Blind adaptive space-time detection method and system suitable for multiple point target scenes

Abstract

The invention provides a blind adaptive space-time detection method and system suitable for a multi-point target scene, and the method comprises the steps: carrying out the filtering, amplification and sampling quantification preprocessing of a plurality of point target echoes in a detection sector, and carrying out the modeling of the obtained echo data; introducing hidden random variables representing a receiving sample unit where a target is located and arrival angle information, and establishing a multi-target echo probability density function under an EM classification framework by using an LVM method; solving a log-likelihood function of a receiving data matrix by utilizing EM optimization iteration to obtain loop optimization estimation expressions of unknown parameters in the step E and the step M, and estimating the positions of multiple targets in a receiving sample unit, the angles of arrival of the multiple targets and the number of the targets; and according to a generalized likelihood ratio detection formula, obtaining independent estimation of each target arrival angle. The method has the advantages that the calculation complexity caused by angle traversal is reduced, the method does not need to depend on steering vector prior information, and the method has higher environmental suitability and robustness.

Description

Technical Field

[0001] The present application belongs to the field of underwater sonar technology, and specifically relates to a blind adaptive space-time detection method and system suitable for multiple point target scenes. Background Art

[0002] In recent years, underwater space-time adaptive detection (STAD) has been vigorously promoted in the field of active sonar detection due to its technical advantages such as excellent reverberation suppression, flexible signal processing, and strong environmental adaptability. At this stage, the clustered operation mode using multiple unmanned aerial vehicle equipment is becoming the development trend of underwater unmanned platforms, which puts higher requirements on the multi-target detection capability of active sonar. Compared with single target detection, the difficulty of multi-target detection lies in the target shielding between multiple strong echoes in the detection window, as well as the estimation of key parameters such as the number of multiple targets and wave arrival angle. How to solve the above problems and realize accurate and reliable multi-target detection of moving sonar platforms has important application value and theoretical significance for underwater multi-target monitoring.

[0003] At present, most underwater multi-target detection methods based on STAD require complete target guidance vector information, that is, the multi-target wave arrival angles are known a priori during the detector design process. However, in actual sonar operation scenarios, accurate wave direction information cannot be obtained before target detection judgment, and the corresponding multi-target guidance vector is biased or even completely unknown, so existing methods cannot work effectively.

[0004] An existing technical solution for the problem of multi-target detection and blind adaptive detection based on STAD is an underwater cluster target detection method based on expectation maximization clustering. This method clusters multiple point targets in the detection sector, assumes that the spatial and temporal structure of the target echo is unknown, and implements the classification and detection of multiple cluster targets by mining the covariance matrix characteristics of the cluster target echo under the energy fluctuation model. This technical solution does not strictly adopt the multiple point target echo model when modeling, but uses the multi-target cluster hypothesis instead, that is, it ignores the inconsistency of the multi-target arrival angles in the sonar detection main lobe. Although it does not rely on the prior knowledge of the multi-target arrival angles to complete target detection, it does not have the ability to independently estimate the multi-target arrival angles.

[0005] Another existing technical solution is a blind adaptive sonar target detection method based on expectation maximization, which is only used for a single extended range target to achieve single extended range target wave arrival angle estimation and detection. This technical solution is only a blind adaptive detection method for a single extended range target and cannot be applied to multiple point target detection scenarios. Summary of the invention

[0006] The purpose of the present application is to overcome the defect that the prior art does not have the ability to independently estimate the arrival angles of multiple targets.

[0007] In order to achieve the above objectives, the present application proposes a blind b-adaptive space-time detection method applicable to multiple point target scenes, comprising:

[0008] Step 1: Model the echo data obtained after filtering, amplification and sampling quantization preprocessing of multiple point target echoes in the detection sector;

[0009] Step 2: Introduce hidden random variables representing the receiving sample unit where the target is located and the wave arrival angle information, and use the LVM method to establish the multi-target echo probability density function under the EM classification framework;

[0010] Step 3: Use EM optimization to iteratively solve the log-likelihood function of the received data matrix, obtain the cyclic optimization estimation expression of the unknown parameters in the E step and the M step, and estimate the position of the multi-target in the receiving sample unit, the arrival angle of each multi-target, and the number of targets based on the maximum a posteriori classification criterion;

[0011] Step 4: Substitute the estimation result of step 3 into the generalized likelihood ratio detection formula to obtain the existence judgment of multiple targets in the distance unit dimension and the independent estimation of the arrival angle of each target.

[0012] As an improvement of the above method, the result of modeling the echo data in step 1 is:

[0013]

[0014] Among them, H0 and H1 represent the null hypothesis that there is no target and the alternative hypothesis that there is a target, respectively; k represents the kth N×1-dimensional echo data vector; CN N (0, M) means that it obeys the zero mean, and the interference covariance matrix is The complex Gaussian distribution of Represents the N×N dimensional complex matrix space, N represents the number of array elements; n k represents an underwater interference signal composed of white noise and reverberation, obeying CN N The complex Gaussian distribution of (0, M) satisfies the independent and identically distributed assumptions among multiple interference signals; α k，n Indicates that the kth echo sample comes from a wave arrival angle of θ n The echo amplitude of the target wave arrival angle is n, and the corresponding ordinal number of the target wave arrival angle in the spatial discrete angle region is v(θ n ) indicates that the kth echo sample comes from the wave arrival angle θ n The target orientation vector of Located in the detection sector Ω θ At the angle sampling grid points within K θ Represents the detection sector Ω θThe number of discrete grid points; Ω = {1, ..., K} represents the subscript index set of K echo samples, where the data sample index set containing the target echo component is represented by Ω T ={k1, ..., k T}, T≥1 indicates the number of multiple targets, and the subscript index of the data vector without target signal components is Ω I =Ω T \Ω,\ represents the set Ω and Ω T The difference between .

[0015] As an improvement of the above method, the probability density function of the multi-target echo in step 2 is:

[0016]

[0017] Among them, c k , k = 1, ... K and e k , k=1,...,K represent K independent and identically distributed hidden random variables, and P(c k =s) = π s , s=0,1 and P(e k =θ n )=p n ,θ n ∈Ω θ For variable c k and e k The corresponding unknown probability mass function, π s Represents the echo sample z k The probability of the target existing or not existing, s = 0 means there is no target, s = 1 means there is a target, p n Represents the echo sample z k The upper target arrival angle is θ n The probability of and Represent the echo samples z k The unknown parameter sets in the above cases with and without targets, represents the transpose operation, represents the vector-valued function corresponding to a particular entry of the matrix M, Represents the target amplitude vector corresponding to different wave angles, Φ k =Φ k，0 ∪Φ k，1 ∪{p n :n=1,...,K θ , π s :s=0,1};

[0018] f(z k |c k=0;Φ k，0 ) and f(z k |c k =1,e k =n;Φ k,1 ) is expressed as follows:

[0019]

[0020]

[0021] Among them, det(.) and Tr[.] represent the matrix determinant and trace operation respectively; e(.) represents the exponential operation with e as the base.

[0022] As an improvement of the above method, step 3 comprises:

[0023] Using EM optimization to iteratively solve the received data matrix The log-likelihood function L(Z; Φ′) of the unknown parameters is obtained:

[0024]

[0025] Where Φ={Φ k : k = 1, ..., K} represents the unknown parameter set of all K echo samples, q k (s), s = 0, 1, and r k (n), n=1, ..., K θ Indicates c k =s and e k =n, the unknown posterior probability in the E-step loop is expressed as:

[0026]

[0027]

[0028] in, represents the m-1th EM iteration result of the corresponding parameter; s=0.1 is the penalty factor based on the generalized information criterion, and ρ is the parameter;

[0029] Using M-step loops to estimate unknown parameters in Φ, the optimization objective function is expressed as:

[0030]

[0031] Applying the Lagrange criterion, we first get π s and p n The estimated result at the mth iteration is:

[0032]

[0033] The mth EM iteration estimation result of parameter M is:

[0034]

[0035] in,

[0036] set up Given that, find the optimization objective function about The maximum estimate of , and the mth EM iteration estimation result is obtained:

[0037]

[0038] in,

[0039]

[0040]

[0041] in,

[0042]

[0043]

[0044]

[0045]

[0046] diag{.} represents diagonal matrix operation; EM iterative estimation for each n=1,...,K θ Do repeated operations;

[0047] When the loop meets the set iteration stop condition, the final estimation result is obtained, which is expressed as in Indicates the number of cycles when the iteration stops; parameter Ω T , T and the wave arrival angle θ n The maximum a posteriori probability decision is estimated as follows:

[0048]

[0049] in, and represents the final estimation result of the corresponding posterior probability, It represents the estimated echo sample subscript set with target component, and the target arrival angle estimation result is Indicates the angle subscript of the target echo in the detection sector, and the target quantity estimation for The length of the collection.

[0050] As an improvement of the above method, step 4 includes:

[0051] Substituting the estimation result of step 3 into the generalized likelihood ratio detection formula, we get:

[0052]

[0053] Wherein, η represents the set threshold;

[0054]

[0055] represent the probability density function of Z under the H1 and H0 assumptions respectively, It represents the estimation result of M under the H0 hypothesis.

[0056] The present application also provides a blind adaptive space-time detection system applicable to multiple point target scenarios, which is implemented based on the above method and includes:

[0057] The echo data modeling module is used to model the echo data obtained after filtering, amplification and sampling quantization preprocessing of multiple point target echoes in the detection sector;

[0058] A multi-target echo probability density function module is established to introduce hidden random variables representing the receiving sample unit and the arrival angle information of the target, and the multi-target echo probability density function is established under the EM classification framework using the LVM method.

[0059] The target parameter estimation module is used to use EM optimization to iteratively solve the log-likelihood function of the received data matrix, obtain the cyclic optimization estimation expression of the unknown parameters in the E step and the M step, and estimate the position of multiple targets in the receiving sample unit, the arrival angles of multiple targets, and the number of targets based on the maximum a posteriori classification criterion;

[0060] The target arrival angle estimation module is used to substitute the estimation result into the generalized likelihood ratio detection formula to obtain the existence judgment of multiple targets in the distance unit dimension and the independent estimation of the arrival angle of each target.

[0061] Compared with the prior art, the advantages of this application are:

[0062] 1. Aiming at the application needs of multi-target detection, the present invention proposes a blind adaptive space-time detection method suitable for multiple point target scenarios. The design of the proposed space-time adaptive detector does not rely on the multi-target guidance vector information. In the case of mismatch or complete unknown of the guidance vector, accurate multi-target classification and multi-arrival angle estimation can be achieved, and it has higher environmental adaptability and robustness.

[0063] 2. The present invention appropriately simplifies the target signal, ignores the target signal distance extension caused by multipath effect, target range migration and echo distortion, and models the multi-targets as multiple idealized point target models without actual size, geometric shape and extension.

[0064] 3. The present invention is based on multiple point target detection scenarios, abandons the unified cluster assumption of multi-target arrival angles, and can achieve independent estimation of different arrival angles of multiple targets, which is closer to actual application needs.

[0065] 4. The present invention is based on the EM algorithm and the improved LVM model. By batch processing sonar echo samples, a two-step maximum a posteriori probability criterion judgment is performed to achieve multi-target identification and joint estimation of multiple arrival angles, which greatly reduces the high amount of calculation caused by the traversal of each echo sample and sector angle, and has low time complexity.

[0066] 5. The present invention adopts the method based on The inner loop iteration realizes the independent optimization estimation of multi-objective amplitudes and interference covariance matrices, which solves the problem of difficult joint estimation of two parameters under this model system.

[0067] 6. The present invention takes into account the amplitude fluctuations of multi-target echoes in different sample units and different arrival angles in modeling. Compared with the Swerling 0 model without echo fluctuations between multiple sampling points, the model proposed by the present invention is closer to the actual underwater echo modeling requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 Shown is a flow chart of a blind adaptive space-time detection method applicable to multiple point target scenarios;

[0069] Figure 2 The figure shows the result of multi-target classification (1 simulation test, SINR = 15dB);

[0070] Figure 3 The figure shows the result of multi-target angle estimation (1 simulation test, SINR = 15dB);

[0071] Figure 4 Shown is a correct classification probability graph (1000 simulation tests, SINR = 15dB);

[0072] Figure 5 The figure shows the false alarm probability of 10 under 1000 Monte Carlo simulation tests. -3 , a comparison chart of the detection probability (Pd) curves of the method proposed in this application and the CMP-GLRT method. DETAILED DESCRIPTION

[0073] The technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0074] Aiming at the problem of active detection and arrival angle estimation of multiple point targets in the absence of guidance vectors, the present invention proposes a blind adaptive space-time detection method and system suitable for multiple point target scenarios. The method uses maximum expectation clustering (EM) and latent variable model (LVM) to achieve the existence judgment of multiple targets in the distance unit dimension and the independent estimation of the arrival angle of each target based on the two-step maximum a posteriori probability criterion, getting rid of the clustering assumption of the echo direction of multiple targets and the dependence on the prior information of the target guidance vector in the prior art, and has a strong application prospect.

[0075] like Figure 1 Shown is a flow chart of a blind adaptive space-time detection method applicable to multiple point target scenarios provided by the present application.

[0076] As an applicable scenario of the method of the present invention, consider an active sonar system equipped with a uniform linear array of N array elements. After filtering, amplification, sampling and quantization preprocessing of multiple point target echoes in the detection sector, K N×1-dimensional echo data vectors z1,...,z K Assume that Ω = {1, ..., K} represents the subscript index set of K echo samples, where the data sample index set containing the target echo component is represented by Ω T ={k1, ..., k T}, T≥1 represents the number of multiple targets, and the subscript index of the data vector without target signal components is defined as Ω I =Ω T \Ω,\ represents the set Ω and Ω T Therefore, the problem of multiple point target detection can be expressed by the following hypothesis test:

[0077]

[0078] Where: H0 and H1 represent the null hypothesis that there is no target and the alternative hypothesis that there is a target, respectively; CN N (0, M) means that it obeys zero mean and the interference covariance matrix is Complex Gaussian distribution; n k It represents the underwater interference signal composed of white noise and reverberation, which obeys zero mean and the interference covariance matrix is The complex Gaussian distribution of , while satisfying the independent and identically distributed (IID) assumption between multiple interference signals, represents the N×N dimensional complex matrix space; α k，n Indicates that the kth echo sample comes from a wave arrival angle of θ n The echo amplitude of the target wave arrival angle is n, and the corresponding ordinal number of the target wave arrival angle in the spatial discrete angle region is v(θ n) represents the k echo samples from the wave arrival angle θ n The target orientation vector of Located in the detection sector Ω θ At the angle sampling grid points inside (3dB beam main lobe), K θ Represents the detection sector Ω θ The number of discrete grid points.

[0079] The above signal model is now extended to blind adaptive detection in multi-target scenarios, that is, when the target index position Ω in the echo sample in equation (1) is T , the number of targets T and the target and interference statistical parameters α k，n ,θ n , when both M and T are unknown, multi-target detection and multi-arrival angle estimation are achieved.

[0080] The blind adaptive space-time detection method for multiple point target scenes provided by the present invention comprises: firstly, based on the maximum a posteriori probability judgment of K echo sample dimensions, the echo samples containing the target components are classified. Ω T On this basis, the maximum a posteriori probability judgment based on the detection sector angle dimension is used to classify the echo samples after the first level classification. The arrival angle recognition is performed, and finally the estimated parameters are used to replace the unknown quantity in the likelihood ratio test (LRT) to achieve target detection under the guidance of classification.

[0081] LVM Modeling:

[0082] Define K IID latent random variables c k , k = 1, ... K and e k , k = 1, ..., K, is the classification mark of the two-step maximum a posteriori probability decision, let P(c k =s) = π s , s=0,1 and P(e k =θ n )=p n ,θ n ∈Ω θ For variable c k and e k The corresponding unknown probability mass function (PMF), where π s Represents the echo sample z k The probability of the target existing or not existing (s = 0 means no target, s = 1 means there is a target), p n Represents the echo sample z k The upper target arrival angle is θ n Based on this, equation (1) is the probability of echo sample z in multi-target scene.k , 1, ..., K, the probability density function is expressed as

[0083]

[0084] In the formula and Represent the echo samples z k The unknown parameter sets in the above cases with and without targets, represents the transpose operation, represents the vector-valued function corresponding to a particular entry of the matrix M, Represents the target amplitude vector corresponding to different wave angles, Φ k =Φ k,0 ∪Φ k,1 ∪{p n :n=1,...,K θ , π s :s=0,1},f(z k |c k =0;Φ k，0 ) and f(z k |c k =1,e k =n;Φ k，1 ) is expressed as follows:

[0085]

[0086]

[0087] In the formula, det(.), Tr[.] and They represent the determinant, trace and transpose operations of the matrix respectively, and e(.) represents the exponential operation with e as the base.

[0088] EM estimation process:

[0089] Next, we use EM optimization to iterate and solve the received data matrix The log-likelihood function L(Z; Φ′) of the above unknown parameters is obtained, where

[0090]

[0091] Where Φ={Φ k : k = 1, ..., K} represents the unknown parameter set of all K echo samples, q k (s), s = 0, 1, and r k (n), n=1,,..,K θ Indicates c k =s and e k =n, the unknown posterior probability in the E-step loop is expressed as

[0092]

[0093] in Indicates the m-1th EM iteration result of the corresponding parameters. s=0.1 is the penalty factor based on the generalized information criterion, and the parameter ρ is defined as ≥1.

[0094] Next, the M-step loop is used to estimate the unknown parameters in Φ. The optimization objective function is expressed as

[0095]

[0096] Applying the Lagrange criterion, we first get π s and p n The estimated result at the mth iteration is

[0097]

[0098]

[0099] Then, the mth EM iteration estimation result of parameter M is

[0100]

[0101] In the formula It is worth noting that the parameters M and α k , k = 1, ..., K joint maximum likelihood estimation is difficult to solve mathematically. We redefine the K-dimensional vector n=1,...,K θ Estimate the target amplitude. Specifically, assume Given that, find the objective function (9) with respect to The maximum estimate of the mth EM iteration estimation result is obtained

[0102]

[0103] In the formula

[0104]

[0105] in

[0106]

[0107]

[0108] diag{.} represents a diagonal matrix operation. Formula (13) is for each n=1,...,K θRepeat the operation. When the above loop meets the iteration stop condition, the final estimation result is obtained, which is expressed as in Indicates the number of cycles when the iteration stops. The iteration stop condition can be reaching the set maximum number of iterations or the log likelihood decreases less than the set threshold. The remaining parameters Ω T , T and the wave arrival angle θ n The maximum a posteriori probability decision is estimated as follows:

[0109]

[0110] in and represents the final estimation result of the corresponding posterior probability, It represents the estimated echo sample subscript set with target component, and the target arrival angle estimation result is Indicates the angle subscript of the target echo in the detection sector, and the target quantity estimation for The length of the collection.

[0111] STAD detector design:

[0112] Substituting the estimated results into the generalized likelihood ratio test (GLRT) test formula, we get

[0113]

[0114] In the formula, η represents the set threshold;

[0115]

[0116] represents the probability density function of Z under the H1 and H0 assumptions, It represents the estimation result of M under the H0 hypothesis.

[0117] Simulation analysis:

[0118] Assume that the sonar is equipped with a uniform linear array with N=8 elements, the number of received echo samples K=24, and the detection sector Ω θ =[-20°:2°:20°], i.e. sector length K θ =21, the locations of multiple targets in the sector are shown in Table 1, and the signal-to-interference-noise ratio (SINR) is expressed as where α k，i and θ k，i Represents the amplitude and arrival angle of the i-th target at the k-th echo sample, and the interference covariance matrix in CNR = 15dB means reverberation noise ratio, R c is the Hermitian positive definite Toeplitz matrix. The EM parameters are initialized as ρ=1, s=0,1, n=1,...,K θ , k=1,...,K.

[0119] Table 1 Multi-objective parameters

[0120] Parameter information Goal 1 Goal 2 Goal 3 Echo sample number 6 13 20 Arrival angle / ° -16 4 12

[0121] Figure 2 and Figure 3 The multi-target classification and arrival angle estimation results of a simulation experiment with SINR = 15dB are given respectively. The results show that the classification results of the proposed method are consistent with the actual category settings. Figure 2 The multi-target echoes are accurately identified at the 6th, 13th and 20th echo samples. Figure 3 In Figure 2 At the corresponding target position, the target arrival angle is accurately estimated, and the estimation result is consistent with the parameter settings in Table 1.

[0122] Figure 4 The correct classification probability (%) of all echo samples and their arrival angle domain under 1000 Monte Carlo simulation tests with SINR = 20dB is given. The results show that the proposed method has an excellent effect in the given parameter settings, with a classification accuracy of nearly 100% for the existence of multiple targets and a classification accuracy of nearly 80% for the arrival angle of multiple targets.

[0123] Figure 5 Given 1000 Monte Carlo simulations, the false alarm probability is 10 -3 Detection probability (Pd) curve of the proposed method and the CMP-GLRT comparison method. It can be seen that the proposed method is better than the comparison method in P d =0.9, it has a performance advantage of about 3dB.

[0124] Aiming at the problem of multiple point target detection and arrival angle estimation under the condition of missing target guidance vector information, the present invention proposes a blind adaptive space-time detection method suitable for multiple point target scenarios. The proposed method uses two-step posterior probability judgment. On the one hand, it abandons the clustering assumption of multi-target echoes in the prior art, and realizes independent estimation of different arrival angles of multiple targets by batch processing sonar echo samples, which can greatly reduce the computational complexity brought by angle traversal. On the other hand, it gets rid of the dependence of the existing multi-target detection technology on the prior information of the guidance vector, and has higher environmental adaptability and robustness when the guidance vector is mismatched or completely unknown.

[0125] The present application also provides a blind adaptive space-time detection system applicable to multiple point target scenarios, which is implemented based on the above method and includes:

[0126] The echo data modeling module is used to model the echo data obtained after filtering, amplification and sampling quantization preprocessing of multiple point target echoes in the detection sector;

[0127] A multi-target echo probability density function module is established to introduce hidden random variables representing the receiving sample unit and the arrival angle information of the target, and the multi-target echo probability density function is established under the EM classification framework using the LVM method.

[0128] The target parameter estimation module is used to use EM optimization to iteratively solve the log-likelihood function of the received data matrix, obtain the cyclic optimization estimation expression of the unknown parameters in the E step and the M step, and estimate the position of multiple targets in the receiving sample unit, the arrival angles of multiple targets, and the number of targets based on the maximum a posteriori classification criterion;

[0129] The target arrival angle estimation module is used to substitute the estimation result into the generalized likelihood ratio detection formula to obtain the existence judgment of multiple targets in the distance unit dimension and the independent estimation of the arrival angle of each target.

[0130] The present application may also provide a computer device, comprising: at least one processor, a memory, at least one network interface and a user interface. The various components in the device are coupled together through a bus system. It is understood that the bus system is used to achieve connection and communication between these components. In addition to the data bus, the bus system also includes a power bus, a control bus and a status signal bus.

[0131] The user interface may include a display, a keyboard or a pointing device, such as a mouse, a trackball, a touch pad or a touch screen.

[0132] It is understood that the memory in the embodiments disclosed in the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus random access memory (DRRAM). The memories described herein are intended to include, but are not limited to, these and any other suitable types of memories.

[0133] In some embodiments, the memory stores the following elements, executable modules or data structures, or a subset thereof, or an extended set thereof: an operating system and applications.

[0134] The operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., which are used to implement various basic services and process hardware-based tasks. The application includes various application programs, such as a media player (Media Player), a browser (Browser), etc., which are used to implement various application services. The program for implementing the method of the embodiment of the present disclosure can be included in the application.

[0135] In the above embodiment, the processor may also call a program or instruction stored in the memory, specifically, a program or instruction stored in an application program, and is used to:

[0136] Execute the steps of the above method.

[0137] The above method can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The above disclosed methods, steps and logic block diagrams can be implemented or executed. The general processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the above disclosed method can be directly embodied as a hardware decoding processor to execute, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0138] It is understood that the embodiments described in the present application can be implemented by hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application or a combination thereof.

[0139] For software implementation, the technology of the present application can be implemented by executing the functional modules (such as procedures, functions, etc.) of the present application. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0140] The present application may also provide a non-volatile storage medium for storing a computer program. When the computer program is executed by a processor, each step in the above method embodiment can be implemented.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application is described in detail with reference to the embodiments, a person skilled in the art should understand that any modification or equivalent replacement of the technical solution of the present application does not depart from the spirit and scope of the technical solution of the present application and should be included in the scope of the claims of the present application.

Claims

1. A blind adaptive space-time detection method suitable for multiple point target scenes, comprising: Step 1: Model the echo data obtained after filtering, amplification and sampling quantization preprocessing of multiple point target echoes in the detection sector; Step 2: Introduce hidden random variables representing the receiving sample unit where the target is located and the wave arrival angle information, and use the LVM method to establish the multi-target echo probability density function under the EM classification framework; Step 3: Use EM optimization to iteratively solve the log-likelihood function of the received data matrix, obtain the cyclic optimization estimation expression of the unknown parameters in the E step and the M step, and estimate the position of the multi-target in the receiving sample unit, the arrival angle of each multi-target, and the number of targets based on the maximum a posteriori classification criterion; Step 4: Substitute the estimation result of step 3 into the generalized likelihood ratio detection formula to obtain the existence judgment of multiple targets in the distance unit dimension and the independent estimation of the arrival angle of each target.

2. The blind adaptive space-time detection method for multiple point target scenes according to claim 1, characterized in that: The result of modeling the echo data in step 1 is: Among them, H0 and H1 represent the null hypothesis that there is no target and the alternative hypothesis that there is a target, respectively; k represents the kth N×1-dimensional echo data vector; CN N (0, M) means that it obeys the zero mean, and the interference covariance matrix is ​​M∈ The complex Gaussian distribution of Represents the N×N dimensional complex matrix space, N represents the number of array elements; n k represents an underwater interference signal composed of white noise and reverberation, obeying CN N The complex Gaussian distribution of (0, M) satisfies the independent and identically distributed assumptions among multiple interference signals; α k，n Indicates that the kth echo sample comes from a wave arrival angle of θ n The echo amplitude of the target wave arrival angle is n, and the corresponding ordinal number of the target wave arrival angle in the spatial discrete angle region is v(θ n ) indicates that the kth echo sample comes from the wave arrival angle θ n The target orientation vector of Located in the detection sector Ω θ At the angle sampling grid points within K θ Represents the detection sector Ω θ The number of discrete grid points; Ω = {1, ..., K} represents the subscript index set of K echo samples, where the data sample index set containing the target echo component is represented by Ω T ={k1, ..., k T }, T≥1 indicates the number of multiple targets, and the subscript index of the data vector without target signal components is Ω I =Ω T \Ω,\ represents the set Ω and Ω T The difference between .

3. The blind adaptive space-time detection method applicable to multiple point target scenes according to claim 2 is characterized in that: The probability density function of the multi-target echo in step 2 is: Among them, c k , k = 1, ... K and e k , k=1,...,K represent K independent and identically distributed hidden random variables, and P(c k =s) = π s , s=0,1 and P(e k =θ n )=p n ,θ n ∈Ω θ For variable c k and e k The corresponding unknown probability mass function, π s Represents the echo sample z k The probability of the target existing or not existing, s = 0 means there is no target, s = 1 means there is a target, p n Represents the echo sample z k The upper target arrival angle is θ n probability; and Represent the echo samples z k The unknown parameter sets in the above cases with and without targets, represents the transpose operation, represents the vector-valued function corresponding to a particular entry of the matrix M, Represents the target amplitude vector corresponding to different wave angles, Φ k =Φ k,0 ∪Φ k，1 ∪{p n :n=1,...,K θ , π s :s=0,1}; f(z k |c k =0;φ k，0 ) and f(z k |c k =1,e k =n;Φ k，1 ) is expressed as follows: Among them, det(.) and Tr[.] represent the matrix determinant and trace operation respectively; e ( . ) Represents exponential operation with base e.

4. The blind adaptive space-time detection method applicable to multiple point target scenes according to claim 3 is characterized in that: The step 3 comprises: Using EM optimization to iteratively solve the received data matrix The log-likelihood function L(Z; Φ′) of the unknown parameters is obtained: Where Φ={Φ k : k = 1, ..., K} represents the unknown parameter set of all K echo samples, q k (s), s = 0, 1, and r k (n), n=1, ..., K θ Indicates c k =s and e k =n, the unknown posterior probability in the E-step loop is expressed as: in, represents the m-1th EM iteration result of the corresponding parameter; s=0.1 is the penalty factor based on the generalized information criterion, and ρ is the parameter; Using M-step loops to estimate unknown parameters in Φ, the optimization objective function is expressed as: Applying the Lagrange criterion, we first get π s and p n The estimated result at the mth iteration is: The mth EM iteration estimation result of parameter M is: in, set up Given that, find the optimization objective function about The maximum estimate of , and the mth EM iteration estimation result is obtained: in, in, represents diagonal matrix operation; EM iterative estimation for each n=1,...,K θ Do repeated operations; When the loop meets the set iteration stop condition, the final estimation result is obtained, which is expressed as in Indicates the number of cycles when the iteration stops; parameter Ω T , T and the wave arrival angle θ n The maximum a posteriori probability decision is estimated as follows: in, and represents the final estimation result of the corresponding posterior probability, 1: k∈Ω} represents the estimated echo sample subscript set with target components, and the target arrival angle estimation result is Indicates the angle subscript of the target echo in the detection sector, and the target quantity estimation for The length of the collection.

5. The blind adaptive space-time detection method applicable to multiple point target scenes according to claim 4 is characterized in that: The step 4 comprises: Substituting the estimation result of step 3 into the generalized likelihood ratio detection formula, we get: Wherein, η represents the set threshold; represent the probability density function of Z under the H1 and H0 assumptions, respectively. It represents the estimation result of M under the H0 hypothesis.

6. A blind adaptive space-time detection system suitable for multiple point target scenes, implemented based on the method described in any one of claims 1 to 5, characterized in that: The system comprises: The echo data modeling module is used to model the echo data obtained after filtering, amplification and sampling quantization preprocessing of multiple point target echoes in the detection sector; A multi-target echo probability density function module is established to introduce hidden random variables representing the receiving sample unit and the arrival angle information of the target, and the multi-target echo probability density function is established under the EM classification framework using the LVM method. A target parameter estimation module is used to use EM optimization to iteratively solve the log-likelihood function of the received data matrix, obtain the cyclic optimization estimation expression of the unknown parameters in the E step and the M step, and estimate the positions of multiple targets in the received sample unit, the arrival angles of multiple targets, and the number of targets based on the maximum a posteriori classification criterion; and The target arrival angle estimation module is used to substitute the estimation result into the generalized likelihood ratio detection formula to obtain the existence judgment of multiple targets in the distance unit dimension and the independent estimation of the arrival angle of each target.

Images