Method and system for maneuvering target parameter estimation and long-time coherent accumulation detection

Abstract

The application provides a motor target parameter estimation and long-time phase correlation accumulation detection method and system, which first carries out radio frequency conditioning and sampling on a radar original echo, carries out pulse compression after obtaining a discrete complex baseband signal, and outputs a range focusing but still containing range migration frequency domain signal. The core step utilizes a second order Keystone transformation to correct the second order range migration of the target, outputs a second order correction signal eliminating the range-azimuth coupling, and then searches the optimal speed through a Bayesian optimization and CPU parallel acceleration algorithm. On this basis, the IAR algorithm is utilized in combination with the optimal speed estimation value to correct the first order range migration, and phase compensation is carried out according to the speed matching phase factor to obtain a phase matching signal. Finally, the matching signal is subjected to a non-uniform fast Fourier transformation in a slow time domain to realize energy focusing, and the complete motion parameters of the target are solved through analyzing the peak position and amplitude characteristics in the Doppler domain.

Description

Technical Field

[0001] This invention relates to the field of long-term coherent accumulation detection methods for maneuvering targets, and more specifically, to a method and system for estimating maneuvering target parameters and detecting long-term coherent accumulation. Background Technology

[0002] With the rapid development of high-speed drones, hypersonic weapons, and other equipment, these targets have small radar cross-sections, making their echo signals easily masked by noise, posing a serious challenge to radar moving target detection. To improve the output signal-to-noise ratio of targets, we usually choose to introduce coherent accumulation to enhance radar's target detection performance. However, over long accumulation periods, target echoes are prone to range migration and Doppler frequency migration, leading to a decrease in the gain of coherent accumulation.

[0003] Non-coherent accumulation methods have a lower implementation threshold, but their focusing performance degrades significantly in low SNR environments, making it impossible to effectively extract target signals from noise. Coherent accumulation methods, on the other hand, can effectively utilize the phase characteristics of the target, resulting in higher accumulation gain and enabling target signal extraction even in low SNR environments. Research on long-term coherent accumulation algorithms mainly falls into two categories: uniformly moving targets and non-uniformly moving targets. One type of coherent accumulation algorithm, represented by the Keystone transform proposed by RP Perry et al. in "SAR imaging of moving targets," models the linear characteristics of uniform target motion and can only adapt to first-order linear range migration generated by uniform targets. When facing non-uniform targets, the target's trajectory and signal characteristics deviate completely from the linear model preset by these algorithms, making it impossible to accurately match the complex motion state of non-uniform targets.

[0004] To address the challenges posed by non-uniform and highly maneuverable targets, existing technologies have also conducted research on related coherent accumulation algorithms. For example, Chinese Patent Application No. CN201910024743.5 discloses a radar coherent accumulation detection method for highly maneuverable targets based on time-reversed non-uniform sampling. This method achieves rapid reduction of the order of high-order phase signals and parameter estimation through radar echo pulse compression, time-reversed matched Fourier transform, second-order phase compensation, non-uniform sampling order reduction, and variable scaling transformation.

[0005] In summary, existing coherent accumulation detection methods still face the following pressing technical challenges when dealing with high-speed, highly maneuverable targets in low signal-to-noise ratio environments: Highly maneuverable targets possess both unknown radial velocity and acceleration, leading to severe range migration (first-order range migration) and range curvature (second-order range migration) in the echo signal, exhibiting extremely complex range-azimuth coupling characteristics. Traditional methods often require a two-dimensional joint exhaustive search for velocity and acceleration when performing multi-order range migration correction, resulting in an exponential increase in computational complexity.

[0006] Therefore, how to propose a detection method that can completely eliminate the range migration of highly maneuvering targets at all levels, achieve perfect energy focusing under non-uniform grids, break through the bottleneck of traditional exhaustive search, and efficiently and accurately calculate the true motion parameters of the target with extremely low computational cost is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for estimating maneuvering target parameters and detecting long-term coherent accumulation.

[0008] According to one aspect of the present invention, a method for estimating maneuvering target parameters and detecting long-term coherent accumulation includes:

[0009] Step S1: Preprocess the original radar radio frequency echo signal to output a discrete complex baseband signal;

[0010] Step S2: Perform pulse compression on the discrete complex baseband signal to output a frequency domain signal that is focused in the range direction but still contains range migration in the azimuth direction;

[0011] Step S3: Process the frequency domain signal using the second-order Keystone transform to correct the target's second-order range migration and output a second-order corrected signal that eliminates range-azimuth coupling;

[0012] Step S4: Based on Bayesian optimization, perform velocity search matching on the second-order correction signal, and output the radial velocity of the candidate target and the corresponding velocity matching phase factor;

[0013] Step S5: Based on the IAR algorithm, combined with the second-order correction signal and the radial velocity of the candidate target, the first-order range migration of the target is corrected, and the first-order correction signal with complete elimination of range-azimuth coupling is output.

[0014] Step S6: Based on the velocity matching phase factor, perform velocity phase compensation on the first-order correction signal and output the phase matching signal after first-order velocity phase cancellation.

[0015] Step S7: Perform a non-uniform fast Fourier transform on the phase-matched signal in the slow time domain to achieve target energy focusing and output the Doppler domain peak signal after slow time focusing;

[0016] Among them, repeat steps S4-S7 until the iteration termination condition is met, and then proceed to step S8;

[0017] Step S8: Analyze the peak position and amplitude characteristics of the Doppler domain peak signal, and calculate the complete motion parameters of the target in combination with the system parameters.

[0018] Preferably, the expression for the frequency domain signal in S2 is:

[0019]

[0020] In the formula, For frequency domain signals, The signal amplitude is in the distance frequency domain and slow time domain. For distance frequency, For carrier frequency, For location, slow time, For signal bandwidth, For the target radial velocity, For the target radial acceleration, At the speed of light, The initial radial distance to the target. It is the imaginary unit.

[0021] Preferably, S3 includes:

[0022] Step S3.1: Construct new slow-time variables ,in, The item includes distance frequency With direction and slow time The coupling term is expressed as follows:

[0023]

[0024] In the formula, For location, slow time, For distance frequency, For carrier frequency, For slow-time variables;

[0025] Step S3.2: Use the new slow-time variable replace , release distance frequency With direction and slow time The quadratic coupling relationship corrects the distance curvature, which varies parabolically with slow time, into a linearly varying distance travel function, i.e., a second-order correction signal, as follows:

[0026]

[0027] Right now:

[0028]

[0029] In the formula, For frequency domain signals, For time-domain signals, For distance frequency, For carrier frequency, For slow time variables, The time delay corresponding to the initial distance. The signal amplitude is in the distance frequency domain and slow time domain. For signal bandwidth, For fast time variables, For the target radial velocity, For the target radial acceleration, At the speed of light, It is the imaginary unit.

[0030] Preferably, step S4 includes:

[0031] Step S4.1: Based on the pulse repetition frequency, determine the baseband unambiguous velocity range, introduce integer ambiguity numbers and blind velocity components, and construct an extended velocity search space containing Doppler ambiguity;

[0032] Step S4.2: By fitting the target function to maximize the focusing index using a Gaussian process as a surrogate model, and based on a preset velocity range, using the acquisition function, adaptively predict the radial velocity of the most promising candidate target in the expanded velocity search space. The expression for the radial velocity of the candidate target is as follows:

[0033]

[0034] In the formula, The fundamental frequency falling within the unambiguous interval; For carrier frequency; Baseband speed; This refers to the radar blind speed component; It is a fuzzy number and is an integer; The radial velocity of the target; The speed of light; The pulse repetition frequency;

[0035] Step S4.3: Based on the radial velocity of the candidate target, generate the corresponding velocity matching phase factor, wherein the expression for the velocity matching phase factor is as follows:

[0036]

[0037] In the formula, The speed of light; The target radial velocity; For carrier frequency; For slow-time variables; It is the imaginary unit.

[0038] Preferably, step S5 includes:

[0039] Step S5.1: Based on the radial velocity of the candidate target, construct a new fast-time variable; wherein the expression of the fast-time variable is as follows:

[0040]

[0041] In the formula, For the original fast time variable, For the new fast time variable, Baseband speed; For radar blind speed component, For slow time variables, The speed of light;

[0042] Step S5.2: Based on the shift property of the Fourier transform, the phase compensation factor containing the new fast time variable is multiplied by the second-order correction signal in the range frequency domain to obtain the first-order correction signal with completely eliminated range-azimuth coupling; wherein, the expression for the phase compensation factor containing the new fast time variable is:

[0043]

[0044] In the formula, For distance frequency, Baseband speed; For radar blind speed component, For slow time variables, At the speed of light, It is the imaginary unit.

[0045] Preferably, step S6 includes:

[0046] Multiplying the first-order correction signal by the velocity matching phase factor yields the first-order velocity phase cancellation phase-matched signal.

[0047] Preferably, step S7 specifically includes:

[0048] Step S7.1: Using non-uniform fast Fourier transform, the phase-matching signal is transformed to a uniform Doppler frequency domain to obtain a focused two-dimensional image;

[0049] Step S7.2: Calculate the focus index of the two-dimensional image, which is the peak intensity of the image; feed the focus index back to the Bayesian optimizer to update the Gaussian process model.

[0050] Preferably, the iteration termination condition includes: reaching a preset number of iterations or the target optimization function reaching a convergence condition.

[0051] Preferably, step S8 specifically includes:

[0052] After the iteration is completed, the Gaussian is updated based on the maximum value of the focusing index and recorded as the optimal velocity. Based on the optimal velocity, the true velocity of the target is determined, and the target's distance, true velocity, acceleration, and image results are output.

[0053] According to another aspect of the present invention, a system for estimating maneuvering target parameters and detecting long-term coherent accumulation includes:

[0054] Module M1: Preprocesses the raw radar radio frequency echo signal and outputs a discrete complex baseband signal;

[0055] Module M2: Performs pulse compression on discrete complex baseband signals and outputs frequency domain signals that are focused in the range direction but still contain range migration in the azimuth direction;

[0056] Module M3: Uses second-order Keystone transform to process frequency domain signals, corrects target second-order range migration, and outputs a second-order corrected signal that eliminates range-azimuth coupling;

[0057] Module M4: Based on Bayesian optimization, it performs velocity search matching on the second-order corrected signal and outputs the radial velocity of the candidate target and the corresponding velocity matching phase factor;

[0058] Module M5: Based on the IAR algorithm, it combines the second-order correction signal and the radial velocity of the candidate target to correct the first-order range migration of the target and outputs a first-order correction signal that completely eliminates range-azimuth coupling;

[0059] Module M6: Based on the velocity matching phase factor, it performs velocity phase compensation on the first-order correction signal and outputs the phase matching signal after first-order velocity phase cancellation;

[0060] Module M7: Performs a non-uniform fast Fourier transform on the phase-matched signal in the slow time domain to achieve target energy focusing and outputs the Doppler domain peak signal after slow-time focusing;

[0061] Among them, modules M4-M7 are repeatedly triggered until the iteration end condition is met, at which point module M8 is executed;

[0062] Module M8: Analyzes the peak position and amplitude characteristics of the Doppler domain peak signal, calculates and outputs the complete motion parameters of the target in combination with system parameters.

[0063] Compared with the prior art, the present invention has the following advantages:

[0064] This invention proposes a long-term coherent accumulation detection method for maneuvering targets based on second-order Keystone-improved axis rotation-non-uniform fast Fourier transform. This method can calculate the acceleration of the target echo, accurately estimate the parameters of maneuvering targets in low SNR environments, and has low computational complexity and excellent computational efficiency.

[0065] This invention significantly improves the detection and parameter estimation performance of high-speed maneuvering targets in low signal-to-noise ratio environments by deeply integrating a series of advanced radar signal processing techniques with intelligent optimization algorithms. The method introduces a second-order Keystone transform to process the pulse-compressed signal, effectively correcting the second-order range migration caused by the target's radial acceleration, thus eliminating the "range curvature" phenomenon. This reduces the complexity of the problem from a complex two-dimensional search that originally required simultaneous searching of velocity and acceleration to a one-dimensional search focused solely on velocity.

[0066] To overcome the enormous computational bottleneck faced by traditional global grid search in extended velocity spaces containing Doppler blur, this invention creatively uses the focusing index of subsequent two-dimensional images as the target evaluation function. Combined with a Gaussian process model of Bayesian optimization algorithm and an acquisition function, and supplemented by a multi-core CPU parallel computing architecture, it achieves adaptive intelligent search for the optimal target radial velocity and blur number. This approach significantly reduces the number of trial-and-error iterations, ensuring the search for the global optimum while meeting the efficiency requirements of engineering processing.

[0067] This invention utilizes the velocity estimate generated by the search and performs complex multiplication in the distance frequency domain using an improved frequency domain axis rotation algorithm. This cleverly replaces the huge computational time and error caused by traditional time domain interpolation, completely eliminating the target's first-order distance movement and achieving horizontal alignment of the trajectory. Finally, addressing the non-uniform time grid deformation problem inevitably introduced by the above physical transformation process, this method employs a non-uniform fast Fourier transform to accurately resample the signal to a uniform Doppler frequency domain, achieving perfect focusing of the target energy.

[0068] This method not only completely eliminates the severe range-azimuth coupling in the echoes of complex maneuvering targets, but also generates extremely sharp strong signal peaks in the Doppler domain at very low computational cost, thereby achieving accurate and efficient motion parameter calculation and long-term coherent accumulation detection of high-speed maneuvering targets. Attached Figure Description

[0069] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0070] Figure 1The flowchart of the long-term coherent accumulation detection method for maneuvering targets based on second-order Keystone-improved axis rotation-non-uniform fast Fourier transform provided in the embodiments of the present invention is shown.

[0071] Figure 2 The diagram shows the results after implementing the traditional processing method and after implementing the present invention. Detailed Implementation

[0072] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0073] Example 1:

[0074] This embodiment provides a method for estimating maneuvering target parameters and detecting long-term coherent accumulation, including:

[0075] Step S1: Preprocess the raw radar radio frequency echo signal to output a discrete complex baseband signal; the preprocessing includes bandpass filtering, IQ demodulation, low-pass filtering and sampling;

[0076] Step S2: Perform pulse compression on the discrete complex baseband signal and transform it to the fast time-distance frequency domain. - Slow time In the time domain, the output is a frequency domain signal that is focused in the range direction but still contains range migration in the azimuth direction. ;

[0077] Step S3: Process the frequency domain signal using the second-order Keystone transform to correct the target's second-order range migration and output a second-order corrected signal that eliminates range-azimuth coupling;

[0078] Step S4: Using a Bayesian optimization and CPU parallel acceleration algorithm, perform velocity search matching on the second-order correction signal, and output the radial velocity of the candidate target and the corresponding velocity matching phase factor.

[0079] Step S5: Using the IAR algorithm, combined with the second-order correction signal and the radial velocity of the candidate target, the first-order range migration of the target is corrected, and the first-order correction signal with completely eliminated range-azimuth coupling is output.

[0080] Step S6: Based on the velocity matching phase factor, perform velocity phase compensation on the first-order correction signal and output the phase matching signal after first-order velocity phase cancellation.

[0081] Step S7: Perform a non-uniform fast Fourier transform on the phase-matched signal in the slow time domain to achieve target energy focusing and output the Doppler domain peak signal after slow time focusing;

[0082] Among them, repeat steps S4-S7 until the iteration termination condition is met, and then proceed to step S8;

[0083] Step S8: Analyze the peak position and amplitude characteristics of the Doppler domain peak signal, and calculate the complete motion parameters of the target in combination with the system parameters.

[0084] In this embodiment, the expression for the frequency domain signal in S2 is:

[0085]

[0086] In the formula, For frequency domain signals, The signal amplitude is in the distance frequency domain and slow time domain. For distance frequency, For carrier frequency, For location, slow time, For signal bandwidth, For the target radial velocity, For the target radial acceleration, At the speed of light, The initial radial distance to the target. It is the imaginary unit.

[0087] In this embodiment, S3 includes:

[0088] Step S3.1: Construct new slow-time variables ,in, The item includes distance frequency With direction and slow time The coupling term is expressed as follows:

[0089]

[0090] In the formula, For location, slow time, For distance frequency, For carrier frequency, For slow-time variables;

[0091] Step S3.2: Use the new slow-time variable replace , release distance frequency With direction and slow time The quadratic coupling relationship corrects the distance curvature, which varies parabolically with slow time, into a linearly varying distance travel function, i.e., a second-order correction signal, as follows:

[0092]

[0093] Right now:

[0094]

[0095] In the formula, For frequency domain signals, For time-domain signals, For distance frequency, For carrier frequency, For slow time variables, The time delay corresponding to the initial distance. The signal amplitude is in the distance frequency domain and slow time domain. For signal bandwidth, For fast time variables, For the target radial velocity, For the target radial acceleration, At the speed of light, It is the imaginary unit.

[0096] It should be noted that the second-order Keystone transform in S3 introduces a scaling factor determined based on the Rayleigh criterion. The scaling factor Used to eliminate double slopes with the same sign after the second-order Keystone transform, indexed by slow time. The trajectory of the boundary jump is abnormal, and the scaling factor For real numbers greater than 1, the theoretical lower bound derived from the Rayleigh criterion is used in engineering implementation. The rounded-up value is preferably an integer power of 2; the scaling factor By reconstructing the velocity search space to reduce the folding velocity step, the new folding velocity is expressed as: ,in The original folding speed is used. At the same time, higher-order phase errors in the second-order Keystone transform are suppressed, the physical distance at the trajectory edge is compressed to less than one distance sampling unit, the odd higher-order phase distortion terms of the asymmetric jump approach 0, the frequency domain signal is simplified to an ideal linear phase, the pulse envelope symmetry characteristics are restored, the physical space of the Fresnel fan-shaped expansion is cleared, and the time reversal effect and bandpass cutoff diffraction distortion are eliminated.

[0097] In this embodiment, step S4 includes:

[0098] Step S4.1: Based on the pulse repetition frequency, determine the baseband unambiguous velocity range, introduce integer ambiguity numbers and blind velocity components, and construct an extended velocity search space containing Doppler ambiguity;

[0099] Step S4.2: Using a Gaussian process as a surrogate model, the objective optimization function that maximizes the focusing index is obtained. Based on a preset velocity range, and using the acquisition function, the radial velocity of the most promising candidate target is adaptively predicted in the expanded velocity search space. The expression for the radial velocity of the candidate target is as follows:

[0100]

[0101] In the formula, The fundamental frequency falling within the unambiguous interval; For carrier frequency; Baseband speed; This refers to the radar blind speed component; It is a fuzzy number and is an integer; The radial velocity of the target; The speed of light; is the pulse repetition frequency.

[0102] Step S4.3: Based on the radial velocity of the candidate target, generate the corresponding velocity matching phase factor, wherein the expression for the velocity matching phase factor is as follows:

[0103]

[0104] In the formula, The speed of light; The target radial velocity; For carrier frequency; For slow time variables, It is the imaginary unit.

[0105] It should be noted that, to avoid the massive computational burden caused by expanding the velocity search space and to improve parameter estimation efficiency, the velocity search matching in S4 specifically employs a Bayesian optimization algorithm instead of the traditional exhaustive search: a Gaussian process is used as a surrogate model to fit the objective function, and the acquisition function is used to guide the next search direction. The core time-consuming part of Bayesian optimization is Gaussian process regression. The complete theoretical complexity of the proposed method is:

[0106]

[0107] In the formula, The theoretical complexity of the proposed method is... This represents the asymptotic upper bound of the algorithm's complexity. Total number of searches For distance to gates, This represents the number of pulses.

[0108] Meanwhile, the peak contrast or peak intensity of the two-dimensional image output by S7 after focusing is used as the optimization objective function. Using a Bayesian optimized surrogate model and acquisition function, the system adaptively predicts and selects the next set of most promising candidate velocity parameters in the expanded velocity search space for iterative calculation, replacing the full-space grid traversal search, thereby quickly converging to the optimal focusing parameters.

[0109] For example, S4 employs a multi-core CPU parallel acceleration strategy, specifically including:

[0110] In the batch processing mode of Bayesian optimization or the local refinement stage of grid search, independent operations such as IAR correction, range-directed IFFT, and azimuth-directed NUFFT under different velocity assumptions are distributed to... Each physical core performs parallel computation; ideally, the parallel speedup is... satisfy:

[0111]

[0112] in, For parallel speedup ratio, This represents the number of physical CPU cores.

[0113] In this embodiment, step S5 includes:

[0114] Step S5.1: Based on the radial velocity of the candidate target, construct a new fast-time variable; wherein the expression of the fast-time variable is as follows:

[0115]

[0116] In the formula, For the original fast time variable, For the new fast time variable, Baseband speed; For radar blind speed component, For slow time variables, The speed of light;

[0117] Step S5.2: Based on the shift property of the Fourier transform, the phase compensation factor containing the new fast time variable is multiplied by the second-order correction signal in the range frequency domain to obtain the first-order correction signal with completely eliminated range-azimuth coupling; wherein, the expression for the phase compensation factor containing the new fast time variable is:

[0118]

[0119] In the formula, For distance frequency, Baseband speed; For radar blind speed component, For slow time variables, At the speed of light, It is the imaginary unit.

[0120] The displacement properties of the Fourier transform are as follows:

[0121]

[0122] In the formula, For theoretical time-domain signals, This is the time-domain shift amount. for The corresponding frequency domain signal, For Fourier transform operators, For time variables, For frequency variables, The imaginary unit;

[0123] The expression for the first-order correction signal is as follows:

[0124]

[0125] In the formula, To eliminate the specific echo time-domain signal of the coupling, i.e., the first-order correction signal, The signal amplitude is in the distance frequency domain and slow time domain. For signal bandwidth, For carrier frequency, The time delay corresponding to the initial distance. For the new fast-time variable, For slow time variables, The radial velocity of the target, The radial acceleration of the target, At the speed of light, It is the imaginary unit.

[0126] It should be noted that a linear time delay in the time domain is equivalent to a linear phase rotation in the frequency domain; therefore, the output of S3, substituted with a new fast-time variable, is obtained by using complex multiplication in the distance frequency domain instead of the complex time-domain operation. Interpolation can be used to obtain a first-order corrected signal.

[0127] In this embodiment, step S6 includes:

[0128] Multiplying the first-order correction signal by the velocity matching phase factor yields the first-order velocity phase-cancelled phase-matched signal, where the expression for the phase-matched signal is as follows:

[0129]

[0130] In the formula, The specific echo time-domain signal after matching, i.e., the phase-matched signal, The signal amplitude is in the distance frequency domain and slow time domain. For signal bandwidth, For carrier frequency, The time delay corresponding to the initial distance. For the new fast-time variable, For slow time variables, The radial acceleration of the target, At the speed of light, It is the imaginary unit.

[0131] In this embodiment, step S7 specifically includes:

[0132] Step S7.1: Using non-uniform fast Fourier transform, the phase-matching signal is transformed to a uniform Doppler frequency domain to obtain a focused two-dimensional image;

[0133] Step S7.2: Calculate the focus index of the two-dimensional image, which is the peak intensity of the image; feed the focus index back to the Bayesian optimizer to update the Gaussian process model.

[0134] The expression for a two-dimensional image is as follows:

[0135]

[0136] In the formula, The signal is the Doppler domain signal after NUFFT focusing, i.e., a two-dimensional image; For sampling point index; The frequency is the Doppler frequency. This is the amplitude attenuation parameter; The signal amplitude; This represents the number of sampling points; For carrier frequency; The time delay corresponding to the initial distance; For signal bandwidth; For fast sampling periods; The number of pulses in the slow time period; The radial acceleration of the target; The speed of light; Index of sampling points for slow time; The pulse repetition period, It is the imaginary unit.

[0137] It should be noted that, regarding the non-uniform deformation problem occurring in the slow-time grid after the second-order Keystone transformation in S3 and the IAR algorithm processing in S5, the specific processing in S7 is as follows:

[0138] The non-uniform sampled data is processed using the NUFFT algorithm. Through interpolation or oversampling convolution, the signal on the non-uniform time grid is resampled to a uniform Doppler frequency grid, thereby achieving accurate coherent accumulation and pulse compression after eliminating second-order and first-order distance migration. The calculation formula of NUFFT in the continuous time domain is as follows:

[0139]

[0140] In the formula, The specific echo time-domain signal after NUFFT transformation. For the new fast time variable, For Doppler frequency, This refers to the specific echo time-domain signal after matching, i.e., the phase-matched signal; For slow time variables, It is the imaginary unit.

[0141] In this embodiment, the iteration termination condition includes: reaching a preset number of iterations or the target optimization function reaching a convergence condition.

[0142] In this embodiment, step S8 specifically includes:

[0143] After the iteration is completed, the Gaussian is updated based on the maximum value of the focusing index and recorded as the optimal velocity. Based on the optimal velocity, the true velocity of the target is determined, and the target's distance, true velocity, acceleration, and image results are output.

[0144] It should be noted that when IAR correctly matches the true velocity containing the blur number, the peak intensity after NUFFT focusing is the largest, thus resolving the true blur number. And the true velocity of highly dynamic targets, achieving defuzzification.

[0145] Example 2

[0146] This embodiment provides a long-term coherent accumulation detection method for maneuvering targets based on second-order Keystone, improved axis rotation, and non-uniform fast Fourier transform. This method addresses the range curvature and range travel issues present in the echoes of highly maneuvering targets by correcting them through second-order Keystone transform and improved frequency-domain axis rotation, and combines this with non-uniform fast Fourier transform to achieve energy focusing. Specifically, it includes the following steps:

[0147] 1. Data Input and Preprocessing: The raw echo data received by the input radar is bandpass filtered, IQ demodulated, and lowpass filtered before pulse compression and transformation to the fast time-range frequency domain. - Slow time Time domain;

[0148] 2. Parameter Space Construction: Define the motion parameter space to be searched, i.e., the first-order velocity. ( (Automatically includes cases where speed is ambiguous);

[0149] 3. SOKT Second-Order Correction: Constructing a New Second-Order Keystone Slow-Time Variable Eliminate acceleration The resulting second-order distance migration, and the release of distance frequency. With slow time variables Coupling;

[0150] 4. Frequency Domain IAR First-Order Correction: Based on the current searched velocity parameters A linear phase compensation factor is constructed in the distance frequency domain, and the axis rotation in the time domain is achieved through equivalent multiplication in the frequency domain, thus eliminating the effect of... First-order distance migration caused by velocity;

[0151] 5. NUFFT Energy Focusing: To address the non-uniform sampling grid resulting from SOKT and IAR transformations, NUFFT is used for slow time-domain focusing to acquire a range-Doppler two-dimensional image;

[0152] 6. Bayesian Iterative Optimization: Using the peak value of the focused image as the objective function, the Bayesian optimization algorithm is used to iteratively search in the parameter space until the optimal focusing parameters and target image are obtained.

[0153] The improved frequency domain axis rotation processing described in step 4 is essentially about avoiding time-domain interpolation. Its fast-time new variable expression is: In the formula For the new fast time variable, For the original fast time variable, The current candidate speed for the search. At the speed of light, For a slow-time variable, its phase compensation function expression is:

[0154]

[0155] In the formula, This is the phase compensation factor. For distance frequency, For the original slow time variable, The current candidate speed for the search. For carrier frequency, For the new slow time variable, At the speed of light, The unit is the imaginary unit. In this embodiment, instead of using traditional Sinc interpolation or linear interpolation, the distance-frequency domain data is directly interpolated with... Multiplication allows the signal envelope to change with slow time. Linear translation (i.e., axis rotation) is used to correct the first-order distance travel.

[0156] The velocity fuzzing process described in step 2 involves fuzzing the velocity in the search space. Defined as:

[0157]

[0158] In the formula, The current candidate speed for the search. The fundamental frequency falling within the unambiguous interval. For carrier frequency, For baseband speed, For radar blind speed, The pulse repetition frequency and the ambiguity number are given. It is an integer. It is the speed of light.

[0159] During the optimization process, based on the search... It automatically obtains the baseband speed and radar blind speed, thereby automatically resolving the ambiguity problem of high-speed targets.

[0160] Example 3:

[0161] The following describes the specific implementation steps of the long-term coherent accumulation detection method for maneuvering targets based on second-order Keystone-improved axis rotation-non-uniform fast Fourier transform in this embodiment. Each implementation step in this embodiment is described below. The simulation was performed on a simulation platform, where the speed of Bayesian optimization was measured. The search matching is accelerated using a multi-core parallel computing pool of CPUs.

[0162] like Figure 1 As shown, the implementation steps of this embodiment include:

[0163] S1: Input radar echo data matrix The signal is subjected to bandpass filtering, IQ demodulation, low-pass filtering, and sampling to obtain discrete echo signals. Simultaneously, a Bayesian optimizer is initialized, with the objective function set to "maximize image peak intensity," and parameter search boundaries defined as: velocity range... .

[0164] S2: The discrete signal obtained in S1 is subjected to pulse compression, and the range energy is focused using matched filtering to improve range resolution. This yields the range-frequency domain-azimuth-time domain signal. .

[0165] S3: Construct a new slow-time variable for the second-order Keystone. Eliminate acceleration The resulting second-order distance migration, and the release of distance frequency. With slow time variables This step eliminates range curvature, making the target trajectory a straight line in the time-frequency plane, but this line may be tilted relative to the azimuth axis (range travel exists). The resulting second-order corrected signal with range-azimuth coupling eliminated is obtained. .

[0166] S4: Speed ​​Search and Parameter Matching

[0167] Input data: Receive the second-order correction signal for eliminating range-azimuth coupling output from step S3. ".

[0168] Specific processing logic:

[0169] 1. Constructing an extended search space: Determining the unambiguous baseband velocity range based on pulse repetition frequency, and introducing integer ambiguity numbers. With blind speed component Construct an extended velocity search space containing Doppler fuzziness, and represent candidate velocities as... In the formula, The current candidate speed for the search. For baseband speed, For radar blind speed.

[0170] 2. Bayesian Optimization Search: Using the peak intensity of the 2D image output from subsequent step S7 as the optimization objective function, and leveraging a Gaussian process surrogate model and acquisition function, the next set of most promising candidate velocity parameters is adaptively predicted in the expanded velocity search space. .

[0171] 3. Parallel acceleration: In iterative search computation, independent operations under different speed assumptions are distributed to... Parallel computing is performed using individual CPU physical cores to improve search efficiency.

[0172] Output: After the iteration converges, the optimal velocity estimate is output. and the corresponding velocity matching phase factor In the formula, At the speed of light, This is the optimal speed estimate. For carrier frequency, For slow time variables, It is the imaginary unit.

[0173] S5: First-order distance migration correction (IAR algorithm)

[0174] Input data: The second-order correction signal output from step S3 "and the "optimal speed estimate" output by step S4 ".

[0175] Specific processing logic: Utilize the optimal velocity estimate obtained in step S4. (Including fuzzy numbers) ), construct new fast time variables In the formula, For the original fast time variable, For the new fast time variable, For baseband speed, For radar blind speed, At the speed of light, For slow-time variables, based on the shift property of the Fourier transform (linear time-domain delay equivalent to linear frequency-domain phase rotation), a phase compensation factor incorporating the new fast-time variable will be used. The second-order correction signal of S3 Complex multiplication is performed in the distance-frequency domain to replace the complex sinc interpolation operation in the time domain, ultimately yielding... In the formula, For distance frequency, For baseband speed, For radar blind speed, At the speed of light, For slow time variables, It is the imaginary unit.

[0176] Output result: First-order correction signal with complete elimination of range-azimuth coupling. ".

[0177] Special note: This step is equivalent to rotating the data matrix in the time domain, correcting the tilted target trajectory to a horizontal line parallel to the azimuth axis. Compared to traditional time-domain interpolation (IAR), this step significantly improves computational efficiency and avoids the amplitude loss inherent in interpolation.

[0178] S6: Compensate for velocity phase based on velocity search results: based on the velocity parameters obtained from the current iteration matching in S4. Calculate the target's velocity phase offset and apply it to the S5 corrected signal. Multiply by phase factor Further phase compensation is performed to further eliminate phase distortion caused by velocity, resulting in... In the formula, At the speed of light, This is the optimal speed estimate. For carrier frequency, For slow time variables, It is the imaginary unit.

[0179] The focus index of the image is calculated and fed back to the Bayesian optimizer to update the Gaussian process model.

[0180] S7: Identify the current non-uniform time grid Using the Non-Uniform Fast Fourier Transform (NUFFT) algorithm, Transformed to a uniform Doppler frequency domain, the focused two-dimensional image is obtained. This step is executed in parallel on a multi-core CPU, using the MATLAB Parallel Computing Toolbox to accelerate matrix operations.

[0181] After completing the energy focusing in S7, the focusing index of the image is calculated and fed back to the Bayesian optimizer to update the Gaussian process model. Then it is determined whether the preset maximum number of iterations (e.g., 30 times) has been reached or whether the objective function has converged: if not, return to S3 to continue the next round of iteration; if so, proceed to S8.

[0182] S8: Extract optimal parameters The corresponding focused image. Based on the optimal speed. Determine the target's true velocity and output the target's distance, true velocity, acceleration, and image results.

[0183] The processing results obtained by the method according to the present invention are as follows Figure 2 As shown.

[0184] Figure 2 (a) shows the radar echo range-Doppler spectrum before processing. Due to severe velocity ambiguity and acceleration, the target energy diverges into a curved bright line, making it undetectable.

[0185] Figure 2 (b) shows the result of performing only traditional slow time-domain FFT processing, where the target is severely out of focus due to uncorrected distance migration.

[0186] Figure 2 (c) in this embodiment shows the intermediate results after SOKT processing. It can be seen that the originally curved trajectory was corrected into a tilted straight line (the second-order term was eliminated), which proves the effectiveness of SOKT.

[0187] Figure 2 (d) in the figure shows the intermediate results after frequency domain IAR processing in this embodiment. The tilted straight line was corrected to a horizontal straight line (eliminating distance shift), verifying that the frequency domain multiplication method is equivalent to the time domain rotation.

[0188] Figure 2 (e) in Figure 2Figure (f) shows the final focused imaging result after convergence via NUFFT and Bayesian optimization. The target energy is perfectly focused into a sharp, strong point, and the previously overlapping multiple moving targets can be clearly distinguished.

[0189] Figure 2 Figure (g) shows the convergence curve of the objective function value during Bayesian optimization. Compared to full-space grid search, this invention utilizes Bayesian optimization to converge to the global optimum in a very small number of iterations (only 320 iterations within the search matching range), significantly reducing the number of iterations compared to the 9,424 iterations required for 0.01 precision in other searches. Therefore, computational time is significantly reduced.

[0190] Performance comparison analysis:

[0191] Compared to traditional IAR algorithms based on time-domain interpolation, the frequency-domain phase processing employed in this invention improves the computation speed by approximately an order of magnitude. Simultaneously, combined with multi-core CPU parallel technology, the processing time for radar data of size is reduced from tens of minutes to minutes, meeting the requirements for near real-time processing.

[0192] The present invention also provides a system for estimating maneuvering target parameters and detecting long-term coherent accumulation. The system can be implemented by executing the process steps of the method for estimating maneuvering target parameters and detecting long-term coherent accumulation. That is, those skilled in the art can understand the method for estimating maneuvering target parameters and detecting long-term coherent accumulation as a preferred embodiment of the system for estimating maneuvering target parameters and detecting long-term coherent accumulation.

[0193] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0194] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for estimating the parameters of a maneuvering target and detecting long-term coherent accumulation, characterized in that, include: Step S1: Preprocess the original radar radio frequency echo signal to output a discrete complex baseband signal; Step S2: Perform pulse compression on the discrete complex baseband signal to output a frequency domain signal that is focused in the range direction but still contains range migration in the azimuth direction; Step S3: Process the frequency domain signal using the second-order Keystone transform to correct the target's second-order range migration and output a second-order corrected signal that eliminates range-azimuth coupling; Step S4: Based on Bayesian optimization, perform velocity search matching on the second-order corrected signal, and output the radial velocity of the candidate target and the corresponding velocity matching phase factor; wherein, step S4 specifically includes: Step S4.1: Based on the pulse repetition frequency, determine the baseband unambiguous velocity range, introduce integer ambiguity numbers and blind velocity components, and construct an extended velocity search space containing Doppler ambiguity; Step S4.2: By fitting the target function to maximize the focusing index using a Gaussian process as a surrogate model, and based on a preset velocity range, using the acquisition function, adaptively predict the radial velocity of the most promising candidate target in the expanded velocity search space. The expression for the radial velocity of the candidate target is as follows: In the formula, The fundamental frequency falling within the unambiguous interval; For carrier frequency; Baseband speed; This refers to the radar blind speed component; It is a fuzzy number and is an integer; The radial velocity of the target; The speed of light; The pulse repetition frequency; Step S4.3: Based on the radial velocity of the candidate target, generate the corresponding velocity matching phase factor, wherein the expression for the velocity matching phase factor is as follows: In the formula, The speed of light; The target radial velocity; For carrier frequency; For slow time variables, The imaginary unit; Step S5: Based on the IAR algorithm, combining the second-order correction signal and the radial velocity of the candidate target, correct the first-order range migration of the target, and output a first-order correction signal with completely eliminated range-azimuth coupling; wherein, step S5 specifically includes: Step S5.1: Based on the radial velocity of the candidate target, construct a new fast-time variable; wherein the expression of the fast-time variable is as follows: In the formula, For the original fast time variable, For the new fast time variable, Baseband speed; For radar blind speed component, For slow time variables, The speed of light; Step S5.2: Based on the shift property of the Fourier transform, the phase compensation factor containing the new fast time variable is multiplied by the second-order correction signal in the range frequency domain to obtain the first-order correction signal with completely eliminated range-azimuth coupling; wherein, the expression for the phase compensation factor containing the new fast time variable is: In the formula, For distance frequency, Baseband speed; For radar blind speed component, For slow time variables, At the speed of light, The imaginary unit; Step S6: Based on the velocity matching phase factor, perform velocity phase compensation on the first-order correction signal and output the phase matching signal after first-order velocity phase cancellation. Step S7: Perform a non-uniform fast Fourier transform on the phase-matched signal in the slow time domain to achieve target energy focusing and output the Doppler domain peak signal after slow time focusing; Among them, repeat steps S4-S7 until the iteration termination condition is met, and then proceed to step S8; Step S8: Analyze the peak position and amplitude characteristics of the Doppler domain peak signal, and calculate the complete motion parameters of the target in combination with the system parameters.

2. The method according to claim 1, characterized in that, The expression for the frequency domain signal in S2 is: In the formula, For frequency domain signals, The signal amplitude is in the distance frequency domain and slow time domain. For distance frequency, For carrier frequency, For location, slow time, For signal bandwidth, For the target radial velocity, For the target radial acceleration, At the speed of light, The initial radial distance to the target. It is the imaginary unit.

3. The method according to claim 2, characterized in that, S3 includes: Step S3.1: Construct new slow-time variables ,in, The item includes distance frequency With direction and slow time The coupling term is expressed as follows: In the formula, For location, slow time, For distance frequency, For carrier frequency, For slow-time variables; Step S3.2: Use the new slow-time variable replace , release distance frequency With direction and slow time The quadratic coupling relationship corrects the distance curvature, which varies parabolically with slow time, into a linearly varying distance travel function, i.e., a second-order correction signal, as follows: Right now: In the formula, For frequency domain signals, For time-domain signals, For distance frequency, For carrier frequency, For slow time variables, The time delay corresponding to the initial distance. The signal amplitude is in the distance frequency domain and slow time domain. For signal bandwidth, For fast time variables, For the target radial velocity, For the target radial acceleration, At the speed of light, It is the imaginary unit.

4. The method according to claim 1, characterized in that, Step S6 includes: Multiplying the first-order correction signal by the velocity matching phase factor yields the first-order velocity phase cancellation phase-matched signal.

5. The method according to claim 1, characterized in that, Step S7 specifically includes: Step S7.1: Using non-uniform fast Fourier transform, the phase-matching signal is transformed to a uniform Doppler frequency domain to obtain a focused two-dimensional image; Step S7.2: Calculate the focus index of the two-dimensional image, which is the peak intensity of the image; feed the focus index back to the Bayesian optimizer to update the Gaussian process model.

6. The method according to claim 1, characterized in that, The iteration termination conditions include: reaching a preset number of iterations or the target optimization function reaching a convergence condition.

7. The method according to claim 1, characterized in that, Step S8 specifically includes: After the iteration is completed, the Gaussian is updated based on the maximum value of the focusing index and recorded as the optimal velocity. Based on the optimal velocity, the true velocity of the target is determined, and the target's distance, true velocity, acceleration, and image results are output.

8. A system for estimating parameters of a maneuvering target and for long-term coherent accumulation detection, characterized in that, include: Module M1: Preprocesses the raw radar radio frequency echo signal and outputs a discrete complex baseband signal; Module M2: Performs pulse compression on discrete complex baseband signals and outputs frequency domain signals that are focused in the range direction but still contain range migration in the azimuth direction; Module M3: Uses second-order Keystone transform to process frequency domain signals, corrects target second-order range migration, and outputs a second-order corrected signal that eliminates range-azimuth coupling; Module M4: Based on Bayesian optimization, it performs velocity search matching on the second-order corrected signal and outputs the radial velocity of candidate targets and the corresponding velocity matching phase factor; wherein, module M4 specifically includes: Module M4.1: Based on the pulse repetition frequency, the baseband unambiguous velocity range is determined, and integer ambiguity numbers and blind velocity components are introduced to construct an extended velocity search space containing Doppler ambiguity; Module M4.2: It uses a Gaussian process as a surrogate model to fit the target optimization function that maximizes the focusing index. Based on a preset velocity range, it adaptively predicts the radial velocity of the most promising candidate target within the expanded velocity search space using a data acquisition function. The expression for the radial velocity of the candidate target is as follows: In the formula, The fundamental frequency falling within the unambiguous interval; For carrier frequency; Baseband speed; This refers to the radar blind speed component; It is a fuzzy number and is an integer; The radial velocity of the target; The speed of light; The pulse repetition frequency; Module M4.3: Based on the radial velocity of the candidate target, generate the corresponding velocity matching phase factor, where the expression for the velocity matching phase factor is as follows: In the formula, The speed of light; The target radial velocity; For carrier frequency; For slow time variables, The imaginary unit; Module M5: Based on the IAR algorithm, combining the second-order correction signal and the radial velocity of the candidate target, it corrects the first-order range migration of the target and outputs a first-order correction signal that completely eliminates range-azimuth coupling; wherein, module M5 specifically includes: Module M5.1: Constructs a new fast-time variable based on the radial velocity of the candidate target; wherein the expression of the fast-time variable is as follows: In the formula, For the original fast time variable, For the new fast time variable, Baseband speed; For radar blind speed component, For slow time variables, The speed of light; Module M5.2: Based on the shift property of the Fourier transform, the phase compensation factor containing the new fast time variable is multiplied by the second-order correction signal in the range frequency domain to obtain the first-order correction signal with completely eliminated range-azimuth coupling; wherein, the expression for the phase compensation factor containing the new fast time variable is: In the formula, For distance frequency, Baseband speed; For radar blind speed component, For slow time variables, At the speed of light, The imaginary unit; Module M6: Based on the velocity matching phase factor, it performs velocity phase compensation on the first-order correction signal and outputs the phase matching signal after first-order velocity phase cancellation; Module M7: Performs a non-uniform fast Fourier transform on the phase-matched signal in the slow-time domain to achieve target energy focusing and outputs the Doppler domain peak signal after slow-time focusing; Among them, modules M4-M7 are repeatedly triggered until the iteration end condition is met, at which point module M8 is executed. Module M8: Analyzes the peak position and amplitude characteristics of the Doppler domain peak signal, calculates and outputs the complete motion parameters of the target in combination with system parameters.

Images