An unmanned aerial vehicle target coherent integration detection method based on Bayesian compressed sensing

Abstract

The application discloses a kind of unmanned plane target coherent accumulation detection methods based on bayesian compressive sensing, comprising: S1, the radar echo of unmanned plane is compressed in distance, and micro-doppler signal is separated using zero space tracking algorithm, obtain the radar echo of unmanned plane;S2, based on the obtained radar echo of unmanned plane, the distance of the envelope of unmanned plane is corrected using Keystone conversion to walk, and the slow-time data of the position of distance envelope peak value is extracted as the observation signal of sparse reconstruction by iteration;S3, design sensing matrix, and the observation signal is reconstructed using bayesian compressive sensing algorithm to estimate the motion parameter of each order of unmanned plane;S4, according to the motion parameter of each order of unmanned plane, construct motion phase compensation filter, the echo of unmanned plane is phase compensated in pulse-range frequency domain, and the frequency domain echo signal after compensation is coherent accumulation and is detected by CFAR.The application has the advantages of high parameter estimation accuracy, low computational complexity, strong robustness and the like.

Description

Technical Field

[0001] This invention relates to the field of radar signal technology, and more specifically, to a method for detecting UAV targets based on Bayesian compressed sensing coherent accumulation. Background Technology

[0002] In recent years, with the rapid development of UAV technology in both military and civilian fields, the detection of highly maneuverable UAVs, posing a significant threat, has gradually become a major challenge in radar detection. UAVs, as typical small, slow-moving targets, generally have extremely low radar scattering coefficients and are significantly affected by ground clutter. Furthermore, their high maneuverability and micro-Doppler effect also impact radar detection, making UAV radar detection in low signal-to-noise ratio (SNR) environments extremely challenging. Generally, coherent accumulation can significantly improve the echo SNR of UAV targets and is the most effective way to achieve better detection and tracking performance. However, for highly maneuverable UAVs (defined here as UAV targets with velocity, acceleration, and jerk), they have the following specific characteristics: firstly, the micro-Doppler signal generated by the UAV rotor interferes with coherent accumulation performance, requiring separation of the micro-Doppler signal before coherent accumulation; secondly, the echo of highly maneuverable UAVs contains high-order motion phases, requiring effective estimation of their motion parameters such as velocity, acceleration, and jerk to achieve phase compensation and coherent accumulation.

[0003] Currently, existing methods for estimating motion parameters and detecting coherent accumulation in highly maneuverable unmanned aerial vehicles (UAVs) mainly fall into two categories: one is parameter search-based methods, represented by the Generalized Radon Fourier Transform (GRFT). These methods estimate motion parameters and compensate for motion phase in highly maneuverable UAVs through multi-dimensional parameter search. However, the main problems with this type of detection method are high computational complexity, excessive computation time, and interference from fuzzy sidelobes. The other type of detection method is coherent accumulation algorithms based on time-frequency transforms, represented by the fractional Fourier transform (FrFT) and Lv's distribution (LVD). The estimation accuracy of this type of method is limited by time-frequency resolution, making it difficult to effectively improve the estimation accuracy and introducing inherent errors. Furthermore, this type of method only estimates parameters for the signal modulation frequency during design, and cannot effectively estimate the third-order motion phase introduced by jerks. Both of the above-mentioned coherent accumulation detection methods have shortcomings. The aforementioned technical problems mean that there is still considerable room for improvement in the estimation of motion parameters and coherent accumulation detection methods for highly maneuverable UAVs. Therefore, it is urgent to design a coherent accumulation detection method for UAV targets based on Bayesian compressed sensing. Summary of the Invention

[0004] The purpose of this invention is to provide a coherent accumulation detection method for UAV targets based on Bayesian compressed sensing, so as to overcome the defects of the existing technology.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for detecting UAV targets based on Bayesian compressed sensing coherent accumulation includes the following steps:

[0007] S1. Perform range compression on the UAV's radar echo and use the null space tracking algorithm to separate the micro-Doppler signal to obtain the UAV's radar echo;

[0008] S2. Based on the obtained UAV radar echo, the distance movement of the UAV envelope is corrected by Keystone transform, and the slow time data of the peak position of the distance envelope is extracted as the observation signal for sparse reconstruction.

[0009] S3. Design the perception matrix and use the Bayesian compressed sensing algorithm to reconstruct the observation signal and estimate the motion parameters of the UAV at each order.

[0010] S4. Construct a motion phase compensation filter based on the motion parameters of the UAV at each order, perform phase compensation on the UAV echo in the pulse-range frequency domain, and perform coherent accumulation and CFAR detection on the compensated frequency domain echo signal.

[0011] Further, step S1 specifically includes:

[0012] S11, the radar echo s(τ,t) of the UAV m The range is compressed using a fast-time Fourier transform to obtain the range-compressed echo s(f). r ,t m );

[0013] S12. Employing a zero-space tracking algorithm to separate the UAV radar echo s(f) r ,t m The fuselage echo signal and micro-Doppler signal in the separated UAV radar echo s′(τ,t) m ) represents the radar echo signal s from the drone's fuselage. f (τ,t m ).

[0014] Furthermore, in step S11, the radar echo s(τ,t) of the UAV m The expression for ) is:

[0015] s(τ,t m ) = s f (τ,t m )+s rot (τ,tm )

[0016] In the formula, τ represents fast time, t m The slow time is represented by m = 1, 2, ..., M, where m represents the pulse number and M represents the number of pulses accumulated by coherence. (s) f (τ,t m ) represents the echo from the drone's fuselage, s rot (τ,t m This indicates the micro-Doppler signal introduced by the UAV rotor.

[0017] The expression for the drone fuselage echo is:

[0018]

[0019] In the formula, A f The amplitude of the echo signal from the fuselage is represented by f0, the center frequency by μ, the modulation frequency of the emitted waveform by c, and the speed of light by r(t). m The distance history of a highly maneuverable unmanned aerial vehicle (UAV) is expressed as follows:

[0020]

[0021] In the formula, r0 represents the initial distance of the drone, v0 represents the speed of the drone, a1 represents the acceleration of the drone, and a2 represents the jerk of the drone.

[0022] Micro-Doppler signal s introduced by UAV rotor r o t (τ,t m The expression for ) is:

[0023]

[0024] In the formula, K represents the total number of rotors in the drone, and I represents the total number of blades in the drone. The echo of the i-th blade on the k-th rotor is expressed as:

[0025]

[0026]

[0027] In the formula, This represents the rotational frequency of the k-th rotor. Let L represent the initial phase of the i-th blade on the k-th rotor, L represent the length of the UAV blade, and β represent the line-of-sight angle of the UAV relative to the radar when the radar is located at the origin.

[0028] The echo s(f) after distance compression r ,t m The expression for ) is:

[0029] s(f r ,t m )=F[s f (τ,t m )+s rot (τ,t m )]=s f (f r ,t m )+s rot (f r ,t m );

[0030] In the formula, F[*] denotes the fast-time Fourier transform, s f (f r ,t m The expression for the fast-time Fourier transform of the UAV fuselage echo is as follows:

[0031]

[0032] Among them, A′ f T represents the amplitude of the fuselage echo signal after distance compression. r λ represents the pulse repetition time, and λ0 represents the carrier wavelength.

[0033] Furthermore, in step S12, the input signal of the null-space tracking algorithm is the distance-compressed UAV echo s(f) r ,t m By traversing all range cells, the micro-Doppler signal introduced by the UAV rotor is separated, and the separated signal is then subjected to a fast-time inverse Fourier transform to obtain the UAV radar echo s′(τ,t) in the time domain. m ), which is the echo signal s from the drone's fuselage. f (τ,t m ).

[0034] Further, step S2 specifically includes:

[0035] S21. Designing slow-time variables in the Keystone transformation.

[0036] S22, Change the slow time variable Substituting the UAV radar echo s′(τ,t) after micro-Doppler signal separation in step S1... m The time-domain UAV echo signal after Keystone transformation is obtained from the data. Time-domain UAV echo signals The distance-frequency domain signal after Keystone transform is obtained by performing a fast-time Fourier transform. Iterate through all distance cells to obtain the distance cell containing the peak value of the UAV's distance envelope, and extract the slow-time data within that distance cell. This slow time data As an observation signal in the sparse reconstruction process.

[0037] Furthermore, the time-domain UAV echo signal in step S22 The expression is:

[0038]

[0039] Keystone transform distance-frequency domain signal The expression is:

[0040]

[0041] In the formula, f′ KT The distance cell representing the location of the peak of the distance envelope is expressed as follows:

[0042]

[0043] Iterate through all distance cells to obtain the distance cell containing the peak value of the UAV's distance envelope, and extract the slow-time data within that distance cell. The expression is:

[0044]

[0045] In the formula, A″ f =A′ f exp(j4πr0 / λ0) represents the complex amplitude including the envelope of the UAV echo signal after Keystone transformation.

[0046] Furthermore, step S3 specifically includes:

[0047] S31. Design and construct the sparse reconstruction perception matrix Φ;

[0048] S32, Based on the sensing matrix Φ and the observed signal Substituting the compressed sensing signal model y = Φx, a one-dimensional sparse signal vector is reconstructed using the Bayesian compressed sensing algorithm.

[0049] S33, Transform a one-dimensional sparse signal vector Transform to 3D matrix The elements in matrix X satisfy:

[0050] [X] n,p,q =x (n-1)×PQ+(p-1 ) ×Q+q ;

[0051] S34, from a three-dimensional matrix The coordinates of the position of the maximum absolute value are estimated and represented as follows: The estimated values ​​of the drone's speed, acceleration, and jerk are obtained based on the location of the maximum value, as follows:

[0052]

[0053] Output the motion parameter estimation results of the UAV

[0054] Further, step S31 specifically includes:

[0055] S311. Define a set representing the range of velocity, acceleration, and jerk parameters:

[0056] The set of velocity parameter ranges Γ v ={(nN / 2)Δv|n=1,2,...,N}, the set of acceleration parameters. Set of ranges of jerkness parameters

[0057] In the formula, N, P, and Q represent the number of elements in the velocity, acceleration, and jerk parameter sets, respectively, and Δv, Δa1, and Δa2 represent the search step size of the velocity, acceleration, and jerk parameters, respectively.

[0058] S312. Search set based on parameters. To design a sparse reconstruction perception matrix

[0059] S313. Construct a perception matrix, representing it as N submatrices, Φ = [Φ1, Φ2, ..., Φ]. N ],in, This represents the nth submatrix, which corresponds to Γ. v The nth element in;

[0060] Submatrix Φ n Represented as P submatrices, Φ n =[Φ n,1 ,Φ n,2 ,...,Φ n,P ],in, Represents the p-th submatrix, corresponding to The p-th element in the submatrix Φ n,p The element in the m-th row and q-th column is represented as:

[0061]

[0062] In the formula, matrix Φ n,p The element [Φ]n,p ] m,q The coordinates in the perception matrix Φ are represented as follows:

[0063] [Φ] m,(n-1)×PQ+(p-1)×Q+q =[Φ n,p ] m,q .

[0064] Furthermore, step S4 specifically includes:

[0065] S41. Construct a motion phase compensation filter based on the estimated UAV speed, acceleration, and jerk speed, and analyze the UAV echo in the pulse-range frequency domain. Perform phase compensation;

[0066] S42. Perform a slow-time Fourier transform on the frequency domain echo signal after phase compensation to obtain the coherent accumulation result in the range-azimuth frequency domain, and perform CFAR detection.

[0067] Furthermore, in step S41, the pulse-range frequency domain pulse echo of the UAV is... The expression is:

[0068]

[0069] Using the estimated motion parameters of the l-th target Construct the compensation function h l (f,t m Its expression is as follows:

[0070]

[0071] The expression for the compensated signal is:

[0072] Compared with existing technologies, the advantages of this invention are as follows: First, it separates the UAV fuselage echo and micro-Doppler signal using the NSP algorithm. Then, it uses Keystone transform to correct the distance travel of the envelope. Next, it achieves high-precision estimation of target motion parameters (including velocity, acceleration, and jerk) using a Bayesian compressed sensing method. Finally, it uses the estimation results to complete phase compensation and coherent accumulation. This invention has advantages such as high parameter estimation accuracy, low computational complexity, and strong robustness. In summary, this invention significantly improves existing coherent accumulation detection and motion parameter estimation technologies for highly maneuverable UAVs. It can be widely applied in the search and surveillance of airborne UAV targets and is effectively suitable for motion parameter estimation and coherent accumulation of highly maneuverable UAVs under low signal-to-noise ratio conditions, showing significant application prospects. Attached Figure Description

[0073] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0074] Figure 1 This is a flowchart of the UAV target coherent accumulation detection method based on Bayesian compressed sensing according to the present invention.

[0075] Figure 2 This is a schematic diagram illustrating the separation of micro-Doppler signals using the null space tracking algorithm of this invention.

[0076] Figure 3 This is a schematic diagram illustrating the use of Keystone transform to correct the distance travel of a UAV echo envelope in this invention.

[0077] Figure 4 This is a comparison chart of the coherent accumulation detection performance of the present invention and existing technologies under simulation data.

[0078] Figure 5 This is a comparison chart of the coherent accumulation detection probability of the present invention and existing technologies under simulation data.

[0079] Figure 6 This is a comparison chart of the coherent accumulation detection performance of the present invention and existing data under measured data.

[0080] Figure 7 This is a graph showing the coherent accumulation results obtained after estimating the motion parameters of the UAV using different methods and compensating for the motion phase.

[0081] Figure 8 This is a quantitative analysis chart of the energy accumulation performance of different methods. Detailed Implementation

[0082] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention.

[0083] The special meanings of the mathematical symbols used below are declared: bold lowercase letters, such as x, represent vectors; bold uppercase letters, such as X, represent matrices. n Let [X] represent the nth element of vector x. m,n Let X represent the element in the m-th row and n-th column of matrix X.

[0084] See Figure 1As shown, this embodiment discloses a method for UAV target coherent accumulation detection based on Bayesian compressed sensing, including the following steps:

[0085] Step S1: Compress the distance of the UAV's radar echo and separate the micro-Doppler signal using the null space tracking algorithm to obtain the UAV's radar echo.

[0086] Specifically, in this embodiment, step S1 includes:

[0087] Step S11: The radar echo s(τ,t) of the UAV... m The range is compressed using a fast-time Fourier transform to obtain the range-compressed echo s(f). r ,t m ).

[0088] Step S12: Separate the UAV radar echo s(f) using the zero-space tracking algorithm. r ,t m The fuselage echo signal and micro-Doppler signal in the separated UAV radar echo s′(τ,t) m ) represents the radar echo signal s from the drone's fuselage. f (τ,t m ).

[0089] Among them, the radar echo s(τ,t) of the UAV in step S11 m The expression for ) is:

[0090] s(τ,t m ) = s f (τ,t m )+s rot (τ,t m )

[0091] In the formula, τ represents fast time, t m The slow time is represented by m = 1, 2, ..., M, where m represents the pulse number and M represents the number of pulses accumulated by coherence. (s) f (τ,t m ) represents the echo from the drone's fuselage, s rot (τ,t m ) represents the micro-Doppler signal introduced by the drone rotor.

[0092] The expression for the drone fuselage echo is:

[0093]

[0094] In the formula, A f The amplitude of the echo signal from the fuselage is represented by f0, the center frequency by μ, the modulation frequency of the emitted waveform by c, and the speed of light by r(t). mThe distance history of a highly maneuverable unmanned aerial vehicle (UAV) is expressed as follows:

[0095]

[0096] In the formula, r0 represents the initial distance of the drone, v0 represents the speed of the drone, a1 represents the acceleration of the drone, and a2 represents the jerk of the drone.

[0097] Micro-Doppler signal s introduced by UAV rotor rot (τ,t m The expression for ) is:

[0098]

[0099] In the formula, K represents the total number of rotors in the drone, and I represents the total number of blades in the drone. The echo of the i-th blade on the k-th rotor is expressed as:

[0100]

[0101] In the formula, This represents the rotational frequency of the k-th rotor. Let L represent the initial phase of the i-th blade on the k-th rotor, L represent the length of the UAV blade, and β represent the line-of-sight angle of the UAV relative to the radar when the radar is located at the origin.

[0102] The echo s(f) after distance compression r ,t m The expression for ) is:

[0103] s(f r ,t m )=F[s f (τ,t m )+s rot (τ,t m )]=s f (f r ,t m )+s rot (f r ,t m );

[0104] In the formula, F[*] denotes the fast-time Fourier transform, s f (f r ,t m The expression for the fast-time Fourier transform of the UAV fuselage echo is as follows:

[0105]

[0106] Among them, A′ fT represents the amplitude of the fuselage echo signal after distance compression. r λ represents the pulse repetition time, and λ0 represents the carrier wavelength.

[0107] UAV echo s(f) after range compression r ,t m s in ) rot (f r ,t m The expression for the fast-time Fourier transform of the micro-Doppler signal is:

[0108]

[0109] Where p is an integer ranging from negative infinity to positive infinity, C q This represents the complex coefficients generated during the Fourier transform. Micro-Doppler signals can create periodic ambiguous sidelobes in the azimuth frequency domain, causing false alarms. Therefore, it is necessary to separate the UAV fuselage echo signal and the micro-Doppler signal before coherent accumulation.

[0110] In this embodiment, the input signal of the null-space tracking algorithm in step S12 is the distance-compressed UAV echo s(f) r ,t m By traversing all range cells, the micro-Doppler signal introduced by the UAV rotor is separated, and the separated signal is then subjected to a fast-time inverse Fourier transform to obtain the UAV radar echo s′(τ,t) in the time domain. m It is approximately equal to the echo signal s from the drone's fuselage. f (τ,t m ).

[0111] Step S2: Based on the obtained UAV radar echo, use Keystone transform to correct the distance movement of the UAV envelope, and iterate through and extract slow-time data of the peak position of the distance envelope as the observation signal for sparse reconstruction.

[0112] Specifically, in this embodiment, step S2 includes:

[0113] Step S21: Design the slow-time variables in the Keystone transformation

[0114] Step S22: Change the slow time variable Substituting the UAV radar echo s′(τ,t) after micro-Doppler signal separation in step S1... m The time-domain UAV echo signal after Keystone transformation is obtained from the data. Time-domain UAV echo signals The distance-frequency domain signal after Keystone transform is obtained by performing a fast-time Fourier transform. Iterate through all distance cells to obtain the distance cell containing the peak value of the UAV's distance envelope, and extract the slow-time data within that distance cell. This slow time data As an observation signal in the sparse reconstruction process.

[0115] Among them, the time-domain UAV echo signal in step S22 The expression is:

[0116]

[0117] Keystone transform distance-frequency domain signal The expression is:

[0118]

[0119] In the formula, f′ KT The distance cell representing the location of the peak of the distance envelope is expressed as follows:

[0120]

[0121] In the above equation, the peak position of the distance envelope is f′ KT Although it still varies with the slow time variable t m While the range changes, due to the limited output power of small UAV platforms, their maximum acceleration a1 and maximum jerk a2 are generally insufficient to introduce an envelope across distance cells during the coherent accumulation time. Therefore, by traversing all distance cells to obtain the distance cell containing the peak value of the UAV's range envelope, slow-time data within that distance cell is extracted. The expression is:

[0122]

[0123] In the formula, A″ f =A′ f exp(j4πr0 / λ0) represents the complex amplitude including the envelope of the UAV echo signal after Keystone transformation.

[0124] In other words, the slow-time data within the range cell where the distance envelope peak of the UAV is located. It will serve as an observation signal during the sparse reconstruction process.

[0125] Step S3: Design the perception matrix and use the Bayesian compressed sensing algorithm to reconstruct the observation signal and estimate the motion parameters of the UAV at each order.

[0126] Specifically, in this embodiment, step S3 includes:

[0127] Step S31: Design and construct the sparse reconstruction perception matrix Φ.

[0128] Step S32: Based on the sensing matrix Φ and the observed signal Substituting the compressed sensing signal model y = Φx, a one-dimensional sparse signal vector is reconstructed using the Bayesian compressed sensing algorithm.

[0129] Step S33: Convert the one-dimensional sparse signal vector Transform to a three-dimensional matrix X∈£ N×P×Q The elements in matrix X satisfy:

[0130]

[0131] Step S34, from the three-dimensional matrix The coordinates of the position of the maximum absolute value are estimated and represented as follows: The estimated values ​​of the drone's speed, acceleration, and jerk are obtained based on the location of the maximum value, as follows:

[0132]

[0133]

[0134]

[0135] Output the motion parameter estimation results of the UAV

[0136] Specifically, step S31 includes:

[0137] Step S311: Define a set representing the range of velocity, acceleration, and jerk parameters:

[0138] The set of velocity parameter ranges Γ v ={(nN / 2)Δv|n=1,2,...,N}, the set of acceleration parameter ranges Γ a1 ={(pP / 2)Δa1|p=1,2,...,P}, the set Γ of the range of jerk parameters. a2 ={(qQ / 2)Δa2|q=1,2,...,Q};

[0139] In the formula, N, P, and Q represent the number of elements in the velocity, acceleration, and jerk parameter sets, respectively, and Δv, Δa1, and Δa2 represent the search step size of the velocity, acceleration, and jerk parameters, respectively.

[0140] Step S312. Search set based on parameters To design a sparse reconstruction perception matrix

[0141] Step S313: Construct the perception matrix, representing it as N sub-matrices, Φ = [Φ1, Φ2, ..., Φ N ], where Φ n ∈￡ M×PQ This represents the nth submatrix, which corresponds to Γ. v The nth element in;

[0142] Submatrix Φ n Represented as P submatrices, Φ n =[Φ n,1 ,Φ n,2 ,...,Φ n,P ],in, Represents the p-th submatrix, corresponding to The p-th element in the submatrix Φ n,p The element in the m-th row and q-th column is represented as:

[0143]

[0144] In the formula, matrix Φ n,p The element [Φ] n,p ] m,q The coordinates in the perception matrix Φ are represented as follows:

[0145]

[0146] Step S4: Construct a motion phase compensation filter based on the motion parameters of the UAV at each order, perform phase compensation on the UAV echo in the pulse-range frequency domain, and perform coherent accumulation and CFAR detection on the compensated frequency domain echo signal.

[0147] Specifically, in this embodiment, step S4 includes:

[0148] Step S41: Construct a motion phase compensation filter based on the estimated UAV speed, acceleration, and jerkness, and analyze the UAV echo in the pulse-range frequency domain. Perform phase compensation;

[0149] Step S42: Perform a slow-time Fourier transform on the phase-compensated frequency domain echo signal to obtain the coherent accumulation result in the range-azimuth frequency domain, and perform CFAR detection.

[0150] In step S41, the pulse-range frequency domain pulse echo of the UAV is compared. The expression is:

[0151]

[0152] Using the estimated motion parameters of the l-th target Construct the compensation function h l (f,tm Its expression is as follows:

[0153]

[0154] The expression for the compensated signal is:

[0155] The invention will be further illustrated below through experimental simulation examples. It is assumed that the system parameters and high-maneuverability UAV parameters from the numerical simulation are shown in Table 1 below:

[0156] Table 1. Simulation system parameters and target parameters for this example.

[0157]

[0158] The system parameters and motion parameters of the high-maneuverability UAV set in the numerical simulation are shown in Table 1. The range compression echo of the high-maneuverability UAV is obtained through simulation using these parameters. Figure 2 As shown in (a), the range-azimuth frequency domain UAV echo signal is obtained by performing an azimuth-direction Fourier transform on the range-compressed UAV echo. Figure 2 As shown in (b), from Figure 2 As shown in (b), the micro-Doppler signal introduced by the UAV rotor produces obvious periodic fuzzy sidelobes in the azimuth frequency domain. The effect of using the NSP algorithm to separate the micro-Doppler signal from the one-dimensional slow-time data within each range cell is as follows: Figure 2 As shown in (c), the echo signal separated by the NSP algorithm only contains the echo component of the UAV fuselage and does not contain the micro-Doppler signal component. By traversing all range cells, the two-dimensional UAV echo signal separated by the NSP algorithm can be obtained, as shown below. Figure 2 As shown in (d). From Figure 2 As can be seen in (d), the azimuth frequency domain ambiguity sidelobes introduced by the micro-Doppler signal are effectively separated, and the UAV echo signal processed by the NSP algorithm can be approximated as the echo signal of the UAV fuselage.

[0159] Figure 3 (a) shows the UAV echo signal after range compression. The figure clearly shows the envelope range movement. Figure 3 (b) shows the UAV echo signal after Keystone transformation. It can be seen that the distance movement of the UAV envelope is corrected by Keystone transformation, and the UAV envelope peak is located in the same distance cell at different slow times.

[0160] The slow-time data within the range cells containing the peak value of the UAV echo envelope is extracted by traversing the range cells. The motion parameters of the highly maneuverable UAV are estimated using the method proposed in this invention, along with existing parameter search and time-frequency transformation methods, resulting in the following... Figure 4The estimation results shown are as follows: Figure 4 (a) The estimation result obtained using the GRFT method in the parameter search algorithm class, from Figure 4 As can be seen in (a), the estimation results obtained using the GRFT method are... Figure 4 As can be seen from (a), although this method can estimate the motion parameters of the high-maneuverability UAV by the position of the maximum peak, the method has high computational complexity and long computation time, with a computation time of 98.43s, and there are obvious fuzzy sidelobes in the estimation results. Figure 4 (b) The estimation results obtained by using the FrFT method in the time-frequency transformation class are shown in the figure. It can be seen from the figure that the FrFT method cannot effectively focus the echo energy of a highly maneuverable UAV with rapid movement in the time-frequency domain and cannot accurately estimate the motion parameters of the target. Figure 4 (c) The estimation results obtained by using the LVD method in the time-frequency transformation class of methods can be seen from the figure. It can be seen that the LVD method also cannot focus the highly maneuverable UAV target on a point in the time-frequency plane, so it cannot accurately estimate the target's motion parameters. Figure 4 (d) To facilitate the display of the estimation results obtained using the method of this invention, the one-dimensional sparse vector reconstructed by the Bayesian compressed sensing algorithm is transformed into a two-dimensional plane for easy illustration. Figure 4 (d) The method of the present invention can effectively estimate the motion parameters of highly maneuverable UAVs, and the estimation results are in complete agreement with the true values. Figure 4 (e) To estimate the motion parameters of the highly maneuverable UAV using the method of this invention, a phase compensation filter is constructed to compensate the UAV echo signal, resulting in coherent accumulation. The figure qualitatively shows that the method of this invention can effectively achieve coherent accumulation of the target energy of the highly maneuverable UAV, and... Figure 4 The GRFT method shown in (a) is unaffected by fuzzy sidelobes.

[0161] The coherent accumulation performance of the four algorithms was quantitatively and statistically analyzed through 500 Monte Carlo simulations under different signal-to-noise ratios. The change curves of detection probability of different methods under different signal-to-noise ratios are shown in the figure. Figure 5 As shown, the false alarm probability was set to 10 in this experiment. -5 .from Figure 5 As can be seen from the above, the accumulation performance of the proposed method is similar to that of the GRFT method based on multidimensional parameter search. However, the proposed method achieves motion parameter estimation through sparse reconstruction, eliminating the need for multidimensional parameter search, and its computational efficiency is significantly better than that of the GRFT method. Compared to FrFT and LVD, two coherent accumulation methods based on time-frequency transform, the proposed method exhibits significantly stronger robustness, meaning that under the same signal-to-noise ratio, the detection probability is significantly higher than that of FrFT and LVD methods. At the detection probability P... dUnder the condition of 0.8, the signal-to-noise ratio of the proposed method is improved by 4dB and 7dB compared with FrFT and LVD methods, respectively. Compared with GRFT method, the proposed method has a significant advantage in computational efficiency. Under the same hardware conditions, the proposed method takes only 1.55s, which is much shorter than GRFT method (98.43s).

[0162] The invention will be further illustrated below through a set of measured data examples. The observation equipment for the measured data is an experimental frequency-modulated continuous wave radar detection system. A schematic diagram of the experimental system and observation scene is shown below. Figure 6 As shown in (a). Its system parameters are shown in Table 2. The drone target used for observation is a DJI Air 2 drone.

[0163] Table 2 Parameters of Frequency Modulated Continuous Wave Radar Observation System

[0164]

[0165]

[0166] Figure 6 (b) The figure shows the echo data of a highly maneuverable UAV measured using the aforementioned frequency-modulated continuous wave radar system. A total of 250 seconds of observation data was recorded during this period. The UAV flew back and forth within a distance range of 20 to 150 meters, remaining within the main lobe of the radar antenna beam throughout the flight. It can be seen that the target experiences acceleration and abrupt changes during the UAV's turn. Therefore, [the following is a selection / selection]... Figure 6 (b) Observation data of a 400ms segment (corresponding to 400 pulses) during the UAV's turning process (e.g. Figure 6 (c) is shown, used to verify the performance advantages of the proposed method compared with existing methods.

[0167] Figure 7 This is the coherent accumulation result obtained after estimating the motion parameters of the UAV using different methods and compensating for the motion phase. Figure 7 (a) shows the coherent accumulation result obtained by the motion target detection (MTD) method. As shown in the figure, although this method is the most commonly used classic method for low-speed moving target detection, it cannot achieve coherent accumulation of echo energy of highly maneuverable UAVs. Figure 7 (b) and Figure 7 (c) The coherent accumulation results obtained by using the FrFT method and LVD method of the time-frequency transformation class are shown in the figure. Although the accumulation effect of the time-frequency transformation class method is better than that of the MTD method, this class of methods still cannot effectively correct the third-order motion phase error introduced by the jerkiness of the high-maneuverability UAV. Therefore, the coherent accumulation results show the phenomenon of main lobe splitting. Figure 7(d) The figure shows the coherent accumulation result obtained using the method proposed in this invention. The method effectively compensates for the higher-order motion phase error of highly maneuverable UAVs, achieving coherent accumulation of highly maneuverable targets. In this experiment, the estimation results of the motion parameters obtained by the method proposed in this invention are the same as those obtained by the GRFT method based on parameter search.

[0168] Figure 8 Quantitative analysis of the energy accumulation performance of different methods. Figure 8 Selected from Figure 7 The one-dimensional azimuth frequency domain profiles of the peak points in the coherent accumulation results of different methods are shown, and they are compared and quantitatively analyzed. As can be seen from the figure, the proposed method improves the signal-to-noise ratio by 3.85 dB compared to the FrFT and LVD methods, and by 13.79 dB compared to the MTD method.

[0169] In summary, the motion parameter estimation and coherent accumulation detection method for highly maneuverable unmanned aerial vehicles (UAVs) presented in this invention has similar coherent accumulation detection performance to the parameter search-based GRFT method, while exhibiting higher computational efficiency and being unaffected by fuzzy sidelobes. Compared to existing time-frequency transform-based methods, such as FrFT and LVD methods, the proposed method offers higher estimation accuracy and robustness, enabling more effective coherent accumulation of highly maneuverable UAV targets.

[0170] Although embodiments of the present invention have been described in conjunction with the accompanying drawings, the patent owner may make various modifications or alterations within the scope of the appended claims, as long as they do not exceed the protection scope described in the claims of the present invention, they shall be within the protection scope of the present invention.

Claims

1. A method for detecting UAV targets based on Bayesian compressed sensing coherent accumulation, characterized in that, Includes the following steps: S1. Perform range compression on the UAV's radar echo and use the null space tracking algorithm to separate the micro-Doppler signal to obtain the UAV's radar echo; S2. Based on the obtained UAV radar echo, the distance movement of the UAV envelope is corrected by Keystone transform, and the slow time data of the peak position of the distance envelope is extracted as the observation signal for sparse reconstruction. S3. Design the perception matrix and use the Bayesian compressed sensing algorithm to reconstruct the observation signal and estimate the motion parameters of the UAV at each order. S4. Construct a motion phase compensation filter based on the motion parameters of the UAV at each order, perform phase compensation on the UAV echo in the pulse-range frequency domain, and perform coherent accumulation and CFAR detection on the compensated frequency domain echo signal. Step S1 specifically includes: S11, transmit the radar echo of the drone Range compression is performed using fast-time Fourier transform to obtain the range-compressed echo. ; S12. Employing a zero-space tracking algorithm to separate UAV radar echoes. The fuselage echo signal and micro-Doppler signal in the separated UAV radar echo Radar echo signal of the drone fuselage .

2. The method for UAV target coherent accumulation detection based on Bayesian compressed sensing according to claim 1, characterized in that, The radar echo of the UAV in step S11 The expression is: In the formula, Indicates a fast time. Indicates slow time. M represents the pulse sequence number, and M represents the number of pulses accumulated by coherence. This represents the echo from the drone's fuselage. This indicates the micro-Doppler signal introduced by the drone's rotor. The expression for the drone fuselage echo is: ； In the formula, Indicates the amplitude of the fuselage echo signal. Indicates the center frequency. Indicates the frequency modulation of the transmitted waveform. Represents the speed of light. The distance-time of a highly maneuverable unmanned aerial vehicle (UAV) is expressed as follows: ； In the formula, Indicates the initial distance of the drone. Indicates the speed of the drone. Indicates the acceleration of the drone. Indicates the swiftness of the drone; Micro-Doppler signals introduced by UAV rotors The expression is: ； In the formula, K represents the total number of rotors in the drone, and I represents the total number of blades in the drone. The echo of the i-th blade on the k-th rotor is expressed as: ； In the formula, This represents the rotational frequency of the k-th rotor. Let L represent the initial phase of the i-th blade on the k-th rotor, and L represent the length of the UAV blade. This indicates the line-of-sight angle of the UAV relative to the radar when the radar is located at the origin of the coordinate system. Echo after distance compression The expression is: ； In the formula, This represents the fast-time Fourier transform. The fast-time Fourier transform of the UAV fuselage echo is expressed as follows: ； in, This indicates the amplitude of the fuselage echo signal after distance compression. Indicates the pulse repetition time. Indicates the carrier wavelength.

3. The method for UAV target coherent accumulation detection based on Bayesian compressed sensing according to claim 2, characterized in that, The input signal for the null-space tracking algorithm in step S12 is the distance-compressed UAV echo. By traversing all range cells, the micro-Doppler signal introduced by the UAV rotor is separated, and the separated signal is then subjected to a fast-time inverse Fourier transform to obtain the UAV radar echo in the time domain. The echo signal from the drone's fuselage .

4. The method for detecting UAV targets based on Bayesian compressed sensing according to claim 3, characterized in that, Step S2 specifically includes: S21. Designing slow-time variables in the Keystone transformation. ; S22, Change the slow time variable Substitute the UAV radar echo after micro-Doppler signal separation in step S1 The time-domain UAV echo signal after Keystone transformation is obtained. For time-domain UAV echo signals The distance-frequency domain signal after Keystone transform is obtained by performing a fast-time Fourier transform. Iterate through all distance cells to obtain the distance cell containing the peak value of the UAV's distance envelope, and extract the slow-time data within that distance cell. This slow time data As an observation signal in the sparse reconstruction process.

5. The UAV target coherent accumulation detection method based on Bayesian compressed sensing according to claim 4, characterized in that, The time-domain UAV echo signal in step S22 The expression is: Keystone transform distance-frequency domain signal The expression is: In the formula, The distance cell representing the location of the peak of the distance envelope is expressed as: Iterate through all distance cells to obtain the distance cell containing the peak value of the UAV's distance envelope, and extract the slow-time data within that distance cell. The expression is: In the formula, This represents the complex amplitude of the UAV echo signal envelope after Keystone transformation.

6. The method for detecting UAV targets based on Bayesian compressed sensing according to claim 5, characterized in that, Step S3 specifically includes: S31. Design and construct a sparse reconstruction perception matrix. ; S32, Based on the perception matrix and observed signals Substitute into the compressed sensing signal model One-dimensional sparse signal vectors are reconstructed using the Bayesian compressed sensing algorithm. ; S33, Transform a one-dimensional sparse signal vector Transform to 3D matrix , where the matrix The elements in satisfy: ； S34, from a three-dimensional matrix The coordinates of the position of the maximum absolute value are estimated and represented as follows: Based on the location of the maximum value, the estimated values ​​of the drone's speed, acceleration, and jerk are obtained, respectively: ； ； ； Output the motion parameter estimation results of the UAV .

7. The method for detecting UAV targets based on Bayesian compressed sensing according to claim 6, characterized in that, Step S31 specifically includes: S311. Define a set representing the range of velocity, acceleration, and jerk parameters: Set of speed parameter ranges The set of acceleration parameter ranges A set of swiftness parameter ranges ; In the formula, N, P, and Q represent the number of elements in the sets of velocity, acceleration, and jerk parameters, respectively. , and These represent the search step sizes for the velocity, acceleration, and jerk parameters, respectively. S312. Search set based on parameters. To design a sparse reconstruction perception matrix ; S313. Construct a perception matrix and represent it as N sub-matrices. ,in, This represents the nth submatrix, which corresponds to The nth element in; submatrix Represented as P submatrices, ,in, Represents the p-th submatrix, corresponding to The p-th element in the submatrix The element in the m-th row and q-th column is represented as: ； In the formula, the matrix elements in In the perception matrix The coordinates in the diagram are represented as follows: 。 8. The method for detecting UAV targets based on Bayesian compressed sensing according to claim 7, characterized in that, Step S4 specifically includes: S41. Construct a motion phase compensation filter based on the estimated UAV speed, acceleration, and jerk speed, and analyze the UAV echo in the pulse-range frequency domain. Perform phase compensation; S42. Perform a slow-time Fourier transform on the frequency domain echo signal after phase compensation to obtain the coherent accumulation result in the range-azimuth frequency domain, and perform CFAR detection.

9. The method for detecting UAV targets based on Bayesian compressed sensing according to claim 8, characterized in that, In step S41, the pulse-range frequency domain is used to analyze the UAV echo. The expression is: ； Using the first l Estimated motion parameters of the target Constructing the compensation function Its expression is as follows: ； The expression for the compensated signal is: .

Images