Target knowledge-assisted intelligent fusion detection method under non-uniform clutter

Abstract

The present invention discloses a target knowledge-assisted intelligent fusion detection method under non-uniform clutter, which belongs to the field of broadband radar signal processing. Aiming at the problem of poor radar range-extended target detection performance under small sample conditions, the skew-symmetric structure prior information of the clutter covariance matrix is ​​fully utilized, and the main data and auxiliary data are jointly used to improve the estimation accuracy of the unknown clutter skew-symmetric covariance matrix structure, and reduce the demand for auxiliary data volume; a subspace skew-symmetric generalized likelihood ratio test detector under non-uniform clutter is constructed, and the detector has a closed form expression, and the detection performance is better than the existing unstructured range-extended target subspace detector and skew-symmetric range-extended target subspace detector while ensuring the constant false alarm rate characteristics; the detection statistic structure of the method of the present invention is simple, easy to implement in engineering, and improves the detection performance of broadband radar for weak targets under small sample conditions, which has the value of popularization and application.

Description

1. Technical Field

[0001] The invention belongs to the field of radar signal processing, and specifically relates to a target knowledge-assisted intelligent fusion detection method under non-uniform clutter. 2. Background Technology

[0002] With the continuous increase of radar bandwidth, the range resolution of broadband radar is constantly improving. It has obvious advantages in anti-interference, anti-reconnaissance, short-range detection, and improving signal-to-noise ratio. It has attracted widespread attention in modern military and civilian fields and has become an important branch of radar technology development. Conventional narrowband radars generally use wide pulses, and the range resolution is very low. The scattered echo signal of the target only occupies one range resolution unit, presenting as a "point" target. However, broadband radar uses pulse compression or frequency agility technology, which makes the radar transmission signal have a large time-bandwidth product, thereby obtaining a higher range resolution. The range resolution of broadband radar can reach sub-meter level. The echo of a general target is distributed in different radial distance units, presenting as a "one-dimensional range image", forming a range-extended target. With the widespread application of broadband radar, the problem of range-extended target detection is causing extensive discussion and research, and has become one of the hot and difficult issues in the field of radar signal processing in recent years.

[0003] Adaptive detection of extended range targets is mainly achieved with the help of auxiliary data. Auxiliary data are generally taken from reference range cells that are spatially adjacent to the range cell to be detected, and are assumed to contain no target signals, but only clutter components that are independent and identically distributed with the main data clutter components of the range cell to be detected. Sufficient auxiliary data can be used to accurately estimate the unknown clutter covariance matrix. However, for the actual radar, the uniformity of the clutter background is destroyed due to the drastic changes in clutter power, discrete clutter, clutter edges and other outliers. It is sometimes difficult to obtain auxiliary data that meets the global uniformity, which seriously affects the adaptive detection performance of extended range targets. In fact, although the global uniformity of the complex clutter background is destroyed, the local uniformity of the clutter is still reflected within a certain radial distance range. At this time, the partially uniform model can be used to model the clutter, that is, the clutter components in the range cell to be detected and the reference range cell have the same covariance matrix structure and different power levels. This model can make full use of the local uniformity of the clutter, but the number of reference range cells that can be used is limited by the actual degree of clutter non-uniformity.

[0004] In addition, in the commonly used rank-one signal target detection model, the target's steering vector is usually assumed to be a known fixed vector, but in practical applications, due to the existence of beam pointing errors and multipath phenomena, the target's steering vector may be mismatched. In order to deal with this problem, it is possible to consider using a subspace model to model the target signal. In the subspace model, the signal is represented as the product of a known subspace matrix and an unknown coordinate matrix. Under the range-extended target subspace model, the Gradient test, generalized likelihood ratio test (GLRT), Rao test, and Wald test can be used to obtain subspace range-extended target detectors based on Gradient, GLRT, Rao, Wald, and two-step methods (abbreviated as S-Gradient, S-GLRT, S-Rao, S-Wald, and S-2SD, respectively). In an actual non-uniform clutter environment, a small sample situation with limited auxiliary data may occur. In a small sample environment, the detection performance of the original detector will be greatly reduced, and it is difficult to achieve an ideal detection effect. It is difficult to obtain enough pure clutter auxiliary data in practical environments. For radar receivers using centrosymmetric linear arrays or centrosymmetric interval pulse trains, their clutter covariance matrices have a special skew-symmetric structure. Based on the Rao and Wald test criteria and considering the skew-symmetric prior information of the clutter covariance matrix, the subspace skew-symmetric Rao and Wald detectors (abbreviated as Per-Rao and Per-Wald, respectively) for range-extended targets can be obtained. Compared with the S-Rao and S-Wald detectors, the performance of the structured detector with skew-symmetric information is improved to a certain extent, but the detection criteria in small sample environments still need to be further optimized.

[0005] At present, most range-extended target detectors are designed mainly for the case where there is sufficient auxiliary data. Their detection performance is degraded in the case of small samples with limited auxiliary data. In view of the poor performance of radar range-extended target detection under small sample conditions, how to make full use of the skew-symmetric structure information, reduce the actual demand for auxiliary data, improve the estimation accuracy of the unknown clutter covariance matrix, and construct a closed-form intelligent range-extended target detector through the optimal inspection criterion, while ensuring the constant false alarm rate (CFAR) characteristics of the detector, and reducing the demand for auxiliary data, is the key to improving the detection capability of broadband radar targets under non-uniform small sample backgrounds, and is also one of the problems that need to be solved urgently. III. Summary of the invention

[0006] 1. Technical problems to be solved

[0007] In the wideband radar detection scenario, in order to solve the problem of range-extended target detection performance degradation in non-uniform small sample background, how to make full use of skew-symmetric structure information, reduce the actual demand for auxiliary data, improve the estimation accuracy of unknown clutter covariance matrix, and construct a closed-form intelligent detector for range-extended targets by optimizing the inspection criteria. While ensuring the constant false alarm rate (CFAR) characteristics of the detector, the demand for auxiliary data is reduced, and the adaptability of wideband radar to non-uniform clutter environment is further improved. The target detection performance of wideband radar in partially uniform environment is improved, especially the detection capability of wideband radar for weak targets in small sample environment is improved.

[0008] 2. Technical solution

[0009] The target knowledge-assisted intelligent fusion detection method under non-uniform clutter described in the present invention includes the following technical measures:

[0010] Step 1: Obtain primary data Z from K distance units to be detected, and obtain auxiliary data Z from R distance units adjacent to the distance unit to be detected. R ; Under the target hypothesis and the non-target hypothesis, use Z and Z respectively R The joint probability density function of is derived from the skew-symmetric covariance matrix structure M of the clutter and set to zero, and the maximum likelihood estimate of the unknown skew-symmetric covariance matrix structure of the clutter under the target assumption is obtained. At the same time, the maximum likelihood estimate of the unknown skew-symmetric covariance matrix structure of the clutter under the no-target assumption is obtained. Using Z and Z under the target assumption, R The joint probability density function of is derived from the target transformation coordinate matrix and set to zero to obtain the maximum likelihood estimate of the unknown target transformation coordinate matrix; then the subspace skew-symmetric generalized likelihood ratio test intermediate statistic is constructed under the condition of known clutter power factor; the specific steps include:

[0011] For a coherent radar system with N space-time joint channels, assuming that the target may occupy K consecutive detection range units, the main data corresponding to its echo signal can be expressed as an N×K dimensional complex matrix Z=[z1,z2,…,z K ],z k represents the N×1 dimensional main data component corresponding to the kth range unit to be detected. Under the assumption H0 that there is no target, Z only contains the N×K dimensional clutter component complex matrix C=[c1,c2,…,c K ], where the N×1 dimensional complex vector c k (k=1,2,...,K) represents the clutter component in the kth distance unit to be detected, which obeys the zero mean covariance matrix and is an N×N dimensional complex matrix M tThe complex circular Gaussian distribution of , and the clutter vectors between different distance units are independent and identically distributed. Under the assumption H1 that there is a target, Z is composed of an N×K-dimensional signal component matrix S and a clutter component matrix C; where the signal component matrix S can be expressed as the product of a known N×r-dimensional multi-rank subspace complex matrix U and an r×K-dimensional unknown complex coordinate matrix B, and B=[b1,b2,…,b K ], b k represents the complex coordinate vector of the r×1-dimensional target subspace in the kth distance unit to be detected, and r represents the rank of the matrix U. In order to estimate the unknown M t , obtain R observation data from R pure clutter reference distance units adjacent to the distance unit to be detected, and the auxiliary data Z corresponding to the echo signal R It can be expressed as an N×R dimensional complex matrix Z R =C R =[c K+1 ,c K+2 ,…,c K+R ], c K+k represents the auxiliary data component corresponding to the K+kth reference distance unit. t (t=K+1,K+2,…,K+R) obeys the complex circular Gaussian distribution with zero mean covariance matrix as N×N dimensional complex matrix M, and the auxiliary data components between different distance units are independent and identically distributed. In a partially uniform clutter environment, there are M t =γM, γ represents the unknown clutter power factor, and M is called the clutter skew-symmetric covariance matrix structure. When the radar receiver uses a centrally symmetric linear array or a centrally symmetric interval pulse train, the clutter covariance matrix M and the signal multi-rank subspace matrix U have a skew-symmetric structure, that is, M=JM * J,U=JU * .in,(·) * represents conjugation, and J represents an N×N dimensional permutation matrix with diagonal elements set to 1 and other elements set to 0. The introduction of the skew-symmetric structure can further improve the estimation accuracy of the skew-symmetric covariance matrix structure of unknown clutter, thereby reducing the demand for auxiliary data and providing favorable conditions for adaptive detection of range-extended targets under small sample conditions.

[0012] Model the range-extended object detection as a binary hypothesis testing problem:

[0013]

[0014] The generalized likelihood ratio test criterion can be expressed as

[0015]

[0016] Where T is the detection threshold, f1(Z,Z R |M,B,γ ) and f0(Z,Z R |M ,γ ) represent Z and Z under assumptions H1 and H0 respectively. R The joint probability density function of .

[0017] Using the skew-symmetric information in the clutter skew-symmetric covariance matrix structure M, the main data Z and auxiliary data Z under the target hypothesis H1 are R The joint probability density function f1(Z,Z R |M,B,γ) can be expressed as

[0018]

[0019] in,

[0020]

[0021] Where j is the imaginary unit, |·| represents the determinant of the square matrix, tr(·) represents the trace of the matrix, Re(·) and Im(·) represent the real and imaginary parts, respectively. and denote the set of real matrices and the set of complex matrices of m×n dimensions, respectively, (·) H represents the conjugate transpose.

[0022] Using the skew-symmetric information in the clutter skew-symmetric covariance matrix structure M, the main data Z and auxiliary data Z under the target-free hypothesis H0 are R The joint probability density function f0(Z,Z R |M,γ) can be expressed as

[0023]

[0024] in,

[0025] Under the target hypothesis H1, using Z and Z R The joint probability density function of is differentiated with respect to the skew-symmetric covariance matrix structure of the clutter and set to zero, that is, Get the maximum likelihood estimate of M under the H1 hypothesis for

[0026]

[0027] The above formula jointly utilizes the main data and auxiliary data to improve the estimation accuracy of the unknown clutter skew-symmetric covariance matrix structure and reduce the demand for the amount of auxiliary data.

[0028] Under the non-target hypothesis H0, using Z and Z R The joint probability density function of is differentiated with respect to the skew-symmetric covariance matrix structure of the clutter and set to zero, that is, Get the maximum likelihood estimate of M under the H0 hypothesis for

[0029]

[0030] The above formula jointly utilizes the main data and auxiliary data to improve the estimation accuracy of the unknown clutter skew-symmetric covariance matrix structure and reduce the demand for the amount of auxiliary data.

[0031] Substituting equations (5) and (6) into equation (2), the detection statistic λ can be equivalently expressed as

[0032]

[0033] Using the targeted hypothesis H1, Z and Z R The joint probability density function of the target transformation coordinate matrix B p Take the derivative and set it to zero, that is, Under the H1 hypothesis, we get p The maximum likelihood estimate of is

[0034]

[0035] Substituting equation (9) into equation (8), we can obtain the intermediate statistic of the subspace skew-symmetric generalized likelihood ratio test under non-uniform clutter:

[0036]

[0037] In the formula, I 2K Represents the 2K×2K-dimensional identity matrix.

[0038] Step 2 obtains the maximum likelihood estimate of the unknown clutter power factor under the target hypothesis and the no-target hypothesis by solving the unique positive solution of the eigenvalue equation under the target hypothesis and the no-target hypothesis, and substitutes the maximum likelihood estimate of the unknown clutter power factor into the subspace skew-symmetric generalized likelihood ratio test intermediate statistic obtained in step 1, replaces the unknown clutter power factor under the target hypothesis and the no-target hypothesis in the intermediate statistic, and constructs the detection statistic λ of the target knowledge-assisted intelligent fusion detector under non-uniform clutter; the specific steps include:

[0039] The eigenvalue equation (11) is constructed for the target hypothesis H1 and the non-target hypothesis H0. By solving the unique positive solution of equation (11), the maximum likelihood estimate of the unknown clutter power factor γ under the target hypothesis can be obtained. At the same time, the maximum likelihood estimation of the unknown clutter power factor γ under the assumption of no target is obtained

[0040] Under the assumption that H i(i=0,1), the maximum likelihood estimate of γ is the only positive solution that satisfies the eigenvalue equation of formula (11):

[0041]

[0042] in, λ k,0 and λ k,1 Respectively represent matrices and The kth eigenvalue of .

[0043] The maximum likelihood estimation of the unknown clutter power factor under assumptions H1 and H0 is and Substituting into equation (10), we can replace the unknown clutter power factor under the target assumption and the target-free assumption, that is, and Replacing γ in the denominator and numerator of formula (10) respectively, after simplified calculation, the detection statistic λ of the target knowledge-assisted intelligent fusion detector under non-uniform clutter can be expressed as

[0044]

[0045] The detector of the method disclosed in this invention is also called the subspace skew-symmetric generalized likelihood ratio test detector under non-uniform clutter (abbreviated as Per-GLRT), which has a closed form expression, does not require iterative operations, and has CFAR characteristics for both M and γ.

[0046] Step 3 is to maintain the CFAR characteristics of the detection method, and set the detection threshold T according to the preset false alarm probability; compare the detection statistic λ with the detection threshold T, if λ≥T, it is determined that there is a target in the current distance unit to be detected, and the main data is not used as auxiliary data for other subsequent distance units to be detected; on the contrary, if λ<T, it is determined that there is no target in the current distance unit to be detected, and the main data is used as auxiliary data for other subsequent distance units to be detected.

[0047] 3. Beneficial effects

[0048] Compared with the background technology, the present invention has the following beneficial effects:

[0049] 1) Make full use of the skew-symmetric structure prior information of the clutter covariance matrix, and jointly use the main data and auxiliary data to improve the estimation accuracy of the unknown clutter skew-symmetric covariance matrix structure, reduce the demand for auxiliary data, and provide favorable support for the adaptive detection of range-extended targets under small sample conditions in non-uniform clutter environments;

[0050] 2) A subspace skew-symmetric generalized likelihood ratio test detector under non-uniform clutter is constructed, and the detector has a closed form expression, and its detection performance is better than the existing unstructured range-extended target subspace detector and skew-symmetric range-extended target subspace detector; the detection statistic structure of the method of the present invention is simple and easy to implement in engineering;

[0051] 3) A subspace distance extended target signal model was constructed to avoid the difficulty of the rank-one signal model in dealing with the target guidance vector mismatch, and improve the robustness of the broadband radar to the target guidance vector mismatch;

[0052] 4) The detector of the method of the present invention can be applied to non-uniform environments such as partially uniform environments. By fully exploring the local uniformity of clutter and using the maximum likelihood estimation of the clutter power factor under the assumptions of target and no target, the intelligent adaptability of the detector to the non-uniform clutter environment is improved;

[0053] 5) The method of the present invention is applicable to some non-wideband radar detection situations, for example, using low / medium resolution radar to detect large targets or to detect spatially adjacent point target groups moving at the same speed (such as ship formations, aircraft formations, vehicle formations, etc.), and has good application prospects. IV. Description of the drawings

[0054] Figure 1 It is a functional module diagram of the target knowledge-assisted intelligent fusion detection method under non-uniform clutter of the present invention;

[0055] Figure 2 is a comparison chart of the detection performance of the method of the present invention and the existing unstructured range-extended target subspace detector;

[0056] Figure 3 It is a comparison diagram of detection performance between the method of the present invention and the existing oblique symmetric distance extended target subspace detector;

[0057] Figure 2 Where N = 12, K = 15, R = 24, r = 3, false alarm probability P fa =10 -3 ;

[0058] Figure 3 Where N = 12, K = 15, R = 12, r = 3, false alarm probability P fa =10 -3 ; V. Specific implementation methods

[0059] The present invention is further described below in conjunction with the accompanying drawings. The embodiments of the present invention are used to explain the present invention rather than to limit the present invention. Any modifications and changes made to the present invention within the spirit of the present invention and the protection scope of the claims fall within the protection scope of the present invention.

[0060] In order to verify the effectiveness of the method described in the present invention, this specific implementation mode provides two embodiments, the first embodiment is aimed at the sea detection environment, and the second embodiment is aimed at the ground detection environment.

[0061] Embodiment 1:

[0062] Refer to the instruction manual Figure 1 The specific implementation of Example 1 is divided into the following steps:

[0063] Step A1: Use sea detection radar to illuminate the sea area to be detected, and obtain the main data Z of K distance units to be detected; use radar to illuminate the target-free range around the sea area to be detected, and obtain the auxiliary data Z of R reference distance units containing only pure sea clutter. R . The main data Z and auxiliary data Z R Sent to the decorrelation processing module, in which the transformed main data Z is obtained according to formula (4) p and the estimation of the clutter skew-symmetric covariance matrix structure based on auxiliary data On the one hand, Z p and Sent to the joint probability density function module of the main data and auxiliary data under the H0 hypothesis, according to Z p and The joint probability density function of the main data and the auxiliary data under the H0 assumption is obtained (5); the joint probability density function of the main data and the auxiliary data under the H0 assumption is sent to the maximum likelihood estimation module of the clutter skew-symmetric covariance matrix structure under the H0 assumption, and the joint probability density function of the main data and the auxiliary data under the targetless assumption is used to derive the clutter skew-symmetric covariance matrix structure M and set it to zero. According to formula (7), the maximum likelihood estimation of M under the H0 assumption is obtained: On the other hand, Z p and Sent to the joint probability density function module of the main data and auxiliary data under the H1 hypothesis, according to Z p and The joint probability density function of the main data and the auxiliary data under the H1 assumption is obtained as formula (3); the joint probability density function of the main data and the auxiliary data under the H1 assumption is sent to the maximum likelihood estimation module of the clutter skew-symmetric covariance matrix structure under the H1 assumption, and the joint probability density function of the main data and the auxiliary data under the target assumption is used to derive the clutter skew-symmetric covariance matrix structure M and set it to zero. According to formula (6), the maximum likelihood estimation of M under the H1 assumption is obtained: Will The coordinate matrix is ​​sent to the maximum likelihood estimation module under the H1 hypothesis, and Z and Z are used under the target hypothesis. RThe joint probability density function of is derived from the target transformation coordinate matrix and set to zero. According to formula (9), the unknown target transformation coordinate matrix B is obtained: p The maximum likelihood estimate of

[0064] It is worth noting that in step A1, the method of the present invention constructs a subspace range extended target signal model, avoids the problem that the rank-one signal model is difficult to cope with the target guidance vector mismatch, and improves the robustness of the broadband radar to the maritime target guidance vector mismatch; in addition, the method of the present invention makes full use of the skew-symmetric structure prior information of the sea clutter covariance matrix, and jointly uses the main data and auxiliary data to improve the estimation accuracy of the unknown sea clutter skew-symmetric covariance matrix structure, reduces the demand for the amount of auxiliary data, and provides favorable support for realizing the adaptive detection of range extended targets under small sample conditions in a non-uniform sea clutter environment; at the same time, the detector of the method of the present invention can be applied to non-uniform sea clutter environments such as partially uniform, and by fully exploring the local uniformity of sea clutter, the maximum likelihood estimation of the clutter power factor under the two assumptions of target and no target is used respectively, thereby improving the intelligent adaptability of the detector to the non-uniform sea clutter environment.

[0065] Step A2 obtains the maximum likelihood estimation of the unknown clutter power factor under the target assumption and the non-target assumption by solving the unique positive solution of the eigenvalue equation, which specifically includes two aspects. It is sent to the maximum likelihood estimation module of the power coefficient under the H0 assumption, and the maximum likelihood estimation of γ under the H0 assumption is obtained according to formula (11): On the other hand, and It is sent to the maximum likelihood estimation module of the power coefficient under the H1 assumption, and the maximum likelihood estimation of γ under the H1 assumption is obtained according to formula (11): Substitute the maximum likelihood estimate of the unknown clutter power factor into the intermediate statistic of the subspace skew-symmetric generalized likelihood ratio test obtained in step A1, and replace the unknown clutter power factor under the target hypothesis and the no-target hypothesis in the intermediate statistic respectively. and It is sent to the target adaptive intelligent detector construction module under non-uniform clutter, and the detection statistic λ of the target knowledge-assisted intelligent fusion detector under non-uniform clutter is obtained according to formula (12), and λ is sent to the detection decision module.

[0066] It is noteworthy that the method of the present invention constructs a subspace skew-symmetric generalized likelihood ratio test detector under non-uniform clutter, and the detector has a closed-form expression, a simple detection statistic structure, and is easy to implement in engineering.

[0067] Step A3 sets the detection threshold T according to the preset false alarm probability: Specifically, the false alarm probability is set to P faAccording to the Monte Carlo method, based on the 100 / P accumulated in the previous period fa The detection threshold T is calculated based on the measured sea clutter data. Considering the difficulty of acquiring sea clutter, if the actual amount of pure sea clutter measured data R is less than 100 / Pfa, the missing 100 / Pfa-R clutter data can be simulated using the sea clutter simulation model, in which the model parameters are reasonably estimated and set based on the obtained pure sea clutter measured data. Further, the detection statistic λ is compared with the detection threshold T. If λ≥T, it is determined that there is a target in the current K distance units to be detected, and the main data is not used as auxiliary data for other subsequent distance units to be detected; on the contrary, if λ<T, it is determined that there is no target in the current K distance units to be detected, and the main data is used as auxiliary data for other subsequent distance units to be detected.

[0068] The detection performance comparison of the matching signal by the method of the present invention and the existing detection method is shown in the attached Figure 2 The results show that compared with the existing unstructured range-extended target subspace detectors (S-Gradient, S-GLRT, S-Wald, S-Rao, S-2SD, etc.), the proposed method has better detection performance for broadband radar weak targets under inhomogeneous sea clutter.

[0069] Embodiment 2:

[0070] Refer to the instruction manual Figure 1 The specific implementation of Example 2 is divided into the following steps:

[0071] Step B1 uses a ground detection radar to illuminate the area to be detected, and obtains the main data Z of K distance units to be detected; the radar is used to illuminate the target-free range around the area to be detected, and obtains R reference distance units containing only pure ground clutter auxiliary data Z R . The main data Z and auxiliary data Z R Sent to the decorrelation processing module, in which the transformed main data Z is obtained according to formula (4) p and the estimation of the clutter skew-symmetric covariance matrix structure based on auxiliary data On the one hand, Z p and Sent to the joint probability density function module of the main data and auxiliary data under the H0 hypothesis, according to Z p and The joint probability density function of the main data and the auxiliary data under the H0 assumption is obtained (5); the joint probability density function of the main data and the auxiliary data under the H0 assumption is sent to the maximum likelihood estimation module of the clutter skew-symmetric covariance matrix structure under the H0 assumption, and the joint probability density function of the main data and the auxiliary data under the targetless assumption is used to derive the clutter skew-symmetric covariance matrix structure M and set it to zero. According to formula (7), the maximum likelihood estimation of M under the H0 assumption is obtained: On the other hand, Z p and Sent to the joint probability density function module of the main data and auxiliary data under the H1 hypothesis, according to Z p and The joint probability density function of the main data and the auxiliary data under the H1 assumption is obtained as formula (3); the joint probability density function of the main data and the auxiliary data under the H1 assumption is sent to the maximum likelihood estimation module of the clutter skew-symmetric covariance matrix structure under the H1 assumption, and the joint probability density function of the main data and the auxiliary data under the target assumption is used to derive the clutter skew-symmetric covariance matrix structure M and set it to zero. According to formula (6), the maximum likelihood estimation of M under the H1 assumption is obtained: Will The coordinate matrix is ​​sent to the maximum likelihood estimation module under the H1 hypothesis, and Z and Z are used under the target hypothesis. R The joint probability density function of is derived from the target transformation coordinate matrix and set to zero. According to formula (9), the unknown target transformation coordinate matrix B is obtained: p The maximum likelihood estimate of

[0072] It is worth noting that in step B1, the method of the present invention constructs a subspace range-extended target signal model, avoids the problem that the rank-one signal model is difficult to cope with the target guidance vector mismatch, and improves the robustness of the broadband radar to the ground target guidance vector mismatch situation; in addition, the method of the present invention makes full use of the skew-symmetric structure prior information of the ground clutter covariance matrix, and jointly uses the main data and auxiliary data to improve the estimation accuracy of the unknown ground clutter skew-symmetric covariance matrix structure, reduces the demand for the amount of auxiliary data, and provides favorable support for realizing the adaptive detection of range-extended targets under small sample conditions in a non-uniform ground clutter environment; 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, and by fully exploring the local uniformity of the ground clutter, the maximum likelihood estimation of the clutter power factor under the two assumptions of target and no target is used respectively, thereby improving the intelligent adaptability of the detector to the non-uniform ground clutter environment.

[0073] Step B2 obtains the maximum likelihood estimation of the unknown clutter power factor under the target assumption and the non-target assumption by solving the unique positive solution of the eigenvalue equation, which specifically includes two aspects. It is sent to the maximum likelihood estimation module of the power coefficient under the H0 assumption, and the maximum likelihood estimation of γ under the H0 assumption is obtained according to formula (11): On the other hand, and It is sent to the maximum likelihood estimation module of the power coefficient under the H1 assumption, and the maximum likelihood estimation of γ under the H1 assumption is obtained according to formula (11): Substitute the maximum likelihood estimate of the unknown clutter power factor into the intermediate statistic of the subspace skew-symmetric generalized likelihood ratio test obtained in step B1, and replace the unknown clutter power factor under the target hypothesis and the no-target hypothesis in the intermediate statistic respectively. and It is sent to the target adaptive intelligent detector construction module under non-uniform clutter, and the detection statistic λ of the target knowledge-assisted intelligent fusion detector under non-uniform clutter is obtained according to formula (12), and λ is sent to the detection decision module.

[0074] It is noteworthy that the method of the present invention constructs a subspace skew-symmetric generalized likelihood ratio test detector under non-uniform clutter, and the detector has a closed-form expression, a simple detection statistic structure, and is easy to implement in engineering.

[0075] Step B3 sets the detection threshold T according to the preset false alarm probability: specifically, the false alarm probability is set to P fa According to the Monte Carlo method, based on the 100 / P accumulated in the previous period fa The detection threshold T is calculated based on the measured ground clutter data. Furthermore, the detection statistic λ is compared with the detection threshold T. If λ≥T, it is determined that there is a target in the current K distance units to be detected, and the main data is not used as auxiliary data for other subsequent distance units to be detected; on the contrary, if λ<T, it is determined that there is no target in the current K distance units to be detected, and the main data is used as auxiliary data for other subsequent distance units to be detected.

[0076] The detection performance comparison between the method of the present invention and the existing oblique symmetric range extended target subspace detector is shown in the attached Figure 3 The results show that compared with the existing skew-symmetric range-extended target subspace detectors (Per-Rao and Per-Wald), the proposed method has better detection performance for broadband radar weak targets under non-uniform ground clutter.

Claims

1. The target knowledge-assisted intelligent fusion detection method under non-uniform clutter is characterized by: The following steps are involved: Step 1: Obtain primary data Z from K distance units to be detected, and obtain auxiliary data Z from R distance units adjacent to the distance unit to be detected. R ; Under the target hypothesis and the non-target hypothesis, use Z and Z respectively R The joint probability density function of is derived from the skew-symmetric covariance matrix structure M of the clutter and set to zero, and the maximum likelihood estimate of the unknown skew-symmetric covariance matrix structure of the clutter under the target assumption is obtained. At the same time, the maximum likelihood estimate of the unknown skew-symmetric covariance matrix structure of the clutter under the no-target assumption is obtained. Using Z and Z under the target assumption, R The joint probability density function of is derived from the target transformation coordinate matrix and set to zero to obtain the maximum likelihood estimate of the unknown target transformation coordinate matrix; then the subspace skew-symmetric generalized likelihood ratio test intermediate statistic is constructed under the condition that the clutter power factor is known; Maximum likelihood estimation of M under the target assumption for in, U represents the known N×r-dimensional signal multi-rank subspace matrix, Re(·) and Im(·) represent the real part and imaginary part respectively; B p =[b e1 ,b e2 ,…,b eK ,b o1 ,b o2 ,…,b oK ], b ek =Re(b k ), b ok =jIm(b k ), j is an imaginary unit, b k represents the complex coordinate vector of the r×1-dimensional target subspace in the k-th distance unit to be detected, Re(·) and Im(·) represent the real part and imaginary part respectively; Z p =[z e1 ,z e2 ,…,z eK ,z o1 ,z o2 ,…,z oK ], z k represents the N×1 dimensional main data component corresponding to the kth range unit to be detected, J represents the N×N dimensional permutation matrix with diagonal elements set to 1 and other elements set to 0; γ represents the clutter power factor, N represents the number of space-time joint channels, r represents the rank of the matrix U, (·) * and(·) H denote conjugate and conjugate transpose respectively; Maximum Likelihood Estimation of M without Target Assumption for Step 2: By solving the unique positive solution of the eigenvalue equation under the target hypothesis and the no-target hypothesis, the maximum likelihood estimate of the unknown clutter power factor under the target hypothesis and the no-target hypothesis is obtained respectively, and the maximum likelihood estimate of the unknown clutter power factor is substituted into the subspace skew-symmetric generalized likelihood ratio test intermediate statistic obtained in step 1, and the unknown clutter power factor under the target hypothesis and the no-target hypothesis is replaced in the intermediate statistic respectively, so as to construct the detection statistic λ of the target knowledge-assisted intelligent fusion detector under non-uniform clutter; Maximum likelihood estimation of unknown clutter power factor γ under the no-target hypothesis H0 and the target hypothesis H1 and is the only positive solution to the following eigenvalue equation: Where i = 0, 1, λ k,0 and λ k,1 Respectively represent matrices and The kth eigenvalue of ; Step 3 is to maintain the constant false alarm rate characteristic of the detection method, and set the detection threshold T according to the preset false alarm probability; compare the detection statistic λ with the detection threshold T, if λ≥T, it is determined that there is a target in the current distance unit to be detected, and the main data is not used as auxiliary data for other subsequent distance units to be detected; on the contrary, if λ<T, it is determined that there is no target in the current distance unit to be detected, and the main data is used as auxiliary data for other subsequent distance units to be detected.

2. The target knowledge-assisted intelligent fusion detection method under non-uniform clutter according to claim 1 is characterized in that: In step 1: The target transformation coordinate matrix B under the target assumption p The maximum likelihood estimate of is 3. The target knowledge-assisted intelligent fusion detection method under non-uniform clutter according to claim 1 is characterized in that: In step 2: The detection statistic λ of the target knowledge-assisted intelligent fusion detector under non-uniform clutter is expressed as in, and denote the maximum likelihood estimation of the clutter power factor γ under the assumption of no target and target, respectively, |·| denotes the determinant of the square matrix, I 2K Represents the 2K×2K-dimensional identity matrix.

Images