Frequency agility radar multi-target detection method based on compressed sensing

By introducing a constant false alarm threshold and a dual-layer matched filter in frequency agile radar using the Adaptive-Dual-Filter-Orthogonal Matched Pursuit (A-TMF-OMP) algorithm, the estimation problem of non-grid point targets in multi-target detection is solved, and efficient multi-target detection and false alarm suppression are achieved.

CN118915046APending Publication Date: 2024-11-08BEIJING INST OF REMOTE SENSING EQUIP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410279317.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-03-12
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

Traditional frequency-agile radars struggle to effectively detect multiple targets, especially since the presence of non-grid point targets degrades the performance of compressed sensing algorithms, and prior information about the number of targets is difficult to obtain in practical engineering.

Method used

An adaptive dual-filter orthogonal matched pursuit algorithm (A-TMF-OMP) is adopted, which introduces a constant false alarm threshold and a dual-layer matched filter structure. By adaptively adjusting the number of targets, the distance-velocity joint estimation of non-grid point targets is achieved.

Benefits of technology

It improves the multi-target detection capability of frequency-agile radar, suppresses false alarms and accelerates convergence speed, and can accurately detect multiple targets without requiring prior information on the number of targets.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915046A_ABST
    Figure CN118915046A_ABST
Patent Text Reader

Abstract

The invention discloses a frequency agility radar multi-target detection method based on compressed sensing. The method comprises the following steps: firstly, introducing speed tolerance and distance tolerance to measure the influence of a non-grid point target on a compressed sensing algorithm; on the basis, the research proposes a structure of a double-layer matched filter, and the distance-Doppler two-dimensional reconstruction of the non-grid target is completed through the rough search of a front-layer matched filter and the precise search of a rear-layer matched filter. In order to enable the algorithm to have adaptive capability, a thresholding method in Stagewise Orthogonal Matching Pursuit (StOMP) is used for reference, and a constant false alarm threshold is introduced into the structure of a double-layer matched filter. A simulation result proves the effectiveness of the new method.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of radar, and in particular to a multi-target detection method of frequency agile radar based on compressed sensing. Background Art

[0002] Agile frequency radar has been widely used due to its superior anti-interference performance and high-resolution distance capability. Due to the high side lobes introduced by the randomness of the carrier frequency, traditional methods are difficult to use for agile frequency radar target detection in multi-target scenarios. Compressed sensing uses the sparsity of radar observation scenes to simultaneously complete range-Doppler reconstruction for multiple targets. However, compressed sensing often requires the number of targets to be known in advance, and this prior information is often difficult to obtain in actual engineering. Not only that, the presence of off-grid targets will also lead to a decrease in the performance of the compressed sensing algorithm and reduce the detection probability of the radar. Summary of the invention

[0003] The purpose of the present invention is to provide a multi-target detection method for frequency-agile radar based on compressed sensing, which solves the problem that compressed sensing often requires the number of targets to be known in advance, but it is often difficult to obtain this prior information in actual engineering. In addition, the existence of off-grid targets will also lead to a decrease in the performance of the compressed sensing algorithm and reduce the detection probability of the radar.

[0004] A frequency agile radar multi-target detection method based on compressed sensing, the method comprising:

[0005] Step 1: Obtain the sensing matrix Φ, echo y, and false alarm rate P at each iteration f ;

[0006] Step 2: Initialize the residual r0, index set Q0 and reconstruction matrix Φ according to the sensing matrix Φ and echo y t ;

[0007] Step 3: Pass the current residual through matched filter 1 and matched filter 2 respectively to obtain the output a of matched filter 1 t (j), output b of matched filter 2 t (j);

[0008] Step 4: t (j) Perform index selection to obtain the rough position of the target, and t (j) performing index selection to obtain the target detailed position;

[0009] Step 5: After iteration, the target rough position set Λ is obtained t , target detailed location set I t and the target vector matrix Φ t ;

[0010] Step 6: According to the target rough position set Λ t , target detailed location set I t and the target vector matrix Φ t , get the target scattering center vector x t ;

[0011] Step 7: According to the target scattering center vector x t , update the residual, and iterate steps 2-7 until constant false alarm detection is performed and matched filter 2 outputs b t (j) If the maximum value is less than the preset threshold, the final target scattering center vector x t Output, according to the target scattering center vector x t Get the multi-target detection results.

[0012] In one embodiment, the residual r0, the index set Q0 and the reconstruction matrix Φ are initialized according to the sensing matrix Φ and the echo y. t include:

[0013] Initialize residual r0 = y, index set The reconstructed matrix is Iteration counter t = 1, is an empty set.

[0014] In one embodiment, the output a of the matched filter 1 is t (j) is expressed as formula (1);

[0015]

[0016] r t-1 Indicates that when the residual of the previous iteration, represents the j-th atom in the perception matrix Φ, where j is a natural number.

[0017] In one embodiment, the output b of the matched filter 2 t (j) is expressed as:

[0018]

[0019] Ψ represents the second layer matched filter, represents the step length and t represents the iteration count.

[0020] In one embodiment, the pair a t (j) Perform index selection to obtain the target rough position, which is expressed as formula (3);

[0021] λ t = argmax j |at (j) | ... Formula (3);

[0022] λ t Indicates the rough position of the target.

[0023] In one embodiment, the pair b t (j) Perform index selection to obtain the target fine position, which is expressed as formula (4):

[0024] μ t = argmax j |b t (j) | ... Formula (4);

[0025] μ t Indicates the target fine position.

[0026] In one embodiment, the target rough position set Λ t Expressed as Λ t =[Λ t-1 ,λ t ],

[0027] Λ t-1 Represents the target coarse position set of the previous iteration.

[0028] In one embodiment, the target fine position set I t Represented as I t =[I t-1 ,μ t ], I t-1 Represents the target fine position set of the previous iteration.

[0029] In one embodiment, the target vector matrix Φ t Represented as Φ t =[Φ t-1 ,ψ j ],Φ t-1 Represents the target vector matrix of the previous iteration

[0030] In one embodiment, the target rough position set Λ t , target detailed location set I t and the target vector matrix Φ t , get the target scattering center vector x t Including: Get the target scattering center vector x through formula (5) t

[0031] x t =argmin x ||v-Φ t ||2

[0032] v represents the target moving in a straight line at a constant speed v along the radial illumination direction of the radar, and ||·||2 is the 2-norm.

[0033] The present invention proposes a multi-target detection method of frequency-agile radar based on compressed sensing, which is an adaptive two-filter orthogonal matching pursuit algorithm (Adaptive Two Matched Filters Orthogonal Matching Pursuit, A-TMF-OMP). The new algorithm introduces a constant false alarm rate (Constant False Alarm Rate) threshold in OMP, which has the ability to adapt to the number of targets. At the same time, the method adopts a double-layer matched filter structure in the iterative process to complete the distance-speed joint estimation for non-grid point targets. Theoretical analysis and simulation results show that A-TMF-OMP has advantages in detection capability, false alarm suppression and convergence speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a schematic diagram of a double-layer matched filter of the present invention;

[0035] Figure 2 It is a schematic diagram of the sparse recovery results of non-grid point targets by A-TMF-OMP of the present invention;

[0036] Figure 3 The present invention is an algorithm flow chart of a frequency-agile radar multi-target detection method based on compressed sensing. DETAILED DESCRIPTION

[0037] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description and claims.

[0038] Although the present invention can be expanded with various forms of modification and substitution, some specific implementation examples are also listed in the specification and described in detail; it should be understood that the inventor's starting point is not to limit the invention to the specific embodiments described, on the contrary, the inventor's starting point is to protect all improvements, equivalent substitutions and modifications made within the spirit or scope defined by this claim.

[0039] A frequency agile radar multi-target detection method based on compressed sensing, the method comprising:

[0040] Step 1: Obtain the sensing matrix Φ, echo y, and false alarm rate P at each iteration f ;

[0041] Step 2: Initialize the residual r0, index set Q0 and reconstruction matrix Φ according to the sensing matrix Φ and echo y t ;

[0042] Step 3: Pass the current residual through matched filter 1 and matched filter 2 respectively to obtain the output a of matched filter 1 t (j), output b of matched filter 2 t (j);

[0043] Step 4: t (j) Perform index selection to obtain the rough position of the target, and t (j) performing index selection to obtain the target detailed position;

[0044] Step 5: After iteration, the target rough position set Λ is obtained t , target detailed location set I t and the target vector matrix Φ t ;

[0045] Step 6: According to the target rough position set Λ t , target detailed location set I t and the target vector matrix Φ t , get the target scattering center vector x t ;

[0046] Step 7: According to the target scattering center vector x t , update the residual, and iterate steps 2-7 until constant false alarm detection is performed and matched filter 2 outputs b t (j) If the maximum value is less than the preset threshold, the final target scattering center vector x t Output, according to the target scattering center vector x t Get the multi-target detection results.

[0047] In one embodiment, the residual r0, the index set Q0 and the reconstruction matrix Φ are initialized according to the sensing matrix Φ and the echo y. t include:

[0048] Initialize residual r0 = y, index set The reconstructed matrix is Iteration counter t = 1, is an empty set.

[0049] In one embodiment, the output a of the matched filter 1 is t (j) is expressed as formula (1);

[0050]

[0051] r t-1 Indicates that when the residual of the previous iteration, represents the j-th atom in the perception matrix Φ, where j is a natural number.

[0052] In one embodiment, the output b of the matched filter 2 t (j) is expressed as:

[0053]

[0054] Ψ represents the second layer matched filter, represents the step length and t represents the iteration count.

[0055] In one embodiment, the pair a t (j) Perform index selection to obtain the target rough position, which is expressed as formula (3);

[0056] λ t = argmax j |a t (j) | ... Formula (3);

[0057] λ t Indicates the rough position of the target.

[0058] In one embodiment, the pair b t (j) Perform index selection to obtain the target fine position, which is expressed as formula (4):

[0059] μ t = argmax j |b t (j) | ... Formula (4);

[0060] μ t Indicates the target fine position.

[0061] In one embodiment, the target rough position set Λ t Expressed as Λ t =[Λ t-1 ,λ t ],

[0062] Λ t-1 Represents the target coarse position set of the previous iteration.

[0063] In one embodiment, the target fine position set I t Represented as I t =[I t-1 ,μ t ], I t-1 Represents the target fine position set of the previous iteration.

[0064] In one embodiment, the target vector matrix Φ t Represented as Φ t =[Φ t-1 ,ψ j ],Φ t-1Represents the target vector matrix of the previous iteration

[0065] In one embodiment, the target rough position set Λ t , target detailed location set I t and the target vector matrix Φ t , get the target scattering center vector x t Including: Get the target scattering center vector x through formula (5) t

[0066] x t =argmin x ||v-Φ t ||2

[0067] v represents the target moving in a straight line at a constant speed v along the radial illumination direction of the radar, and ||·||2 is the 2-norm.

[0068] The present invention proposes a multi-target detection method of frequency-agile radar based on compressed sensing, which is an adaptive two-filter orthogonal matching pursuit algorithm (Adaptive Two Matched Filters Orthogonal Matching Pursuit, A-TMF-OMP). The new algorithm introduces a constant false alarm rate (Constant False Alarm Rate) threshold in OMP, which has the ability to adapt to the number of targets. At the same time, the method adopts a double-layer matched filter structure in the iterative process to complete the distance-speed joint estimation for non-grid point targets. Theoretical analysis and simulation results show that A-TMF-OMP has advantages in detection capability, false alarm suppression and convergence speed.

[0069] Based on the orthogonal matching pursuit algorithm, this paper proposes an adaptive two-layer filter orthogonal matching pursuit algorithm (Adaptive Two Matched Filters Orthogonal Matching Pursuit, A-TMF-OMP). This paper first introduces speed tolerance and distance tolerance to measure the impact of non-grid point targets on the compressed sensing algorithm. On this basis, this study proposes a two-layer matched filter structure, and completes the range-Doppler two-dimensional reconstruction of non-grid targets through a rough search of the front-layer matched filter and a precise search of the back-layer matched filter. In order to make the algorithm adaptive, this paper draws on the thresholding method in the stagewise orthogonal matching pursuit (StOMP) and introduces a constant false alarm threshold in the structure of the two-layer matched filter. The simulation results confirm the effectiveness of the new method.

[0070] Assuming that the frequency-agile radar system transmits N pulses within a coherent process interval (CPI), and the carrier frequency of each pulse changes randomly within a given bandwidth B, the nth transmitted pulse can be expressed as:

[0071]

[0072] Here n=0,1…N-1,d n is a frequency agile code sequence that is uniformly distributed between 0 and 1. r is the pulse repetition period, T p is the pulse width, f0 is the starting carrier frequency. rect(·) represents the following envelope function:

[0073]

[0074] Assuming there is a target point within the radar antenna beam, the echo of the nth pulse is

[0075]

[0076] Where c is the speed of light in a vacuum, and r(t) represents the radial distance of the point target from the radar at time t. Assuming that the point target moves in a straight line at a constant speed v along the radial illumination direction of the radar, then r(t) = r(0) + vt, and the echo can be expressed as:

[0077]

[0078] After down-converting the received echo, the echo signal becomes

[0079]

[0080] Sampling is performed at the Nyquist sampling rate, i.e., f s =1 / T p ,

[0081]

[0082] To simplify the formula, we can ignore the notation l and define R d (n): = R d (n,nT r +l / f s ). Assuming that there are K scattering points in the lth CRR unit, the echo signal can be expressed as:

[0083]

[0084] where β gis the complex scattering intensity of the gth scattering center, r g and v g represent the distance and speed of the g-th target respectively.

[0085] Grid discretization of the velocity-distance plane

[18] . The velocity resolution is ΔV = c / 2f0NT r Divided into N units, covering the speed range [0, V blind ]. blind It is the first blind speed of radar

[0086] V blind =c / 2f0T r

[0087] For pulse radar, the coarse range resolution (CRR) is T P C / 2, high range resolution (HRR) is ΔR = 2B / c. The number of HRR units in a CRR unit is defined as

[0088]

[0089] Thus, the coarse resolution distance unit is subdivided into M high resolution distance units.

[0090] The entire speed-distance plane is evenly divided into M×N grids. Assume that the speed parameter and distance parameter of the target fall exactly on the grid, that is,

[0091]

[0092]

[0093] The gridded echo signal is

[0094]

[0095] Among them, w n is zero and has a variance of Gaussian white noise. The perception matrix is ​​constructed as:

[0096]

[0097] here is the atom in the perception matrix and can be expressed as:

[0098]

[0099] k=0,1…N-1,m=0,1…M-1. The echo signal is vectorized as follows:

[0100] y=[y(0),y(1),…y(N-1)] T

[0101] and generate the scattered intensity vector

[0102] x=[x0,x1,…x MN-1 ] T

[0103] The elements in x can be written as:

[0104]

[0105] Finally, the matrix form of the echo signal can be written as:

[0106] y=Φx+w

[0107] Since the number of columns of the measurement matrix Φ is greater than the number of rows, that is, MN>N, the above problem is transformed into the solution of an underdetermined linear equation system, that is, recovering the vector x from the observation y. The sparsity of the target is introduced to solve the reconstruction problem, that is,

[0108] min‖x‖0 sty=Φx

[0109] ‖x‖0 represents the number of non-zero elements in vector x.

[0110] This paper increases the number of adaptive thresholds and matched filter layers in the OMP algorithm, which can achieve more accurate recovery of non-grid point targets.

[0111] In one embodiment, the A-TMF-OMP algorithm does not need to know the prior information of the number of targets, but automatically stops the iteration process through the constant false alarm threshold. t By the noise level σ t and false alarm rate P f The noise level can be estimated from the residual,

[0112]

[0113] The threshold can be set to

[0114]

[0115] When the energy in all units is less than the threshold, it indicates that the target no longer exists in the residual, and the A-TMF-OMP algorithm stops automatically.

[0116] The A-TMF-OMP algorithm uses the first layer of matched filters to select the speed-distance unit with the strongest energy, and it can be considered that the unit is most likely to contain a target. Secondly, the algorithm generates matched filter 2 in the corresponding unit. Since the spectral peak position generated by matched filter 1 is different in each iteration, the search range of matched filter 2 is dynamic.

[0117] Figure 1 The structure of the multi-layer matched filter is shown. In MF1, the range-velocity plane is discretized into M×N grids. The distance from the target to the nearest grid is greater than VT or RT. In MF2, the corresponding range-velocity unit is divided into finer grids so that the distance from the target to the adjacent grid is less than VT or RT. After the residual signal passes through the first layer of matched filters, the distance index and velocity index can be obtained respectively. and The second layer of matched filters will A fine grid is generated near the , and the width of the fine grid is h1ΔR and h2ΔV. The step lengths h1 and h2 are constrained by the speed tolerance and distance tolerance, that is,

[0118]

[0119] If h1 is equal to h2, the above formula simplifies to

[0120] h1=h2<0.31

[0121] According to the above definition, the second layer matched filter can be expressed as

[0122]

[0123] here

[0124] Figure 2 The results of 2D restoration of non-grid point targets using A-TMF-OMP are shown. The algorithm accurately estimates the distance, velocity, and amplitude of six non-grid point targets.

[0125] The present invention proposes a multi-target detection method of frequency-agile radar based on compressed sensing, which is an adaptive two-filter orthogonal matching pursuit algorithm (Adaptive Two Matched Filters Orthogonal Matching Pursuit, A-TMF-OMP). The new algorithm introduces a constant false alarm rate (Constant False Alarm Rate) threshold in OMP, which has the ability to adapt to the number of targets. At the same time, the method adopts a double-layer matched filter structure in the iterative process to complete the distance-speed joint estimation for non-grid point targets. Theoretical analysis and simulation results show that A-TMF-OMP has advantages in detection capability, false alarm suppression and convergence speed.

Claims

1. A frequency-agile radar multi-target detection method based on compressed sensing, characterized in that: The method comprises: Step 1: Obtain the sensing matrix Φ, echo y, and false alarm rate P at each iteration f ; Step 2: Initialize the residual r0, index set Q0 and reconstruction matrix Φ according to the sensing matrix Φ and echo y t ; Step 3: Pass the current residual through matched filter 1 and matched filter 2 respectively to obtain the output a of matched filter 1 t (j), output b of matched filter 2 t (j); Step 4: t (j) Perform index selection to obtain the rough position of the target, and t (j) performing index selection to obtain the target detailed position; Step 5: After iteration, the target rough position set Λ is obtained t , target detailed location set I t and the target vector matrix Φ t ; Step 6: According to the target rough position set Λ t , target detailed location set I t and the target vector matrix Φ t , get the target scattering center vector x t ; Step 7: According to the target scattering center vector x t , update the residual, and iterate steps 2-7 until constant false alarm detection is performed and matched filter 2 outputs b t (j) If the maximum value is less than the preset threshold, the final target scattering center vector x t Output, according to the target scattering center vector x t Get the multi-target detection results.

2. The method according to claim 1, characterized in that According to the sensing matrix Φ, the echo y, the residual r0, the index set Q0 and the reconstruction matrix Φ are initialized. t include: Initialize residual r0 = y, index set The reconstructed matrix is Iteration counter t = 1, is an empty set.

3. The method according to claim 2, characterized in that The output a of the matched filter 1 t (j) is expressed as formula (1); r t-1 Indicates that when the residual of the previous iteration, represents the j-th atom in the perception matrix Φ, where j is a natural number.

4. The method according to claim 3, characterized in that The output b of the matched filter 2 t (j) is expressed as: Ψ represents the second layer matched filter, represents the step length and t represents the iteration count.

5. The method according to claim 4, characterized in that The pair t (j) Perform index selection to obtain the target rough position, which is expressed as formula (3); λ t = argmax j |a t (j) | ... Formula (3); λ t Indicates the rough position of the target.

6. The method according to claim 5, characterized in that The pair b t (j) Perform index selection to obtain the target fine position, which is expressed as formula (4): μ t = argmax j |b t (j) | ... Formula (4); μ t Indicates the target fine position.

7. The method according to claim 6, characterized in that The target rough position set Λ t Expressed as Λ t =[Λ t-1 ,λ t ],Λ t-1 Represents the target coarse position set of the previous iteration.

8. The method according to claim 7, characterized in that The target detailed position set I t Represented as I t =[I t-1 , μ t ], I t-1 Represents the target fine position set of the previous iteration.

9. The method according to claim 8, characterized in that The target vector matrix Φ t Represented as Φ t =[Φ t-1 , ψ j ],Φ t-1 Matrix representing the target vector for the previous iteration.

10. The method according to claim 9, characterized in that According to the target rough position set Λ t , target detailed location set I t and the target vector matrix Φ t , get the target scattering center vector x t Including: Get the target scattering center vector x through formula (5) t x t =argmin x ||v-Φ t ||2 v represents the target moving in a straight line at a constant speed v along the radial illumination direction of the radar, and ||·||2 is the 2-norm.