An underwater target recognition method and system based on fusion of acoustic, flow and electrical characteristics

Abstract

The application discloses an underwater target recognition method and system based on sound, flow and electric characteristic fusion, extracts sound field and electric field signal characteristic frequency through spectrum analysis, can suppress symmetrically distributed random noise such as Gaussian noise, strengthen the fundamental frequency component of harmonic signals, remove the harmonic quantity of uncoupled phase, and is favorable to extracting the fundamental frequency component of signals; can more comprehensively extract target flow field signal characteristics through time domain analysis of pressure difference data of two axisymmetric flow field sensors; and extracts underwater target sound, flow and electric signals through a wavelet packet decomposition method, on one hand, the wavelet packet decomposition overcomes the difficulty in characteristic extraction of traditional signal processing methods when processing nonlinear and non-stationary signals, and on the other hand, compared with wavelet decomposition, the wavelet packet decomposition can simultaneously decompose low-frequency and high-frequency parts of signals, and can more completely extract the characteristics of each frequency segment of signals.

Description

Technical Field

[0001] This invention belongs to the field of underwater target recognition, specifically relating to an underwater target recognition method and system based on the fusion of acoustic, flow, and electrical features. Background Technology

[0002] The ocean has become a crucial arena for global cooperation and competition, and its vast resources are of paramount importance to national development. The ability to accurately identify unidentified targets is key to the performance of underwater equipment, and is of utmost significance for marine resource protection and maritime security.

[0003] However, the mechanisms and types of physical field signals generated by targets such as ships are complex, posing challenges to feature extraction and target identification. How to effectively extract information reflecting the characteristics of a target from its physical field signals and classify and identify it accordingly is a difficult point and an urgent problem to be solved in underwater target identification technology research. At the same time, traditional single-physical-field detection methods all have certain limitations. For example, acoustic detection is greatly affected by the marine environment and has near-field detection blind spots; current field detection has a short range; and electric field detection has a short range and low accuracy when measuring moving targets. Traditional single-physical-field target identification methods have limited application scenarios and low accuracy, making them unsuitable for the complex marine environment and unable to meet the accuracy requirements of underwater target identification. Summary of the Invention

[0004] The purpose of this invention is to provide an underwater target recognition method and system based on the fusion of acoustic, flow and electrical features, so as to overcome the problems of low accuracy and redundant features affecting classification and recognition efficiency in the existing technology of underwater target recognition using a single physical field feature.

[0005] An underwater target identification method based on the fusion of acoustic, flow, and electrical features includes the following steps:

[0006] S1 uses a composite acoustic, flow, and electrical detection array to collect acoustic field simulation signals, electric field simulation signals, and flow field simulation signals of underwater targets;

[0007] S2, adopts Spectral analysis, wavelet packet decomposition, and time-domain analysis were used to extract features from the simulated sound field signal; Spectral analysis, wavelet packet decomposition, and time-domain analysis are used to extract features from the electric field simulation signal; wavelet packet decomposition and time-domain analysis are used to extract features from the flow field simulation signal; then, the features extracted from each physical field are fused within that physical field to obtain the feature vector of each physical field.

[0008] S3, Based on the feature layer fusion method, the feature vectors of each physical field obtained in step S2 are fused to obtain the target multi-physics field serial fusion feature;

[0009] S4. The PCA dimensionality reduction algorithm is used to reduce the dimensionality of the target multi-physics field cascade fusion features and remove redundant features.

[0010] S5, constructing a dataset using the dimensionality-reduced target multiphysics field cascaded feature data;

[0011] S6. Construct a support vector machine model and use the K-fold cross-validation algorithm to optimize the parameters in the support vector machine model;

[0012] S7. The support vector machine model is trained using the dataset constructed by the multi-physics field fusion feature of the reduced-dimensional target to obtain the globally optimal support vector machine model.

[0013] S8 acquires multi-physics field signals of underwater targets in real time, and uses a globally optimal support vector machine model to identify the multi-physics field signals of underwater targets and output the corresponding identification results.

[0014] Preferably, the characteristic frequency of the target sound field signal, the energy characteristics of the target sound field, and the time-domain characteristics of the sound field are combined to obtain the characteristic vector of the target sound field; the characteristic frequency of the electric field signal, the energy characteristics of the electric field signal, and the time-domain characteristics of the electric field are combined to obtain the characteristic vector of the target electric field; and the energy characteristics of the flow field and the time-domain characteristics of the flow field are combined to obtain the characteristic vector of the target flow field.

[0015] Preferably, the eigenvectors of the target electric field, the target flow field, and the target sound field are arranged into a u-row, v-column matrix; each row of the matrix is ​​zero-mean; then the covariance matrix is ​​calculated; the eigenvalues ​​and corresponding eigenvectors of the covariance matrix are obtained, and the cumulative contribution rate of the features is calculated; the eigenvectors are arranged into a matrix from top to bottom according to the size of their corresponding eigenvalues, and the minimum cumulative contribution rate is set according to the requirements. The first z rows are taken to form the matrix, which is the feature matrix after dimensionality reduction to z dimensions.

[0016] Preferably, the third-order cumulant c of the simulated sound field signal s(t) is calculated. 3s (τ s1 ,τ s2 ):

[0017] c 3s (τ s1 ,τ s2 )=cum{s(t s ),s(t s +τ s1 ),s(t s +τ s2 )}

[0018] Where t is time, τ si (i = 1, 2) represent different time windows;

[0019] Let τ s1 =τ s2 =τ s The third-order cumulant c is obtained. 3s (τ s1 ,τ s2 diagonal slice c 3s (τ s ,τ s );

[0020] For c 3s (τ s ,τ s Perform a Fourier transform to obtain s(t). spectral density C(ω) s ):

[0021]

[0022] according to spectral density C(ω) s Extracting the feature frequency vector W s =(ω s1 ,ω s2 ,...,ω sk );

[0023] Where, ω si (i = 1, 2, ..., k) represent the different characteristic frequencies of the sound field.

[0024] Preferably, the third-order cumulant c of the simulated electric field signal e(t) is calculated. 3e (τ e1 ,τ e2 ):

[0025] c 3e (τ e1 ,τ e2 )=cum{e(t e ),e(t e +τ e1 ),e(t e +τ e2 )}

[0026] Where t is time, τ ei (i = 1, 2) represent different time windows;

[0027] Let τ e1 =τ e2 =τ e The third-order cumulant c is obtained. 3e (τ e1 ,τ e2 diagonal slice c 3e (τ e ,τe );

[0028] For c 3e (τ e ,τ e Perform a Fourier transform to obtain e(t). spectral density C(ω) e ):

[0029]

[0030] according to spectral density C(ω) e Extracting the feature frequency vector W e =(ω e1 ,ω e2 ,...,ω ek );

[0031] Where, ω ej (j=1,2,...,n) represent the different characteristic frequencies of the sound field.

[0032] Preferably, time-domain feature information is extracted from the sound field simulation signal, electric field simulation signal and flow field simulation signal respectively using time-domain analysis to obtain the sound field time-domain feature, flow field time-domain feature and electric field time-domain feature.

[0033] An underwater target recognition system based on the fusion of acoustic, fluid, and electrical features includes a simulation module, a feature vector calculation module, a feature fusion module, a dimensionality reduction module, a dataset module, a vector machine module, an optimization module, and a recognition module.

[0034] The simulation module uses a composite acoustic, flow, and electrical detection array to collect acoustic field simulation signals, electric field simulation signals, and flow field simulation signals of underwater targets.

[0035] The feature vector calculation module adopts... Spectral analysis, wavelet packet decomposition, and time-domain analysis were used to extract features from the simulated sound field signal; Spectral analysis, wavelet packet decomposition, and time-domain analysis are used to extract features from the electric field simulation signal; wavelet packet decomposition and time-domain analysis are used to extract features from the flow field simulation signal; then, the features extracted from each physical field are fused within that physical field to obtain the feature vector of each physical field.

[0036] The feature fusion module fuses the feature vectors of each physical field obtained in step S2 based on the feature layer fusion method to obtain the target multi-physics field cascaded fusion feature;

[0037] The dimensionality reduction module uses the PCA dimensionality reduction algorithm to reduce the dimensionality of the target multi-physics field cascaded fusion features and remove redundant features.

[0038] The dataset module constructs a dataset using the dimensionality-reduced target multiphysics field cascaded feature data.

[0039] The Support Vector Machine module constructs a Support Vector Machine model and uses the K-fold cross-validation algorithm to optimize the parameters in the Support Vector Machine model.

[0040] The optimization module uses the training set data constructed from the multi-physics field fusion features of the reduced-dimensional target to train the support vector machine model and obtain the globally optimal support vector machine model.

[0041] The identification module acquires multi-physics field signals of underwater targets in real time, and uses a globally optimal support vector machine model to identify the multi-physics field signals of underwater targets and output the corresponding identification results.

[0042] Preferably, the characteristic frequency of the target sound field signal, the energy characteristics of the target sound field, and the time-domain characteristics of the sound field are combined to obtain the characteristic vector of the target sound field; the characteristic frequency of the electric field signal, the energy characteristics of the electric field signal, and the time-domain characteristics of the electric field are combined to obtain the characteristic vector of the target electric field; and the energy characteristics of the flow field and the time-domain characteristics of the flow field are combined to obtain the characteristic vector of the target flow field.

[0043] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the underwater target recognition method based on the fusion of acoustic, fluid, and electrical features described above.

[0044] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described underwater target identification method that integrates acoustic, fluid, and electrical features.

[0045] Compared with the prior art, the present invention has the following beneficial technical effects:

[0046] This invention provides an underwater target identification method based on the fusion of acoustic, flow, and electrical features. It extracts features with strong target discrimination capabilities from each of the three single physical field signals (acoustic, flow, and electrical), solving the problem of ineffective features resulting from extracting the same features from different physical field signals. Spectral analysis extracts characteristic frequencies of acoustic and electric field signals, which can suppress symmetrically distributed random noise such as Gaussian noise, enhance the fundamental frequency component of harmonic signals, and remove harmonic quantities of non-coupled phases, thus facilitating the extraction of the fundamental frequency components of the signal. Time-domain analysis using pressure difference data from two axisymmetric flow field sensors enables a more comprehensive extraction of target flow field signal characteristics. By employing wavelet packet decomposition to extract features of underwater target acoustic, flow, and electrical signals, wavelet packet decomposition overcomes the difficulty of feature extraction in processing nonlinear and non-stationary signals using traditional signal processing methods. Furthermore, compared to wavelet decomposition, it can simultaneously decompose the low-frequency and high-frequency components of the signal, enabling a more complete extraction of features across each frequency band.

[0047] This invention comprehensively considers wavelet packet energy, time domain, and frequency domain useful information. The extracted feature information includes 12 indicators, enabling a comprehensive multi-domain representation of the differential characteristics of underwater targets. This solves the problems of existing algorithms using single physical field feature indicators, which are incomplete and inadequate in describing the differential characteristics of underwater targets. Addressing the issue of varying sensitivity among features, this method employs PCA dimensionality reduction to filter features and construct a multi-domain feature set sensitive to underwater target signals, thus resolving the feature redundancy problem.

[0048] In another aspect, this invention improves the accuracy and efficiency of subsequent underwater target recognition models based on support vector machines (SVMs). K-fold cross-validation allows for repeated training and validation using data samples even with limited data, effectively avoiding overfitting and underfitting, and yielding highly convincing results. The SVM algorithm is simple in principle, fast in training, and can achieve high classification accuracy even with a small number of samples. Furthermore, it can map samples from low-dimensional space to high-dimensional space through kernel functions, exhibiting strong generalization performance. Therefore, the underwater target recognition model constructed based on the SVM algorithm can achieve accurate and efficient identification of underwater targets.

[0049] This invention extracts features based on the characteristics of various physical fields, resulting in stronger feature extraction capabilities. It also addresses the problem that existing methods using single-type feature indicators are incomplete and inadequate in describing the differences in underwater target characteristics. Compared to existing methods, this method integrates acoustic, current, and electrical multi-physical field signal features, providing a more comprehensive and advantageous characterization of underwater target properties. This allows for more accurate and efficient classification and identification of underwater targets, thus demonstrating promising application prospects in underwater target feature extraction and intelligent classification and identification. (See attached figures.)

[0050] Figure 1 This is a flowchart of an underwater target recognition method based on the fusion of acoustic, flow, and electrical features in an embodiment of the present invention.

[0051] Figure 2This is a schematic diagram of wavelet packet decomposition and energy values ​​of each sub-band in an embodiment of the present invention.

[0052] Figure 3 This is a schematic diagram of the series fusion of multi-physics field features in an embodiment of the present invention.

[0053] Figure 4 This is a flowchart of the K-fold cross-validation process in an embodiment of the present invention.

[0054] Figure 5 This is a schematic diagram of input data partitioning in K-fold cross-validation in an embodiment of the present invention.

[0055] Figure 6 The data represents the acoustic, current, and electrical simulation signals of a ship target collected by the composite detection array in this embodiment of the invention.

[0056] Figures 7(a), 7(b), 7(c), and 7(d) show the results of single physical field feature identification and fusion feature identification of acoustic, fluid, and electrical characteristics of three types of ship targets, respectively. Detailed Implementation

[0057] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0058] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0059] like Figure 1 As shown, this invention provides an underwater target identification method based on the fusion of acoustic, flow, and electrical features, comprising the following steps:

[0060] S1 utilizes a composite acoustic, current, and electrical detection array to collect multi-physics field simulation signals of underwater targets;

[0061] The multiphysics simulation signals for underwater targets include acoustic field simulation signals, electric field simulation signals, and flow field simulation signals.

[0062] S2, adopts Spectral analysis, wavelet packet decomposition, and time-domain analysis were used to extract features from the simulated sound field signal; Spectral analysis, wavelet packet decomposition, and time-domain analysis are used to extract features from the electric field simulation signal; wavelet packet decomposition and time-domain analysis are used to extract features from the flow field simulation signal; then, the features extracted from each physical field are fused within that physical field to obtain the feature vector of each physical field.

[0063] S3, Based on the feature layer fusion method, the feature vectors of each physical field obtained in step S2 are fused to obtain the target multi-physics field serial fusion feature;

[0064] S4. The PCA dimensionality reduction algorithm is used to reduce the dimensionality of the target multi-physics field cascade fusion features and remove redundant features.

[0065] S5: Construct a dataset using the dimensionality-reduced target multiphysics field cascaded feature data, and divide it into training and testing sets;

[0066] S6. Construct a support vector machine model and use the K-fold cross-validation algorithm to optimize the parameters in the support vector machine model;

[0067] S7. The support vector machine model is trained using the training set data constructed by the multi-physics field fusion feature of the target after dimensionality reduction, and the globally optimal support vector machine model is obtained.

[0068] S8 acquires multi-physics field signals of underwater targets in real time, and uses a globally optimal support vector machine model to identify the multi-physics field signals of underwater targets and output the corresponding identification results.

[0069] Specifically, this application is approved 3D spectral analysis extracts the characteristic frequencies W of the target sound field signal from the sound field simulation signal and the electric field simulation signal, respectively. S and the characteristic frequency W of the electric field signal E The specific steps are as follows:

[0070] (2-1-1-1) Calculate the third-order cumulant c of the simulated sound field signal s(t). 3s (τ s1 ,τ s2 ):

[0071] c 3s (τ s1 ,τ s2 )=cum{s(t s ),s(ts +τ s1 ),s(t s +τ s2 )}

[0072] Where t is time, τ si (i = 1, 2) represent different time windows.

[0073] (2-1-2-1) Let τ s1 =τ s2 =τ s The third-order cumulant c is obtained. 3s (τ s1 ,τ s2 diagonal slice c 3s (τ s ,τ s );

[0074] (2-1-3-1) for c 3s (τ s ,τ s Perform a Fourier transform to obtain s(t). spectral density C(ω) s ):

[0075]

[0076] (2-1-4-1) According to spectral density C(ω) s Extracting the feature frequency vector W s =(ω s1 ,ω s2 ,...,ω sk ).

[0077] Where, ω si (i = 1, 2, ..., k) represent the different characteristic frequencies of the sound field.

[0078] (2-1-1-2) Calculate the third-order cumulant c of the simulated electric field signal e(t). 3e (τ e1 ,τ e2 ):

[0079] c 3e (τ e1 ,τ e2 )=cum{e(t e ),e(t e +τ e1 ),e(t e +τ e2 )}

[0080] Where t is time, τ ei(i = 1, 2) represent different time windows.

[0081] (2-1-2-2) Let τ e1 =τ e2 =τ e The third-order cumulant c is obtained. 3e (τ e1 ,τ e2 diagonal slice c 3e (τ e ,τ e );

[0082] (2-1-3-2) for c 3e (τ e ,τ e Perform a Fourier transform to obtain e(t). spectral density C(ω) e ):

[0083]

[0084] (2-1-4-2) According to spectral density C(ω) e Extracting the feature frequency vector W e =(ω e1 ,ω e2 ,...,ω ek ).

[0085] Where, ω ej (j=1,2,...,n) represent the different characteristic frequencies of the sound field.

[0086] Specifically, such as Figure 2 As shown, the target sound field energy feature T is extracted from the sound field simulation signal, electric field simulation signal, and flow field simulation signal respectively by wavelet packet decomposition. S Flow field energy characteristics T L and electric field signal energy characteristics T E The specific steps are as follows:

[0087] (2-2-1) First, select the wavelet type and entropy type based on the characteristics of the acquired target signal;

[0088] (2-2-2) Determine the number of wavelet packet decomposition levels j;

[0089] (2-2-3) For signal x m (t)(j) m Layer wavelet packet decomposition, yielding Sub-signals in different frequency bands;

[0090] Where m = 1, 2, 3..., m1 is the sound field signal, m2 is the flow field signal, and m3 is the electric field signal; j m Different wavelet packet decomposition layers are used.

[0091] (2-2-4) Calculate the wavelet packet energy value of each sub-band.

[0092] (2-2-5) The energy of all sub-bands constitutes the wavelet packet energy spectrum as an identification feature;

[0093] (2-2-6) Normalize the energy characteristics to obtain the sound field, flow field, and electric field identification feature vector T. S T L T E :

[0094]

[0095]

[0096] Step S2 describes using time-domain analysis to extract time-domain feature information from the sound field simulation signal, electric field simulation signal, and flow field simulation signal to obtain the sound field time-domain feature D. S Flow field time domain characteristics D S and electric field time-domain characteristics D E The time-domain feature information specifically includes 10 time-domain statistical feature parameters: mean, standard deviation, variance, root mean square, maximum value, skewness index, kurtosis index, peak index, impulse index, and margin index.

[0097] Based on the above steps, the characteristic frequency W of the target sound field signal is obtained. S Target sound field energy characteristics T S Harmony sound field temporal characteristics D S ; characteristic frequency W of electric field signal E Electric field signal energy characteristics T E and electric field time-domain characteristics D E Flow field energy characteristics T L and flow field time domain characteristics D S The characteristics of each physical field are combined, such as combining the characteristic frequency W of the target sound field signal. S Target sound field energy characteristics T S Harmony sound field temporal characteristics D S The characteristic vector of the target sound field is obtained by combining the components, and the characteristic frequency W of the electric field signal is obtained. E Electric field signal energy characteristics T E and electric field time-domain characteristics D E The eigenvectors of the target electric field are obtained by combining them, and the flow field energy characteristics T are obtained by combining them. L and flow field time domain characteristics DS The eigenvectors of the target flow field are obtained by combining them.

[0098] The feature-layer fusion method fuses the feature vectors of the target electric field, the target flow field, and the target sound field extracted in step S2 to obtain the target multi-physics field cascaded fusion feature. The steps are as follows:

[0099] (3-1) The extracted feature vectors of the target electric field, the target flow field, and the target acoustic field are denoted as F, respectively. S F L F E

[0100] (3-2) The eigenvectors of the target electric field, the target flow field, and the target sound field are fused and spliced ​​together to form an eigenvector F = (F S ,F L ,F E This enables the fusion of multi-physics field features.

[0101] The steps for using the PCA dimensionality reduction algorithm to perform dimensionality reduction processing on the target multiphysics field cascaded fusion features are as follows:

[0102] Suppose there are u one-dimensional eigenvectors, and each eigenvector contains v eigenvalues.

[0103] (4-1) Arrange the eigenvectors of the target electric field, the target flow field, and the target sound field into a v-row, u-column matrix X;

[0104] (4-2) Zero mean is applied to each row of matrix X, that is, the mean of that row is subtracted;

[0105] (4-3) Calculate the covariance matrix

[0106] (4-4) Find the eigenvalues ​​and corresponding eigenvectors of the covariance matrix, and calculate the cumulative contribution rate of the features;

[0107] (4-5) Arrange the eigenvectors into a matrix from top to bottom according to the size of the corresponding eigenvalues. Set the minimum cumulative contribution rate according to the requirements and take the first z rows to form matrix R.

[0108] (4-6)Y=(RX) T This is the feature matrix after dimensionality reduction to k dimensions.

[0109] like Figure 3 As shown, the specific steps for optimizing the parameters in the support vector machine model using the K-fold cross-validation algorithm are as follows:

[0110] (6-1) First, set the variation range and step size of the penalty parameter c and the RBF kernel parameter g;

[0111] (6-2) Divide the input data into K equal parts;

[0112] (6-3) Initialize parameters c and g;

[0113] (6-4) Use each part as the test set and the rest as the training set;

[0114] (6-5) Train the model and calculate the model's accuracy on the test set;

[0115] (6-6) Repeat (6-4) and (6-5) K times each time using a different part as the test set;

[0116] (6-7) The average accuracy is taken as the final accuracy of this training;

[0117] (6-8) Save the highest training accuracy and its corresponding parameters c and g;

[0118] (6-9) Update parameters c and g;

[0119] (6-10) Repeat (6-2) to (6-9) until all combinations of parameters c and g are traversed.

[0120] An underwater target recognition system based on the fusion of acoustic, fluid, and electrical features includes a simulation module, a feature vector calculation module, a feature fusion module, a dimensionality reduction module, a dataset module, a vector machine module, an optimization module, and a recognition module.

[0121] The simulation module uses a composite acoustic, flow, and electrical detection array to collect acoustic field simulation signals, electric field simulation signals, and flow field simulation signals of underwater targets.

[0122] The feature vector calculation module adopts... Spectral analysis, wavelet packet decomposition, and time-domain analysis were used to extract features from the simulated sound field signal; Spectral analysis, wavelet packet decomposition, and time-domain analysis are used to extract features from the electric field simulation signal; wavelet packet decomposition and time-domain analysis are used to extract features from the flow field simulation signal; then, the features extracted from each physical field are fused within that physical field to obtain the feature vector of each physical field.

[0123] The feature fusion module fuses the feature vectors of each physical field obtained in step S2 based on the feature layer fusion method to obtain the target multi-physics field cascaded fusion feature;

[0124] The dimensionality reduction module uses the PCA dimensionality reduction algorithm to reduce the dimensionality of the target multi-physics field cascaded fusion features and remove redundant features.

[0125] The dataset module constructs a dataset using the dimensionality-reduced target multiphysics field cascaded feature data.

[0126] The Support Vector Machine module constructs a Support Vector Machine model and uses the K-fold cross-validation algorithm to optimize the parameters in the Support Vector Machine model.

[0127] The optimization module uses the training set data constructed from the multi-physics field fusion features of the reduced-dimensional target to train the support vector machine model and obtain the globally optimal support vector machine model.

[0128] The identification module acquires multi-physics field signals of underwater targets in real time, and uses a globally optimal support vector machine model to identify the multi-physics field signals of underwater targets and output the corresponding identification results.

[0129] In one embodiment of this application, the characteristic frequency of the target sound field signal, the energy characteristics of the target sound field, and the time-domain characteristics of the sound field are combined to obtain the characteristic vector of the target sound field; the characteristic frequency of the electric field signal, the energy characteristics of the electric field signal, and the time-domain characteristics of the electric field are combined to obtain the characteristic vector of the target electric field; and the energy characteristics of the flow field and the time-domain characteristics of the flow field are combined to obtain the characteristic vector of the target flow field.

[0130] In another embodiment of the present invention, a terminal device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of an underwater target recognition method based on the fusion of acoustic, flow, and electrical features.

[0131] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory). This computer-readable storage medium is a memory device in a terminal device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and extended storage media supported by the terminal device. The computer-readable storage medium provides storage space containing the terminal's operating system. Furthermore, this storage space also contains one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the underwater target recognition method based on the fusion of acoustic, flow, and electrical features in the above embodiments.

[0132] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0133] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0134] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0135] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0136] Table 1 shows the normalized target sound field energy characteristics T obtained by performing three-level wavelet packet decomposition on the sound field signal to obtain eight sub-frequency bands and then calculating them. S

[0137] Table 1. Eigenvalues ​​of Normalized Wavelet Packet Energy of Sound Field

[0138] Feature number <![CDATA[T S1 ]]> <![CDATA[T S2 ]]> <![CDATA[T S3 ]]> <![CDATA[T S4 ]]> <![CDATA[T S5 ]]> <![CDATA[T S6 ]]> <![CDATA[T S7 ]]> <![CDATA[T S8 ]]> numerical values 0.8833 0.0171 0.0164 0.0167 0.0160 0.0178 0.01618 0.0167

[0139] The time-domain feature information extraction specifically includes 10 time-domain statistical feature parameters: mean, standard deviation, variance, root mean square value, maximum value, skewness index, kurtosis index, peak value, impulse index, and margin index. Detailed descriptions of each feature parameter are shown in Table 2 below.

[0140] Table 2. Description of Feature Parameter Information

[0141]

[0142] To verify the advantages of the acoustic, flow, and electrical feature fusion identification method used in this invention, this method uses a composite detection array to simulate the original acoustic, flow, and electrical signals of three types of ship targets, extracts features from the signal data of each physical field, and uses a support vector machine model to classify single physical field feature targets and fused feature targets, thereby verifying the performance of the identification model. Figure 7(a) shows the classification results of the identification model trained using only the feature set of the sound field for three types of ship targets, with an accuracy of 88.89%. Figure 7(b) shows the classification results of the identification model trained using only the feature set of the flow field for three types of ship targets, with an accuracy of 68.89%. Figure 7(c) shows the classification results of the identification model trained using only the feature set of the electric field for three types of ship targets, with an accuracy of 84.44%. Figure 7(d) shows the classification results of the identification model trained using only the fused feature set of the sound field, flow field, and electric field for three types of ship targets, with an accuracy of 97.78%. The results show that the underwater target identification model trained using the multi-physics field fusion feature set of acoustic, fluid, and electrical fields has a higher recognition accuracy for the three types of ship targets. The method of this invention is more advantageous than the method that only uses a single physics field.

Claims

1. A method for underwater target recognition based on fusion of acoustic, flow and electrical characteristics, characterized in that, The method comprises the following steps: S1, collecting the acoustic field simulation signal, the electric field simulation signal and the flow field simulation signal of the underwater target by using the acoustic, flow and electric composite detection array; S2, using The characteristic extraction is respectively performed on the sound field simulation signal by using the Wigner-Ville distribution, wavelet packet decomposition and time domain analysis. The characteristic extraction is respectively performed on the electric field simulation signal by using the Wigner-Ville distribution, wavelet packet decomposition and time domain analysis. The characteristic extraction is respectively performed on the flow field simulation signal by using the wavelet packet decomposition and time domain analysis. S3, fusing the feature vectors of each physical field obtained in step S2 to obtain the target multi-physical field series fusion feature based on a feature layer fusion method; S4, performing dimension reduction processing on the target multi-physical field series fusion feature by using a PCA dimension reduction algorithm to remove redundant features; S5, constructing a data set by using the target multi-physical field series fusion feature data after dimension reduction; S6, constructing a support vector machine model and optimizing the parameters in the support vector machine model by using a K-fold cross-validation algorithm; S7, training the support vector machine model by using the data set constructed by the target multi-physical field series fusion feature after dimension reduction to obtain a globally optimal support vector machine model; S8, obtaining the multi-physical field signal of the underwater target in real time, and identifying the multi-physical field signal of the underwater target by using the globally optimal support vector machine model to output a corresponding identification result. The target acoustic field signal feature frequency, the target acoustic field energy feature and the acoustic field time domain feature are combined to obtain the feature vector of the target acoustic field, the electric field signal feature frequency, the electric field signal energy feature and the electric field time domain feature are combined to obtain the feature vector of the target electric field, and the flow field energy feature and the flow field time domain feature are combined to obtain the feature vector of the target flow field. The eigenvector of the target electric field, the eigenvector of the target flow field and the eigenvector of the target sound field are combined into a column matrix row column matrix; each row of the matrix is zero-mean; then a covariance matrix is calculated; the eigenvalue of the covariance matrix and the corresponding eigenvector are calculated, and the cumulative contribution rate of the eigenvalue is calculated; The eigenvectors are arranged in a matrix from top to bottom according to corresponding eigenvalues, and a lowest cumulative contribution rate is set according to requirements, and the first rows are taken to form a matrix, which is a characteristic matrix after dimension reduction to ​ Computing a sound field simulation signal of a third order cumulant : wherein, t is time, is a different time window; make The third-order cumulant is obtained. diagonal slice ; To Fourier transform, we get the dimensional spectrum : According to Spectra Extracting a feature frequency vector ; wherein, are different characteristic frequencies of the sound field.

2. The underwater target recognition method based on fusion of acoustic, flow and electrical characteristics according to claim 1, characterized in that, Computing an electric field simulation signal of the third order : wherein, is time, is a different time window; Let , obtain the diagonal slice of the third-order cumulant ; To perform a Fourier transform, we get the spectrum : According to spectrum extracting a feature frequency vector ; wherein, are different characteristic frequencies of the sound field.

3. The underwater target recognition method based on fusion of acoustic, flow and electrical characteristics according to claim 1, characterized in that, The time domain analysis is used to extract the time domain feature information from the acoustic field simulation signal, the electric field simulation signal and the flow field simulation signal to obtain the acoustic field time domain feature, the flow field time domain feature and the electric field time domain feature.

4. An underwater target recognition system based on fusion of acoustic, flow, and electrical signatures for use in the method of claim 1, characterized in that, The simulation module, the feature vector calculation module, the fusion feature module, the dimension reduction module, the data set module, the vector machine module, the optimization module and the identification module are included. The simulation module collects the acoustic field simulation signal, the electric field simulation signal and the flow field simulation signal of the underwater target by using the acoustic, flow and electric composite detection array. The feature vector calculation module adopts The characteristic extraction is respectively performed on the sound field simulation signal by using the three-dimensional spectrum analysis, wavelet packet decomposition and time domain analysis. The characteristic extraction is respectively performed on the electric field simulation signal by using the three-dimensional spectrum analysis, wavelet packet decomposition and time domain analysis. The characteristic extraction is respectively performed on the flow field simulation signal by using the wavelet packet decomposition and time domain analysis. Then, the characteristics extracted from the physical fields are fused in the physical field to obtain the feature vectors of the physical fields. The fusion feature module fuses the feature vectors of each physical field obtained in step S2 to obtain the target multi-physical field series fusion feature based on a feature layer fusion method. The dimension reduction module performs dimension reduction processing on the target multi-physical field series fusion feature by using a PCA dimension reduction algorithm to remove redundant features. The data set module constructs a data set by using the target multi-physical field series fusion feature data after dimension reduction. The vector machine module constructs a support vector machine model and optimizes the parameters in the support vector machine model by using a K-fold cross-validation algorithm. The optimization module trains the support vector machine model by using the training set data constructed by the target multi-physical field series fusion feature after dimension reduction to obtain a globally optimal support vector machine model. The identification module obtains the multi-physical field signal of the underwater target in real time, and identifies the multi-physical field signal of the underwater target by using the globally optimal support vector machine model to output a corresponding identification result.

5. The underwater target recognition system based on fusion of acoustic, flow and electrical characteristics according to claim 4, characterized in that, The target acoustic field signal feature frequency, the target acoustic field energy feature and the acoustic field time domain feature are combined to obtain the feature vector of the target acoustic field, the electric field signal feature frequency, the electric field signal energy feature and the electric field time domain feature are combined to obtain the feature vector of the target electric field, and the flow field energy feature and the flow field time domain feature are combined to obtain the feature vector of the target flow field.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The computer program is executed by the processor to implement the steps of the underwater target identification method based on sound, flow and electric feature fusion according to any one of claims 1 to 3.

7. A computer-readable storage medium storing a computer program, wherein the computer program comprises the following steps of: The computer program is executed by the processor to implement the steps of the underwater target identification method based on sound, flow and electric feature fusion according to any one of claims 1 to 3.

Images