An adaptive target detection method and system suitable for a reverberation edge environment

Abstract

The present application relates to the field of active sonar target detection, and relates to a kind of adaptive target detection method and system suitable for reverberation edge environment.The method of the present application comprises: selecting the data of unit to be detected as main data, and selecting the received data of multiple distance units on both sides of the unit to be detected as overall auxiliary data;Overall auxiliary data is distributed in the three uniform reverberation areas formed by the left reverberation edge and the right reverberation edge of the unit to be detected, and is divided into three auxiliary data subsets accordingly;Modeling target adaptive detection as a binary hypothesis testing problem including no-target hypothesis and target hypothesis;Obtain the joint probability density function of main data and overall auxiliary data under no-target hypothesis and target hypothesis;Based on the modified two-step generalized likelihood ratio test criterion, the joint probability density function and the probability density function of main data and auxiliary data subsets are processed in turn, and the final detection expression is obtained to realize target adaptive detection.

Description

Technical Field

[0001] This invention relates to the field of active sonar target detection, and in particular to an adaptive target detection method and system suitable for reverberant edge environments. Background Technology

[0002] For shallow-sea active sonar, reverberation, besides inherent underwater environmental noise, is the main background interference affecting target detection performance. When the sonar carrier moves, reverberation in different directions exhibits different Doppler frequency shifts, causing the reverberation spectrum to display spatiotemporal coupling characteristics. Traditional cascaded filtering methods, which first target the spatial domain and then the temporal domain, are insufficient to effectively suppress reverberation. Spatiotemporal adaptive detection technology, with spatiotemporal joint processing as its framework and target detection as its objective, utilizes received data to form detection statistics. By comparing these statistics with a detection threshold, the presence or absence of a target can be determined, achieving the integration of reverberation suppression and target detection. Compared to cascaded methods that first suppress reverberation and then perform constant false alarm rate (CFAR) detection, the spatiotemporal adaptive detection process is simpler, more flexible in design, and can utilize observation data more efficiently, resulting in superior detection performance.

[0003] In recent years, spatiotemporal adaptive detection has received increasingly extensive research in the field of active sonar, particularly for point target detection in Gaussian reverberation environments. Classical detection methods such as the Generalized Likelihood Ratio Test (GLRT), Adaptive Matched Filter (AMF), and Adaptive Coherence Estimator (ACE) can all be used to solve this type of problem. These classic adaptive detection methods are all proposed in homogeneous or partially homogeneous environments, both of which assume the existence of a set of uniform auxiliary data for target detection, with all auxiliary data exhibiting identical statistical properties. However, due to factors such as interface fluctuations and temperature variations, the actual underwater environment may exhibit non-uniformity. A commonly used model to describe the non-uniformity of the background environment is the reverberation edge environment. This scenario assumes that there are reverberation energy jumps in the auxiliary data, and the auxiliary data is divided into multiple uniform reverberation sub-regions by the reverberation edge, leading to a significant performance degradation of classic detection methods.

[0004] To address the aforementioned issues, Researcher Xu Da from the Institute of Acoustics, Chinese Academy of Sciences, proposed a method for re-filtering auxiliary data using reverberation edge location information in his doctoral dissertation, "Research on Multi-channel Sonar Reverberation Edge Detection and Target Detection Methods." This method retains only auxiliary data with the same reverberation covariance matrix as the target data, discarding the rest. Based on the two-step GLRT criterion, an effective solution was proposed, termed the T-AMF method. Research results show that the T-AMF method can improve target detection performance in reverberation edge environments. However, the biggest drawback of this method is that it does not consider the potential relationships between uniform reverberation sub-regions. The removal of auxiliary data leads to a significant decrease in data utilization, greatly limiting the target detection performance of active sonar. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing methods by modeling the reverberation energy jumps in auxiliary data as energy factors that differ between the reverberation covariance matrices in different reverberation regions. Based on the modified two-step GLRT criterion, detection statistics are calculated using all auxiliary data to achieve adaptive target detection in reverberation edge environments.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution.

[0007] This invention proposes an adaptive target detection method suitable for reverberant edge environments, the method comprising:

[0008] Data from a certain range cell received by the active sonar is selected as the data of the cell to be detected and is called the master data. At the same time, the received data from multiple range cells on both sides of the cell to be detected are selected as the overall auxiliary data. The overall auxiliary data is distributed in three uniform reverberation regions separated by the left and right reverberation edges of the cell to be detected and is divided into three auxiliary data subsets accordingly.

[0009] The adaptive target detection in the reverberant edge environment is modeled as a binary hypothesis testing problem that includes both the no-target assumption and the target assumption; wherein, the reverberant energy jump in the overall auxiliary data is modeled as the energy factor that differs between the reverberation covariance matrices in different reverberant regions;

[0010] Obtain the probability density function of the main data under no target hypothesis and with target hypothesis, as well as the probability density function of each auxiliary data subset, and further obtain the joint probability density function of the main data and the overall auxiliary data under the two hypotheses;

[0011] Based on the modified two-step generalized likelihood ratio test criterion, the joint probability density function and the probability density functions of the main data and auxiliary data subsets are processed successively to obtain the final detection expression, thereby achieving adaptive detection of the presence or absence of the target.

[0012] As an improvement to the above technical solution, the range cell data received by the active sonar are all N-dimensional complex Gaussian random vectors, where N represents the spatial, temporal, or spatiotemporal dimension.

[0013] As an improvement to the above technical solution, the expression for the binary hypothesis testing problem is as follows:

[0014]

[0015] Where H0 and H1 represent the no-target hypothesis and the target hypothesis, respectively; This represents the data of the unit to be detected, also known as master data. Represents the set of complex numbers; zk Let K represent the k-th auxiliary data; K1 and K2 represent the positions of the left and right reverberation edges of the unit to be detected, respectively, and assume that K2-K1≥N; n, Let n represent the background reverberation vectors of the unit to be detected and the three uniform reverberation regions, respectively, and let n be the background reverberation vectors of the three uniform reverberation regions. Represent n and It follows a complex Gaussian distribution with a mean of 0 and a covariance matrix of R, where γ1 and γ2 represent the unknown energy factors on both sides of the reverberation edge; The target signal's complex amplitude is an unknown deterministic parameter; Let v(θ) be the target nominal guiding vector, and θ be the target direction. When θ is known, v(θ) is abbreviated as v.

[0016] As one improvement to the above technical solution, the joint probability density function f(z, Z) of the master data and the overall auxiliary data under the assumptions of no target and with target is... K ;lα,γ1,γ2,R,H l The expression is:

[0017]

[0018] in, This represents the sample covariance matrix of the overall auxiliary data. This represents the sample covariance matrix of the auxiliary data to the left of the reverberation edge K1. This represents the sample covariance matrix of auxiliary data between reverberation edges K1 and K2. Z represents the sample covariance matrix of the auxiliary data to the right of the reverberation edge K2. K Z represents the overall auxiliary data. K =[z1…z K ], Z K1 This represents auxiliary data to the left of the reverberation edge K1. Z K2 This represents auxiliary data between the reverberation edges K1 and K2. Z K3 This represents auxiliary data to the right of the reverberation edge K2. Tr(·) denotes the trace of the matrix, (·) -1 The symbol (·) represents the inverse operation. H The conjugate transpose operation is represented by `det(·)`, and the determinant operation is represented by `det(·)`.

[0019] As one improvement to the above technical solution, the joint probability density function f(z, Z) of the master data and the overall auxiliary data under the assumptions of no target and with target is... K ;lα,γ1,γ2,R,H l The acquisition process includes:

[0020] Obtain the probability density function f(z; lα, R, H) of the master data under the hypotheses H0 and H1. l ):

[0021]

[0022] Obtain auxiliary data Z to the left of reverberation edge K1 K1 The probability density function f(Z) K1 ;γ1,R):

[0023]

[0024] Obtain auxiliary data Z between reverberation edges K1 and K2 K2 The probability density function f(Z) K2 ;R):

[0025]

[0026] Obtain auxiliary data Z to the right of the reverberation edge K2 K3 The probability density function f(Z) K3 ;γ2,R):

[0027]

[0028] Then the master data z and the overall auxiliary data Z K The joint probability density function f(z, Z) under the assumptions H0 and H1 K ;lα,γ1,γ2,R,H l )for:

[0029] f(z,Z K ;lα,γ1,γ2,R,H l )=f(z;lα,R,H l )f(Z K1 ;γ1,R)f(Z K2 ;R)f(Z K3 ;γ2,R),l=0,1.

[0030] As an improvement to the above technical solution, the modified two-step generalized likelihood ratio test criterion processes the joint probability density function and the probability density functions of the main data and auxiliary data subsets respectively to obtain the final detection expression, specifically including:

[0031] First, assuming that the energy factors γ1 and γ2 on both sides of the reverberation edge and the complex amplitude α of the target signal are known, the unknown parameter R is estimated by processing the joint probability density function, and a preliminary detection expression is obtained.

[0032] Then process Z separately.K1 Z K3 By taking the probability density function of z and the probability density function of z under the H1 hypothesis, we obtain the maximum likelihood estimates of γ1, γ2 and α. Substituting these estimates into the aforementioned detection expression to replace the true values, we obtain the final detection expression.

[0033] As one improvement to the above technical solution, the assumption that the energy factors γ1 and γ2 on both sides of the reverberation edge and the complex amplitude α of the target signal are known is used to estimate the unknown parameter R by processing the joint probability density function, and to obtain a preliminary detection expression, specifically including:

[0034] When γ1, γ2, and α are known, based on z and Z K The detection expression is:

[0035]

[0036] Where η represents the expected false alarm probability P fa The set detection threshold;

[0037] According to the following inequality:

[0038] log det(A)≤Tr(A)-N

[0039] get:

[0040]

[0041] Where A is an arbitrary N-dimensional matrix with non-negative eigenvalues. The estimated value of the unknown parameter R is calculated as follows:

[0042] Therefore, the preliminary detection expression can be obtained as follows:

[0043]

[0044] As one improvement to the above technical solution, the separate processing of Z K1 Z K3 The probability density function of z and the probability density function of z under the H1 hypothesis are used to obtain the maximum likelihood estimates of γ1, γ2 and α. These estimates are then substituted into the aforementioned detection expression to replace the true values, resulting in the final detection expression, which specifically includes:

[0045] First, calculate the logarithmic function lnf(Z). K1 ;γ1,R) and lnf(Z K3 Take the derivatives of γ2, R) with respect to γ1 and γ2 and set them to zero. Then replace R with S2 / (K2-K1) to obtain the maximum likelihood estimates of γ1 and γ2:

[0046]

[0047]

[0048] Find the derivative of the logarithmic function lnf(z; α, R, H1) with respect to α and set it to zero, then use... Substituting R, we obtain the maximum likelihood estimate of α:

[0049]

[0050] Will and Substituting γ1, γ2, and α into the preliminary detection expression yields the final detection expression:

[0051]

[0052] This invention also proposes an adaptive target detection system suitable for reverberant edge environments, the system comprising:

[0053] The data acquisition module is used to select data from a certain range cell received by the active sonar as the data of the unit to be detected, which is called the main data. At the same time, it selects the received data from multiple range cells on both sides of the unit to be detected as the overall auxiliary data. The overall auxiliary data is distributed in three uniform reverberation regions formed by the left and right reverberation edges of the unit to be detected, and is correspondingly divided into three auxiliary data subsets.

[0054] The problem modeling module is used to model the adaptive detection of targets in a reverberant edge environment as a binary hypothesis testing problem that includes both the assumption of no target and the assumption of a target; wherein, the reverberant energy jumps in the overall auxiliary data are modeled as the energy factor that differs between the reverberation covariance matrices in different reverberant regions;

[0055] The joint probability density function acquisition module is used to obtain the probability density functions of the main data under both the objective and negative hypotheses, as well as the probability density functions of each subset of auxiliary data. It further obtains the joint probability density function of the main data and the overall auxiliary data under both hypotheses.

[0056] The adaptive detection module is used to process the joint probability density function and the probability density functions of the main data and auxiliary data subsets respectively based on the modified two-step generalized likelihood ratio test criterion to obtain the final detection expression, thereby realizing adaptive detection of the presence or absence of the target.

[0057] The advantages of this invention compared to the prior art are:

[0058] 1. Compared with existing methods, which have low utilization of auxiliary data, resulting in loss of target detection performance and limited practical value, this invention models the reverberation energy jump as the energy factor between the reverberation covariance matrices in different reverberation regions, and realizes covariance matrix estimation based on all auxiliary data, thereby significantly improving the target detection performance in reverberation edge environments.

[0059] 2. This invention establishes a mathematical relationship between reverberation covariance matrices in different regions by modeling the reverberation edge as the energy factor between the reverberation covariance matrices in different uniform reverberation regions, so that all auxiliary data can be used for target detection.

[0060] 3. This invention is based on a modified two-step GLRT criterion. First, it assumes that the energy factor and the target echo complex amplitude are known. Then, it derives the GLRT by combining the master data and the overall auxiliary data to obtain a preliminary detection expression. Then, it obtains the maximum likelihood estimates of the above parameters based on the partial auxiliary data and the master data respectively. The estimated values ​​are substituted into the preliminary detection expression to replace the true values, and finally, a fully adaptive detection method is obtained.

[0061] 4. This invention proposes a novel adaptive target detection method suitable for reverberant edge environments, which significantly improves target detection performance in reverberant edge environments by utilizing all auxiliary data. Attached Figure Description

[0062] Figure 1 This is a flowchart of the method of the present invention;

[0063] Figure 2 This is a schematic diagram of a double reverberation edge environment;

[0064] Figure 3 P for the two detection methods d Curves showing the variation of SRNR (K1=8, K2=30);

[0065] Figure 4 P for the two detection methods d Curves showing the variation of SRNR (K1=6, K2=32). Detailed Implementation

[0066] To address the shortcomings of existing technologies, this invention proposes an adaptive target detection method suitable for reverberant edge environments. It models the reverberation energy jumps in auxiliary data as energy factors representing the differences in reverberation covariance matrices between different reverberation regions. Based on a modified two-step GLRT criterion, it calculates detection statistics using all auxiliary data, achieving adaptive target detection in reverberant edge environments. By improving the utilization rate of auxiliary data, it aims to enhance target detection performance.

[0067] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0068] Example 1

[0069] like Figure 1 The diagram shown is a flowchart of the method of the present invention. Figure 2 The image shown is a schematic diagram of a double reverberation edge environment.

[0070] Adaptive target detection in reverberant edge environments is a binary hypothesis testing problem, consisting of two parts: the assumption of no target and the assumption of a target. For reverberant edge environments, a more general case is considered: a reverberant edge exists in the auxiliary data on both sides of the target unit, dividing the auxiliary data into three uniform reverberant sub-regions. This invention models the jump in reverberation energy on both sides of the edge as an energy factor differing between the reverberation covariance matrices of the three uniform reverberation regions. By establishing the mathematical relationship between the reverberation covariance matrices, effective utilization of all auxiliary data is achieved. Based on the modified two-step GLRT criterion, a new adaptive target detection method suitable for reverberant edge environments is proposed, called the M2SGLRT method. The detailed design process of this invention is as follows:

[0071] 1. Problem Description

[0072] Assuming the received data from an active sonar system is a series of N-dimensional complex Gaussian random vectors, where N represents the spatial, temporal, or spatiotemporal dimension, point target detection in a reverberant edge environment can be modeled as the following binary hypothesis testing problem.

[0073]

[0074] in

[0075] H0 and H1 represent the no-target hypothesis and the target hypothesis, respectively;

[0076] · This represents the data of the unit to be detected, also known as master data; This represents the k-th auxiliary data;

[0077] K1 and K2 represent the positions of the left and right reverberation edges of the unit to be detected, respectively, and it is assumed that K2-K1≥N;

[0078] ·n, Let n represent the background reverberation vectors of the unit to be detected and the three uniform reverberation regions, respectively, and let n be the background reverberation vectors of the three uniform reverberation regions. γ1 and γ2 represent the energy factors on either side of the reverberation edge;

[0079] · The target signal's complex amplitude is an unknown deterministic parameter;

[0080] · Let v(θ) be the target nominal guiding vector, and θ be the target direction. Assuming θ is known, v(θ) can be simplified to v.

[0081] Based on the above assumptions, the probability density function of the main data under assumptions H0 and H1 can be written as follows:

[0082]

[0083] K1 Left Side Auxiliary Data The probability density function is

[0084]

[0085] Auxiliary data between K1 and K2 The probability density function is

[0086]

[0087] K2 right-side auxiliary data The probability density function is

[0088]

[0089] in, Tr(·) denotes the trace of the matrix, (·) -1 The symbol (·) represents the inverse operation. H `det(·)` represents the conjugate transpose operation, and `det(·)` represents the determinant operation of a matrix. Then, the main data and the overall auxiliary data Z... K =[z1…z K The joint probability density function under the assumptions H0 and H1 can be written as follows:

[0090]

[0091] in,

[0092] 2. Detection Method Design

[0093] To solve problem (1), an adaptive target detection method suitable for reverberant edge environments is designed based on the modified two-step GLRT criterion. The design process is divided into two steps: first, assuming that γ1, γ2, and α are known, the unknown parameter R is estimated by combining the master data and auxiliary data; then, based on Z... K1 Z K3 We obtain the maximum likelihood estimates of γ1, γ2 and α from z, and use the estimated values ​​to replace the true values ​​to finally obtain a fully adaptive detection method.

[0094] Based on the first step, when γ1, γ2, and α are known, based on z and Z K The detection expression is

[0095]

[0096] Where η represents the expected false alarm probability (P) fa The detection threshold is set.

[0097] According to the following inequality

[0098] logdet(A)≤Tr(A)-N,(8)

[0099] Where A is an arbitrary N-dimensional matrix with non-negative eigenvalues, we can obtain

[0100]

[0101] in,

[0102]

[0103] Substituting equation (9) into equation (7) and rearranging, we can obtain the preliminary detection expression.

[0104]

[0105] According to the second step, firstly based on Z K1 and Z K3 Find the maximum likelihood estimates of γ1 and γ2, i.e.

[0106]

[0107]

[0108] Find the logarithmic function lnf(Z) respectively. K1 ;γ1,R) and lnf(Z K3 Taking the derivatives of γ2, R) with respect to γ1 and γ2 and setting them to zero, and then replacing R with S2 / (K2-K1), we can obtain...

[0109]

[0110]

[0111] Then, based on z, the maximum likelihood estimate of α is obtained, i.e.

[0112]

[0113] Find the derivative of the logarithmic function lnf(z; α, R, H1) with respect to α and set it to zero, then use... By substituting R, we can obtain

[0114]

[0115] By substituting equations (14), (15), and (17) for γ1, γ2, and α respectively, equation (11) can be rewritten as follows:

[0116]

[0117] In the subsequent simulation, we refer to this detection method as the M2SGLRT method.

[0118] 3. Performance Analysis

[0119] Monte Carlo simulation was used to analyze the detection performance of the M2SGLRT method and compared it with the T-AMF method proposed by researcher Xu Da. To ensure the reliability of the simulation results, the number of simulations for calculating the detection threshold η and the detection probability Pd was 100 / P. fa and 10 4 The false alarm probability P fa =10 -4 The reverberation model employs a commonly used exponentially correlated complex Gaussian model, with the reverberation covariance matrix... Where (n1, n2) are the coordinates of the matrix elements. The reverberation power can be determined based on the noise-mixing ratio. calculate, This represents the normalized noise power. Energy factors γ1 and γ2 are based on the reverberation power ratio RPR1 = 10log[…]. 10 γ1 and RPR2 = 10log 10 γ² is calculated. The signal-to-mix ratio is defined as SRR = |α| 2 v H R -1 v, where the target orientation vector (·) T This indicates the transpose operation.

[0120] Figure 3 and Figure 4 The P values ​​of the two detection methods are given for two sets of (K1, K2) values ​​when N=16, K=40, RPR1=10dB, and RPR2=13dB, respectively. d The curves showing the change in SRR. It can be seen that the M2SGLRT method significantly outperforms the T-AMF method, and the auxiliary data length between the two reverberation edges (denoted as K) is also significantly better. s The smaller the value of K1, the more significant the performance advantage. For example, when K1 = 6 and K2 = 32 (at this time, K... s =26), the M2SGLRT method in P d At K = 0.9, the detection gain relative to the T-AMF method is 3.1 dB; while when K1 = 8 and K2 = 30 (at this time K... S =22), this gain increases to 5.3dB.

[0121] In summary, the M2SGLRT method proposed in this invention improves the utilization rate of auxiliary data by using all auxiliary data for target detection, thereby significantly enhancing the target detection performance in reverberant edge environments.

[0122] Example 2

[0123] An adaptive target detection system suitable for reverberant edge environments according to the present invention, the system comprising:

[0124] The data acquisition module is used to select data from a certain range cell received by the active sonar as the data of the unit to be detected, which is called the main data. At the same time, it selects the received data from multiple range cells on both sides of the unit to be detected as the overall auxiliary data. The overall auxiliary data is distributed in three uniform reverberation regions formed by the left and right reverberation edges of the unit to be detected, and is correspondingly divided into three auxiliary data subsets.

[0125] The problem modeling module is used to model the adaptive detection of targets in a reverberant edge environment as a binary hypothesis testing problem that includes both the assumption of no target and the assumption of a target; wherein, the reverberant energy jumps in the overall auxiliary data are modeled as the energy factor that differs between the reverberation covariance matrices in different reverberant regions;

[0126] The joint probability density function acquisition module is used to obtain the probability density functions of the main data under both the objective and negative hypotheses, as well as the probability density functions of each subset of auxiliary data. It further obtains the joint probability density function of the main data and the overall auxiliary data under both hypotheses.

[0127] The adaptive detection module is used to process the joint probability density function and the probability density functions of the main data and auxiliary data subsets respectively based on the modified two-step generalized likelihood ratio test criterion to obtain the final detection expression, thereby realizing adaptive detection of the presence or absence of the target.

[0128] In summary, this invention improves target detection performance in reverberant edge environments by increasing the utilization rate of auxiliary data.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An adaptive target detection method suitable for reverberant edge environments, the method comprising: Data from a certain range cell received by the active sonar is selected as the data of the cell to be detected and is called the master data. At the same time, the received data from multiple range cells on both sides of the cell to be detected are selected as the overall auxiliary data. The overall auxiliary data is distributed in three uniform reverberation regions separated by the left and right reverberation edges of the cell to be detected and is divided into three auxiliary data subsets accordingly. The adaptive target detection in the reverberant edge environment is modeled as a binary hypothesis testing problem that includes both the no-target assumption and the target assumption; wherein, the reverberant energy jump in the overall auxiliary data is modeled as the energy factor that differs between the reverberation covariance matrices in different reverberant regions; Obtain the probability density function of the main data under no target hypothesis and with target hypothesis, as well as the probability density function of each auxiliary data subset, and further obtain the joint probability density function of the main data and the overall auxiliary data under the two hypotheses; Based on the modified two-step generalized likelihood ratio test criterion, the joint probability density function and the probability density functions of the main data and auxiliary data subsets are processed successively to obtain the final detection expression, thereby achieving adaptive detection of the presence or absence of the target.

2. The adaptive target detection method for reverberant edge environments according to claim 1, characterized in that, The data received by the active sonar from each range cell are N-dimensional complex Gaussian random vectors, where N represents the spatial, temporal, or spatiotemporal dimension.

3. The adaptive target detection method for reverberant edge environments according to claim 2, characterized in that, The expression for the binary hypothesis testing problem is: Where H0 and H1 represent the no-target hypothesis and the target hypothesis, respectively; This represents the data of the unit to be detected, also known as master data. Represents the set of complex numbers; z k Let K represent the k-th auxiliary data; K1 and K2 represent the positions of the left and right reverberation edges of the unit to be detected, respectively, and assume that... Let represent the background reverberation vectors of the unit to be detected and the three uniform reverberation regions, respectively, and have . Represent n and It follows a complex Gaussian distribution with a mean of 0 and a covariance matrix of R, where γ1 and γ2 represent the unknown energy factors on both sides of the reverberation edge; The target signal's complex amplitude is an unknown deterministic parameter; Let v(θ) be the target nominal guiding vector, and θ be the target direction. When θ is known, v(θ) is abbreviated as v.

4. The adaptive target detection method for reverberant edge environments according to claim 3, characterized in that, The joint probability density function f(z,Z) of the master data and overall auxiliary data under both no-target and target-target assumptions. K ;lα,γ1,γ2,R,H l The expression is: in, This represents the sample covariance matrix of the overall auxiliary data. This represents the sample covariance matrix of the auxiliary data to the left of the reverberation edge K1. This represents the sample covariance matrix of auxiliary data between reverberation edges K1 and K2. Z represents the sample covariance matrix of the auxiliary data to the right of the reverberation edge K2. K Z represents the overall auxiliary data. K =[z1…z K ], Z K1 This represents auxiliary data to the left of the reverberation edge K1. Z K2 This represents auxiliary data between the reverberation edges K1 and K2. Z K3 This represents auxiliary data to the right of the reverberation edge K2. Tr(·) denotes the trace of the matrix, (·) -1 The symbol (·) represents the inverse operation. H The conjugate transpose operation is represented by `det(·)`, and the determinant operation is represented by `det(·)`.

5. The adaptive target detection method for reverberant edge environments according to claim 4, characterized in that, The joint probability density function f(z,Z) of the master data and overall auxiliary data under both no-target and target-target assumptions. K ;lα,γ1,γ2,R,H l The acquisition process includes: Obtain the probability density function f(z; lα,R,H) of the main data under the hypotheses H0 and H1. l ): Obtain auxiliary data Z to the left of reverberation edge K1 K1 The probability density function f(Z) K1 ;γ1,R): Obtain auxiliary data Z between reverberation edges K1 and K2 K2 The probability density function f(Z) K2 ;R): Obtain auxiliary data Z to the right of the reverberation edge K2 K3 The probability density function f(Z) K3 ;γ2,R): Then the master data z and the overall auxiliary data Z K The joint probability density function f(z, Z) under the assumptions H0 and H1 K ;lα,γ1,γ2,R,H l )for: f(z,Z K ;lα,γ1,γ2,R,H l )=f(z;lα,R,H l )f(Z K1 ;γ1,R)f(Z K2 ;R)f(Z K3 ;γ2,R),l=0,1。 6. The adaptive target detection method for reverberant edge environments according to claim 4, characterized in that, The modified two-step generalized likelihood ratio test criterion processes the joint probability density function and the probability density functions of the main data and auxiliary data subsets respectively to obtain the final detection expression, specifically including: First, assuming that the energy factors γ1 and γ2 on both sides of the reverberation edge and the complex amplitude α of the target signal are known, the unknown parameter R is estimated by processing the joint probability density function, and a preliminary detection expression is obtained. Then process Z separately. K1 Z K3 By taking the probability density function of z and the probability density function of z under the H1 hypothesis, we obtain the maximum likelihood estimates of γ1, γ2 and α. Substituting these estimates into the aforementioned detection expression to replace the true values, we obtain the final detection expression.

7. The adaptive target detection method for reverberant edge environments according to claim 6, characterized in that, The assumption is that the energy factors γ1 and γ2 on both sides of the reverberation edge and the complex amplitude α of the target signal are known. The unknown parameter R is estimated by processing the joint probability density function, and a preliminary detection expression is obtained, specifically including: When γ1, γ2, and α are known, based on z and Z K The detection expression is: Where η represents the expected false alarm probability P fa The set detection threshold; According to the following inequality: log det(A)≤Tr(A)-N get: Where A is an arbitrary N-dimensional matrix with non-negative eigenvalues. The estimated value of the unknown parameter R is calculated as follows: Therefore, the preliminary detection expression can be obtained as follows:

8. The adaptive target detection method for reverberant edge environments according to claim 7, characterized in that, The separate processing of Z K1 Z K3 The probability density function of z and the probability density function of z under the H1 hypothesis are used to obtain the maximum likelihood estimates of γ1, γ2 and α. These estimates are then substituted into the aforementioned detection expression to replace the true values, resulting in the final detection expression, which specifically includes: First, calculate the logarithmic function lnf(Z). K1 ;γ1,R) and lnf(Z K3 Take the derivatives of γ2, R) with respect to γ1 and γ2 and set them to zero. Then replace R with S2 / (K2-K1) to obtain the maximum likelihood estimates of γ1 and γ2: Find the derivative of the logarithmic function lnf(z; α, R, H1) with respect to α and set it to zero, then use... Substituting R, we obtain the maximum likelihood estimate of α: Will and Substituting γ1, γ2, and α into the preliminary detection expression yields the final detection expression:

9. An adaptive target detection system suitable for reverberant edge environments, the system comprising: The data acquisition module is used to select data from a certain range cell received by the active sonar as the data of the unit to be detected, which is called the main data. At the same time, it selects the received data from multiple range cells on both sides of the unit to be detected as the overall auxiliary data. The overall auxiliary data is distributed in three uniform reverberation regions formed by the left and right reverberation edges of the unit to be detected, and is correspondingly divided into three auxiliary data subsets. The problem modeling module is used to model the adaptive detection of targets in a reverberant edge environment as a binary hypothesis testing problem that includes both the assumption of no target and the assumption of a target; wherein, the reverberant energy jumps in the overall auxiliary data are modeled as the energy factor that differs between the reverberation covariance matrices in different reverberant regions; The joint probability density function acquisition module is used to obtain the probability density functions of the main data under both the objective and negative hypotheses, as well as the probability density functions of each subset of auxiliary data. It further obtains the joint probability density function of the main data and the overall auxiliary data under both hypotheses. The adaptive detection module is used to process the joint probability density function and the probability density functions of the main data and auxiliary data subsets respectively based on the modified two-step generalized likelihood ratio test criterion to obtain the final detection expression, thereby realizing adaptive detection of the presence or absence of the target.

Images