameters relies on manual trial and error or local search algorithm, which has the problems of slow convergence and easy to fall into local optimum, resulting in insufficient resolution and anti-interference ability in complex scenes. In the strong ground object, sea clutter or dense interference environment, the suppression ability of the traditional moving target display filter is limited, and the fixed parameters are difficult to adapt to the changing scene, resulting in high false alarm rate and serious weak target missing detection.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a signal acquisition and processing method based on a multifunctional radar, which comprises:

[0008] The quantum genetic algorithm is used to optimize the radar transmission waveform parameters, including center frequency, bandwidth and frequency modulation slope, and a nonlinear frequency modulation waveform is generated by FPGA hardware;

[0009] The non-uniform sparse array is used to receive target echo signals, and time domain random interval sampling and frequency domain pseudo-random frequency point selection are implemented synchronously to form time-space-frequency three-dimensional compressed sensing observation data;

[0010] The time-space-frequency three-dimensional compressed sensing observation data is subjected to trilinear tensor joint sparse reconstruction, the polarization scattering matrix eigenvalue is decomposed from the reconstructed signal, and the coupling characteristic quantity of the eigenvalue entropy and the micro-Doppler frequency is calculated;

[0011] The coupling characteristic quantity is input into a deep reinforcement learning model, and a constant false alarm detection threshold, a moving target display filter order and a resource allocation weight are dynamically output;

[0012] The reconstructed signal is subjected to null space projection filtering based on the constant false alarm detection threshold, the moving target display filter order and the resource allocation weight, and an improved joint probability data association algorithm is used to generate target trajectory information;

[0013] The target trajectory information and the environmental state are recorded and fed back to the step of optimizing the radar transmission waveform, and the fitness function weight of the waveform optimization algorithm is adjusted.

[0014] As a preferred scheme of the signal acquisition and processing method based on the multifunctional radar, the quantum genetic algorithm is used to optimize the radar transmission waveform parameters, wherein the parameters include the center frequency, the bandwidth and the frequency modulation slope, and the non-linear frequency modulation waveform is generated by the FPGA hardware, and specifically includes:

[0015] The optimization of the waveform parameters is performed under the quantum genetic algorithm framework, and the value range of the center frequency, the bandwidth and the frequency modulation slope is respectively mapped to 、 and to constitute a three-dimensional search space;

[0016] Each individual is represented by a quantum bit code, and the quantum state is , wherein and represent the probability amplitude of the bit being 0 or 1 respectively;

[0017] A population of individuals is initialized, and each individual is decoded into actual waveform parameters after being measured and collapsed into a classical binary string;

[0018] The waveform is transmitted by a radar simulation system, the echo is received, and the distance autocorrelation function is calculated, and the maximum sidelobe level , the integrated sidelobe level and the peak-to-average power ratio of the signal in the region outside the main lobe are extracted;

[0019] The fitness function is used to comprehensively evaluate the waveform performance , and the expression is:

[0020] ;

[0021] , wherein , , The weight coefficient can be set according to the task type;

[0022] The population is updated according to the fitness value, the quantum state is adjusted using a quantum rotation gate, and the direction is determined by looking up a table according to the comparison result of the current individual and the global optimum;

[0023] After multiple generations of optimization, the optimal waveform parameter combination is obtained , which is used to generate a nonlinear frequency modulation signal in an FPGA, and the time domain expression is:

[0024] ;

[0025] wherein, is a nonlinear coefficient determined by the optimization process, is the pulse width.

[0026] As a preferred scheme of the signal acquisition and processing method based on a multifunctional radar, the non-uniform sparse array is used to receive target echo signals, time-domain random interval sampling and frequency-domain pseudo-random frequency point selection are synchronously implemented, time-space-frequency three-dimensional compressed sensing observation data are formed, and the method specifically comprises the following steps:

[0027] The receiving end adopts a non-uniform sparse array structure, and the number of array elements is , and the positions of the array elements are determined by non-periodic intervals;

[0028] Supposing that the working wavelength of the radar is , the position of the mth array element is , wherein, , m is a positive integer different from each other, and is used to break the periodicity of the array and suppress spatial grating lobes; The analog echo signal received by each array element is first subjected to time-domain random sampling;

[0029] The number of sampling points is

[0030] , which is much smaller than the Nyquist sampling point number , wherein the sampling time is controlled by a pseudo-random sequence to form a reduced dimension time sequence ;

[0031] Each is subjected to short-time Fourier transform (STFT) to obtain a frequency spectrum matrix , and pseudo-random frequency point selection is performed in the frequency domain;

[0032] Randomly select points from frequency points, and the selection sequence is determined after being generated by a Logistic chaotic mapping and normalized, to obtain compressed frequency domain data​ ;

[0033] Stacking all the array elements dimension data by spatial position, constructing a three-dimensional data body with the dimensions of array element number , time domain sampling points and frequency domain sampling points , which is the time-space-frequency three-dimensional compressed sensing observation data.

[0034] As a preferred scheme of the signal acquisition and processing method based on the multifunctional radar, wherein: the three-linear tensor joint sparse reconstruction of the time-space-frequency three-dimensional compressed sensing observation data is performed, the eigenvalue of the polarization scattering matrix is decomposed from the reconstructed signal, and the coupling characteristic quantity of the eigenvalue entropy and the micro-Doppler frequency is calculated, which specifically includes:

[0035] In order to recover the complete signal from the compressed observation data , a three-linear tensor sparse reconstruction model is adopted, and the real signal tensor can be decomposed as:

[0036] ;

[0037] Wherein, is a spatial mode vector, is a time mode vector, is a frequency mode vector, is a tensor rank, represents an outer product;

[0038] The observation process satisfies , wherein, is a joint observation operator composed of an array structure, a time domain sampling matrix and a frequency domain selection matrix, is noise;

[0039] In order to solve , an optimization problem containing a sparse regularization term is constructed, and the expression is:

[0040] ;

[0041] Wherein, is a row sparse norm, is a regularization parameter, and an alternating direction multiplier method ADMM is used for iteration to obtain the reconstructed tensor;

[0042] The complex signals of the HH, HV, VH and VV polarization channels are extracted from , and the polarization scattering matrix of the target unit is constructed;

[0043] Eigenvalue decomposition is performed on , and two non-negative eigenvalues are obtained and , and ≥ , and the normalized probability distribution is obtained;

[0044] The eigenvalue entropy is calculated, and the expression is as follows:

[0045] ;

[0046] Meanwhile, the time pattern vector is subjected to spectrum analysis, periodic frequency modulation caused by the micro-motion component is extracted, and a micro-Doppler frequency is obtained.

[0047] The eigenvalue entropy is multiplied by the micro-Doppler frequency to obtain a coupling feature quantity.

[0048] As a preferred scheme of the signal acquisition and processing method based on the multifunctional radar, the coupling feature quantity is input into a deep reinforcement learning model, and a constant false alarm detection threshold, a moving target indication filter order and a resource allocation weight are dynamically output, and the specific process includes the following steps.

[0049] The coupling feature quantity calculated is combined with noise power estimation and target quantity estimation of the current environment to form a state vector, which is input into a deep Q network model DQN.

[0050] The action space of the deep Q network model includes three adjustable parameters: a constant false alarm detection threshold, a moving target indication filter order and a resource allocation weight vector.

[0051] The deep Q network model obtains an optimal strategy through training, and a reward function is designed as follows.

[0052] ;

[0053] Wherein, is a detection probability, is a false alarm probability, is a resource utilization rate, , , is a weight coefficient.

[0054] As a preferred scheme of the signal acquisition and processing method based on the multifunctional radar, the reconstructed signal is subjected to zero space projection filtering based on the constant false alarm detection threshold, the moving target indication filter order and the resource allocation weight, and an improved joint probability data association algorithm is used to generate target trajectory information, and the specific process includes the following steps.

[0055] The time sequence of the reconstructed signal is subjected to moving target indication filtering by using the MTI filter order output in the foregoing, wherein the filter structure is a stage canceller, and the transfer function thereof is as follows:

[0056] ;

[0057] The null-space projection method is employed. By analyzing the background region signal, the clutter covariance matrix is ​​estimated, and eigenvalue decomposition is performed. The eigenvectors corresponding to the smallest eigenvalues ​​are used to form the null-space basis, and the projection operator is constructed. The expression is as follows:

[0058] ;

[0059] The signal after MTI filtering is projected to obtain a further purified signal;

[0060] Target detection is performed on the range-Doppler graph using a constant false alarm threshold (CFAT). The detection logic is that if the energy of a given cell is higher than the average energy of its neighborhood, the target is detected. If the value is multiple times higher, it is considered a valid target;

[0061] For the detected measurements, an improved joint probabilistic data association algorithm is used to perform track association, the expression of which is:

[0062] ;

[0063] in, For the first The covariance matrix of Gaussian components is used to calculate the correlation probability based on this likelihood, and the track state is updated in a weighted manner to finally generate the target trajectory information.

[0064] As a preferred embodiment of the signal acquisition and processing method based on multi-functional radar described in this invention, the step of recording and feeding back the target trajectory information and environmental state to optimize the radar transmission waveform, and adjusting the fitness function weights of the waveform optimization algorithm, specifically includes:

[0065] The target trajectory information generated above is statistically analyzed, and the velocity sequence of each trajectory is extracted. Calculate the variance of the velocity change:

[0066] ;

[0067] in, As an indicator of trajectory stability;

[0068] Simultaneously acquire environmental conditions, including signal-to-noise ratio, clutter intensity, and target density;

[0069] according to Determine the scene type based on target density;

[0070] like and They believed that the targets were densely packed and moving stably, requiring improved discrimination capabilities.

[0071] If If If

[0072] Accordingly, the fitness function weight in waveform optimization is adjusted. Accordingly, the fitness function weight in waveform optimization is adjusted.

[0073] In a second aspect, the present application provides a signal acquisition and processing system based on a multifunctional radar, comprising: In a second aspect, the present application provides a signal acquisition and processing system based on a multifunctional radar, comprising:

[0074] A waveform optimization module, a compressed sensing module, a tensor reconstruction module, an intelligent decision module, a signal filtering module and a trajectory generation module. A waveform optimization module, a compressed sensing module, a tensor reconstruction module, an intelligent decision module, a signal filtering module and a trajectory generation module.

[0075] The waveform optimization module is configured to optimize the center frequency, bandwidth and frequency modulation slope of the radar transmitting waveform by using a quantum genetic algorithm, and generate a nonlinear frequency modulation signal through FPGA hardware. The waveform optimization module is configured to optimize the center frequency, bandwidth and frequency modulation slope of the radar transmitting waveform by using a quantum genetic algorithm, and generate a nonlinear frequency modulation signal through FPGA hardware.

[0076] The compressed sensing module is configured to implement time-domain random interval sampling and frequency-domain pseudo-random frequency point selection on the target echo signal synchronously under the condition of a non-uniform sparse array to form time-space-frequency three-dimensional compressed sensing observation data. The compressed sensing module is configured to implement time-domain random interval sampling and frequency-domain pseudo-random frequency point selection on the target echo signal synchronously under the condition of a non-uniform sparse array to form time-space-frequency three-dimensional compressed sensing observation data.

[0077] The tensor reconstruction module is configured to perform three-linear tensor joint sparse reconstruction on the three-dimensional compressed sensing observation data, extract the eigenvalue of the polarization scattering matrix from the reconstructed signal, calculate the eigenvalue entropy, and generate coupled characteristic quantities in combination with the micro-Doppler frequency. The tensor reconstruction module is configured to perform three-linear tensor joint sparse reconstruction on the three-dimensional compressed sensing observation data, extract the eigenvalue of the polarization scattering matrix from the reconstructed signal, calculate the eigenvalue entropy, and generate coupled characteristic quantities in combination with the micro-Doppler frequency.

[0078] The intelligent decision module is configured to input the coupled characteristic quantities into a deep reinforcement learning model, and dynamically output a constant false alarm detection threshold, a moving target indication filter order and a resource allocation weight in combination with the environment state. The intelligent decision module is configured to input the coupled characteristic quantities into a deep reinforcement learning model, and dynamically output a constant false alarm detection threshold, a moving target indication filter order and a resource allocation weight in combination with the environment state.

[0079] The signal filtering module is configured to sequentially perform moving target indication filtering and null space projection filtering on the reconstructed signal based on the parameters output by the intelligent decision module to suppress clutter interference. The signal filtering module is configured to sequentially perform moving target indication filtering and null space projection filtering on the reconstructed signal based on the parameters output by the intelligent decision module to suppress clutter interference.

[0080] The trajectory generation module is configured to perform constant false alarm detection based on the filtered signal, complete the association of measurement and track by using an improved joint probabilistic data association algorithm, generate target trajectory information, and feed back the trajectory and the environment state to the waveform optimization module to realize dynamic adjustment of the fitness function weight. The trajectory generation module is configured to perform constant false alarm detection based on the filtered signal, complete the association of measurement and track by using an improved joint probabilistic data association algorithm, generate target trajectory information, and feed back the trajectory and the environment state to the waveform optimization module to realize dynamic adjustment of the fitness function weight.

[0081] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any step of the signal acquisition and processing method based on a multifunctional radar according to the first aspect of the present application is realized. In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any step of the signal acquisition and processing method based on a multifunctional radar according to the first aspect of the present application is realized.

[0082] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any step of the signal acquisition and processing method based on a multifunctional radar according to the first aspect of the present application is realized.​​​​​​​​​​​​In a fourth aspect, the present application provides a computer readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements any step of the signal acquisition and processing method based on a multifunctional radar according to the first aspect of the present application.

[0083] The present application has the following beneficial effects: the radar waveform parameters are optimized by the quantum genetic algorithm, high-performance nonlinear frequency modulation signal generation is realized by combining with FPGA, the data acquisition and transmission burden is significantly reduced by using a non-uniform sparse array and space-time-frequency joint compressed sensing, target information is effectively recovered by using a three-linear tensor sparse reconstruction technology at the receiving end, the coupling characteristic quantity of the polarization scattering eigenvalue entropy and the micro-Doppler frequency is extracted, dynamic intelligent decision of the detection threshold, filtering parameters and resource allocation is further realized by a deep reinforcement learning model, the target detection and tracking accuracy in a complex environment is improved by combining zero space projection and an improved JPDA algorithm, high resolution, strong robustness and adaptive sensing ability of the radar system to multiple types of targets under low sampling rate conditions are realized, and the comprehensive performance and intelligent level of the multifunctional radar in a complex electromagnetic environment and a resource-limited scene are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0084] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0085] Figure 1 The flowchart of the signal acquisition and processing method based on a multifunctional radar in embodiment 1.

[0086] Figure 2 The schematic diagram of the signal acquisition and processing system based on a multifunctional radar in embodiment 1. DETAILED DESCRIPTION

[0087] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings in the specification.

[0088] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0089] Second, the "one embodiment" or "an embodiment" described herein as including a particular feature, structure, or characteristic, but not every embodiment necessarily includes the particular feature, structure, or characteristic. The appearances of "in one embodiment" or "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment.

[0090] Embodiments, with reference to Figure 1 and Figure 2 , are embodiments of the present application, which provide a signal acquisition and processing method based on a multifunctional radar, comprising the following steps:

[0091] S1, the quantum genetic algorithm is used to optimize the radar transmitting waveform parameters, wherein the parameters include the center frequency, the bandwidth and the frequency modulation slope, and the non-linear frequency modulation waveform is generated by the FPGA hardware;

[0092] Further, the optimization of the waveform parameters is carried out under the quantum genetic algorithm framework, and the value ranges of the center frequency, the bandwidth and the frequency modulation slope are respectively mapped to , and to form a three-dimensional search space;

[0093] Each individual is represented by a quantum bit code, and the quantum state is , wherein and respectively represent the probability amplitude of the bit being 0 or 1;

[0094] A population of individuals is initialized, and each individual is decoded into actual waveform parameters after being measured and collapsed into a classical binary string;

[0095] The waveform is transmitted by a radar simulation system, the echo is received, and the distance autocorrelation function is calculated, and the maximum sidelobe level , the integrated sidelobe level and the peak-to-average power ratio of the signal in the region outside the main lobe are extracted;

[0096] The fitness function is used to comprehensively evaluate the waveform performance , and the expression is:

[0097] ;

[0098] , wherein , , are weight coefficients, which can be set according to the task type;

[0099] The population is updated according to the fitness value, the quantum state is adjusted using the quantum rotation gate, and the direction is determined by looking up the table according to the comparison result of the current individual and the global optimum.

[0100] After multiple generations of optimization, the optimal combination of waveform parameters was obtained. This is used to generate nonlinear frequency-modulated signals in an FPGA, and its time-domain expression is:

[0101] ;

[0102] in, These are nonlinear coefficients, determined by the optimization process. The pulse width;

[0103] It should be noted that by optimizing waveform parameters through quantum genetic algorithms, the superposition and entanglement characteristics of quantum states are fully utilized, significantly improving search efficiency and optimization accuracy, avoiding getting trapped in local optima, and obtaining nonlinear frequency-modulated waveforms with lower sidelobe levels and better ambiguity function characteristics, thereby enhancing the radar's range resolution and anti-interference performance. FPGA hardware implementation ensures the real-time performance and high stability of waveform generation, meeting the dual requirements of multifunctional radar for the flexibility and reliability of transmitted signals in complex mission scenarios.

[0104] S2. A non-uniform sparse array is used to receive the target echo signal, and time-domain random interval sampling and frequency-domain pseudo-random frequency point selection are carried out simultaneously to form time-space-frequency three-dimensional compressed sensing observation data.

[0105] Furthermore, the receiver employs a non-uniform sparse array structure with a number of array elements. Its position is determined by the non-periodic spacing;

[0106] Assume the radar operating wavelength is , No. The position of each element is ,in, These are distinct positive integers used to break the periodicity of the array and suppress spatial grating lobes;

[0107] The analog echo signal received by each array element is first randomly sampled in the time domain;

[0108] The number of sampling points is Much smaller than the number of Nyquist sampling points The sampling time is controlled by a pseudo-random sequence, forming a dimensionality-reduced time series. ;

[0109] For each Perform a short-time Fourier transform (STFT) to obtain the spectrum matrix. Pseudo-random frequency selection is performed in the frequency domain;

[0110] from Randomly selected from 1 frequency point A point, select the sequence from the Logistic chaotic mapping After generating and normalizing, the compressed frequency domain data is determined ;

[0111] Stack all the array elements Dimensional data by spatial position, build a three-dimensional data body, whose dimensions are the number of array elements , time domain sampling points And frequency domain sampling points , the data body is the time-space-frequency three-dimensional compressed sensing observation data;

[0112] It should be noted that the non-uniform sparse array reduces the hardware cost and system power consumption while effectively suppressing the spatial grating lobe and improving the angle resolution. Combined with the time domain random sampling and the frequency domain pseudo-random selection, the three-dimensional compressed sensing structure is constructed, which greatly reduces the data acquisition rate and transmission bandwidth demand, breaks through the traditional Nyquist sampling limit, ensures the integrity of the target information while realizing efficient sparse sampling, and provides technical support for the lightweight and low-power operation of the radar system.

[0113] S3, three-linear tensor joint sparse reconstruction is performed on the time-space-frequency three-dimensional compressed sensing observation data, the polarization scattering matrix eigenvalue is decomposed from the reconstructed signal, and the coupling characteristic quantity of the eigenvalue entropy and the micro-Doppler frequency is calculated;

[0114] Further, in order to recover the complete signal from the compressed observation data , a three-linear tensor sparse reconstruction model is adopted, and the real signal tensor can be decomposed as:

[0115] ;

[0116] Among them, is a spatial mode vector, is a time mode vector, is a frequency mode vector, is a tensor rank, represents the outer product;

[0117] The observation process satisfies , wherein is a joint observation operator composed of the array structure, the time domain sampling matrix and the frequency domain selection matrix, is noise;

[0118] In order to solve , an optimization problem containing a sparse regularization term is constructed, and the expression is:

[0119] ;

[0120] Among them, is sparse norm, is a regularization parameter, and the alternating direction method of multipliers (ADMM) is used to iteratively solve the reconstruction tensor;

[0121] The complex signals of the HH, HV, VH and VV polarization channels are extracted from the polarization scattering matrix of the target unit is constructed;

[0122] The eigenvalue decomposition is performed on the , to obtain two non-negative eigenvalues and , and ≥ The probability distribution is obtained after normalization;

[0123] The eigenvalue entropy is calculated, and the expression is:

[0124] ;

[0125] At the same time, the time pattern vector is subjected to spectral analysis, and the periodic frequency modulation caused by the micro-motion component is extracted to obtain the micro-Doppler frequency;

[0126] The eigenvalue entropy is multiplied by the micro-Doppler frequency to obtain the coupling feature quantity;

[0127] It should be noted that the joint sparse reconstruction using the tri-linear tensor model fully preserves the inherent correlation structure of the signal in the spatial, temporal and frequency dimensions, significantly improving the reconstruction accuracy and robustness compared to the traditional vectorization method. By extracting the coupling feature quantity of the eigenvalue entropy and the micro-Doppler frequency of the polarization scattering matrix, the material, structure and micro-motion characteristics of the target are fused, enhancing the classification and recognition ability of complex targets and improving the perception accuracy of the radar in a mixed target scene.

[0128] S4, input the coupling feature quantity into the deep reinforcement learning model to dynamically output the constant false alarm detection threshold, the moving target display filter order and the resource allocation weight;

[0129] Further, the calculated coupling feature quantity is combined with the noise power estimation and target quantity estimation of the current environment to form a state vector, which is input into the deep Q network model (DQN);

[0130] The action space of the deep Q network model includes three adjustable parameters: the constant false alarm detection threshold, the moving target display filter order, and the resource allocation weight vector;

[0131] The deep Q network model obtains the optimal strategy through training, and the reward function is designed as:

[0132] ;

[0133] in, For detection probability, This represents the probability of a false alarm. For resource utilization, , , These are the weighting coefficients;

[0134] It should be noted that by inputting coupled feature quantities into the deep reinforcement learning model, intelligent dynamic configuration of radar detection and resource management parameters is achieved, overcoming the poor adaptability problem caused by traditional fixed thresholds or experience-based settings. The model can autonomously learn the optimal strategy according to environmental changes, improving the detection probability while effectively suppressing false alarms, achieving synergistic optimization of detection performance and resource utilization, and significantly enhancing the autonomous decision-making ability of the radar system in unknown or dynamic environments.

[0135] S5. Based on the constant false alarm rate detection threshold, the order of the moving target display filter and the resource allocation weight, the reconstructed signal is subjected to null space projection filtering, and the target trajectory information is generated by an improved joint probability data association algorithm.

[0136] Furthermore, using the MTI filter order output above, a moving target display filter is applied to the time series of the reconstructed signal. The filter structure is a canceller with the following transfer function:

[0137] ;

[0138] The null-space projection method is employed. By analyzing the background region signal, the clutter covariance matrix is ​​estimated, and eigenvalue decomposition is performed. The eigenvectors corresponding to the smallest eigenvalues ​​are used to form the null-space basis, and the projection operator is constructed. The expression is as follows:

[0139] ;

[0140] The signal after MTI filtering is projected to obtain a further purified signal;

[0141] Target detection is performed on the range-Doppler graph using a constant false alarm threshold (CFAT). The detection logic is that if the energy of a given cell is higher than the average energy of its neighborhood, the target is detected by... If the value is multiple times higher, it is considered a valid target;

[0142] For the detected measurements, an improved joint probabilistic data association algorithm is used to perform track association, the expression of which is:

[0143] ;

[0144] in, For the first The covariance matrix of each Gaussian component is used to calculate the correlation probability based on this likelihood, and the track state is updated in a weighted manner to finally generate the target trajectory information.

[0145] It should be noted that the null space projection filtering utilizes the orthogonality of the clutter subspace to efficiently suppress the strong residual clutter, significantly improves the visibility of the moving target in the strong interference background, combines the improved joint probability data association algorithm, and more accurately describes the non-Gaussian measurement distribution through the Gaussian mixture model, reduces the track association error rate, especially in the target dense or cross motion scene, maintains the track continuity and stability, and thus improves the overall tracking performance and system reliability.

[0146] S6, record and feedback the target trajectory information and the environment state to the step of optimizing the radar transmission waveform, and adjust the fitness function weight of the waveform optimization algorithm;

[0147] Further, the target trajectory information generated in the foregoing is statistically analyzed, for each trajectory, the speed sequence is extracted, and the variance of the speed change is calculated:

[0148]

[0149] wherein, is a trajectory stability index;

[0150] Meanwhile, the environment state is obtained, including the signal-to-noise ratio, the clutter intensity and the target density;

[0151] According to and the target density, the scene type is determined;

[0152] If and , it is considered that the target is dense and the motion is stable, and the resolution capability needs to be improved;

[0153] If , it is considered that the environment is bad, and the detection capability needs to be enhanced;

[0154] Accordingly, the fitness function weight in the waveform optimization is adjusted;

[0155] It should be noted that by feeding back the target trajectory information and the environment state to the waveform optimization link, a closed-loop adaptive regulation mechanism is constructed, so that the radar transmission waveform can dynamically adjust the design target according to the actual detection effect, such as enhancing the resolution capability in the target dense scene and improving the detection sensitivity in the low signal-to-noise ratio environment, realizing the intelligent evolution of perception-learning-optimization, and significantly improving the long-term adaptability and comprehensive combat effectiveness of the radar system in the complex and changeable task environment.

[0156] The embodiment also provides a signal acquisition and processing system based on a multifunctional radar, which comprises:

[0157] ​The waveform optimization module, the compressive sensing module, the tensor reconstruction module, the intelligent decision module, the signal filtering module and the trajectory generation module;

[0158] The waveform optimization module is configured to optimize parameters of a center frequency, a bandwidth and a frequency modulation slope of a radar transmitting waveform by using a quantum genetic algorithm, and generate a nonlinear frequency modulation signal by using an FPGA hardware;

[0159] The compressive sensing module is configured to implement time-domain random interval sampling and frequency-domain pseudo-random frequency point selection on a target echo signal synchronously under a non-uniform sparse array receiving condition, and form time-space-frequency three-dimensional compressive sensing observation data.

[0160] The tensor reconstruction module is configured to perform three-linear tensor joint sparse reconstruction on the three-dimensional compressive sensing observation data, extract a polarization scattering matrix eigenvalue from the reconstructed signal, calculate an eigenvalue entropy, and generate a coupled feature quantity in combination with a micro-Doppler frequency.

[0161] The intelligent decision module is configured to input the coupled feature quantity into a deep reinforcement learning model, and dynamically output a constant false alarm detection threshold, a moving target display filter order and a resource allocation weight in combination with an environment state.

[0162] The signal filtering module is configured to sequentially perform moving target display filtering and null space projection filtering on the reconstructed signal based on parameters output by the intelligent decision module, and suppress clutter interference.

[0163] The trajectory generation module is configured to implement constant false alarm detection based on the filtered signal, complete measurement and track association by using an improved joint probability data association algorithm, generate target trajectory information, and feed back the trajectory and the environment state to the waveform optimization module to realize dynamic adjustment of a fitness function weight.

[0164] The embodiment also provides a computer device suitable for the signal acquisition and processing method based on the multifunctional radar, which comprises a memory and a processor.

[0165] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved by WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, a trackball or a touchpad arranged on the shell of the computer device, or an external keyboard, a touchpad or a mouse, etc.

[0166] The embodiment also provides a storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the signal acquisition and processing method based on the multi-functional radar. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk or an optical disk.

[0167] To sum up, the application optimizes radar waveform parameters by quantum genetic algorithm, realizes high-performance nonlinear frequency modulation signal generation combined with FPGA, significantly reduces data acquisition and transmission burden by using non-uniform sparse array and space-time-frequency joint compression sensing, effectively recovers target information by adopting three-linear tensor sparse reconstruction technology at the receiving end, extracts the coupling characteristic quantity of polarization scattering eigenvalue entropy and micro-Doppler frequency, further realizes dynamic intelligent decision of detection threshold, filtering parameter and resource allocation through a deep reinforcement learning model, improves target detection and tracking accuracy in complex environments by combining zero space projection and improved JPDA algorithm, realizes high resolution, strong robustness and adaptive sensing ability of the radar system for multiple types of targets under low sampling rate conditions, and significantly improves the comprehensive performance and intelligent level of the multifunctional radar in complex electromagnetic environments and resource limited scenarios.

[0168] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A signal acquisition and processing method based on multi-functional radar, characterized in that: include: A quantum genetic algorithm is used to optimize radar transmission waveform parameters, including center frequency, bandwidth and frequency modulation slope, and a nonlinear frequency modulation waveform is generated by FPGA hardware. A non-uniform sparse array is used to receive the target echo signal, and time-domain random interval sampling and frequency-domain pseudo-random frequency point selection are carried out simultaneously to form time-space-frequency three-dimensional compressed sensing observation data. A trilinear tensor joint sparse reconstruction is performed on the time-space-frequency three-dimensional compressed sensing observation data. The eigenvalues ​​of the polarization scattering matrix are decomposed from the reconstructed signal, and the coupling characteristic quantity of eigenvalue entropy and micro-Doppler frequency is calculated. The coupled feature quantities are input into the deep reinforcement learning model, and the constant false alarm rate detection threshold, the order of the moving target display filter, and the resource allocation weights are dynamically output. Null-space projection filtering is performed on the reconstructed signal based on constant false alarm rate detection threshold, moving target display filter order and resource allocation weight, and target trajectory information is generated by an improved joint probabilistic data association algorithm. The target trajectory information and environmental status are recorded and fed back into the step of optimizing the radar transmission waveform, and the fitness function weights of the waveform optimization algorithm are adjusted.

2. The signal acquisition and processing method based on multi-functional radar as described in claim 1, characterized in that: The process employs a quantum genetic algorithm to optimize radar transmission waveform parameters, including center frequency, bandwidth, and frequency modulation slope, and generates a nonlinear frequency-modulated waveform using FPGA hardware. Specifically, this includes: Waveform parameter optimization is performed within a quantum genetic algorithm framework, mapping the ranges of center frequency, bandwidth, and frequency modulation slope to... , and This constitutes a three-dimensional search space; Each individual is represented by a qubit encoding, and its quantum state is... ,in and These represent the probability amplitudes of the individual being 0 or 1, respectively. Initialization generation A population composed of individuals, each individual is measured and collapsed into a classical binary string, then decoded into actual waveform parameters; The waveform is transmitted using a radar simulation system, the echo is received, and its range autocorrelation function is calculated. The maximum sidelobe level in the region outside the main lobe is then extracted. Integral sidelobe level Peak-to-average power ratio of the signal ; The fitness function is used to comprehensively evaluate waveform performance. Its expression is: ; in, , , This is a weighting coefficient, which can be set according to the task type; The population is updated based on fitness values, and quantum states are adjusted using quantum rotation gates. The direction is determined by looking up a table based on the comparison between the current individual and the global optimum. After multiple generations of optimization, the optimal combination of waveform parameters was obtained. This is used to generate nonlinear frequency-modulated signals in an FPGA, and its time-domain expression is: ; in, These are nonlinear coefficients, determined by the optimization process. This represents the pulse width.

3. The signal acquisition and processing method based on multi-functional radar as described in claim 2, characterized in that: The method employs a non-uniform sparse array to receive target echo signals, simultaneously implementing time-domain random interval sampling and frequency-domain pseudo-random frequency point selection to form time-space-frequency three-dimensional compressed sensing observation data, specifically including: The receiver employs a non-uniform sparse array structure with the number of array elements being... Its position is determined by the non-periodic spacing; Assume the radar operating wavelength is , No. The position of each element is ,in, These are distinct positive integers used to break the periodicity of the array and suppress spatial grating lobes; The analog echo signal received by each array element is first randomly sampled in the time domain; The number of sampling points is Much smaller than the number of Nyquist sampling points The sampling time is controlled by a pseudo-random sequence, forming a dimensionality-reduced time series. ; For each Perform a short-time Fourier transform (STFT) to obtain the spectrum matrix. Pseudo-random frequency selection is performed in the frequency domain; from Randomly selected from 1 frequency point The selected sequence is derived from the Logistic chaotic mapping. After generation and normalization, compressed frequency domain data is obtained. ; All array elements 3D data are stacked according to their spatial location to construct a 3D data volume, the dimension of which is the number of matrix elements. Time domain sampling points and frequency domain sampling points This data volume is the time-space-frequency three-dimensional compressed sensing observation data.

4. The signal acquisition and processing method based on multi-functional radar as described in claim 3, characterized in that: The process of performing trilinear tensor joint sparse reconstruction on the spatiotemporal-frequency three-dimensional compressed sensing observation data, decomposing the eigenvalues ​​of the polarization scattering matrix from the reconstructed signal, and calculating the coupled eigenvalue entropy and micro-Doppler frequency features specifically includes: To compress observation data To recover the complete signal, a trilinear tensor sparse reconstruction model is used, where the true signal tensor is... It can be decomposed into: ; in, For spatial pattern vectors, For time pattern vectors, For frequency mode vectors, Represents the rank of a tensor; The observation process satisfies ,in, It is a joint observation operator consisting of an array structure, a time-domain sampling matrix, and a frequency-domain selection matrix. For noise; To solve Construct an optimization problem containing sparse regularization terms, expressed as: ; in, For the sparse norm, The regularization parameter is solved iteratively using the Alternating Direction Multiplier Method (ADMM) to obtain the reconstructed tensor. from The complex signals of the four polarization channels HH, HV, VH, and VV are extracted to construct the polarization scattering matrix of the target unit; right Eigenvalue decomposition yields two non-negative eigenvalues. and ,and ≥ After normalization, the probability distribution is obtained; The eigenvalue entropy is calculated using the following expression: ; Simultaneously, spectral analysis is performed on the time pattern vector to extract the periodic frequency modulation caused by the micro-movement component, thus obtaining the micro-Doppler frequency; Multiplying the eigenvalue entropy by the micro-Doppler frequency yields the coupled characteristic quantity.

5. The signal acquisition and processing method based on multi-functional radar as described in claim 4, characterized in that: The process of inputting coupled feature quantities into a deep reinforcement learning model and dynamically outputting a constant false alarm rate (CFAR) detection threshold, a moving target indication filter order, and resource allocation weights specifically includes: The calculated coupling features are combined with the noise power estimate and target quantity estimate of the current environment to form a state vector, which is then input into the deep Q-network model DQN. The action space of the deep Q-network model includes three adjustable parameters: constant false alarm rate (CFAR) threshold, moving target display filter order, and resource allocation weight vector. The deep Q-network model obtains the optimal policy through training, and its reward function is designed as follows: ； in, For detection probability, This represents the probability of a false alarm. For resource utilization, , , These are the weighting coefficients.

6. The signal acquisition and processing method based on multi-functional radar as described in claim 5, characterized in that: The process of performing null-spatial projection filtering on the reconstructed signal based on the constant false alarm rate (CFAR) detection threshold, the order of the moving target display filter, and resource allocation weights, and generating target trajectory information using an improved joint probabilistic data association algorithm, specifically includes: Using the MTI filter order output above, a moving target display filter is applied to the time series of the reconstructed signal. The filter structure is a canceller of order, and its transfer function is: ; The null-space projection method is employed. By analyzing the background region signal, the clutter covariance matrix is ​​estimated, and eigenvalue decomposition is performed. The eigenvectors corresponding to the smallest eigenvalues ​​are used to form the null-space basis, and the projection operator is constructed. The expression is as follows: ； The signal after MTI filtering is projected to obtain a further purified signal; Target detection is performed on the range-Doppler graph using a constant false alarm threshold (CFAT). The detection logic is that if the energy of a given cell is higher than the average energy of its neighborhood, the target is detected. If the value is multiple times higher, it is considered a valid target; For the detected measurements, an improved joint probabilistic data association algorithm is used to perform track association, the expression of which is: ; in, For the first The covariance matrix of Gaussian components is used to calculate the correlation probability based on this likelihood, and the track state is updated in a weighted manner to finally generate the target trajectory information.

7. The signal acquisition and processing method based on multi-functional radar as described in claim 6, characterized in that: The step of recording target trajectory information and environmental status and feeding it back to optimize radar transmission waveform, and adjusting the fitness function weights of the waveform optimization algorithm, specifically includes: The target trajectory information generated above is statistically analyzed, and the velocity sequence of each trajectory is extracted. Calculate the variance of the velocity change: ; in, As an indicator of trajectory stability; Simultaneously acquire environmental conditions, including signal-to-noise ratio, clutter intensity, and target density; according to Determine the scene type based on target density; like and They believed that the targets were densely packed and moving stably, requiring improved discrimination capabilities. like They believed the environment was harsh and that testing capabilities needed to be strengthened. Adjust the fitness function weights in waveform optimization accordingly.

8. A signal acquisition and processing system based on multi-functional radar, based on the signal acquisition and processing method based on multi-functional radar according to any one of claims 1 to 7, characterized in that: include: The system includes a waveform optimization module, a compressed sensing module, a tensor reconstruction module, an intelligent decision-making module, a signal filtering module, and a trajectory generation module. The waveform optimization module is used to optimize the center frequency, bandwidth and frequency modulation slope of the radar transmitted waveform using a quantum genetic algorithm, and to generate a nonlinear frequency modulation signal through FPGA hardware. The compressed sensing module is used to synchronously perform time-domain random interval sampling and frequency-domain pseudo-random frequency point selection on the target echo signal under non-uniform sparse array receiving conditions to form time-space-frequency three-dimensional compressed sensing observation data. The tensor reconstruction module is used to perform trilinear tensor joint sparse reconstruction on three-dimensional compressed sensing observation data, extract polarization scattering matrix eigenvalues ​​from the reconstructed signal, calculate eigenvalue entropy, and generate coupled eigenvalues ​​by combining micro-Doppler frequency. The intelligent decision-making module is used to input coupled feature quantities into a deep reinforcement learning model and dynamically output constant false alarm detection threshold, moving target display filter order, and resource allocation weights in combination with environmental conditions. The signal filtering module is used to perform moving target display filtering and null space projection filtering on the reconstructed signal sequentially based on the parameters output by the intelligent decision module, so as to suppress clutter interference. The trajectory generation module is used to perform constant false alarm rate (CFAR) detection based on the filtered signal, and to complete the association between measurement and trajectory using an improved joint probabilistic data association algorithm to generate target trajectory information. At the same time, the trajectory and environmental status are fed back to the waveform optimization module to realize dynamic adjustment of the fitness function weights.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the signal acquisition and processing method based on multi-functional radar as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the signal acquisition and processing method based on any one of claims 1 to 7.

Images