Target intelligent robust detection method and system under multi-disturbance background

Abstract

The invention relates to the technical field of broadband radar signals, and provides a target intelligent robust detection method and system under a multi-disturbance background, and the method comprises the steps: carrying out the unitary transformation of test data and a parameter matrix set based on the skew symmetry characteristic of a clutter covariance matrix, and constructing a robust detection frame; performing correlation calculation estimation under the condition of no target hypothesis and the condition of target hypothesis, and constructing detection statistics of the target intelligent robust detection method under the multi-disturbance background; and determining a detection threshold according to a preset false alarm probability, comparing the detection statistic with the detection threshold, and determining whether the current to-be-detected distance unit has a distance expansion target according to a comparison result. According to the method, the dependence on the training data volume is remarkably reduced, meanwhile, the estimation precision of the unknown clutter covariance matrix is greatly improved, powerful support is provided for target self-adaptive detection under the small sample condition, and the method has application and popularization value.

Description

Technical Field

[0001] The present invention relates to the technical field of broadband radar signals, and in particular to an intelligent robust target detection method and system under multi-disturbance backgrounds. Background Art

[0002] As radar bandwidth increases, its range resolution further improves. Wideband radar is widely used in modern military and civilian applications, including anti-interference, counter-reconnaissance, precision detection and imaging, high-precision tracking, and target identification. Adaptive detection of range-extended targets with wideband radar has become a hot topic in the radar community. Unlike narrowband radar, where target echo signals typically occupy only one range resolution unit, with wideband radar, target scattering energy can spread across adjacent range units, appearing as a "one-dimensional range profile," creating a range-extended target. If the point target detection method is still used to detect the target on the echo signal of a single range unit, and the statistical characteristics of the background clutter are estimated by sampling the adjacent range units; on the one hand, the energy of the strong scattering point of the range-extended target is easy to leak into the adjacent range units, resulting in signal pollution, and further forming a masking effect on the target signal of the single range unit to be detected, making the point target detection method ineffective; on the other hand, in practical applications, radar detection faces a complex electromagnetic environment, which may have natural or man-made interference sources such as electronic countermeasure signals or various civilian electromagnetic signals. In addition, the environment in which the target is located is complex and changeable, which enhances the non-uniformity of the background clutter. The number of pure clutter training data that meets the independent and identical distribution is relatively limited. Compared with narrowband radar, this problem is particularly prominent in the target detection scenario of wideband radar, making it difficult for existing range-extended target detection methods to achieve ideal detection results.

[0003] Furthermore, in commonly used rank-one signal target detection models, the target's steering vector is typically assumed to be a known fixed vector. However, in practical applications, due to beam pointing errors and multipath, the target's steering vector may be mismatched. To address this issue, a subspace model can be used to model the target signal. In a subspace model, the signal is represented as the product of a known subspace matrix and an unknown coordinate matrix. Furthermore, given the difficulty in obtaining sufficient pure clutter training data in real environments, the clutter covariance matrix of radar receivers using centrosymmetric linear arrays or centrosymmetric spaced pulse trains exhibits a unique skew-symmetric structure. Utilizing this skew-symmetric structure can often improve the detector's detection performance and reduce the amount of training data required. Under the premise of considering skew-symmetric prior information, if subspace modeling is adopted for the target and interference signals based on the overall data set consisting of test data and training data of multiple range units to be detected, and the detection statistics are constructed using the GLRT test criterion, a subspace GLRT detector (abbreviated as P2S-GLRT-HE) for intelligent and robust detection of targets under multiple perturbations can be obtained. However, the construction process of the GLRT detector involves solving the maximum likelihood estimation of unknown parameters under both the target hypothesis and the no-target hypothesis. The detector construction process is complex and the computational complexity is high. If the Rao test criterion is used to construct the detection statistics, a subspace Rao detector (abbreviated as P2S-Rao-HE) for intelligent and robust detection of targets under multiple perturbations can be obtained. However, in the actual implementation process, it faces the challenge of solving the Fisher information matrix, and the detector construction process is also relatively complex.

[0004] In the presence of external interference and a small amount of uniform training data, the detection of range-extended targets by multi-channel broadband radars is difficult to balance between computational complexity and detection performance. How to make full use of the skew-symmetric structure information, reduce the actual demand for training data volume, improve the estimation accuracy of the unknown clutter covariance matrix, and then construct a closed-form intelligent robust detection method for range-extended targets. While maintaining the constant false alarm rate (CFAR) characteristics, it effectively suppresses interference signals and strikes an effective balance between computational complexity and detection performance. This is the key to improving the detection capability of broadband radars in complex interference environments, and is also one of the difficulties faced by adaptive detection of range-extended targets by multi-channel broadband radars. Summary of the Invention

[0005] The purpose of the present invention is to solve at least one technical problem in the background technology and provide a method and system for intelligent robust detection of targets under multi-disturbance backgrounds.

[0006] To achieve the above objectives, the present invention provides a method for intelligent and robust target detection in a multi-disturbance background, comprising:

[0007] A robust detection framework is constructed by constructing test data Z consisting of echo data from K range cells to be measured and training data Y collected from R adjacent reference cells. Under the condition that the clutter covariance matrix M, the target signal coordinate matrix P, and the interference signal coordinate matrix Q are all unknown, a unitary transformation is performed on the test data Z and the parameter matrix set {M, P, Q} based on the skew symmetry of the clutter covariance matrix.

[0008] In the absence of target assumption, solve the maximum likelihood estimation of the clutter covariance matrix M and the interference coordinate matrix Q; in the presence of target assumption, solve the target parameter vector Θ rp-1s The maximum likelihood estimate of the target parameter vector Θ is obtained by using the complex Gaussian joint probability density function of the test data Z and the training data Y. rp-1s Obtain the partial derivative, then substitute the maximum likelihood estimate of the clutter covariance matrix M and the interference coordinate matrix Q obtained under the no-target assumption into the derivative result to obtain the calculation result; finally, substitute the calculation result and the maximum likelihood estimate of the target parameter vector under the target assumption into the range-extended target gradient detection statistic to construct the detection statistic λ of the intelligent robust target detection method under multi-disturbance background;

[0009] The detection threshold T is determined according to the preset false alarm probability. The detection statistic λ is compared with the detection threshold T, and the comparison result is used to determine whether there is a range-extended target in the current range unit to be detected.

[0010] According to one aspect of the present invention, constructing test data Z consisting of echo data of K range cells to be measured and training data Y collected from R adjacent reference cells includes:

[0011] The test data Z is collected from K adjacent distance units, and after sampling and organization, it forms an N×1 dimensional vector, which is expressed as z t ∈￡ N×1 ,t=1,2,...,K, where N represents the product of the number of antenna elements and the number of pulses, represents a set of N×1 dimensional complex matrices; the disturbance in the test data Z consists of clutter components and interference components, which are represented by c t ∈￡ N×1 and j t ∈￡ N×1 ,t=1,2,...,K;set the interference component j t ,t=1,2,...,K is modeled as a deterministic subspace signal, belonging to the known multi-rank subspace J∈￡ N×q , denoted as j t =Jq t , where q t ∈￡ q×1, t=1,2,...,K represents the unknown complex coordinate vector of the interference signal; and the clutter component c t , t=1,2,...,K are independent and identically distributed among different distance units and obey the zero-mean circularly symmetric complex Gaussian distribution of the covariance matrix M, that is, c t ~CN(0 N×1 ,M), where M is an unknown Hermitian positive definite matrix; set the target signal s t ,t=1,2,...,K is also modeled as a deterministic subspace signal, belonging to the known multi-rank subspace H∈￡ N×p , denoted as s t =Hp t , where p t ∈￡ p×1 , t=1,2,...,K represents the unknown complex coordinate vector of the target signal;

[0012] To facilitate derivation, define the augmented matrix In order to estimate the unknown clutter covariance matrix M, it is assumed that there is a set of training data Y containing only clutter components, which is usually collected from R distance units adjacent to the distance unit to be measured, and is recorded as Y = [y1, y2, ..., y R ]∈￡ N×R , where y l =c l ,l=1,2,...,R is also independent and identically distributed and satisfies c l ~CN(0 N×1 ,M).

[0013] According to one aspect of the present invention, based on the skew symmetry of the clutter covariance matrix, a unitary transformation is performed on the test data Z and the parameter matrix set {M, P, Q} to construct a robust detection framework, including:

[0014] Based on the Gradient test criterion, an intelligent robust detection framework for targets under multi-disturbance background is constructed. The one-step Gradient test decision formula of complex signals is expressed as:

[0015]

[0016] Among them, λ 1S-Gradient and T 1S-Gradient Represent the test statistic and threshold respectively; test data Z=[z1,z2,...,z K ]∈￡ N×K , target signal coordinate matrix P=[p1,p2,...,p K ]∈￡ p×K , interference signal coordinate matrix Q=[q1,q2,...,q K ]∈￡ q×K;Unknown parameter set The relevant parameter Θ r =vec(P)∈￡ pK×1 , interference parameters represents the maximum likelihood estimate of Θ under the H0 hypothesis, Represents Θ r The maximum likelihood estimate under the H1 assumption, Θ r0 Represents Θ r True value under the H0 hypothesis; superscript (·) T represents transpose, and the vec(·) function realizes the vectorization of the matrix; under the H0 assumption, the target does not exist, so Θ r0 =0 pK×1 ;f i (Z,Y|M,iP,Q) represents the joint conditional probability density function of the test data Z and the training data Y under the H0 or H1 hypothesis; based on the statistical independence between the test data and the training data, the joint conditional probability density function can be expressed as:

[0017] f i (Z,Y|M,iP,Q)={π N(K+R) |M| K+R} -1 ×exp[-tr(M -1 T i )],i=0,1 (2)

[0018] Among them, T i =S+(ZB i D)(ZB i D) H , sample covariance matrix S = YY H , and define the augmented matrix B i =[iH,J] and D = [P T ,Q T ] T ; |·| represents the determinant of a square matrix, the tr(·) function represents the trace of a square matrix, the exp(·) function represents the exponential operation, and the superscript (·) H represents the conjugate transpose; the condition p+q≤N must be satisfied to ensure that B1=[H,J] is a full-rank augmented matrix; in addition, by utilizing the skew symmetry of the clutter covariance matrix M, it can be derived that:

[0019] tr(M -1 T i )=tr[D N (M * ) -1 D N T i ]=tr(M -1 D NT i * D N ),i=0,1 (3)

[0020] Among them, the superscript (·) * represents conjugation; therefore, Equation (2) can be rewritten as:

[0021] f i (Z,Y|M,iP,Q)={π N(K+R) |M| K+R} -1 ×exp[-tr(M -1 T ip )],i=0,1 (4)

[0022] in, and

[0023]

[0024] In the above formula, and represent the real and imaginary parts respectively, and

[0025]

[0026] According to one aspect of the present invention, in the absence of a target hypothesis, the maximum likelihood estimation of the clutter covariance matrix M and the interference coordinate matrix Q is solved; in the presence of a target hypothesis, the target parameter vector Θ is solved. rp-1s The maximum likelihood estimate of the target parameter vector Θ is obtained by using the complex Gaussian joint probability density function of the test data Z and the training data Y. rp-1s Obtain the partial derivative, and then substitute the maximum likelihood estimate of the clutter covariance matrix M and the interference coordinate matrix Q obtained under the no-target assumption into the derivative result to obtain the calculation result; finally, substitute the calculation result and the maximum likelihood estimate of the target parameter vector under the target assumption into the range-extended target gradient detection statistic to construct the detection statistic λ of the intelligent robust target detection method under multi-disturbance background, including:

[0027] Let the unknown parameter set after unitary transformation be The relevant parameters after unitary change Θ rp-1s =vec(P p )∈￡ 2pK×1 , interference parameters The skew-symmetric one-step gradient test decision formula in formula (1) can be rewritten as:

[0028]

[0029] in Represents Θp-1s The maximum likelihood estimation under the H0 hypothesis is: Represents Θ rp-1s Maximum likelihood estimation under the H1 hypothesis;

[0030] Under the H1 hypothesis, the natural logarithm of equation (4) is about Θ rp-1s Taking partial derivatives, we can get:

[0031]

[0032] Under the H0 assumption, let the derivative of Equation (4) with respect to the clutter covariance matrix M be equal to zero, and we can obtain p The maximum likelihood estimate of M under the conditions of and γ:

[0033]

[0034] The formula (9) obtained Replacing the unknown true covariance matrix M in equation (4), we can obtain:

[0035]

[0036] Among them, I m represents the m×m dimensional identity matrix; let (10) be the value of Q p Find the partial derivative and set the result to zero, and you can get Q under the H0 hypothesis p The maximum likelihood estimate of :

[0037]

[0038] in, Substituting Equation (11) into Equation (9) and using the matrix inversion lemma, we can obtain The inverse matrix of :

[0039]

[0040] in, Represents projection to matrix Then, substituting equations (11) and (12) into equation (8), we can obtain:

[0041]

[0042] in,

[0043] Under the H1 assumption, let equation (4) find the partial derivative of M and set the result to zero, and we can get p The maximum likelihood estimate of M under the condition:

[0044]

[0045] Then substitute formula (14) into formula (4), we can get:

[0046]

[0047] Formula (15) for D p Find the partial derivative and set the result to zero, and we can get D under the assumption H1 p The maximum likelihood estimate of :

[0048]

[0049] in, According to formula (5), P p The maximum likelihood estimation under the assumption H1 can be derived as:

[0050]

[0051] Therefore, we can directly get Θ under the H1 assumption rp-1s The maximum likelihood estimate of is:

[0052]

[0053] Substituting equations (13) and (18) into equation (7), we can obtain the Gradient test statistic in a uniform clutter environment:

[0054]

[0055] in, Represents projection to matrix The orthogonal projection matrix of the column space.

[0056] According to one aspect of the present invention, a detection threshold T is determined based on a preset false alarm probability, a detection statistic λ is compared with the detection threshold T, and a determination is made based on the comparison result whether a range-extended target exists in the current range unit to be detected, including:

[0057] If λ≥T, it is determined that there is a range extension target in the current range unit to be detected, and the test data Z does not participate in the subsequent training data update of other range units;

[0058] If λ<T, it is determined that there is no range extension target in the current range unit to be detected, and the current test data Z is included in the training data set for subsequent detection.

[0059] To achieve the above objectives, the present invention further provides an intelligent robust target detection system under multiple disturbance backgrounds, comprising:

[0060] The robust detection framework construction module constructs test data Z consisting of echo data from K range cells to be measured and training data Y collected from R adjacent reference cells. Under the condition that the clutter covariance matrix M, the target signal coordinate matrix P, and the interference signal coordinate matrix Q are all unknown, a unitary transformation is performed on the test data Z and the parameter matrix set {M, P, Q} based on the skew symmetry of the clutter covariance matrix to build a robust detection framework.

[0061] The detection statistics construction module solves the maximum likelihood estimation of the clutter covariance matrix M and the interference coordinate matrix Q in the absence of target assumptions; solves the target parameter vector Θ in the presence of target assumptions rp-1s The maximum likelihood estimate of the target parameter vector Θ is obtained by using the complex Gaussian joint probability density function of the test data Z and the training data Y. rp-1s Obtain the partial derivative, then substitute the maximum likelihood estimate of the clutter covariance matrix M and the interference coordinate matrix Q obtained under the no-target assumption into the derivative result to obtain the calculation result; finally, substitute the calculation result and the maximum likelihood estimate of the target parameter vector under the target assumption into the range-extended target gradient detection statistic to construct the detection statistic λ of the intelligent robust target detection method under multi-disturbance background;

[0062] The target detection and judgment module determines the detection threshold T according to the preset false alarm probability, compares the detection statistic λ with the detection threshold T, and determines whether there is a range-extended target in the current range unit to be detected based on the comparison result.

[0063] To achieve the above-mentioned objectives, the present invention also provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the intelligent robust detection method for targets in a multi-disturbance background as described above.

[0064] To achieve the above objectives, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for intelligent and robust detection of targets in a multi-disturbance background as described above is implemented.

[0065] According to the solution of the present invention, the present invention can achieve the following beneficial effects:

[0066] 1) Innovative detector design: We propose an intelligent robust detector for targets under multi-perturbation backgrounds, which has a closed-form analytical expression, avoids the complex iterative calculation process in traditional methods, and significantly improves detection efficiency.

[0067] 2) Efficient implementation advantage: Compared with traditional Rao detection methods, the proposed method does not require explicit calculation of the complex Fisher information matrix and can be implemented with only low-complexity operations, which significantly reduces the difficulty of algorithm implementation and has more practicality and deployment advantages in engineering applications.

[0068] 3) Clutter Covariance Matrix Optimization: By deeply exploiting the skew symmetry of the clutter covariance matrix, the accuracy of matrix estimation is significantly improved. This improvement effectively reduces the dependence on the number of training samples and provides reliable technical support for adaptive detection of extended-range targets under small sample conditions.

[0069] 4) Structured interference suppression capability: In complex interference environments, the detector can adaptively distinguish and suppress interference signals of different intensities, demonstrating excellent intelligent anti-interference performance, especially suitable for practical application scenarios with strong interference.

[0070] 5) Wide applicability: This method is not only applicable to traditional broadband radar detection, but can also effectively meet the detection needs of some non-broadband radars. For example, it can use low- / medium-resolution radars to accurately detect large-scale targets or target groups moving at the same speed (such as ship formations, aircraft formations, etc.), and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A flowchart schematically illustrates a method for intelligent and robust target detection under multiple disturbance backgrounds according to an embodiment of the present invention;

[0072] Figure 2 1 is a comparison chart of detection performance between the method of Example 1 of the present invention and an existing detection method when there is sufficient training data;

[0073] Figure 3 This is a comparison chart of the detection performance of the method of Example 2 of the present invention and the existing detection method when training data is scarce. DETAILED DESCRIPTION

[0074] The present invention will now be discussed with reference to exemplary embodiments. It should be understood that the embodiments discussed are only intended to enable those skilled in the art to better understand and implement the present invention, rather than to imply any limitation on the scope of the present invention.

[0075] As used herein, the term "including" and variations thereof are to be interpreted as open-ended terms meaning "including, but not limited to." The term "based on" is to be interpreted as "based, at least in part, on." The terms "one embodiment" and "an embodiment" are to be interpreted as "at least one embodiment."

[0076] Figure 1The flowchart of the method for intelligent and robust target detection under multi-disturbance background according to one embodiment of the present invention is schematically shown. Figure 1 As shown, in this embodiment, the intelligent robust detection method of targets under multi-disturbance background includes:

[0077] A robust detection framework is constructed by constructing test data Z consisting of echo data from K range cells to be measured and training data Y collected from R adjacent reference cells. Under the condition that the clutter covariance matrix M, the target signal coordinate matrix P, and the interference signal coordinate matrix Q are all unknown, a unitary transformation is performed on the test data Z and the parameter matrix set {M, P, Q} based on the skew symmetry of the clutter covariance matrix.

[0078] In the absence of target assumption, solve the maximum likelihood estimation of the clutter covariance matrix M and the interference coordinate matrix Q; in the presence of target assumption, solve the target parameter vector Θ rp-1s The maximum likelihood estimate of the target parameter vector Θ is obtained by using the complex Gaussian joint probability density function of the test data Z and the training data Y. rp-1s Obtain the partial derivative, then substitute the maximum likelihood estimate of the clutter covariance matrix M and the interference coordinate matrix Q obtained under the no-target assumption into the derivative result to obtain the calculation result; finally, substitute the calculation result and the maximum likelihood estimate of the target parameter vector under the target assumption into the range-extended target gradient detection statistic to construct the detection statistic λ of the intelligent robust target detection method under multi-disturbance background;

[0079] The detection threshold T is determined according to the preset false alarm probability. The detection statistic λ is compared with the detection threshold T, and the comparison result is used to determine whether there is a range-extended target in the current range unit to be detected.

[0080] Furthermore, according to one embodiment of the present invention, constructing test data Z consisting of echo data of K range cells to be measured and training data Y collected from R adjacent reference cells includes:

[0081] The test data Z is collected from K adjacent distance units, and after sampling and organization, it forms an N×1 dimensional vector, which is expressed as z t ∈￡ N×1 ,t=1,2,...,K, where N represents the product of the number of antenna elements and the number of pulses, represents a set of N×1 dimensional complex matrices; the disturbance in the test data Z consists of clutter components and interference components, which are represented by c t ∈￡ N×1 and j t ∈￡ N×1 ,t=1,2,...,K;set the interference component j t,t=1,2,...,K is modeled as a deterministic subspace signal, belonging to the known multi-rank subspace J∈￡ N×q , denoted as j t =Jq t , where q t ∈￡ q×1 , t=1,2,...,K represents the unknown complex coordinate vector of the interference signal; and the clutter component c t , t=1,2,...,K are independent and identically distributed among different distance units and obey the zero-mean circularly symmetric complex Gaussian distribution of the covariance matrix M, that is, c t ~CN(0 N×1 ,M), where M is an unknown Hermitian positive definite matrix; set the target signal s t ,t=1,2,...,K is also modeled as a deterministic subspace signal, belonging to the known multi-rank subspace H∈￡ N×p , denoted as s t =Hp t , where p t ∈￡ p×1 , t=1,2,...,K represents the unknown complex coordinate vector of the target signal;

[0082] To facilitate derivation, define the augmented matrix In order to estimate the unknown clutter covariance matrix M, it is assumed that there is a set of training data Y containing only clutter components, which is usually collected from R distance units adjacent to the distance unit to be measured, and is recorded as Y = [y1, y2, ..., y R ]∈￡ N×R , where y l =c l ,l=1,2,...,R is also independent and identically distributed and satisfies c l ~CN(0 N×1 ,M).

[0083] Furthermore, according to one embodiment of the present invention, based on the skew symmetry of the clutter covariance matrix, a unitary transformation is performed on the test data Z and the parameter matrix set {M, P, Q} to construct a robust detection framework, including:

[0084] Based on the Gradient test criterion, an intelligent robust detection framework for targets under multi-disturbance background is constructed. The one-step Gradient test decision formula of complex signals is expressed as:

[0085]

[0086] Among them, λ 1S-Gradient and T 1S-Gradient Represent the test statistic and threshold respectively; test data Z=[z1,z2,...,z K]∈￡ N×K , target signal coordinate matrix P=[p1,p2,...,p K ]∈￡ p×K , interference signal coordinate matrix Q=[q1,q2,...,q K ]∈￡ q×K ;Unknown parameter set The relevant parameter Θ r =vec(P)∈￡ pK×1 , interference parameters represents the maximum likelihood estimate of Θ under the H0 hypothesis, Represents Θ r The maximum likelihood estimate under the H1 assumption, Θ r0 Represents Θ r True value under the H0 hypothesis; superscript (·) T represents transpose, and the vec(·) function realizes the vectorization of the matrix; under the H0 assumption, the target does not exist, so Θ r0 =0 pK×1 ;f i (Z,Y|M,iP,Q) represents the joint conditional probability density function of the test data Z and the training data Y under the H0 or H1 hypothesis; based on the statistical independence between the test data and the training data, the joint conditional probability density function can be expressed as:

[0087] f i (Z,Y|M,iP,Q)={π N(K+R) |M| K+R} -1 ×exp[-tr(M -1 T i )],i=0,1 (2)

[0088] Among them, T i =S+(ZB i D)(ZB i D) H , sample covariance matrix S = YY H , and define the augmented matrix B i =[iH,J] and D = [P T ,Q T ] T ; |·| represents the determinant of a square matrix, the tr(·) function represents the trace of a square matrix, the exp(·) function represents the exponential operation, and the superscript (·) H represents the conjugate transpose; the condition p+q≤N must be satisfied to ensure that B1=[H,J] is a full-rank augmented matrix; in addition, by utilizing the skew symmetry of the clutter covariance matrix M, it can be derived that:

[0089] tr(M -1 Ti )=tr[D N (M * ) -1 D N T i ]=tr(M -1 D N T i * D N ),i=0,1 (3)

[0090] Among them, the superscript (·) * represents conjugation; therefore, Equation (2) can be rewritten as:

[0091] f i (Z,Y|M,iP,Q)={π N(K+R) |M| K+R} -1 ×exp[-tr(M -1 T ip )],i=0,1 (4)

[0092] in, and

[0093]

[0094] In the above formula, and represent the real and imaginary parts respectively, and

[0095]

[0096] Further, according to an embodiment of the present invention, in the absence of a target assumption, the maximum likelihood estimation of the clutter covariance matrix M and the interference coordinate matrix Q is solved; in the presence of a target assumption, the target parameter vector Θ is solved. rp-1s The maximum likelihood estimate of the target parameter vector Θ is obtained by using the complex Gaussian joint probability density function of the test data Z and the training data Y. rp-1s Obtain the partial derivative, and then substitute the maximum likelihood estimate of the clutter covariance matrix M and the interference coordinate matrix Q obtained under the no-target assumption into the derivative result to obtain the calculation result; finally, substitute the calculation result and the maximum likelihood estimate of the target parameter vector under the target assumption into the range-extended target gradient detection statistic to construct the detection statistic λ of the intelligent robust target detection method under multi-disturbance background, including:

[0097] Let the unknown parameter set after unitary transformation be The relevant parameters after unitary change Θ rp-1s =vec(P p )∈￡ 2pK×1 , interference parameters The skew-symmetric one-step gradient test decision formula in formula (1) can be rewritten as:

[0098]

[0099] in Represents Θ p-1s The maximum likelihood estimation under the H0 hypothesis is: Represents Θ rp-1s Maximum likelihood estimation under the H1 hypothesis;

[0100] Under the H1 hypothesis, the natural logarithm of equation (4) is about Θ rp-1s Taking partial derivatives, we can get:

[0101]

[0102] Under the H0 assumption, let the derivative of Equation (4) with respect to the clutter covariance matrix M be equal to zero, and we can obtain p The maximum likelihood estimate of M under the conditions of and γ:

[0103]

[0104] The formula (9) obtained Replacing the unknown true covariance matrix M in equation (4), we can obtain:

[0105]

[0106] Among them, I m represents the m×m dimensional identity matrix; let (10) be the value of Q p Find the partial derivative and set the result to zero, and you can get Q under the H0 hypothesis p The maximum likelihood estimate of :

[0107]

[0108] in, Substituting Equation (11) into Equation (9) and using the matrix inversion lemma, we can obtain The inverse matrix of :

[0109]

[0110] in, Represents projection to matrix Then, substituting equations (11) and (12) into equation (8), we can obtain:

[0111]

[0112] in,

[0113] Under the H1 assumption, let equation (4) find the partial derivative of M and set the result to zero, and we can get p The maximum likelihood estimate of M under the condition:

[0114]

[0115] Then substitute formula (14) into formula (4), we can get:

[0116]

[0117] Formula (15) for D p Find the partial derivative and set the result to zero, and we can get D under the assumption H1 p The maximum likelihood estimate of :

[0118]

[0119] in, According to formula (5), P p The maximum likelihood estimation under the assumption H1 can be derived as:

[0120]

[0121] Therefore, we can directly get Θ under the H1 assumption rp-1s The maximum likelihood estimate of is:

[0122]

[0123] Substituting equations (13) and (18) into equation (7), we can obtain the Gradient test statistic in a uniform clutter environment:

[0124]

[0125] in, Represents projection to matrix The orthogonal projection matrix of the column space.

[0126] Furthermore, according to an embodiment of the present invention, a detection threshold T is determined based on a preset false alarm probability, and the detection statistic λ is compared with the detection threshold T. Then, based on the comparison result, it is determined whether there is a range-extended target in the current range unit to be detected, including:

[0127] If λ≥T, it is determined that there is a range extension target in the current range unit to be detected, and the test data Z does not participate in the subsequent training data update of other range units;

[0128] If λ<T, it is determined that there is no range extension target in the current range unit to be detected, and the current test data Z is included in the training data set for subsequent detection.

[0129] According to the above-mentioned scheme of the present invention, the existing wideband radar range-extended target adaptive detector under multiple disturbance backgrounds has difficulty in taking into account the CFAR characteristics, detection performance and computational complexity. At the same time, considering the problem that the non-uniformity of actual clutter makes it difficult to obtain pure clutter training data, how to fully exploit the structural information of the clutter covariance matrix, further reduce the demand for the amount of training data, and improve the estimation accuracy of the unknown clutter covariance matrix, and then construct a closed-form intelligent robust detection method for targets under multiple disturbance backgrounds. While ensuring the CFAR characteristics, it takes into account the multiple requirements of the intelligent anti-interference, computational complexity and detection performance of the range-extended target adaptive detection algorithm, thereby improving the adaptive detection performance of multi-channel wideband radar for weak targets in complex interference environments.

[0130] The present invention constructs an intelligent robust detector for targets under multiple disturbance backgrounds. As can be seen from formula (19), the proposed intelligent robust detection method for targets under multiple disturbance backgrounds has a closed-form detection statistic expression and does not require iterative operations. Compared with the P2S-Rao-HE detector, the proposed method does not need to calculate the Fisher information matrix during the solution process, which significantly reduces the difficulty of algorithm implementation and improves the feasibility of engineering applications. It is also worth noting that compared with the P2S-GLRT-HE detector, the intelligent robust detection method for targets under multiple disturbance backgrounds has lower algorithm computational complexity and has stronger detection performance for weak targets. In summary, the intelligent robust detection method for targets under multiple disturbance backgrounds of the present invention can effectively balance the reasonable balance between algorithm computational complexity, intelligent anti-interference and detection performance while maintaining the CFAR characteristics.

[0131] Furthermore, to achieve the above-mentioned object, the present invention also provides an intelligent robust target detection system under multiple disturbance backgrounds, comprising:

[0132] The robust detection framework construction module constructs test data Z consisting of echo data from K range cells to be measured and training data Y collected from R adjacent reference cells. Under the condition that the clutter covariance matrix M, the target signal coordinate matrix P, and the interference signal coordinate matrix Q are all unknown, a unitary transformation is performed on the test data Z and the parameter matrix set {M, P, Q} based on the skew symmetry of the clutter covariance matrix to build a robust detection framework.

[0133] The detection statistics construction module solves the maximum likelihood estimation of the clutter covariance matrix M and the interference coordinate matrix Q in the absence of target assumptions; solves the target parameter vector Θ in the presence of target assumptions rp-1sThe maximum likelihood estimate of the target parameter vector Θ is obtained by using the complex Gaussian joint probability density function of the test data Z and the training data Y. rp-1s Obtain the partial derivative, then substitute the maximum likelihood estimate of the clutter covariance matrix M and the interference coordinate matrix Q obtained under the no-target assumption into the derivative result to obtain the calculation result; finally, substitute the calculation result and the maximum likelihood estimate of the target parameter vector under the target assumption into the range-extended target gradient detection statistic to construct the detection statistic λ of the intelligent robust target detection method under multi-disturbance background;

[0134] The target detection and judgment module determines the detection threshold T according to the preset false alarm probability, compares the detection statistic λ with the detection threshold T, and determines whether there is a range-extended target in the current range unit to be detected based on the comparison result.

[0135] The intelligent robust detection system for targets under multiple disturbance backgrounds according to the present invention can implement the intelligent robust detection method for targets under multiple disturbance backgrounds. The specific process steps are as described above and will not be repeated here.

[0136] Furthermore, to achieve the above-mentioned objectives, the present invention also provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the method for intelligent and robust target detection in a multi-disturbance background as described above is implemented.

[0137] Furthermore, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for intelligent and robust detection of targets in a multi-disturbance background as described above is implemented.

[0138] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiment described herein is only an optimal embodiment of the present invention and is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0139] Example 1

[0140] Reference Figure 1 The intelligent robust target detection method under multiple disturbance backgrounds in Example 1 is divided into the following steps:

[0141] Step A1: Use the sea detection radar to illuminate the sea area to be detected and obtain the test data Z of K distance units to be detected; and illuminate the target-free range around the sea area to be detected and obtain the training data Y of R reference distance units containing only pure sea clutter. The test data Z and the training data Y are transformed according to equations (5) and (6) to obtain the skew-symmetric transformation test data Z p ; Transform the skew symmetric test data Z p And the training data Y are obtained according to formula (9) and formula (11) respectively under the H0 hypothesis M and Q p The maximum likelihood estimate of According to formula (13), we can obtain the test data Z under the H1 hypothesis. p The complex Gaussian joint probability density function of the training data Y is the target parameter vector Θ r The derivative result of Θ under the H1 hypothesis is obtained according to formula (18). rp-1s The maximum likelihood estimate of

[0142] It is worth noting that in step A1, considering that external interference may exist in the actual ocean environment and adversely affect the adaptive detection of range-extended targets, the intelligent robust target detection method under multiple disturbances of the present invention incorporates external interference into the detector design process and uses subspace signals to model the interference to reduce the potential mismatch effect of the interference signal. In environments with subspace-structured interference, the intelligent gradient target detection method of the present invention can effectively suppress interference signals of varying intensities, demonstrating excellent intelligent anti-interference capabilities. Furthermore, the detector of the present invention is applicable to non-uniform sea clutter environments, such as partially uniform ones. By fully exploiting the local uniformity of sea clutter and utilizing maximum likelihood estimation of the clutter power factor under the no-target assumption, the detector's intelligent adaptability to non-uniform sea clutter environments is enhanced.

[0143] In step A2, the results obtained from the maximum likelihood estimation solution under the H0 assumption, the derivative of the probability density function under the H1 assumption, and the maximum likelihood estimation solution under the H1 assumption are used to construct the detection statistic λ of the target intelligent robust detection method under multi-disturbance background according to formula (19).

[0144] Notably, in step A2, the proposed method achieves superior detection performance compared to detectors such as P2S-GLRT-HE and P2S-Rao-HE for range-extended targets, particularly in the absence of sufficient training data. Furthermore, the proposed intelligent robust detector for targets in multi-perturbation environments has a closed-form expression. Compared to existing adaptive detection methods for range-extended targets, it maintains the CFAR characteristic while striking a reasonable balance between detection performance and computational complexity, thereby enhancing the adaptive detection capabilities of multi-channel broadband radars for small, weak targets on the sea surface in complex electromagnetic environments.

[0145] Step A3: Set the detection threshold T according to the preset false alarm probability: Specifically, set the false alarm probability to P fa According to the Monte Carlo method, based on the 100 / P accumulated in the early stage fa The detection threshold T is calculated based on the measured sea clutter data. Considering the difficulty of obtaining sea clutter, if the actual amount of pure sea clutter measured data R is less than 100 / P fa , then the missing 100 / P fa -R clutter data can be simulated using a sea clutter simulation model, where the model parameters are reasonably estimated and set based on the obtained pure sea clutter measured data. Furthermore, the detection statistic λ is compared with the detection threshold T. If λ≥T, it is determined that the current K range cells to be detected have a range-extended target, and the test data is not used as training data for the subsequent range cells to be detected. Conversely, if λ<T, it is determined that the current K range cells to be detected do not have a range-extended target, and the test data is used as training data for the subsequent range cells to be detected.

[0146] The performance comparison results of the detector when there is sufficient training data are shown in the attached Figure 2 The results show that compared with existing detectors for range extended targets such as P2S-GLRT-HE and P2S-Rao-HE, the detector proposed by the proposed method has better detection performance when there is sufficient training data.

[0147] Example 2

[0148] Reference Figure 1 The intelligent robust detection method for targets under multiple disturbance backgrounds in Example 2 is divided into the following steps:

[0149] Step B1: Use the ground detection radar to illuminate the area to be detected and obtain test data Z of K range units to be detected; then illuminate the target-free range around the area to be detected and obtain training data Y of R reference range units containing only pure ground clutter. The test data Z and training data Y are sent to the data transformation module; in the data transformation module, the skew-symmetric transformation test data Z is obtained according to equations (5) and (6). p ; Transform the skew symmetric test data Z p And the training data Y are obtained according to formula (9) and formula (11) respectively under the H0 hypothesis M and Q p The maximum likelihood estimate of According to formula (13), we can obtain the test data Z under the H1 hypothesis. p The complex Gaussian joint probability density function of the training data Y is the target parameter vector Θ r The derivative result of Θ under the H1 hypothesis is obtained according to formula (18). rp-1s The maximum likelihood estimate of

[0150] It is worth noting that in step B1, considering that there may be external interference in the actual geographical environment that has an adverse effect on the adaptive detection of distance-extended targets, the intelligent robust detection method for targets under multi-disturbance backgrounds of the present invention takes external interference into account in the detector design process and uses subspace signals to model the interference to reduce the possible mismatch effects of the interference signals. For interference environments with subspace structure, the intelligent detection method for gradient targets of distance-extended targets of the present invention can effectively suppress interference signals of different intensities and has good intelligent anti-interference performance. At the same time, the detector of the method of the present invention can be applied to non-uniform ground clutter environments such as partially uniform ground clutter. By fully exploiting the local uniformity of ground clutter and using the maximum likelihood estimation of the clutter power factor under the assumption of no target, the intelligent adaptability of the detector to non-uniform ground clutter environments is improved.

[0151] In step B2, the results obtained from the maximum likelihood estimation solution under the H0 assumption, the derivative of the probability density function under the H1 assumption, and the maximum likelihood estimation solution under the H1 assumption are used to construct the detection statistic λ of the target intelligent robust detection method under multi-disturbance background according to formula (19).

[0152] Notably, in step B2, the proposed method achieves superior detection performance compared to detectors such as P2S-GLRT-HE and P2S-Rao-HE for range-extended targets, particularly in the absence of sufficient training data. Furthermore, the proposed intelligent robust detector for targets in multi-perturbation environments has a closed-form expression. Compared to existing adaptive detection methods for range-extended targets, it maintains the CFAR characteristic while striking a reasonable balance between detection performance and computational complexity, thereby enhancing the adaptive detection capabilities of multi-channel broadband radars for small, weak ground targets in complex electromagnetic environments.

[0153] Step B3, setting the detection threshold T according to the preset false alarm probability: Specifically, setting the false alarm probability to P fa According to the Monte Carlo method, based on the 100 / P accumulated in the early stage fa The detection threshold T is calculated based on the measured ground clutter data. Considering the difficulty of obtaining ground clutter, if the amount of pure ground clutter measured data R actually obtained is less than 100 / P fa , then the missing 100 / P fa-R clutter data can be simulated using a ground clutter simulation model, where the model parameters are reasonably estimated and set based on the obtained pure ground clutter measured data. Furthermore, the detection statistic λ is compared with the detection threshold T. If λ≥T, it is determined that the current K range cells to be detected have a range-extended target, and the test data is not used as training data for the subsequent range cells to be detected. Conversely, if λ<T, it is determined that the current K range cells to be detected do not have a range-extended target, and the test data is used as training data for the subsequent range cells to be detected.

[0154] The comparison results of detector performance when training data is scarce are shown in the attached Figure 3 The results show that compared with existing detectors for range extended targets such as P2S-GLRT-HE and P2S-Rao-HE, the detector proposed in this paper has better detection performance when training data is scarce.

[0155] Those skilled in the art will appreciate that the modules and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0156] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and equipment can refer to the corresponding processes in the aforementioned method implementation methods and will not be repeated here.

[0157] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is merely a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0158] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the objectives of the embodiments of the present invention.

[0159] In addition, each functional module in the embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0160] If the functions are implemented as software modules and sold or used as standalone products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the energy-saving signal transmission / reception method according to various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, ROM, RAM, a magnetic disk, or an optical disk.

[0161] The above description is merely a preferred embodiment of the present application and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the inventive concept. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this application.

[0162] It should be understood that the size of the serial numbers of each step in the content of the invention and the implementation methods of the present invention does not absolutely mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the implementation methods of the present invention.

Claims

1. An intelligent robust target detection method under multiple disturbance backgrounds, characterized by: include: A robust detection framework is constructed by constructing test data Z consisting of echo data from K range cells to be measured and training data Y collected from R adjacent reference cells. Under the condition that the clutter covariance matrix M, the target signal coordinate matrix P, and the interference signal coordinate matrix Q are all unknown, a unitary transformation is performed on the test data Z and the parameter matrix set {M, P, Q} based on the skew symmetry of the clutter covariance matrix. In the absence of target assumption, solve the maximum likelihood estimation of the clutter covariance matrix M and the interference coordinate matrix Q; in the presence of target assumption, solve the target parameter vector Θ rp-1s The maximum likelihood estimate of the target parameter vector Θ is obtained by using the complex Gaussian joint probability density function of the test data Z and the training data Y. rp-1s Obtain the partial derivative, then substitute the maximum likelihood estimate of the clutter covariance matrix M and the interference coordinate matrix Q obtained under the no-target assumption into the derivative result to obtain the calculation result; finally, substitute the calculation result and the maximum likelihood estimate of the target parameter vector under the target assumption into the range-extended target gradient detection statistic to construct the detection statistic λ of the intelligent robust target detection method under multi-disturbance background; The detection threshold T is determined according to the preset false alarm probability. The detection statistic λ is compared with the detection threshold T, and the comparison result is used to determine whether there is a range-extended target in the current range unit to be detected.

2. The intelligent robust detection method for targets under multi-disturbance background according to claim 1 is characterized in that: Construct test data Z consisting of echo data from K range cells to be measured and training data Y collected from R adjacent reference cells, including: The test data Z is collected from K adjacent distance units, and after sampling and organization, it forms an N×1 dimensional vector, which is expressed as z t ∈￡ N×1 ,t=1,2,...,K, where N represents the product of the number of antenna elements and the number of pulses, represents a set of N×1 dimensional complex matrices; the disturbance in the test data Z consists of clutter components and interference components, which are represented by c t ∈￡ N×1 and j t ∈￡ N×1 ,t=1,2,...,K;set the interference component j t ,t=1,2,...,K is modeled as a deterministic subspace signal, belonging to the known multi-rank subspace J∈￡ N×q , denoted as j t =Jq t , where q t ∈￡ q×1 , t=1,2,...,K represents the unknown complex coordinate vector of the interference signal; and the clutter component c t , t=1,2,...,K are independent and identically distributed among different distance units and obey the zero-mean circularly symmetric complex Gaussian distribution of the covariance matrix M, that is, c t ~CN(0 N×1 ,M), where M is an unknown Hermitian positive definite matrix; set the target signal s t ,t=1,2,...,K is also modeled as a deterministic subspace signal, belonging to the known multi-rank subspace H∈￡ N×p , denoted as s t =Hp t , where p t ∈￡ p×1 , t=1,2,...,K represents the unknown complex coordinate vector of the target signal; To facilitate derivation, define the augmented matrix In order to estimate the unknown clutter covariance matrix M, it is assumed that there is a set of training data Y containing only clutter components, which is usually collected from R distance units adjacent to the distance unit to be measured, and is recorded as Y = [y1, y2, ..., y R ]∈￡ N×R , where y l =c l ,l=1,2,...,R is also independent and identically distributed and satisfies c l ~CN(0 N×1 ,M).

3. The intelligent robust detection method for targets under multi-disturbance background according to claim 2 is characterized in that: Based on the skew symmetry of the clutter covariance matrix, a unitary transformation is performed on the test data Z and the parameter matrix set {M, P, Q} to construct a robust detection framework, including: Based on the Gradient test criterion, an intelligent robust detection framework for targets under multi-disturbance background is constructed. The one-step Gradient test decision formula of complex signals is expressed as: Among them, λ 1S-Gradient and T 1S-Gradient Represent the test statistic and threshold respectively; test data Z=[z1,z2,...,z K ]∈￡ N×K , target signal coordinate matrix P=[p1,p2,...,p K ]∈￡ p×K , interference signal coordinate matrix Q=[q1,q2,...,q K ]∈￡ q×K ;Unknown parameter set , where the relevant parameter Θ r =vec(P)∈￡ pK×1 , interference parameters ; represents the maximum likelihood estimate of Θ under the H0 hypothesis, Represents Θ r The maximum likelihood estimate under the H1 assumption, Θ r0 Represents Θ r True value under the H0 hypothesis; superscript (·) T represents the transpose, and the vec(·) function realizes the vectorization of the matrix; under the H0 assumption, the target does not exist, so Θ r0 =0 pK×1 ;f i (Z,Y|M,iP,Q) represents the joint conditional probability density function of the test data Z and the training data Y under the H0 or H1 hypothesis; based on the statistical independence between the test data and the training data, the joint conditional probability density function can be expressed as: f i (Z,Y|M,iP,Q)={π N(K+R) |M| K+R } -1 ×exp[-tr(M -1 T i )],i=0,1 (2) Among them, T i =S+(ZB i D)(ZB i D) H , sample covariance matrix S = YY H , and define the augmented matrix B i =[iH,J] and D = [P T ,Q T ] T ; |·| represents the determinant of a square matrix, the tr(·) function represents the trace of a square matrix, the exp(·) function represents the exponential operation, and the superscript (·) H represents the conjugate transpose; the condition p+q≤N must be satisfied to ensure that B1=[H,J] is a full-rank augmented matrix; in addition, by utilizing the skew symmetry of the clutter covariance matrix M, it can be derived that: tr(M -1 T i )=tr[D N (M * ) -1 D N T i ]=tr(M -1 D N T i * D N ),i=0.1 (3) Among them, the superscript (·) * represents conjugation; therefore, Equation (2) can be rewritten as: f i (Z,Y|M,iP,Q)={π N(K+R) |M| K+R } -1 ×exp[-tr(M -1 T ip )],i=0,1 (4) in, and In the above formula, and represent the real and imaginary parts respectively, and 4. The intelligent robust detection method for targets under multi-disturbance background according to claim 3 is characterized in that: In the absence of target assumption, solve the maximum likelihood estimation of the clutter covariance matrix M and the interference coordinate matrix Q; in the presence of target assumption, solve the target parameter vector Θ rp-1s The maximum likelihood estimate of the target parameter vector Θ is obtained by using the complex Gaussian joint probability density function of the test data Z and the training data Y. rp-1s Obtain the partial derivative, and then substitute the maximum likelihood estimate of the clutter covariance matrix M and the interference coordinate matrix Q obtained under the no-target assumption into the derivative result to obtain the calculation result; finally, substitute the calculation result and the maximum likelihood estimate of the target parameter vector under the target assumption into the range-extended target gradient detection statistic to construct the detection statistic λ of the intelligent robust target detection method under multi-disturbance background, including: Let the unknown parameter set after unitary transformation be The relevant parameters after unitary change Θ rp-1s =vec(P p )∈￡ 2pK×1 , interference parameters The skew-symmetric one-step gradient test decision formula in formula (1) can be rewritten as: in Represents Θ p-1s The maximum likelihood estimation under the H0 hypothesis is: Represents Θ rp-1s Maximum likelihood estimation under the H1 hypothesis; Under the H1 hypothesis, the natural logarithm of equation (4) is about Θ rp-1s Taking partial derivatives, we can get: Under the H0 assumption, let the derivative of Equation (4) with respect to the clutter covariance matrix M be equal to zero, and we can obtain p The maximum likelihood estimate of M under the conditions of and γ: The formula (9) obtained Replacing the unknown true covariance matrix M in equation (4), we can obtain: Among them, I m represents the m×m dimensional identity matrix; let (10) be the value of Q p Find the partial derivative and set the result to zero, and you can get Q under the H0 hypothesis p The maximum likelihood estimate of : in, Substituting Equation (11) into Equation (9) and using the matrix inversion lemma, we can obtain The inverse matrix of : in, Represents projection to matrix Then, substituting equations (11) and (12) into equation (8), we can obtain: in, Under the H1 assumption, let equation (4) find the partial derivative of M and set the result to zero, and we can get p The maximum likelihood estimate of M under the condition: Then substitute formula (14) into formula (4), we can get: Formula (15) for D p Find the partial derivative and set the result to zero, and we can get D under the assumption H1 p The maximum likelihood estimate of : in, According to formula (5), P p The maximum likelihood estimation under the assumption H1 can be derived as: Therefore, we can directly get Θ under the H1 assumption rp-1s The maximum likelihood estimate of is: Substituting equations (13) and (18) into equation (7), we can obtain the gradient test statistic in a uniform clutter environment: in, Represents projection to matrix The orthogonal projection matrix of the column space.

5. The method for intelligent robust target detection under multi-disturbance background according to any one of claims 1 to 4, characterized in that: The detection threshold T is determined based on the preset false alarm probability. The detection statistic λ is compared with the detection threshold T. Based on the comparison result, it is determined whether there is a range-extended target in the current range unit to be detected, including: If λ≥T, it is determined that there is a range extension target in the current range unit to be detected, and the test data Z does not participate in the subsequent training data update of other range units; If λ<T, it is determined that there is no range extension target in the current range unit to be detected, and the current test data Z is included in the training data set for subsequent detection.

6. Intelligent robust target detection system under multiple disturbance backgrounds, characterized by: include: The robust detection framework construction module constructs test data Z consisting of echo data from K range cells to be measured and training data Y collected from R adjacent reference cells. Under the condition that the clutter covariance matrix M, the target signal coordinate matrix P, and the interference signal coordinate matrix Q are all unknown, a unitary transformation is performed on the test data Z and the parameter matrix set {M, P, Q} based on the skew symmetry of the clutter covariance matrix to build a robust detection framework. The detection statistics construction module solves the maximum likelihood estimation of the clutter covariance matrix M and the interference coordinate matrix Q in the absence of target assumptions; solves the target parameter vector Θ in the presence of target assumptions rp-1s The maximum likelihood estimate of the target parameter vector Θ is obtained by using the complex Gaussian joint probability density function of the test data Z and the training data Y. rp-1s Obtain the partial derivative, then substitute the maximum likelihood estimate of the clutter covariance matrix M and the interference coordinate matrix Q obtained under the no-target assumption into the derivative result to obtain the calculation result; finally, substitute the calculation result and the maximum likelihood estimate of the target parameter vector under the target assumption into the range-extended target gradient detection statistic to construct the detection statistic λ of the intelligent robust target detection method under multi-disturbance background; The target detection and judgment module determines the detection threshold T according to the preset false alarm probability, compares the detection statistic λ with the detection threshold T, and determines whether there is a range-extended target in the current range unit to be detected based on the comparison result.

7. An electronic device, characterized in that The invention comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the method for intelligent and robust detection of targets in a multi-disturbance background according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for intelligent and robust detection of targets under multi-disturbance backgrounds according to any one of claims 1 to 5 is implemented.

Images