A target detection method and device based on an inverse diagonal joint robust detector

Abstract

This invention provides a target detection method and apparatus based on an inverse diagonal joint robust detector, relating to the field of radar detection technology. The method includes: constructing a target steering vector model using a fully incremental form, and building a binary detection model based on the target steering vector for joint target detection and azimuth estimation; upon target detection, solving the first and second final detection formulas using the cyclic coordinate method to adaptively output the target cosine offset, thereby mitigating the sharp decline in the detection performance of the two-dimensional array radar when the target steering vector mismatch occurs; deriving the first and second preset detection formulas using the inverse diagonal matrix, the first unitary matrix, and the second unitary matrix; utilizing the inverse diagonal structure, which allows the covariance matrix of an unknown target to be characterized with only half the parameters, significantly reducing the dependence on the amount of training data, resulting in high target detection sensitivity even with reduced training data and steering vector mismatch.

Description

Technical Field

[0001] This invention relates to the field of radar detection technology, and in particular to a target detection method and apparatus based on an inverse diagonal joint robust detector. Background Technology

[0002] Multi-channel phased array radar systems are widely used on many platforms due to their wide detection range and strong anti-jamming capabilities. Adaptive detection technology can significantly enhance the target detection and early warning capabilities of multi-channel radar systems. Adaptive detection primarily utilizes training data to estimate the covariance matrix of interference and noise in the detection scenario. It detects target signals from noise when training data is sufficient and the steering vector is accurate. However, in real-world scenarios, factors such as spatial variations in distance and geographical environment, increased electromagnetic complexity, special array configurations, and improved radar resolution make the background conditions for target detection extremely demanding and complex. In complex detection scenarios, the training samples are often contaminated with unwanted components, resulting in limited usable training data. This is particularly true for two-dimensional array radars with large-scale arrays, where sufficient training samples are often difficult to obtain. Furthermore, in reality, targets rarely align with the nominal point direction, leading to a mismatch between the nominal and actual target steering vectors. Additionally, when the radar beam turns towards the target, a target component is generated. Therefore, researching robust detectors under conditions of insufficient training samples and steering vector mismatch is crucial for improving the detection capabilities of two-dimensional array radars.

[0003] Currently, the Generalized Likelihood Ratio Test (GLRT) is commonly used for target detection in multi-channel phased array radars. However, the GLRT is only suitable for applications with sufficient radar training data and accurate steering vectors. The GLRT is not applicable when training data is reduced or steering vector mismatch occurs, resulting in lower target detection sensitivity under these conditions. Summary of the Invention

[0004] The purpose of this invention is to provide a target detection method and apparatus based on an inverse diagonal joint robust detector, which solves the problem of low sensitivity of target detection when training data is reduced and steering vector mismatch occurs.

[0005] To address the aforementioned technical problems, the embodiments of the present invention provide the following technical solutions:

[0006] The first aspect of this invention provides a target detection method based on an inverse diagonal joint robust detector, the target detection method comprising:

[0007] Upon receiving the radar echo, a target guidance vector model is constructed using a fully incremental form based on the target's three-dimensional spatial position and the radar's estimated target azimuth position.

[0008] Based on the target-oriented vector model, the noise component in the test data, the noise component in the training data, and the signal amplitude, a binary detection model is constructed. The covariance matrix of the target is determined based on the noise component in the test data and the noise component in the training data. The binary detection model is used to determine the test data and the training data.

[0009] Based on the inverse diagonal matrix, the target covariance matrix, the estimated covariance matrix, the signal amplitude matrix, the first unitary matrix, and the second unitary matrix, the joint probability density function of the test data matrix and the training data matrix is ​​determined. The test data matrix is ​​determined by the test data, and the training data matrix is ​​determined by the training data.

[0010] Based on the joint probability density function and the generalized likelihood ratio test criterion, the first and second preset detection formulas for the inverse diagonal parameters are determined, and the first and second preset detection formulas are both optimized by scores to obtain the corresponding first and second final detection formulas.

[0011] The first and second final detection formulas are solved using the cyclic coordinate method to construct the first and second robust detectors.

[0012] The presence of a target is detected using the Monte Carlo method, a first robust detector, and a second robust detector.

[0013] A second aspect of the present invention provides a target detection device based on an inverse diagonal joint robust detector, the target detection device comprising:

[0014] The first construction module is used to construct a target guidance vector model in a fully incremental form based on the target's three-dimensional spatial position and the target's estimated azimuth position by the radar when the radar echo is received.

[0015] The second construction module is used to construct a binary detection model based on the target guidance vector model, the noise component in the test data, the noise component in the training data, and the signal amplitude, and to determine the covariance matrix of the target based on the noise component in the test data and the noise component in the training data. The binary detection model is used to determine the test data and the training data.

[0016] The determination module is used to determine the joint probability density function of the test data matrix and the training data matrix based on the inverse diagonal matrix, the target covariance matrix, the estimated covariance matrix, the signal amplitude matrix, the first unitary matrix, and the second unitary matrix. The test data matrix is ​​determined by the test data, and the training data matrix is ​​determined by the training data.

[0017] The score optimization module is used to determine the first and second preset detection formulas of the inverse diagonal parameters based on the joint probability density function and the generalized likelihood ratio test criterion, and to perform score optimization on both the first and second preset detection formulas to obtain the corresponding first and second final detection formulas.

[0018] The third construction module is used to solve the first final detection formula and the second final detection formula according to the cyclic coordinate method, so as to construct the first robust detector and the second robust detector.

[0019] The detection module is used to detect the presence of a target based on the Monte Carlo method, a first robust detector, and a second robust detector.

[0020] Compared to existing technologies, the target detection method and apparatus based on an inverse diagonal joint robust detector provided by this invention, upon receiving a radar echo, constructs a target steering vector model using a full-incremental form based on the target's three-dimensional spatial position and the radar-estimated target azimuth position; based on the target steering vector model, noise components in the test data, noise components in the training data, and signal amplitude, a binary detection model is constructed, and the target's covariance matrix is ​​determined based on the noise components in the test data and the noise components in the training data. The binary detection model is used to determine the test data and training data; based on the inverse diagonal matrix, the target's covariance matrix, the estimated covariance matrix, ... The signal amplitude matrix, the first unitary matrix, and the second unitary matrix are used to determine the joint probability density function of the test data matrix and the training data matrix. Based on the joint probability density function and the generalized likelihood ratio test criterion, the first and second preset detection formulas with inverse diagonal parameters are determined, and both the first and second preset detection formulas are optimized by scores to obtain the corresponding first and second final detection formulas. The first and second final detection formulas are solved using the cyclic coordinate method to construct the first and second robust detectors. The presence of a target is detected using the Monte Carlo method, the first robust detector, and the second robust detector. In this way, by constructing a target guidance vector model in a fully incremental form, a binary detection model based on the joint target detection and azimuth estimation is built. When a target is detected, the first and second final detection formulas are solved by using the cyclic coordinate method, and the target cosine offset can be adaptively output, thereby mitigating the sharp decline in the detection performance of the two-dimensional array radar when the target guidance vector is mismatched. The first and second preset detection formulas are derived using the inverse diagonal matrix, the first unitary matrix, and the second unitary matrix. The use of the inverse diagonal structure allows the covariance matrix of the unknown target to be characterized with half of the parameters, which greatly reduces the dependence on the amount of training data and makes the target detection sensitivity high even when the training data is reduced and the guidance vector is mismatched. Attached Figure Description

[0021] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, with the same or corresponding reference numerals denoteing the same or corresponding parts, wherein:

[0022] Figure 1 A flowchart illustrating a target detection method based on an inverse diagonal joint robust detector is shown.

[0023] Figure 2 A schematic diagram illustrates a three-dimensional image of the cosine similarity between the target guidance vector and the actual guidance vector;

[0024] Figure 3 The probability curves of each detector are schematically shown when the training data is sufficient and insufficient.

[0025] Figure 4 The mean-square error (MSE) curve of the detector of the present invention is schematically shown.

[0026] Figure 5 A schematic diagram of a target detection device based on an inverse diagonal joint robust detector is shown. Detailed Implementation

[0027] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art.

[0028] It should be noted that, unless otherwise stated, the technical or scientific terms used in this invention should have the ordinary meaning as understood by one of ordinary skill in the art.

[0029] The methods described in the embodiments of the present invention will be explained in detail below.

[0030] Figure 1 A flowchart illustrating a target detection method based on an inverse diagonal joint robust detector according to an embodiment of the present invention is shown in the figure. See [link to flowchart illustration]. Figure 1 As shown, the target detection method may include:

[0031] S101. Upon receiving a radar echo, a target guidance vector model is constructed using a fully incremental form based on the target's three-dimensional spatial position and the radar's estimated target azimuth position.

[0032] The radar-estimated target bearing position includes a first estimated target bearing position and a second estimated target bearing position.

[0033] Specifically, based on the target's three-dimensional spatial position and the radar-estimated target azimuth position, a target guidance vector model is constructed using a full-incremental approach, including:

[0034] Step A1: Upon receiving the radar echo, convert the target's three-dimensional spatial position to cosine space to obtain the true target direction position.

[0035] The true target direction position includes the first true target direction position and the second true target direction position.

[0036] Suppose a two-dimensional array radar with N×M uniformly rectangularly distributed array elements receives a radar echo reflected from a target with range R, azimuth angle θ, and elevation angle φ′. First, the target's three-dimensional spatial position is transformed into cosine space, i.e., cosine uv space:

[0037] u=sin(θ)cos(φ′), v=sin(θ)sin(φ′);

[0038] Where u is the first true target direction position, which is one component of the target direction in the radar array plane and is related to the direction on the horizontal plane of the array. v is the second true target direction position, which is another component of the target direction in the radar array plane and is related to the vertical direction of the array.

[0039] Cosine space is the space in which cosine offsets are calculated.

[0040] Step A2: Based on the first true target direction position, the second true target direction position, and the number of row array elements and column array elements in the uniform rectangular distribution array, form the target's preset guidance vector.

[0041] The expression for the target's preset guidance vector is:

[0042]

[0043] Where g(u,v) is the target's preset guidance vector, g u (u) is the weighted response of each antenna element to the signal from the u direction, g v (v) is the weighted response of each antenna element to the signal from the v direction. Let Z be a complex matrix space of dimension Z×1. Z is the Hadamard product, where Z is the product of the number of row elements and the number of column elements, Z = NM, where N is the number of row elements, M is the number of column elements, u is the first true target orientation position, and v is the second true target orientation position.

[0044] g u (u) and g v The expressions for (v) are as follows:

[0045]

[0046] Where λ0 is the wavelength of the radar transmitted wave, j is the imaginary unit, (·) T Let u be the transpose symbol, v be the first true target orientation position, and y be the second true target orientation position. Assume the coordinate center is located at the center of the two-dimensional array, x0, x1, ..., x N-1 and y0, y1, ..., y M-1 These are the coordinates of the row and column elements in a two-dimensional array, respectively, x. N-1 Let y be the coordinate position of the (N-1)th row element in the two-dimensional array. M-1 Let M be the coordinate position of the (M-1)th column element in the two-dimensional array. For a complex space of dimension N×1, Let be a complex space of dimension M×1.

[0047] Step A3: Determine the first direction cosine offset based on the first true target direction position and the first estimated target azimuth position, and determine the second direction cosine offset based on the second true target direction position and the second estimated target azimuth position.

[0048] In real-world scenarios, the target's arrival direction is usually not consistent with the array's steering vector. Therefore, a direction cosine offset (Δu, Δv) is needed to replace the target's true angular position in (u, v), which is:

[0049]

[0050] Where Δu is the cosine offset in the first direction. Let Δv be the first estimated target azimuth position, and Δv be the second direction cosine offset. This is the second estimated target location. This is the target's estimated azimuth position for radar.

[0051] Step A4: Construct a target guidance vector model based on the first estimated target azimuth position, the second estimated target azimuth position, the target's preset guidance vector, the first direction cosine offset, and the second direction cosine offset.

[0052] The target's preset guidance vector includes the first true target direction position and the second true target direction position.

[0053] Steps A41 and A42 below are specific operations for linearizing the array around the direction of the guide vector.

[0054] Specifically, based on the first estimated target azimuth position, the second estimated target azimuth position, the target's preset guidance vector, the first direction cosine offset, and the second direction cosine offset, a target guidance vector model is constructed, including:

[0055] Step A41: Replace the first true target direction position in the target's preset guidance vector with the first estimated target orientation position, and replace the second true target direction position with the second estimated target orientation position to obtain the target's estimated guidance vector.

[0056] Replace u in the expression for the target's preset guidance vector in step A2 with v is replaced with The estimated steering vector of the target can then be obtained.

[0057] Step A42: Using the first direction cosine offset and the second direction cosine offset, perform Taylor expansion on the first estimated target azimuth position and the second estimated target azimuth position in the estimated target guidance vector to obtain the target guidance vector model.

[0058] Specifically, using the first and second direction cosine offsets, a Taylor expansion is performed on the first and second estimated target azimuth positions in the estimated target steering vector to obtain the target steering vector model, including:

[0059] Using the first direction cosine offset and the second direction cosine offset (Δu, Δv), the first estimated target azimuth position and the second estimated target azimuth position in the estimated target guidance vector are obtained. Perform a Taylor expansion to obtain a full-increment form, and use the following formula to obtain the target-oriented vector model:

[0060]

[0061]

[0062] Among them, g a (Δu,Δv) is the target-oriented vector model. Let Δu be the estimated guide vector for the target, Δv be the cosine offset in the first direction, and Δv be the cosine offset in the second direction. The first estimated target location. This is the second estimated target location. Let Z be a complex matrix space of dimension Z×1, where Z is the product of the number of row elements and the number of column elements, and Z = NM, where N is the number of row elements and M is the number of column elements. Take the partial derivative of the target-oriented vector model with respect to the first true target orientation position u. For the target-oriented vector model, take the partial derivative with respect to the second true target orientation position v, where j is the imaginary unit and λ0 is the radar transmitted wave wavelength. For each antenna element pair from Weighted response of the signal in the direction, x N-1 Let be the coordinates of the (N-1)th row element in the two-dimensional array. For each antenna element pair from The weighted response of the directional signal, ⊙ is the Kronecker product, y M-1 Let M be the coordinate position of the (M-1)th column element in the two-dimensional array. It is the Hadamard product.

[0063] After step A42, to make the symbols concise, Represented as g, and They are respectively represented as and The target-oriented vector model can then be represented as:

[0064]

[0065] in, g a (Δu, Δv) represents the target guidance vector model, where Δu is the cosine offset in the first direction and Δv is the cosine offset in the second direction. The guide vector for estimating the target. The first estimated target location. This is the second estimated target location. Take the partial derivative of the target-oriented vector model with respect to the first true target orientation position u. The partial derivative of the target guidance vector model with respect to the second true target direction position v is calculated, G is the spliced ​​partial guidance matrix, and Δθ is the target cosine deviation angle information.

[0066] S102. Based on the target guidance vector model, the noise component in the test data, the noise component in the training data, and the signal amplitude, construct a binary detection model, and determine the target's covariance matrix based on the noise component in the test data and the noise component in the training data.

[0067] The binary detection model is used to determine the test data and training data.

[0068] The binary detection model is a detection model that combines target detection and orientation estimation.

[0069] Based on the target steering vector model in the two-dimensional array obtained in step S102, assuming that the useful radar target echo signal is submerged in Gaussian interference with unknown spectral characteristics, and assuming that there are K ≥ N uniform training samples (i.e., data vectors that do not contain useful target echo signals), the detection problem of a target located at (u,v) in the cosine uv space can be transformed into a binary assumption problem. Based on the target steering vector model, the noise component in the test data, the noise component in the training data, and the signal amplitude, a binary detection model is constructed:

[0070]

[0071] in, The condition is that there is no target echo signal. Given the condition of having a target echo signal, K is the number of training unit data points, k is any one of the training unit data points, l is the test data, and i is the noise component in the test data. k For training data, i k Let be the noise component in the training data, 'a' be the signal amplitude (which is unknown), and 'g' be the signal amplitude. a (Δu,Δv) is the target guidance vector model, where Δu is the cosine offset in the first direction and Δv is the cosine offset in the second direction.

[0072] In addition, the noise component i in the training data k The test data l is modeled as an independent complex circular Gaussian random vector, and the covariance matrix of the target in the space containing noise and interference information is:

[0073]

[0074] Where S1 is the covariance matrix of the target, which is an unknown covariance matrix. For matrix Expectations Let i be the conjugate transpose of the test data i. For the noise component i in the training data k The conjugate transpose of .

[0075] S103. Based on the inverse diagonal matrix, the target covariance matrix, the estimated covariance matrix, the signal amplitude matrix, the first unitary matrix, and the second unitary matrix, determine the joint probability density function of the test data matrix and the training data matrix.

[0076] The test data matrix is ​​determined by the test data, and the training data matrix is ​​determined by the training data.

[0077] Specifically, based on the inverse diagonal matrix, the target covariance matrix, the estimated covariance matrix, the signal amplitude matrix, the first unitary matrix, and the second unitary matrix, the joint probability density function of the test data matrix and the training data matrix is ​​determined, including:

[0078] Step B1: Determine the true covariance matrix of the target based on the conjugate matrix corresponding to the inverse diagonal matrix and the covariance matrix of the target.

[0079] Before step B1, it is necessary to construct an inverse diagonal matrix. The inverse diagonal structure indicates that the main diagonal and secondary diagonal of the covariance matrix are symmetric, i.e.:

[0080]

[0081] Where J is the inverse diagonal matrix, N is the number of row elements, and M is the number of column elements.

[0082] When a two-dimensional uniform linear array is symmetrical and has the same spacing in the spatial domain, both the target's covariance matrix and the target steering vector model have an inverse diagonal structure. Based on the conjugate matrix corresponding to the inverse diagonal matrix and the target's covariance matrix, the true target covariance matrix is ​​determined.

[0083] S = JS1 * J;

[0084] Where S is the target true covariance matrix, S1 * Let J be the conjugate of the covariance matrix of the target, and J be the inverse diagonal matrix.

[0085] Correspondingly, g a (Δu,Δv)=Jg a (Δu,Δv) * , where g a (Δu,Δv) is the target-oriented vector model, g a (Δu,Δv) * The conjugate of the target-oriented vector model.

[0086] Step B2: Determine the test data matrix based on the test data, and determine the training data matrix based on the training data.

[0087] The test data matrix was determined based on the test data. The training data matrix is ​​determined based on the training data. Where R is the test data matrix, R s Let l be the training data matrix, and l be the test data. k Let K be the number of training data points. Let Z be a complex space of dimension Z×K.

[0088] Step B3: Determine the preset joint probability density function of the test data matrix and the training data matrix based on the target true covariance matrix, the estimated covariance matrix, the adjacency matrix corresponding to the test data matrix, and the adjacency matrix corresponding to the signal amplitude matrix.

[0089] Based on the target's true covariance matrix, the estimated covariance matrix, the adjacency matrix corresponding to the test data matrix, and the adjacency matrix corresponding to the signal amplitude matrix, construct the test data matrix R and the training data matrix R under the two-dimensional array. s The preset joint probability density function is:

[0090]

[0091] In order to simplify the symbolic representation, g will be... a =g a (Δu,Δv), For R and R s The preset joint probability density function, R is the test data matrix, R s For the training data matrix, The condition is that there is no target echo signal. Given the condition of having a target echo signal, K is the number of training data, N is the number of row array elements, S is the true covariance matrix of the target, and H... i Estimate the information matrix for the test data, where |·| is the determinant of the matrix, tr(·) is the rank of the matrix, exp[·] is the exponent with base e, and (·) -1 It is the inverse of the matrix.

[0092] Test data estimation information matrix H i Represented as:

[0093]

[0094] in, To estimate the covariance matrix, R p To find the adjacency matrices of the real and imaginary parts of the test data matrix, a p To find the adjacency matrix of the real and imaginary parts of the signal amplitude, g a For target-oriented vector model, To find the conjugate transpose of the adjacency matrix for the real and imaginary parts of the signal amplitude, and... R p a p They can be represented as:

[0095]

[0096] Among them, R sFor the training data matrix, J is the conjugate transpose of the training data matrix, and J is the inverse diagonal matrix. Let R be a matrix space of dimension Z×Z. e R is the real part of the test data matrix R. o To test the imaginary part of the data matrix R, Let a be a matrix space of dimension Z×2. e Let a be the real part of the signal amplitude a. o Let a be the imaginary part of the signal amplitude a. Given a matrix space of dimension 2×1, The conjugate transpose of a matrix is ​​denoted as . The conjugate transpose of a real matrix is ​​its transpose matrix, and the conjugate transpose of a complex matrix is ​​the transpose of the conjugate of all its elements.

[0097] R e R o a e a o This can be expressed in detail as follows:

[0098]

[0099]

[0100] Where R is the test data matrix, R * The conjugate of the test data matrix, where a is the signal amplitude. It is a matrix space of dimension Z×1.

[0101] Step B4: Using the first and second unitary matrices, perform unitary transformations on the adjacency matrix corresponding to the signal amplitude matrix, the adjacency matrix corresponding to the test data matrix, the estimated covariance matrix, and the target covariance matrix, respectively, to obtain the unitarily transformed real amplitude matrix, the unitarily transformed real test cell data matrix, the unitarily transformed real estimated covariance matrix, and the unitarily transformed real steering vector.

[0102] Define two unitary matrices D P and V P They are respectively:

[0103]

[0104]

[0105] Among them, D p V is the first unitary matrix, representing a unitary matrix of dimension Z×Z. p Let I be the second unitary matrix, representing a unitary matrix of dimension 2. N Let j be an N x N identity matrix, where j is the imaginary unit.

[0106] Using the first unitary matrix D p Second unitary matrix V p For the adjacency matrix corresponding to the signal amplitude matrix and the adjacency matrix R corresponding to the test data matrix, respectively... p Estimating the covariance matrix The covariance matrix g of the target a Performing a unitary transformation, we obtain:

[0107]

[0108] Among them, a r R is the real magnitude matrix after unitary transformation. r This is the data matrix of the actual test unit after unitary transformation. Let g be the real estimated covariance matrix after unitary transformation. r The real steering vector after unitary transformation. The symbol for the unitary transformation is... It is the conjugate transpose of the second unitary matrix. Let Z be a real matrix space of dimension Z×2. To estimate the covariance matrix, It is the conjugate transpose of the first unitary matrix. Let Z be a real matrix space of dimension Z×Z. Let Z be a real matrix space of dimension Z×1. In order to seek the truth, To find the imaginary part.

[0109] Step B5: Substitute the real amplitude matrix after unitary transformation, the real test unit data matrix after unitary transformation, the real estimated covariance matrix after unitary transformation, and the real steering vector after unitary transformation into the preset joint probability density function to obtain the joint probability density function.

[0110] Specifically, the real amplitude matrix after unitary transformation, the real test unit data matrix after unitary transformation, the real estimated covariance matrix after unitary transformation, and the real steering vector after unitary transformation are substituted into the preset joint probability density function to obtain the joint probability density function, including:

[0111] Substituting the unitarily transformed real amplitude matrix, the unitarily transformed real test unit data matrix, the unitarily transformed real estimated covariance matrix, and the unitarily transformed real steering vector into the preset joint probability density function, the joint probability density function is obtained using the following formula:

[0112]

[0113]

[0114]

[0115] in, Let i be the joint probability density function, and i = 0 indicate the transformed... The joint probability density function under the assumption, where i=1 indicates the transformed... The joint probability density function is assumed, where R is the test data matrix, and R0 is the joint probability density function. s For the training data matrix, The condition is that there is no target echo signal. For the condition that there is a target echo signal, S r Given the real covariance matrix, Given the known test data information matrix, Z is the product of the number of row elements and the number of column elements, Z = NM, where N is the number of row elements, M is the number of column elements, K is the number of training unit data, and D... p Let S be the first unitary matrix, and S be the target true covariance matrix. This is the conjugate transpose of the matrix. To find the real part, J is an inverse diagonal matrix. To find the imaginary part, R is the real estimated covariance matrix after unitary transformation. r G is the actual test unit data matrix after unitary transformation. r The real steering vector after unitary transformation. This is the transpose of the real amplitude matrix after the unitary transformation.

[0116] S104. Based on the joint probability density function and the generalized likelihood ratio test criterion, determine the first and second preset detection formulas for the inverse diagonal parameters, and perform fractional optimization on both the first and second preset detection formulas to obtain the corresponding first and second final detection formulas.

[0117] Step S104 is an operation based on GLRT to transform the inverse diagonal parameter detection expressions under the one-step and two-step methods into fractional optimization problems. The one-step method refers to jointly solving for all unknown parameters using the GLRT criterion, while the two-step method refers to first solving for the known real covariance matrix S... r Under the assumptions, the GLRT decision statistics are derived; then, the real estimated covariance matrix after unitary transformation is... Substitute the decision statistics obtained in the first step into the original known real covariance matrix.

[0118] Specifically, based on the joint probability density function, the detection decision formula based on the one-step maximum likelihood criterion is obtained as follows:

[0119]

[0120] in, for The joint probability density function under the assumption, for The joint probability density function under the assumption, The condition is that there is no target echo signal. Given a target echo signal, ar is the real amplitude matrix after unitary transformation, Sr is the known real covariance matrix, Δu is the cosine offset in the first direction, Δv is the cosine offset in the second direction, R is the test data matrix, and R s The training data matrix is ​​ξ, which is the set detection threshold used to ensure constant false alarm rate. α is the first constraint parameter and β is the second constraint parameter. α and β are constraint parameters to ensure that the target steering vector, i.e. the approximate steering vector, tends to be consistent with the actual steering vector. They are generally set within the 3dB beamwidth of the radar.

[0121] Because the real magnitude matrix a after unitary transformation r The cosine offset (Δu, Δv) and the known real covariance matrix S r Since the value is unknown, it needs to be replaced by the value with the largest likelihood ratio obtained sequentially through GLRT. Simultaneously, S0 is calculated for the numerator and denominator of the obtained detection formula. r The minimum value can be obtained by further simplification:

[0122]

[0123] Among them, R r This is the data matrix of the actual test unit after unitary transformation. Let g be the real estimated covariance matrix after unitary transformation, ξ be the set detection threshold used to ensure constant false alarm rate, and g be the real estimated covariance matrix after unitary transformation. r The real steering vector after unitary transformation. Let |·| be the transpose of the real amplitude matrix after unitary transformation, and |·| be the determinant of the matrix.

[0124] Minimizing the denominator of the above equation yields:

[0125]

[0126] Among them, a r The real magnitude matrix after unitary transformation. This is the transpose of the test unit data matrix after unitary transformation.

[0127] The desired α r Substituting into the detection formula, we obtain the following parameterized detection formula:

[0128]

[0129] in, φ is a variable. Let be the transpose of the real oriented vector after the unitary transformation. To facilitate solving the optimization problem, some parameters in the above equation are whitened, i.e., let . The final score detection optimization problem under the one-step GLRT method, i.e., the first final detection formula, is expressed as:

[0130]

[0131] in, The actual whitening guide vector for the target. Here, Δθ represents the whitening bias guidance matrix, and Δθ represents the target cosine deviation angle information. The first whitening covariance matrix, This is the transpose of the first whitening covariance matrix.

[0132]

[0133] D p Let J be the first unitary matrix, and g be the inverse diagonal matrix. * The conjugate of the guide vector for estimating the target. G is the real estimated covariance matrix after unitary transformation. * R is the conjugate of the concatenated partial orientation matrix. r Here, φ is the actual test unit data matrix after unitary transformation, φ is the variable, and I is the identity matrix. Δu is the transpose of the test unit data matrix after unitary transformation, α is the first constraint parameter, β is the second constraint parameter, Δu is the first direction cosine offset, and Δv is the second direction cosine offset.

[0134] In a two-step GLRT detector, assuming the real covariance matrix S is known... r If it is known, then the two-step GLRT detection formula is:

[0135]

[0136] in, Let be the joint probability density function under the no-objective assumption. Let S be the joint probability density function under the objective assumption, R be the test data matrix, and S be the... r Given the real covariance matrix, a r Let be the real amplitude matrix after unitary transformation, Δθ be the target cosine deviation angle information, ξ be the set detection threshold used to ensure constant false alarm rate, α be the first constraint parameter, and β be the second constraint parameter. The condition is that there is no target echo signal. Given the condition of having a target echo signal, Δu is the cosine offset in the first direction, and Δv is the cosine offset in the second direction.

[0137] Wherein, the real covariance matrix S is known. r for:

[0138]

[0139] Among them, R s For the training data matrix, J is the transpose of the training data matrix, and J is the inverse diagonal matrix.

[0140] Similar to the one-step simplification, the two-step GLRT detection formula is simplified based on the maximum likelihood criterion as follows:

[0141]

[0142] in, This is the transpose of the second whitening covariance matrix. The partial guidance matrix is ​​the projection matrix based on the guidance vector. This is the second whitening covariance matrix, i.e., the covariance matrix of the two-step whitening method. The actual whitening guide vector for the target. This refers to the actual guiding vector for the second whitening of the target, i.e., the actual guiding vector for the two-step whitening method of the target. The actual guiding vector for the second whitening of the target The conjugate transpose of R r S is the actual test unit data matrix after unitary transformation. r Let ξ be the known real covariance matrix, and ξ be the set detection threshold used to ensure constant false alarm rate.

[0143] Simplifying the above equation into a fractional optimization problem, we get:

[0144]

[0145] in, g is the transpose of the test unit data matrix after unitary transformation. r The real steering vector after unitary transformation. It is the transpose of the real directional vector after the unitary transformation.

[0146] Similarly, the above equation can be whitened to obtain the second final detection equation, which is expressed as:

[0147]

[0148] in, The second whitening covariance matrix, S is the transpose of the second whitening covariance matrix. r Given the real covariance matrix, D p G is the first unitary matrix. * The conjugate of the spliced ​​partial orientation matrix, For whitening biased guidance matrix, Δθ represents the actual whitening guide vector of the target, and Δθ represents the cosine deviation angle information of the target.

[0149] S105. Solve the first and second final detection formulas respectively using the cyclic coordinate method to construct the first robust detector and the second robust detector.

[0150] Specifically, the first and second final detection equations are solved using the cyclic coordinate method to construct the first and second robust detectors, including:

[0151] Step C1: Solve the first final detection formula using the cyclic coordinate method; that is, solve the parameter optimization problem obtained by the one-step method using the cyclic coordinate method. First, initialize the iteration count n = 0; initialize the optimal azimuth angle θ. (n) =[Δu (n) ,Δv (n) ] T =[0,0] T , where θ (n) Represents the azimuth angle after the nth iteration; initializes the deflection vector. in Will and Δθ (n) Substituting into the one-step detection formula, i.e., the first final detection formula, we obtain the nth detection value t. (n) :

[0152]

[0153] in, Δu (n) Let Δv be the value of the first cosine offset in the nth iteration. (n) This is the value of the second cosine offset in the nth iteration. The actual whitening guide vector for the target. The whitening biased guidance matrix, Δθ (n) For the target cosine deviation angle information in the nth iteration, The first whitening covariance matrix, is the transpose of the first whitening covariance matrix.

[0154] Step C2: After initialization, let n = n + 1 represent the nth iteration. Assuming Δv is an invariant, calculate the optimal value of the first direction cosine offset Δu:

[0155]

[0156] in, Δu * The optimal value of the cosine offset Δu in the first direction is given, and argmax is the parameter Δu that returns the maximum value of the right-hand side.* , The biased Δv steering vector after the (n-1)th iteration. The first whitening covariance matrix, D is the actual target steering vector containing only the cosine component u in the direction. p It is the first unitary matrix. Find the conjugate of the partial derivative of the target-oriented vector model with respect to the first true target direction position u.

[0157] Step C3: Calculate the biased Δu steering vector after the nth iteration. Assuming Δu is an invariant, the optimal value of the second-direction cosine offset Δv is calculated as follows:

[0158]

[0159] Where, Δv * The optimal value of the second-direction cosine offset Δv is... The biased Δu steering vector after the nth iteration. This is the actual target guidance vector that contains only the cosine component v in the direction.

[0160] Step C4: Calculate the biased Δv steering vector after the nth iteration. Update azimuth θ (n) =[Δu * ,Δv * ] T Update the nth detection value under the one-step detection method.

[0161] Step C5: Assume the threshold for determining the end of the loop in the cyclic coordinate method is... If The first optimal azimuth angle, i.e., the optimal azimuth angle under the one-step method, is then obtained as follows: like Then return to step C2 to calculate the iteration.

[0162] t (n-1) This is the value from the (n-1)th test.

[0163] Step C6: Then, the optimal azimuth angle obtained in the one-step method is optimized using the cyclic coordinate method. The second optimal azimuth angle is the optimal azimuth angle under the two-step method. The expression for the first robust detector, i.e., the robust detector that combines target detection and orientation estimation in the one-step method, is:

[0164]

[0165] in, R r This is the data matrix of the actual test unit after unitary transformation. This is the transpose of the test unit data matrix after unitary transformation. The real estimated covariance matrix after unitary transformation. This refers to the optimized first target actual guidance vector, i.e., the target actual guidance vector optimized by the one-step method. This represents the first target magnitude, i.e., the target magnitude estimated using the one-step method. The condition for the existence of a target echo signal. Let ξ be the set detection threshold used to ensure constant false alarm rate, and D be the condition where no target echo signal exists. p Let J be the first unitary matrix, J be the inverse diagonal matrix, g be the estimated steering vector of the target, and G be the concatenated partial steering matrix. S is the first optimal azimuth angle. r Given the real covariance matrix, This is the conjugate transpose of the optimized first target actual guiding vector, i.e., the conjugate transpose of the target actual guiding vector after one-step optimization.

[0166] Similarly, the expression for the second robust detector, namely the robust detector that combines target detection and orientation estimation in the two-step method, is:

[0167]

[0168] in, This is the transpose of the second whitening covariance matrix. This is the optimized second partial derivative projection matrix, i.e., the partial derivative projection matrix optimized by the two-step method. Let be the second whitening covariance matrix, Δu be the cosine offset in the first direction, and Δv be the cosine offset in the second direction. This refers to the optimized second target actual guidance vector, i.e., the target actual guidance vector optimized by the two-step method. This is the transpose of the optimized second target's actual guiding vector. This is the second optimal azimuth angle, i.e., the azimuth angle optimized by the two-step method.

[0169] The two-step optimization method follows the same steps as the one-step method, the only difference being the different optimization detection methods.

[0170] S106. Detect the presence of a target using the Monte Carlo method, the first robust detector, and the second robust detector.

[0171] Specifically, based on the Monte Carlo method, the first robust detector, and the second robust detector, the presence of a target is detected, including:

[0172] Step C1: In the absence of a target echo signal, generate noise data using the Monte Carlo method.

[0173] exist That is, assuming the false alarm probability of the experiment is P, where no target exists. fa Therefore, the total number of Monte Carlo experiments is 100 / P. fa Next, using simulation software, a set of noise data was randomly generated for each Monte Carlo method experiment.

[0174] Step C2: Input the noise data into the first robust detector and the second robust detector to output multiple detection thresholds.

[0175] 100 / P fa After the Monte Carlo experiment, 100 / P was obtained. fa Each detection threshold.

[0176] Step C3: Select a preset detection threshold from multiple detection thresholds.

[0177] Specifically, it will eventually reach 100 / P fa The detection thresholds are sorted from smallest to largest, and the 100th detection threshold is selected as the detection threshold of the current detector.

[0178] Step C4: Given the presence of a target echo signal, generate noise data with the target steering vector using the Monte Carlo method.

[0179] exist This refers to the detection statistic of the detector calculated under targeted conditions, and then 100 / P is applied. fa In each Monte Carlo experiment, a set of noisy data with a target guidance vector is randomly generated.

[0180] Step C5: Input the noise data with the target guidance vector into the first robust detector and the second robust detector to output the detection value.

[0181] Step C6: Determine whether the detected value is greater than the preset detection threshold. If yes, the target is detected; otherwise, the target is not detected.

[0182] The detected value is compared with the corresponding preset detection threshold. If the detected value is greater than the preset detection threshold, the target is detected; otherwise, the target is not detected.

[0183] Figure 2 A schematic diagram illustrates a 3D image of the cosine similarity between the target guidance vector and the actual guidance vector, where Δu is the cosine offset in the first direction, Δv is the cosine offset in the second direction, Δu and Δv are the cosine offsets of the target guidance vector, and the z-axis in the 3D coordinate system represents the cosine similarity between the target guidance vector and the actual guidance vector. See also... Figure 2As shown, when Δu and Δv are less than a certain threshold (i.e., the 3dB width of the radar transmitted beam), the target steering vector has a high degree of similarity with the actual steering vector.

[0184] Figure 3 The diagram illustrates the probability curves for each detector when training data is sufficient and insufficient. Figure 3 (a) and Figure 3 (b) is a comparison curve of radar detection performance under different mismatch conditions with sufficient training data in a two-dimensional array radar. The amount of training data is selected as twice the dimension of the two-dimensional array system, i.e., K = 2NM = 32. Figure 3 (c) and Figure 3 (d) is a graph comparing the radar detection performance under different mismatch conditions with insufficient training data in a two-dimensional array radar. The amount of training data K is the system dimension K = NM = 16. The array elements are set to N = M = 4, where N is the number of row elements and M is the number of column elements, with constraint parameters α = β = 0.5. Figure 3 (a) and Figure 3 In (c), Δu = Δv = 0 represents the case without mismatch, where Δu is the cosine offset in the first direction and Δv is the cosine offset in the second direction. Figure 3 (b) and Figure 3 In (d), Δu = Δv = 0.3 represents the guide vector mismatch. The vertical coordinate P in the two-dimensional image... d The x-axis represents the detection probability value, and the y-axis represents the signal-to-interference-plus-noise ratio (SINR). The SINR is modeled as... Where a is the signal amplitude, g a For the target-oriented vector model, S is the true covariance matrix of the target.

[0185] Figure 3 (a) and Figure 3In (b), under the condition of no mismatch, the first robust detector proposed in this invention, namely the One-step Persymmetric Coordinate Detector (PerCD-1S), and the second robust detector, namely the Two-step Persymmetric Coordinate Detector (PerCD-2S), have the best detection probabilities. Their detection probabilities differ from those of other contrast detectors at the same signal-to-interference-plus-noise ratio (SINR) by a maximum of about 0.3. The detection probabilities of GLRT and the Adaptive Matched Filter (AMF) are next, while the detection performance of the Spatial Domain Detector (SD) and the Spatial Domain Adaptive Matched Filter (SD-AMF) is the lowest. Under the condition of mismatch, the detectors PerCD-1S and PerCD-2S proposed in this invention still have the best detection performance. The detection performance curves of the mismatch-free detectors GLRT and AMF begin to decline due to the mismatch of the steering vector, and are lower than those of the Spatial Domain Detectors SD and SD-AMF. Figure 3 (c) and Figure 3 In (d), regardless of whether there is a mismatch or not, the detection performance of the one-step joint detector PerCD-1S and the two-step joint detector PerCD-2S of this invention is optimal. Their detection probabilities differ from other contrast detectors by a maximum of approximately 0.9. Furthermore, compared to the detection probability when training data is sufficient, the detection probability of the detectors proposed in this invention at the same SINR with K=16 only decreases by a maximum of approximately 0.3. Because the amount of training data is reduced to half of its original value, the detection performance of the mismatch-free detectors GLRT and AMF, and the subspace detectors SD and SD-AMF, are all at a relatively low level. Therefore... Figure 3 This demonstrates that even with insufficient training data and mismatched steering vectors, the first and second robust detectors of this invention can still maintain good detection performance.

[0186] Figure 4 The MSE curve of the detector of the present invention is schematically shown, specifically the MSE curve of the first robust detector PerCD-1S under different training data, where the angle error is defined. Where R M The number of Monte Carlo experiments is given, and Δθ = [Δu, Δv] is the cosine offset of the true target. This is the estimated target cosine offset. Figure 4(a) It can be seen that, under the condition that the steering vector is not mismatched, the estimation performance curve of the detector proposed in this invention decreases faster as the number of training samples increases; Figure 4 (b) When the steering vector mismatch is shown, the MSE curve is higher than that under the no-mismatch condition at the same SINR, but the MSE gradually decreases as the amount of training data increases. Therefore, in summary... Figure 4 As can be seen from the MSE curve, the detector proposed in this invention has good estimation performance in orientation estimation, and the more training samples there are, the faster the convergence speed of its MSE curve.

[0187] Based on the above Figure 1 As can be seen from the implementation method, when the radar echo is received, the embodiment of the present invention constructs a target steering vector model in a fully incremental form based on the target's three-dimensional spatial position and the target's estimated azimuth position by the radar. Based on the target steering vector model, noise components in the test data, noise components in the training data, and signal amplitude, a binary detection model is constructed. The target's covariance matrix is ​​determined based on the noise components in the test data and the noise components in the training data. The binary detection model is used to determine the test data and training data. The joint probability density function of the test data matrix and the training data matrix is ​​determined based on the inverse diagonal matrix, the target's covariance matrix, the estimated covariance matrix, the signal amplitude matrix, the first unitary matrix, and the second unitary matrix. Based on the joint probability density function and the generalized likelihood ratio test criterion, a first preset detection formula and a second preset detection formula for the inverse diagonal parameters are determined. Both the first and second preset detection formulas are fractionally optimized to obtain the corresponding first and second final detection formulas. The first and second final detection formulas are solved using the cyclic coordinate method to construct a first robust detector and a second robust detector. The presence of a target is detected using the Monte Carlo method, the first robust detector, and the second robust detector. In this way, by constructing a target guidance vector model in a fully incremental form, a binary detection model based on the joint target detection and azimuth estimation is built. When a target is detected, the first and second final detection formulas are solved by using the cyclic coordinate method, and the target cosine offset can be adaptively output, thereby mitigating the sharp decline in the detection performance of the two-dimensional array radar when the target guidance vector is mismatched. The first and second preset detection formulas are derived using the inverse diagonal matrix, the first unitary matrix, and the second unitary matrix. The use of the inverse diagonal structure allows the covariance matrix of the unknown target to be characterized with half of the parameters, which greatly reduces the dependence on the amount of training data and makes the target detection sensitivity high even when the training data is reduced and the guidance vector is mismatched.

[0188] Based on the same inventive concept, as an implementation of the above-mentioned target detection method based on an inverse diagonal joint robust detector, this embodiment of the invention also provides a target detection device based on an inverse diagonal joint robust detector. Figure 5 This is a structural diagram of the device in an embodiment of the present invention. See also: Figure 5 As shown, the device may include:

[0189] The first construction module 501 is used to construct a target guidance vector model in a fully incremental form based on the target's three-dimensional spatial position and the target's estimated azimuth position by the radar when the radar echo is received.

[0190] The second construction module 502 is used to construct a binary detection model based on the target guidance vector model, the noise component in the test data, the noise component in the training data, and the signal amplitude, and to determine the covariance matrix of the target based on the noise component in the test data and the noise component in the training data. The binary detection model is used to determine the test data and the training data.

[0191] The determination module 503 is used to determine the joint probability density function of the test data matrix and the training data matrix based on the inverse diagonal matrix, the target covariance matrix, the estimated covariance matrix, the signal amplitude matrix, the first unitary matrix, and the second unitary matrix. The test data matrix is ​​determined by the test data, and the training data matrix is ​​determined by the training data.

[0192] The score optimization module 504 is used to determine the first and second preset detection formulas of the inverse diagonal parameters based on the joint probability density function and the generalized likelihood ratio test criterion, and to perform score optimization on both the first and second preset detection formulas to obtain the corresponding first and second final detection formulas.

[0193] The third construction module 505 is used to solve the first final detection formula and the second final detection formula according to the cyclic coordinate method, so as to construct the first robust detector and the second robust detector.

[0194] The detection module 506 is used to detect the presence of a target based on the Monte Carlo method, a first robust detector, and a second robust detector.

[0195] The first construction module 501 is specifically used to, upon receiving a radar echo, transform the three-dimensional spatial position of the target into cosine space to obtain the true target orientation position, which includes a first true target orientation position and a second true target orientation position; based on the first true target orientation position, the second true target orientation position, and the number of row array elements and column array elements in the uniform rectangular distribution array, a preset guidance vector of the target is formed; based on the first true target orientation position and the first estimated target azimuth position, a first direction cosine offset is determined, and based on the second true target orientation position and the second estimated target azimuth position, a second direction cosine offset is determined; based on the first estimated target azimuth position, the second estimated target azimuth position, the preset guidance vector of the target, the first direction cosine offset, and the second direction cosine offset, a target guidance vector model is constructed; the radar-estimated target azimuth position includes the first estimated target azimuth position and the second estimated target azimuth position.

[0196] The first construction module 501 constructs a target guidance vector model based on the first estimated target azimuth position, the second estimated target azimuth position, the target's preset guidance vector, the first direction cosine offset, and the second direction cosine offset. This includes: replacing the first true target direction position in the target's preset guidance vector with the first estimated target azimuth position, and replacing the second true target direction position with the second estimated target azimuth position, to obtain the target's estimated guidance vector; and using the first and second direction cosine offsets, performing a Taylor expansion on the first and second estimated target azimuth positions in the target's estimated guidance vector to obtain the target guidance vector model. The target's preset guidance vector includes the first true target direction position and the second true target direction position.

[0197] The first construction module 501 uses a first direction cosine offset and a second direction cosine offset to perform a Taylor expansion on the first estimated target azimuth position and the second estimated target azimuth position in the estimated target guidance vector to obtain a target guidance vector model. This includes: using the first direction cosine offset and the second direction cosine offset to perform a Taylor expansion on the first estimated target azimuth position and the second estimated target azimuth position in the estimated target guidance vector, and using the following formula to obtain the target guidance vector model:

[0198]

[0199]

[0200] Among them, g a (Δu,Δv) is the target-oriented vector model. Let Δu be the estimated guide vector for the target, Δv be the cosine offset in the first direction, and Δv be the cosine offset in the second direction. The first estimated target location. This is the second estimated target location. Let Z be a complex matrix space of dimension Z×1, where Z is the product of the number of row elements and the number of column elements, and Z = NM, where N is the number of row elements and M is the number of column elements. Take the partial derivative of the target-oriented vector model with respect to the first true target orientation position u. For the target-oriented vector model, take the partial derivative with respect to the second true target orientation position v, where j is the imaginary unit and λ0 is the radar transmitted wave wavelength. For each antenna element pair from Weighted response of the signal in the direction, x N-1 Let be the coordinates of the (N-1)th row element in the two-dimensional array. For each antenna element pair from The weighted response of the directional signal, ⊙ is the Kronecker product, y M-1 Let M be the coordinate position of the (M-1)th column element in the two-dimensional array. For Hadama accumulation.

[0201] The determination module 503 is specifically used to determine the true covariance matrix of the target based on the conjugate matrix corresponding to the inverse diagonal matrix and the covariance matrix of the target; determine the test data matrix based on the test data; determine the training data matrix based on the training data; determine the preset joint probability density function of the test data matrix and the training data matrix based on the true covariance matrix of the target, the estimated covariance matrix, the adjacency matrix corresponding to the test data matrix, and the adjacency matrix corresponding to the signal amplitude matrix; use the first unitary matrix and the second unitary matrix to perform unitary transformation on the adjacency matrix corresponding to the signal amplitude matrix, the adjacency matrix corresponding to the test data matrix, the estimated covariance matrix, and the covariance matrix of the target, respectively, to obtain the unitarily transformed real amplitude matrix, the unitarily transformed real test unit data matrix, the unitarily transformed real estimated covariance matrix, and the unitarily transformed real steering vector; substitute the unitarily transformed real amplitude matrix, the unitarily transformed real test unit data matrix, the unitarily transformed real estimated covariance matrix, and the unitarily transformed real steering vector into the preset joint probability density function to obtain the joint probability density function.

[0202] The first final detection expression in the score optimization module 504 is as follows:

[0203]

[0204] in, The actual whitening guide vector for the target. Here, Δθ represents the whitening bias guidance matrix, and Δθ represents the target cosine deviation angle information. The first whitening covariance matrix, This is the transpose of the first whitening covariance matrix.

[0205] D p Let J be the first unitary matrix, and g be the inverse diagonal matrix. * The conjugate of the guide vector for estimating the target. G is the real estimated covariance matrix after unitary transformation. * R is the conjugate of the concatenated partial orientation matrix. r Here, φ is the actual test unit data matrix after unitary transformation, φ is the variable, and I is the identity matrix. α is the transpose of the test unit data matrix after unitary transformation, β is the first constraint parameter, Δu is the first direction cosine offset, and Δv is the second direction cosine offset.

[0206] The second final detection formula is expressed as:

[0207]

[0208] in, The second whitening covariance matrix, S is the transpose of the second whitening covariance matrix. r Given the real covariance matrix, D p It is the first unitary matrix.

[0209] The third building block 505, the expression for the first robust detector is:

[0210]

[0211] in, R r This is the data matrix of the actual test unit after unitary transformation. This is the transpose of the test unit data matrix after unitary transformation. The real estimated covariance matrix after unitary transformation. The optimized first target actual guidance vector, For the first target amplitude, The condition for the existence of a target echo signal. Let ξ be the set detection threshold used to ensure constant false alarm rate, and D be the condition where no target echo signal exists. p Let J be the first unitary matrix, J be the inverse diagonal matrix, g be the estimated steering vector of the target, and G be the concatenated partial steering matrix. S is the first optimal azimuth angle. r Given the real covariance matrix, The conjugate transpose of the optimized first target's actual guiding vector;

[0212] The expression for the second robust detector is:

[0213]

[0214] in, This is the transpose of the second whitening covariance matrix. The optimized second partial derivative projection matrix, Let be the second whitening covariance matrix, Δu be the cosine offset in the first direction, and Δv be the cosine offset in the second direction. The optimized actual guiding vector for the second target. This is the transpose of the optimized second target's actual guiding vector. This is the second optimal azimuth angle.

[0215] The detection module 506 is specifically used to generate noise data using the Monte Carlo method when there is no target echo signal; input the noise data into a first robust detector and a second robust detector to output multiple detection thresholds; select a preset detection threshold from the multiple detection thresholds; generate noise data with a target steering vector using the Monte Carlo method when there is a target echo signal; input the noise data with the target steering vector into the first robust detector and the second robust detector to output a detection value; determine whether the detection value is greater than the preset detection threshold. If it is, a target is detected; otherwise, a target is detected.

[0216] It should be noted that the above description of the target detection device embodiment based on the inverse diagonal joint robust detector is similar to the description of the target detection method embodiment based on the inverse diagonal joint robust detector, and has similar beneficial effects. For any technical details not disclosed in the embodiments of the target detection device based on the inverse diagonal joint robust detector of this invention, please refer to the description of the target detection method embodiment based on the inverse diagonal joint robust detector of this invention for understanding.

[0217] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A target detection method based on an inverse diagonal joint robust detector, characterized in that, The target detection method includes: Upon receiving the radar echo, a target guidance vector model in a two-dimensional array is constructed using a fully incremental approach, based on the target's three-dimensional spatial position and the radar's estimated target azimuth position. Based on the target guidance vector model, the noise component in the test data, the noise component in the training data, and the signal amplitude, a binary detection model is constructed, and the covariance matrix of the target is determined based on the noise component in the test data and the noise component in the training data. The binary detection model is used to determine the test data and the training data. Based on the inverse diagonal matrix, the covariance matrix of the target, the estimated covariance matrix, the signal amplitude matrix, the first unitary matrix, and the second unitary matrix, the joint probability density function of the test data matrix and the training data matrix is ​​determined, wherein the test data matrix is ​​determined by the test data, and the training data matrix is ​​determined by the training data. Based on the joint probability density function and the generalized likelihood ratio test criterion, the first and second preset detection formulas for the inverse diagonal parameters are determined, and the first and second preset detection formulas are both optimized by scores to obtain the corresponding first and second final detection formulas. The first and second final detection formulas are solved using the cyclic coordinate method to construct the first and second robust detectors. The presence of the target is detected using the Monte Carlo method, the first robust detector, and the second robust detector. The step of solving the first final detection formula and the second final detection formula according to the cyclic coordinate method to construct the first robust detector and the second robust detector includes: Step C1: First initialize the number of iterations. Initialize the optimal azimuth angle Initialize the bias guide vector ,in, , Representing the The azimuth angle after the next iteration For the first The value of the first cosine offset in the next iteration. For the first The value of the second cosine offset in the next iteration. The actual whitening guide vector for the target. It is the first unitary matrix. The target-oriented vector model is used to determine the second true target orientation and position. Find the conjugate of the partial derivatives. For only cosine components The actual guiding vector of the target direction; Step C2: After initialization, make Representing the The next iteration, assuming As an invariant, calculate the cosine offset in the first direction. The optimal value is: ； in, , Cosine offset in the first direction The optimal value, To return the parameter containing the maximum value of the right-hand side. , For the first The bias after the second iteration Guide vector, The first whitening covariance matrix, For only cosine components The actual guiding vector of the target direction. The target-oriented vector model is used to determine the first true target orientation and position. Find the conjugate of the partial derivatives; Step C3: Calculate the first... The bias after the second iteration Guide vector ,assumed As an invariant, calculate the cosine offset in the second direction. optimal value ; Step C4: Calculate the first... The bias after the second iteration Guide vector Update azimuth Update the one-step detection formula for the first step. The value of the second test; Step C5: If Then the first optimal azimuth angle is obtained. ;like If so, return to step C2 and continue iterating; Step C6: Optimize the optimal azimuth angle obtained in the one-step method using the cyclic coordinate method. The first and second optimal azimuth angles are used to construct the first and second robust detectors.

2. The target detection method according to claim 1, characterized in that, The radar-estimated target azimuth position includes a first estimated target azimuth position and a second estimated target azimuth position. The step of constructing a target guidance vector model using a full-incremental form based on the target's three-dimensional spatial position and the radar-estimated target azimuth position includes: Upon receiving the radar echo, the three-dimensional spatial position of the target is transformed into cosine space to obtain the true target direction position, which includes a first true target direction position and a second true target direction position. Based on the first true target direction position, the second true target direction position, and the number of row array elements and column array elements in the uniform rectangular distribution array, a preset guidance vector for the target is formed; Based on the first true target direction position and the first estimated target orientation position, a first direction cosine offset is determined, and based on the second true target direction position and the second estimated target orientation position, a second direction cosine offset is determined. The target guidance vector model is constructed based on the first estimated target azimuth position, the second estimated target azimuth position, the target's preset guidance vector, the first direction cosine offset, and the second direction cosine offset.

3. The target detection method according to claim 2, characterized in that, The preset guidance vector of the target includes the first true target direction position and the second true target direction position. The step of constructing the target guidance vector model based on the first estimated target azimuth position, the second estimated target azimuth position, the preset guidance vector of the target, the first direction cosine offset, and the second direction cosine offset includes: The first true target direction position in the preset guide vector of the target is replaced with the first estimated target orientation position, and the second true target direction position is replaced with the second estimated target orientation position to obtain the estimated guide vector of the target; Using the first direction cosine offset and the second direction cosine offset, Taylor expansion is performed on the first estimated target azimuth position and the second estimated target azimuth position in the estimated steering vector of the target to obtain the target steering vector model.

4. The target detection method according to claim 3, characterized in that, The step of performing Taylor expansion on the first estimated target azimuth position and the second estimated target azimuth position in the estimated target steering vector using the first direction cosine offset and the second direction cosine offset to obtain the target steering vector model includes: Using the first direction cosine offset and the second direction cosine offset, a Taylor expansion is performed on the first estimated target azimuth position and the second estimated target azimuth position in the estimated target steering vector, and the target steering vector model is obtained using the following formula: ； ； in, The target-oriented vector model, The estimated guide vector for the target. This is the cosine offset in the first direction. This is the cosine offset in the second direction. The first estimated target location. This is the second estimated target location. for Complex matrix space of dimension 1 It is the product of the number of row elements and the number of column elements. , The number of elements in the row array, The number of elements in the array is [number]. The target guidance vector model is positioned relative to the first real target direction. Find the partial derivative. The target guidance vector model is used to determine the orientation position of the second real target. Find the partial derivative. The imaginary unit, The wavelength of the radar transmitted wave. For each antenna element pair from The weighted response of directional signals. For the second dimensional array The coordinates of each row array element. For each antenna element pair from The weighted response of directional signals, For Kronecker product, For the second dimensional array The coordinates of each array element. For Hadama accumulation.

5. The target detection method according to claim 1, characterized in that, The step of determining the joint probability density function of the test data matrix and the training data matrix based on the inverse diagonal matrix, the covariance matrix of the target, the estimated covariance matrix, the signal amplitude matrix, the first unitary matrix, and the second unitary matrix includes: The true covariance matrix of the target is determined based on the conjugate matrix corresponding to the inverse diagonal matrix and the covariance matrix of the target. The test data matrix is ​​determined based on the test data, and the training data matrix is ​​determined based on the training data. Based on the target true covariance matrix, the estimated covariance matrix, the adjacency matrix corresponding to the test data matrix, and the adjacency matrix corresponding to the signal amplitude matrix, a preset joint probability density function of the test data matrix and the training data matrix is ​​determined; Using the first unitary matrix and the second unitary matrix, unitary transformations are performed on the adjacency matrix corresponding to the signal amplitude matrix, the adjacency matrix corresponding to the test data matrix, the estimated covariance matrix, and the covariance matrix of the target, respectively, to obtain the unitarily transformed real amplitude matrix, the unitarily transformed real test unit data matrix, the unitarily transformed real estimated covariance matrix, and the unitarily transformed real steering vector. Substituting the real amplitude matrix after unitary transformation, the real test unit data matrix after unitary transformation, the real estimated covariance matrix after unitary transformation, and the real steering vector after unitary transformation into the preset joint probability density function yields the joint probability density function.

6. The target detection method according to claim 5, characterized in that, The step of substituting the unitarily transformed real amplitude matrix, the unitarily transformed real test unit data matrix, the unitarily transformed real estimation covariance matrix, and the unitarily transformed real steering vector into the preset joint probability density function to obtain the joint probability density function includes: Substituting the unitarily transformed real amplitude matrix, the unitarily transformed real test unit data matrix, the unitarily transformed real estimation covariance matrix, and the unitarily transformed real steering vector into the preset joint probability density function, the joint probability density function is obtained using the following formula: ； ； ； in, Let be the joint probability density function. The test data matrix, The training data matrix, The condition is that there is no target echo signal. For the condition that there is a target echo signal, Given the real covariance matrix, Given the information matrix of the actual test data, It is the product of the number of row elements and the number of column elements. , The number of elements in the row array, The number of elements in the array is [number]. For the amount of training data, Let be the first unitary matrix. Let be the true covariance matrix of the target. This is the conjugate transpose of the matrix. In order to seek the truth, Let be the inverse diagonal matrix. To find the imaginary part, Let be the real estimated covariance matrix after the unitary transformation. This is the actual test unit data matrix after the unitary transformation. The real steering vector after the unitary transformation. This is the transpose of the real amplitude matrix after the unitary transformation.

7. The target detection method according to claim 6, characterized in that, The first final detection formula is expressed as: ； in, , , , The actual whitening guide vector for the target. For whitening biased guidance matrix, For the target cosine deviation angle information, The first whitening covariance matrix, Let be the transpose of the first whitening covariance matrix. , , , , Let be the first unitary matrix. Let be the inverse diagonal matrix. The conjugate of the guide vector for estimating the target. Let be the real estimated covariance matrix after the unitary transformation. The conjugate of the spliced ​​partial orientation matrix, This is the actual test unit data matrix after the unitary transformation. As variables, It is the identity matrix. This is the transpose of the unitary transformed test unit data matrix. The first constraint parameter, This is the second constraint parameter. For the first direction cosine offset, This is the cosine offset in the second direction; The second final detection formula is expressed as: ； in, , , , The second whitening covariance matrix, This is the transpose of the second whitening covariance matrix. Let be the known real covariance matrix. Let be the first unitary matrix.

8. The target detection method according to claim 7, characterized in that, The expression for the first robust detector is: ； in, , , This is the actual test unit data matrix after the unitary transformation. This is the transpose of the unitary transformed test unit data matrix. Let be the real estimated covariance matrix after the unitary transformation. The optimized first target actual guidance vector, For the first target amplitude, The condition for the existence of a target echo signal. The condition is that there is no target echo signal. To ensure a constant false alarm rate, a detection threshold is set. Let be the first unitary matrix. Let be the inverse diagonal matrix. The guide vector for estimating the target. This is the spliced ​​biased orientation matrix. The first optimal azimuth angle. Let be the known real covariance matrix. The conjugate transpose of the optimized first target's actual guiding vector; The expression for the second robust detector is: ； in, , , This is the transpose of the second whitening covariance matrix. The optimized second partial derivative projection matrix, The second whitening covariance matrix is... This is the cosine offset in the first direction. This is the cosine offset in the second direction. The optimized actual guiding vector for the second target. This is the transpose of the optimized second target's actual guiding vector. This is the second optimal azimuth angle.

9. The target detection method according to claim 1, characterized in that, The step of detecting the presence of the target based on the Monte Carlo method, the first robust detector, and the second robust detector includes: Noise data is generated using the Monte Carlo method when there is no target echo signal. The noise data is input into the first robust detector and the second robust detector to output multiple detection thresholds; Select a preset detection threshold from the plurality of detection thresholds; Given the presence of the target echo signal, noise data with the target steering vector is generated using the Monte Carlo method. The noise data with the target guidance vector is input into the first robust detector and the second robust detector to output the detection value; Determine whether the detected value is greater than the preset detection threshold. If yes, the target is detected to exist; otherwise, the target is detected to not exist.

10. A target detection device based on an inverse diagonal joint robust detector, characterized in that, The device includes: The first construction module is used to construct a target guidance vector model in a two-dimensional array in a fully incremental form based on the target's three-dimensional spatial position and the target's estimated azimuth position by the radar when the radar echo is received. The second construction module is used to construct a binary detection model based on the target guidance vector model, the noise component in the test data, the noise component in the training data, and the signal amplitude, and to determine the covariance matrix of the target based on the noise component in the test data and the noise component in the training data. The binary detection model is used to determine the test data and the training data. The determination module is used to determine the joint probability density function of the test data matrix and the training data matrix based on the inverse diagonal matrix, the covariance matrix of the target, the estimated covariance matrix, the signal amplitude matrix, the first unitary matrix, and the second unitary matrix, wherein the test data matrix is ​​determined by the test data, and the training data matrix is ​​determined by the training data. The score optimization module is used to determine the first preset detection formula and the second preset detection formula of the inverse diagonal parameter according to the joint probability density function and the generalized likelihood ratio test criterion, and to perform score optimization on both the first preset detection formula and the second preset detection formula to obtain the corresponding first final detection formula and the second final detection formula. The third construction module is used to solve the first final detection formula and the second final detection formula according to the cyclic coordinate method, so as to construct the first robust detector and the second robust detector. The detection module is used to detect the presence of the target based on the Monte Carlo method, the first robust detector, and the second robust detector; The third construction module is specifically used to initialize the number of iterations first. Initialize the optimal azimuth angle Initialize the bias guide vector ,in, , Representing the The azimuth angle after the next iteration For the first The value of the first cosine offset in the next iteration. For the first The value of the second cosine offset in the next iteration. The actual whitening guide vector for the target. It is the first unitary matrix. The target-oriented vector model is used to determine the second true target orientation and position. Find the conjugate of the partial derivatives. For only cosine components The target's actual guidance vector; after initialization, make... Representing the The next iteration, assuming As an invariant, calculate the cosine offset in the first direction. The optimal value is: ； in, , Cosine offset in the first direction The optimal value, To return the parameter containing the maximum value of the right-hand side. , For the first The bias after the second iteration Guide vector, The first whitening covariance matrix, For only cosine components The actual guiding vector of the target direction. The target-oriented vector model is used to determine the first true target orientation and position. Find the conjugate of the partial derivatives; Calculate the first The bias after the second iteration Guide vector ,assumed As an invariant, calculate the cosine offset in the second direction. optimal value ; Calculate the first The bias after the second iteration Guide vector Update azimuth Update the one-step detection formula for the first step. The second detection value; if Then the first optimal azimuth angle is obtained. ;like Then return after initialization is complete. Representing the The operation of the next iteration; the optimal azimuth angle in the one-step method is optimized by the cyclic coordinate method. The first and second optimal azimuth angles are used to construct the first and second robust detectors.

Images