A Joint Statistical Method for Multidimensional Features of Underwater Target Echo Bright Spots Based on the Fusion of Experimental and Simulation Data

Abstract

A joint statistical method for multidimensional features of underwater vehicle echo highlights based on the fusion of experimental and simulation data is proposed. This method constructs a highlight feature fusion dataset that combines realism and large data scale by fusing experimentally measured echo data and model simulation echo data from underwater vehicles. By combining Hilbert transform envelope processing and peak-to-peak energy extraction, it accurately extracts multidimensional features such as the number, intensity, relative time delay difference, echo flicker characteristics, and target scale of target echo highlights. An EH distribution model is established to describe the highlight intensity characteristics, and the Parzen window estimation method is used to analyze the highlight time delay difference characteristics. A dynamic statistical framework within the spatial solid angle range is introduced to reveal the spatial orientation statistical distribution law of the multidimensional features. This invention overcomes the limitations of traditional target characteristic analysis methods, such as limited data sources, single feature dimensions, insufficient orientation statistical range, and insufficient adaptability of statistical models, and is applicable to target identification and echo simulation.

Description

Technical Field

[0001] This invention relates to the field of underwater acoustic target characteristic analysis and identification technology, specifically a joint statistical method for multi-dimensional features of underwater target echo highlights based on the fusion of experimental and simulation data. This method is applicable to underwater target detection and identification, acoustic decoy design and echo simulation, and is especially suitable for high-precision, multi-dimensional and comprehensive statistical modeling and analysis of the echo characteristics of complex targets such as underwater vehicles. Background Technology

[0002] Underwater target echo characteristic analysis is a key technology for underwater acoustic detection and target identification. Echoes, influenced by the target and the underwater acoustic environment, typically exhibit azimuth-dependent variations. Existing echo characteristic analysis methods usually only perform target intensity (TS) statistics for the horizontal azimuth. However, the echo characteristics of actual underwater targets exhibit complex fluctuations with spatial solid angles (including azimuth and pitch angles). This single-dimensional statistical approach cannot comprehensively reflect the spatial scattering characteristics of the target, leading to incomplete feature extraction and affecting identification accuracy.

[0003] Existing echo characteristics mostly focus on the static feature of overall target intensity, resulting in a relatively singular feature dimension. They neglect multi-dimensional dynamic information such as the number of bright spots contained in the echo, the intensity of each bright spot, the relative time delay difference, and scintillation characteristics. These features are directly related to the acoustic scattering characteristics of the target substructure (such as bow, sail, and rudder) and structural form (such as single or double shell), and are crucial for refined target identification. However, existing technologies lack effective joint statistical methods.

[0004] Current research often relies on experimental data or simulation data for modeling. While experimental data (such as on-site measurements on a lake) is authentic, it is limited by cost and environment, resulting in a limited amount of data and insufficient spatial coverage. Simulation data (such as calculations using the plate element method) can be generated in batches, but its accuracy depends on the modeling assumptions. Using the two separately makes it difficult to balance authenticity and data completeness.

[0005] Traditional statistical methods (such as simple mean or normal distribution assumptions) are difficult to describe the azimuth fluctuations of bright spot intensity (such as EH distribution characteristics) or the nonlinear distribution of time delay difference (requiring Parzen window estimation), resulting in insufficient matching between the model and the actual echo characteristics. Summary of the Invention

[0006] To overcome the shortcomings and gaps in existing technologies, this invention proposes a joint statistical method for multi-dimensional features of underwater vehicle echo bright spots based on the fusion of experimental and simulation data. By establishing a dynamic statistical framework within a spatial solid angle range and combining it with multi-dimensional joint modeling techniques such as the Copula function, collaborative analysis of features such as bright spot intensity, quantity, and time delay is achieved. Simultaneously, experimental data is used to calibrate simulation results, constructing a feature database that combines realism and scalability, providing more accurate technical support for underwater target identification and acoustic decoy design.

[0007] The technical solution of the present invention is as follows:

[0008] A joint statistical method for multidimensional features of underwater vehicle echo bright spots based on the fusion of experimental and simulation data is characterized by the following steps:

[0009] Step 1. Acquisition of spatial azimuth echo dataset:

[0010] Establish a spherical coordinate system with the center of the underwater target as the origin, and define the spatial orientation of sound wave incidence and reception. By using two complementary approaches—experimental measurement and simulation calculation—echo data were acquired within the spatial solid angle range covered by a set typical spatial orientation and its corresponding cone angle α, and a spatial orientation echo dataset was constructed.

[0011] Step 2. Multidimensional feature extraction and feature fusion:

[0012] The echo data is envelope-processed to extract multidimensional features, including the number of echo bright spots N(θ), the intensity of each bright spot E(θ), the relative time delay difference τ(θ) between adjacent bright spots, the target horizontal scale features, and the echo scintillation features based on the cross-correlation coefficient of adjacent azimuth echoes; and the similar features extracted from the experimental and simulation data are fused to form a multidimensional feature fusion dataset.

[0013] Step 3. Statistical analysis of echo multidimensional features:

[0014] Based on the multidimensional feature fusion dataset obtained in step (2), statistical modeling is performed on the fused multidimensional features, including:

[0015] - Statistical histogram distribution of the number of bright spots and the target horizontal scale;

[0016] - The intensity characteristics of bright spots are statistically modeled using the EH distribution model;

[0017] - The Parzen window kernel density estimation method is used to statistically model the nonlinear distribution of the relative time delay difference characteristics of bright spots;

[0018] -Statistical evaluation of echo scintillation characteristics was performed by calculating normalized cross-correlation coefficients;

[0019] Step 4. Joint statistical method based on Copula function:

[0020] The Copula function is used to perform joint distribution modeling on the multidimensional features, analyze the correlation structure between the feature variables, and complete the joint statistical analysis of the multidimensional features of the echo bright spots.

[0021] The steps to obtain the spatial azimuth echo dataset through simulation calculation are as follows:

[0022] Using the far-field distance r as the radius, three-dimensional coordinate points are uniformly extracted on a three-dimensional sphere centered on the coordinate system.

[0023]

[0024] Where n is the number of points, and i is the coordinate point of the i-th spherical point.

[0025] Converted to Cartesian coordinates:

[0026]

[0027] Since numerical computation has the advantage of large data volume and can calculate the echo at any point, the acoustic scattering transfer function of the azimuth of uniformly distributed points on a three-dimensional sphere is calculated based on the plate element method, and then the corresponding echo data is calculated using the frequency domain indirect method.

[0028] Based on the spherical coordinate points covered by the cylindrical solid angle corresponding to the typical spatial orientation, the echo data of the uniformly distributed spherical orientation are divided into typical spatial orientations for subsequent statistical analysis.

[0029] (2) Multidimensional feature extraction and feature fusion

[0030] Hilbert transform was used to extract the envelope of the echo data. The resulting echo signal envelope exhibited multi-peak bright spot characteristics. The target echo signals from different spatial orientations had different bright spot distribution characteristics. The multi-dimensional feature extraction and fusion method is as follows:

[0031] The echo signal envelope results are normalized and peak-to-peak value is extracted. A bright spot amplitude discrimination threshold is reasonably set. When determining the bright spot discrimination threshold, the method of using the inter-class variance and intra-class variance of the normalized amplitude of the bright spot as the judgment criteria is adopted. The number of echo bright spots N(θ) and the characteristics of bright spot amplitude in each direction are extracted.

[0032] Based on the target intensity values ​​and the normalized amplitude distribution of bright spots in various directions, the intensity characteristics E(θ) of each bright spot can be obtained from the perspective of energy distribution.

[0033] The absolute time distribution of each bright spot is extracted in the time domain, and the absolute time values ​​of two adjacent bright spots are subtracted to obtain the relative time delay difference feature τ(θ) between two adjacent bright spots.

[0034] The horizontal scale characteristics of the target are calculated based on the total pulse width expansion length of the echo signal in the time domain and the spatial orientation of the target model.

[0035] Cross-correlation calculations are performed on echo signal data at adjacent angles within a small azimuth range in the time domain to obtain the normalized cross-correlation coefficient between each pair of adjacent signals. The azimuth fluctuation of the target echo is evaluated based on the cross-correlation characteristics of echo signals at adjacent azimuths.

[0036] Based on the echo bright spot features of the target in various directions extracted from experimental and simulated echo data, feature data fusion is performed according to the type of bright spot feature to obtain a fused dataset for each bright spot feature.

[0037] (3) Multidimensional Statistical Characteristics of Echo

[0038] Statistical analysis of the multidimensional characteristics of the target echo is performed, including the number of target echo bright spots, intensity, time delay difference characteristics, target scale, and echo flicker characteristics, to obtain feature values ​​of the target model with statistical significance of azimuth fluctuations in typical azimuths. The specific method is as follows:

[0039] 1) Based on the target echo azimuth and its corresponding solid angle range, statistical modeling is performed on the characteristics of the number of bright spots and the horizontal scale of the target to obtain histograms of the distribution of the number of echo bright spots and the horizontal scale of the target in each azimuth.

[0040] 2) Based on the target echo azimuth and its corresponding solid angle range, statistical modeling and analysis are performed on the intensity characteristics of the echo bright spots generated by each substructure. The specific approach to the statistical analysis of bright spot intensity characteristics is as follows:

[0041] For a single echo bright spot, there is a relationship between its bright spot energy value and the normalized amplitude of each bright spot:

[0042]

[0043] Where P j (θ) represents the energy value of the j-th bright spot, j∈1,2,...,N(θ), P total (θ) represents the average total energy of the target model in the vicinity of this orientation, and its relationship with the target intensity can be expressed by formula (4):

[0044]

[0045] in This represents the average target intensity within the azimuth angle range. Assuming the contribution of each echo bright spot is considered from the perspective of energy superposition, the intensity characteristics of the echo bright spot can be expressed by formula (5):

[0046] E i (θ)=10lg(P i (θ)) (5)

[0047] Where E i (θ) represents the intensity value of the i-th echo point. For the sonar scattering cross-section random variable σ, χ 2 The distribution model is as follows:

[0048]

[0049] Where k is the half-degree-of-freedom parameter of the model, Let σ be the mean value of the target scattering cross-section. Considering that each bright spot in the target echo is generated by the corresponding substructure of the model, the intensity value of each bright spot can be approximated as satisfying the following relationship:

[0050]

[0051] Where, σ i (θ) represents the scattering cross-sectional area of ​​the substructure corresponding to each bright spot.

[0052] Substituting formula (7) into formula (6) and rearranging, we can obtain the probability density function formula for the intensity of the echo bright spot:

[0053]

[0054] in, Let ψ(k) be the mean intensity of the i-th bright spot, and let ψ(k) = Γ′(k) / Γ(k). The statistical parameter k in the model is used to characterize the intensity fluctuations in the azimuth of the bright spot intensity. Here, the bright spot intensity probability distribution model obtained by formula (8) is defined as the EH distribution model, i.e.

[0055] 3) Based on the target echo azimuth and its corresponding solid angle range, the Parzen window estimation theory is used to statistically model and analyze the time delay difference characteristics of the echo bright spots generated by each substructure. The specific approach is as follows:

[0056] Assuming the bright spot delay difference data τ(θ) satisfies an unknown probability density function, its probability within a finite region R can be expressed as:

[0057]

[0058] Assume the relative time delay difference τ between the (i+1)th and ith bright spots at the θ azimuth position is...i (θ) If K out of a total of N data points fall into region R, then it should follow a binomial distribution:

[0059]

[0060] From probability theory, when the sample size N is sufficiently large, we have:

[0061] K≈N·P(11)

[0062] Suppose V is the space of region R. When region R is sufficiently small, we have:

[0063] P i (θ)≈p(τ i (θ))·V (12)

[0064] Combining equations (11) and (12), we can obtain the expression for probability density estimation of the relative time delay difference characteristic data of bright spots:

[0065]

[0066] Assuming the region R remains constant, i.e., V remains constant, the probability density function can be estimated by determining the size of K. Using the kernel density estimation method, the probability density function can be obtained as follows:

[0067]

[0068] 4) To assess the azimuth fluctuation characteristics of target echoes, a method for evaluating target echo scintillation features is established. Near typical azimuth angles, the cross-correlation coefficients between echo signals at adjacent azimuth angles are normalized, and the formula is defined as follows:

[0069]

[0070] Where θ1 and θ2 represent two adjacent azimuth angle values, and y1 and y2 represent echo signal data at two azimuth angles, respectively. The cross-correlation function representing two sets of signal data. and These represent the values ​​of the autocorrelation function of the two sets of signal data at zero delay. The closer the value is to 0, the stronger the flicker of the echo spot at the corresponding azimuth.

[0071] (4) Joint statistical method based on Copula function

[0072] This study performs joint statistical analysis on the multidimensional features of echo bright spots from the perspective of multi-parameter joint statistical characteristics. Based on the echo bright spot features extracted from the echo bright spot fusion dataset, the Copula function can be used to perform joint statistical modeling on the bright spot features and to conduct statistical analysis on the correlation between the bright spot features.

[0073] One of the major advantages of Copula functions is that they can function F(x1),…,F(x) for N random variables belonging to any different distribution types. n ),,F(x N The construction of the joint distribution function of multiple variables is not affected by the distribution type of each random variable.

[0074] The specific method is as follows:

[0075] For a given univariate marginal distribution F1,…,F N For any joint distribution function, there must exist a Copula function C that satisfies:

[0076] F(x1,…,x n ,…,x N )=C(F(x1),…,F(x n ),…,F(x N (16)

[0077] Where, if F1,…,F N If F1,…,F N If the marginal distribution of the random variable is given, then the function F defined by formula (13) is F1,…,F N The joint distribution function.

[0078] Obviously, if u n =F n (x n Let ) be a one-dimensional random variable, and u n ∈[0,1], n=1,2,…,N, then C(F(x1),…,F(x n ),…,F(x N Copula is a multivariate distribution function with uniform margins in the N-dimensional [0,1] space, also known as the Copula function.

[0079] The density function f of the N-ary distribution function can be obtained using the density function c of the Copula function and the marginal density function, i.e.:

[0080]

[0081] Among them, f n It is a marginal distribution F n The probability density function of the Copula function, c, can be expressed as:

[0082]

[0083] Compared with the prior art, the present invention has the following technical effects:

[0084] 1. An echo bright spot feature fusion dataset was obtained by extracting echo bright spot feature data from underwater vehicle test and model simulation echo data. This dataset not only has the authenticity of actual results, but also has been greatly expanded by taking advantage of the easy availability of simulation data, which can provide a more realistic and complete dataset reference for the design of scale array parameters.

[0085] 2. Based on simulation methods, the echo signal data of the target within a small spatial solid angle was simulated and calculated, and the echo signal data of the target within a small spatial solid angle was obtained, laying the foundation for highlight feature analysis from the perspective of spatial orientation statistical significance.

[0086] 3. Statistical characteristic analysis was conducted on the multidimensional echo bright spot features of underwater targets, revealing the spatial orientation statistical distribution law satisfied by each bright spot feature, which can provide a more accurate reference for target detection and identification and acoustic simulation scale array parameter design.

[0087] 4. The joint statistical method of multidimensional features of underwater target echo highlights can quickly obtain the statistical results of highlight features of the target in all directions, comprehensively grasp the statistical characteristics of target echoes, and improve the fidelity of target echo simulation. Attached Figure Description

[0088] Figure 1 This is a flowchart of the joint statistical method for multidimensional features of echo bright spots in an embodiment of the present invention;

[0089] Figure 2 The present invention provides the target echo results from experiments and simulations, wherein (a) is the echo signal data result obtained by simulating the target model at the bow azimuth, and (b) is the echo signal data result obtained by simulating the target model at the bow azimuth.

[0090] Figure 3 These are the statistical results of the bright spot intensity in the embodiments of the present invention;

[0091] Figure 4 This is a statistical result of the number of bright spots in the embodiments of the present invention;

[0092] Figure 5 These are the highlights and latency statistics in the embodiments of the present invention;

[0093] Figure 6 This refers to the statistical results of the target echo scintillation characteristics in this embodiment of the invention;

[0094] Figure 7 This is the joint statistical result of multidimensional features in the embodiments of the present invention. Detailed Implementation

[0095] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0096] This embodiment takes a certain type of underwater vehicle as the target and uses a joint statistical method based on the multi-dimensional features of underwater vehicle echo bright spots, which is derived from the fusion of experimental and simulation data. Figure 1 As shown, it includes the following steps:

[0097] Step 1: Acquisition of Spatial Azimuth Echo Dataset

[0098] 1.1 Coordinate System Establishment: A spherical coordinate system is established with the center of the underwater target as the origin of the three-dimensional coordinate system, the length of the vessel as the x-axis, the vertical direction as the z-axis, and the positive transverse direction as the y-axis. The spatial orientation of the sound wave incident and received waves is defined as follows: Among them, azimuth angle The pitch angle θ is the angle between the x and z axes after projection onto the xy plane; a typical spatial orientation is set, with the coordinate center as the vertex of the cone and the vector containing the spatial orientation as the central axis of the cone. The cone angle α is defined as the spatial angle range corresponding to the spatial orientation, so as to obtain the spatial orientation echo data of the spatial angle range corresponding to the cone angle.

[0099] 1.2 Data Acquisition:

[0100] The steps to obtain the spatial azimuth echo dataset through experiments are as follows:

[0101] Conduct underwater target sound scattering test on lake surface, set a limited typical spatial orientation according to the test conditions, and deploy receiving hydrophones according to the spatial angle range of the typical spatial orientation to obtain spatial orientation echo data under the condition of combined transmission and reception.

[0102] The measured echo data were denoised and pulsed using signal processing methods such as bandpass filtering and matched filtering, respectively, to obtain echoes with high signal-to-noise ratio and clear bright peaks. A dataset of measured echoes of the target in typical orientations was constructed, such as... Figure 2 As shown in (a), the target model simulates the echo signal data at the bow azimuth. The target echo signal exhibits different scale broadening at different azimuths, and the echo bright spots show spatial distribution characteristics, which can characterize the scale characteristics of the target in that direction.

[0103] 1.3 Simulation Data Acquisition:

[0104] The steps to obtain the spatial azimuth echo dataset through simulation calculation are as follows:

[0105] Using the far-field distance r as the radius, three-dimensional coordinate points are uniformly extracted on a three-dimensional sphere centered on the coordinate system to ensure spatial uniformity.

[0106]

[0107] Where n is the number of points, and i is the coordinate point of the i-th spherical point.

[0108] Converted to Cartesian coordinates:

[0109]

[0110] The acoustic scattering transfer function of the target in all directions is calculated using the plate element method. In the frequency domain, the spectrum of the incident signal is multiplied by the acoustic scattering transfer function to obtain the spectrum of the echo signal. Finally, the inverse Fourier transform method is used to obtain the time-domain echo signal results in all directions, such as... Figure 2 Figure (b) shows the simulated echo signal data of the target model at the bow azimuth.

[0111] Based on the spherical coordinate points covered by the cylindrical solid angle corresponding to the typical spatial orientation, the echo data of the uniformly distributed spherical orientation are divided into typical spatial orientations for subsequent statistical analysis.

[0112] Step 2: Multidimensional Feature Extraction and Feature Fusion

[0113] 2.1 Envelope Extraction: Perform Hilbert transform on the echo signal from each direction to obtain the signal envelope, eliminate the influence of the carrier frequency, highlight the bright spot structure, and present multi-peak bright spot characteristics. Figure 2 (a) and (b) show that the target echo signals from different spatial orientations have different bright spot distribution characteristics.

[0114] 2.2 Bright Spot Count Extraction:

[0115] The echo signal envelope results are normalized and peak-to-peak value is extracted. A reasonable threshold for bright spot amplitude discrimination is set. When determining the bright spot discrimination threshold, the method of using the inter-class variance and intra-class variance of the normalized amplitude of the bright spot as the judgment criteria is adopted. The number of echo bright spots N(θ) and the amplitude of the bright spots in each direction are extracted.

[0116] 2.3 Bright Spot Intensity Extraction:

[0117] Based on the target intensity values ​​and the normalized amplitude distribution of bright spots in various directions, the intensity characteristics E(θ) of each bright spot can be obtained from the perspective of energy distribution.

[0118] 2.4 Relative Delay Difference Extraction:

[0119] The absolute time distribution of each bright spot is extracted in the time domain, and the absolute time values ​​of two adjacent bright spots are subtracted to obtain the relative time delay difference feature τ(θ) between two adjacent bright spots.

[0120] The horizontal scale characteristics of the target are calculated based on the total pulse width expansion length of the echo signal in the time domain and the spatial orientation of the target model.

[0121] Cross-correlation calculations are performed on echo signal data at adjacent angles within a small azimuth range in the time domain to obtain the normalized cross-correlation coefficient between each pair of adjacent signals. The azimuth fluctuation of the target echo is evaluated based on the cross-correlation characteristics of echo signals at adjacent azimuths.

[0122] Based on the echo bright spot features of the target in various directions extracted from experimental and simulated echo data, feature data fusion is performed according to the type of bright spot feature to obtain a fused dataset for each bright spot feature.

[0123] Step 3: Statistical Analysis of Echo Multidimensional Features

[0124] Statistical analysis of the multidimensional characteristics of the target echo is performed, including the number of target echo bright spots, intensity, time delay difference characteristics, target scale, and echo flicker characteristics, to obtain feature values ​​of the target model with statistical significance of azimuth fluctuations in typical azimuths. The specific method is as follows:

[0125] 3.1 Based on the target echo azimuth and its corresponding solid angle range, statistical modeling is performed on the characteristics of the number of bright spots and the horizontal scale of the target to obtain histograms of the distribution of the number of echo bright spots and the horizontal scale of the target in each azimuth.

[0126] Taking the echo data of the target model near the bow azimuth as an example for analysis, the mode feature is used as the statistical feature value of the number of target bright spots in the corresponding azimuth, such as... Figure 4 The figure shows the characteristics of the number of echo bright spots in each typical orientation obtained from statistical analysis.

[0127] 3.2 Based on the target echo azimuth and its corresponding solid angle range, statistical modeling and analysis are performed on the intensity characteristics of the echo bright spots generated by each substructure.

[0128] According to the probability density function formula (8) of the echo bright spot intensity, the statistical parameter k in the model is used to characterize the intensity of the azimuth fluctuation of the bright spot intensity, that is, the bright spot intensity probability distribution model - EH distribution model. Figure 3 The results are statistical results of the intensity characteristics of the bright spots. Each echo bright spot is generated by the substructure corresponding to the underwater vehicle.

[0129] 3.3 Based on the target echo azimuth and its corresponding solid angle range, the Parzen window estimation theory is used to perform statistical modeling and analysis on the time delay difference characteristics of the echo bright spots generated by each substructure.

[0130] Based on the probability density function formula (14) of the relative time delay difference characteristics of the bright spot, taking the echo data of the target model near the bow azimuth as an example for analysis, the Parzen window estimation method can be used to obtain the following results: Figure 5 The statistical distribution results of the relative time delay difference characteristics of the bright spots are shown.

[0131] 3.4 To address the azimuth fluctuation characteristics of target echoes, a target echo scintillation feature evaluation method is established, and the cross-correlation coefficient between echo signals at adjacent azimuth angles near typical azimuth angles is normalized. Figure 6 This represents the scintillation sensitivity of the target echo relative to the surrounding azimuth under typical azimuth conditions. The greater the scintillation fluctuation, the more it deviates from the value of 1.

[0132] Step 4: Joint Statistical Method Based on Copula Functions

[0133] Joint statistical analysis of multidimensional features of echo bright spots is performed from the perspective of multi-parameter joint statistical characteristics. Based on the echo bright spot features extracted from the echo bright spot fusion dataset, the Copula function is used to perform joint statistical modeling of the bright spot features, and statistical analysis is conducted on the correlation between the bright spot features.

[0134] Copula functions can be used to apply F(x1), ..., F(x) to N random variables belonging to arbitrary different distribution types. n ),…,F(x N The construction of the joint distribution function of multiple variables is unaffected by the distribution type of each random variable. Its probability density function c can be expressed as:

[0135]

[0136] The application verification results are as follows:

[0137] This invention achieves modeling of the correlation structure among multiple features of underwater target echoes, enabling flexible description of nonlinear and asymmetric dependencies between features and uncovering deep information that cannot be found through independent statistics. This provides unprecedented statistical basis for understanding complex acoustic scattering mechanisms and constructing more accurate target identification and simulation models. Based on the joint statistical model obtained by the method of this invention, the classification accuracy of underwater target automatic identification systems and the simulation fidelity (realism) of acoustic decoy / echo simulators can be significantly improved. It can generate simulated echoes with statistical characteristics closer to real targets and provide identification algorithms with more discriminative and information-rich joint feature vectors, thus producing positive and quantifiable technical effects in practical applications.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A joint statistical method for multi-dimensional features of underwater vehicle echo highlights based on the fusion of experimental and simulation data, characterized in that, It includes the following steps: Step 1. Acquisition of spatial azimuth echo dataset: A spherical coordinate system is established with the center of the underwater target as the origin, and the spatial azimuth (φ, θ) of sound wave incidence and reception is defined; through two complementary approaches of experimental measurement and simulation calculation, echo data within the spatial solid angle covered by the set typical spatial azimuth and its corresponding cone angle α are obtained respectively, and a spatial azimuth echo dataset is constructed. Step 2. Multidimensional feature extraction and feature fusion: The echo data is subjected to envelope processing to extract multidimensional features including the number of echo highlights N(θ), the intensity E(θ) of each highlight, the relative time delay difference τ(θ) between adjacent highlights, the target horizontal scale feature, and the echo scintillation feature based on the cross-correlation coefficient of adjacent azimuth echoes. And the same type of features extracted from experimental and simulation data are fused to form a multidimensional feature fusion dataset. Step 3. Analysis of statistical characteristics of echo multidimensional features: Based on the multidimensional feature fusion dataset obtained in Step 2, statistical modeling is performed on the fused multidimensional features, including: - Conducting distribution histogram statistics on the number of highlights and the target horizontal scale; - Using the EH distribution model to perform statistical modeling on the highlight intensity feature; - Using the Parzen window kernel density estimation method to perform statistical modeling on the non-linear distribution of the relative time delay difference feature of highlights; - Statistically evaluating the echo scintillation feature by calculating the normalized cross-correlation coefficient; Step 4. Joint statistical method based on Copula function: The Copula function is used to perform joint distribution modeling on the multidimensional features, analyze the correlation structure between each feature variable, and complete the joint statistical analysis of the multidimensional features of echo highlights. The extraction of the number of echo highlights N(θ) in Step 2 specifically includes: performing amplitude normalization processing on the echo envelope signal; adaptively setting the highlight discrimination threshold using a method with the between-class variance and within-class variance of the highlight normalized amplitude as the judgment criterion; performing peak extraction based on the threshold, and statistically obtaining the number of echo highlights at each azimuth, and this quantity feature directly reflects the complexity of the target structure. The probability density function of the EH distribution model described in step 3 is , where is the mean value of the intensity of the i-th bright spot, and the value of k characterizes the degree of azimuthal fluctuation of the bright spot intensity characteristics; The use of the Parzen window kernel density estimation method to perform statistical modeling on the relative time delay difference feature of highlights in Step 3, and its probability density estimation expression is: Wherein, N represents the sample number of the relative time delay difference of the bright spot in the azimuth angle range, it is assumed that K data in N data fall into the time delay difference data range of region R, and V represents the data space of region R, τ i The relative time delay difference sample value of the i th bright spot is represented, and the probability density function curve of the relative time delay difference feature of the bright spot can be obtained according to the above formula, so as to realize statistical modeling of the feature. The echo data obtained by the test measurement in step (1) specifically includes: arranging receiving hydrophones at a set typical spatial orientation, obtaining echo data of the transmitting and receiving combination, and performing band-pass filtering and matched filtering processing on the echo data.

2. The joint statistical method for multi-dimensional features of underwater vehicle echo highlights based on the fusion of test and simulation data according to claim 1, wherein In Step (1), the acquisition of echo data through simulation calculation specifically includes: based on the plate element method, uniformly sampling a large number of spatial azimuth points on the far-field spherical surface, calculating its acoustic scattering transfer function, generating an omnidirectional time-domain echo signal through the frequency-domain indirect method, and classifying it according to the solid angle corresponding to the typical spatial azimuth.

3. The joint statistical method for multi-dimensional features of underwater vehicle echo highlights based on the fusion of experimental and simulation data according to claim 1, wherein The extraction of the intensity E(θ) of each highlight in Step (2) specifically includes: calculating the intensity contribution value of each highlight from the perspective of energy distribution based on the target strength value of the target at each azimuth and the highlight normalized amplitude distribution, and this intensity feature is directly related to the scattering ability of the target sub-structure.

4. The joint statistical method for multi-dimensional features of underwater vehicle echo highlights based on the fusion of test and simulation data according to claim 1, wherein The extraction of the echo scintillation feature in Step (2) specifically includes: selecting adjacent angle echo signal data within a small azimuth angle range in the time domain; calculating the normalized cross-correlation coefficient between two adjacent signals, and the cross-correlation coefficient calculation formula is: where, θ1 and θ2 respectively represent two adjacent azimuth angle values, and respectively represent echo signal data at two azimuth angles, represents the cross-correlation function of two groups of signal data, and respectively represent the values of the autocorrelation functions of two groups of signal data at zero delay. The closer the value is to 0, the stronger the scintillation of the echo highlights at the corresponding azimuth.

Images