Passive sonar target detection and track tracing method based on CLEAN algorithm

Abstract

The present invention relates to a passive sonar target detection and track tracing method based on the CLEAN algorithm. The output results of the SAKT algorithm are whitened, and a reference azimuth history diagram is generated in real time based on the position of the maximum response value. The reference azimuth history diagram is subjected to the same scale transformation so that the amplitude and phase between the real-time generated reference response and the observation result maintain a linear relationship. The slow-frequency Fourier inverse transform of the observation result diagram is transformed into the slow-time dimension. The reference response is used as the measurement atom in the corrected azimuth history spectrum, and azimuth-by-azimuth cancellation is performed using the matching pursuit algorithm. The above process is repeated, and the impulse response distribution matrix is obtained through multiple iterations. Target detection and track tracing are achieved based on the response parameters. The present invention is based on the CLEAN algorithm to achieve target detection and track tracing.

Description

Technical Field

[0001] The present invention relates to the field of sonar signal processing, and in particular to a passive sonar target detection and track tracing method based on the CLEAN algorithm. Background Art

[0002] The detection principle of passive sonar relies on the noise radiated by the target during navigation. Currently, due to the development of integrated noise reduction technologies, the stealth of submarines is becoming increasingly improved, and the detection capabilities of conventional passive sonar are becoming increasingly limited. To counter underwater threats, the capabilities of passive sonar in underwater acoustic warfare can be improved at two levels: hardware and signal processing algorithms. The hardware level primarily involves improving hydrophone sensitivity, reducing self-noise levels, and expanding array apertures. The signal processing algorithm level primarily involves array signal processing techniques and time coherent integration technology. Long-term coherent integration technology efficiently accumulates the azimuth spectrum of energy radiated by a target over a period of time, achieving coherent gain and significantly improving sonar detection capabilities.

[0003] Conventional coherent integration technology is often used in radar signal detection and radar target imaging. For example, patent CN109507669A uses coherent integration to accurately estimate the parameters of weak ground moving targets; patent CN111257844A uses coherent integration gain to characterize the target's fluctuation characteristics. In passive sonar, since the target radiates noise signals and the conventional output is an energy spectrum, direct coherent integration cannot be performed. However, after processing with the split array cross-correlation beamforming algorithm, the output azimuth spectrum is a low-pass complex signal. This type of azimuth spectrum can be time-coherently integrated (Split Array Keystone Transform, SAKT algorithm). During the coherent integration process, due to the truncation effect of the slow time dimension and energy leakage in the azimuth dimension, the two-dimensional response of the focused target has sidelobes. When the dynamic range of energy between sonar targets is large, the sidelobes of the focused response of the strong target will interfere with the detection and parameter estimation of the weak target.

[0004] The conventional processing method is to suppress sidelobe energy through windowing. However, after windowing, the amplitude of the output signal will be weakened, introducing a certain signal-to-noise ratio loss. Another method is deconvolution. Among these methods, the CLEAN algorithm is widely used, such as CN110568439A and CN109655802B. The CLEAN algorithm models the observation results as the convolution of the brightness distribution and the observation response. When the observation response is a known response, the observation result can be deconvolved with the observation response. The deconvolved observation result graph becomes a set of two-dimensional functions with different peak values, and the target can be detected faster and better from this result graph. However, because the SAKT algorithm performs a scale transformation in the slow time dimension, the two-dimensional spectrum of the observation result graph and the observation response are no longer a simple convolution relationship, so the two-dimensional impulse response result can no longer be obtained through simple deconvolution. Summary of the Invention

[0005] To solve the existing technical problems, the present invention provides a passive sonar target detection and track tracing method based on the CLEAN algorithm.

[0006] The specific contents of the present invention are as follows: a passive sonar target detection and track tracing method based on the CLEAN algorithm, which first whitens the output results of the SAKT algorithm to eliminate the influence of the phase factor of the sonar target radiation noise, and generates a reference azimuth history diagram in real time according to the position of the maximum response value. The reference azimuth history diagram is then subjected to the same scale transformation so that the amplitude and phase between the real-time generated reference response and the observation result maintain a linear relationship. In addition, considering that the track of the target in the corrected azimuth history spectrum is a straight line along a specific azimuth, the slow frequency Fourier inverse transform of the observation result diagram is converted to a slow time dimension, and the reference response is used as the measurement atom in the corrected azimuth history spectrum, and the azimuths are offset one by one by the matching pursuit algorithm. The above process is repeated, and the impulse response distribution matrix can be obtained through multiple iterations. The target detection and track tracing can be achieved according to the response parameters.

[0007] The specific implementation steps are:

[0008] 1) First, the frequency on the array at time τ is f c The received signal is split into two sub-arrays, each of which is aligned with the azimuth θ for conventional beamforming. Then, the beam outputs of the two sub-arrays are cross-correlated, and the phase of the cross-correlation is output (equivalent to whitening processing). The cross-correlation azimuth spectrum ψ(θ,τ) at time τ is obtained.

[0009] 2) Fourier transform ψ(θ,τ) along the azimuth dimension θ to obtain γ(ω,τ). The transformation pair is And perform Keystone transformation within the frequency band ω∈(0,F0] to correct the azimuth change, where L is the array aperture, c is the reference sound velocity during beamforming. The corrected γ'(ω,τ) is inversely transformed to ω to obtain the azimuth history spectrum ψ Key (θ,τ).

[0010] 3) The obtained ψ Key (θ, τ) is Fourier transformed along the observation time dimension to obtain the coherent cumulative energy spectrum Ψ(θ, f), and the transformation pair is

[0011] 4) Initialize T(θ,f)=0 to store the response sequence; Φ(θ,τ)=0 to store the target track.

[0012] 5) Execute the kth iteration. The coherent cumulative residual energy spectrum in the kth iteration is Ψ k (θ,f). Traverse Ψ k (θ,f) retrieves the direction θ corresponding to the maximum response amplitude k and angular frequency f k , and record T(θ k ,f k )=a k .

[0013] 6) According to the response direction θ k and angular frequency f k , combined with the array's steering vector to regenerate the split array cross-correlation azimuth process spectrum And record the track corresponding to the target energy response The azimuth frequency range is in right Perform Keystone transformation to obtain the reference azimuth history spectrum

[0014] 7) k (θ,f) is inversely transformed along the f dimension to the slow time dimension ψ Key,k (θ,τ), and As a reference, the Matching Pursuit (MP) algorithm is used to offset ψ along the azimuth dimension. Key,k The energy of the azimuth process in (θ,τ). in is the energy normalization factor.

[0015] 8) Change ψ Key,k+1 (θ,τ) is transformed along the slow time Fourier transform to obtain Ψ k+1 (θ,f).

[0016] Repeat steps 5) to 8) until the number of cycles or ψ is satisfied KeyThe maximum residual in (θ, τ) is less than the given threshold. After the loop is completed, T(θ, f) is the two-dimensional impulse response matrix after target coherent accumulation, and Φ(θ, τ) is the "clean" sonar target direction history map.

[0017] This invention applies sonar azimuth energy coherent accumulation spectra to achieve target detection, track extraction, and backtracking. If storage space permits, step 6) can be pre-calculated and stored. This avoids repeated generation of azimuth history spectra during the entire algorithm run, thereby improving algorithm execution efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The specific embodiments of the present invention will be further explained below with reference to the accompanying drawings.

[0019] Figure 1 This is a processing flow chart of the passive sonar target detection and track tracing method based on the CLEAN algorithm of this application;

[0020] Figure 2 This is a simulation diagram of the three-target orientation process;

[0021] Figure 3 is the corrected position history spectrum;

[0022] Figure 4 This is the SAKT focusing result diagram;

[0023] Figure 5 is the original focus response projection;

[0024] Figure 6 is the focus response projection after windowing;

[0025] Figure 7 Focus response after processing for this technical solution;

[0026] Figure 8 This is the backtracking trajectory of this technical solution. DETAILED DESCRIPTION

[0027] like Figure 1As shown, the passive sonar target detection and track tracing method based on the CLEAN algorithm of the present invention first whitens the output result of the SAKT algorithm to eliminate the influence of the phase factor of the sonar target radiation noise, and generates a reference azimuth history diagram in real time according to the position of the maximum response value. The reference azimuth history diagram is then subjected to the same scale transformation so that the amplitude and phase between the real-time generated reference response and the observation result maintain a linear relationship. In addition, considering that the track of the target in the corrected azimuth history spectrum is a straight line along a specific azimuth, the slow frequency Fourier inverse transform of the observation result diagram is converted to the slow time dimension, and the reference response is used as the measurement atom in the corrected azimuth history spectrum, and the matching pursuit algorithm is used to offset each azimuth in turn. Repeat the above process and iterate multiple times to obtain the impulse response distribution matrix, and the target detection and track tracing can be achieved according to the response parameters.

[0028] The specific steps are as follows:

[0029] 1) First, the frequency on the array at time τ is f c The received signal is split into two sub-arrays, each of which is aligned with the azimuth θ for conventional beamforming. Then, the beam outputs of the two sub-arrays are cross-correlated, and the phase of the cross-correlation is output (equivalent to whitening processing). The cross-correlation azimuth spectrum ψ(θ,τ) at time τ is obtained.

[0030] 2) Fourier transform ψ(θ,τ) along the azimuth dimension θ to obtain γ(ω,τ). The transformation pair is And perform Keystone transformation within the frequency band ω∈(0,F0] to correct the azimuth change, where L is the array aperture, c is the reference sound velocity during beamforming. The corrected γ'(ω,τ) is inversely transformed to ω to obtain the azimuth history spectrum ψ Key (θ,τ).

[0031] 3) The obtained ψ Key (θ, τ) is Fourier transformed along the observation time dimension to obtain the coherent cumulative energy spectrum Ψ(θ, f), and the transformation pair is

[0032] 4) Initialize T(θ,f)=0 to store the response sequence; Φ(θ,τ)=0 to store the target track.

[0033] 5) Execute the kth iteration. The coherent cumulative residual energy spectrum in the kth iteration is Ψ k (θ,f). Traverse Ψ k (θ,f) retrieves the direction θ corresponding to the maximum response amplitude k and angular frequency f k , and record T(θ k ,f k )=ak .

[0034] 6) According to the response direction θ k and angular frequency f k , combined with the array's steering vector to regenerate the split array cross-correlation azimuth process spectrum And record the track corresponding to the target energy response The azimuth frequency range is in right Perform Keystone transformation to obtain the reference azimuth history spectrum

[0035] 7) k (θ,f) is inversely transformed along the f dimension to the slow time dimension ψ Key,k (θ,τ), and As a reference, the Matching Pursuit (MP) algorithm is used to offset ψ along the azimuth dimension. Key,k The energy of the azimuth process in (θ,τ). in is the energy normalization factor.

[0036] 8) Change ψ Key,k+1 (θ,τ) is transformed along the slow time Fourier transform to obtain Ψ k+1 (θ,f).

[0037] Repeat steps 5) to 8) until the number of cycles or ψ is satisfied Key The maximum residual in (θ, τ) is less than the given threshold. After the loop is completed, T(θ, f) is the two-dimensional impulse response matrix after target coherent accumulation, and Φ(θ, τ) is the "clean" sonar target direction history map.

[0038] The specific implementation scheme of the present invention is further illustrated by a simulation example.

[0039] The simulation lasts 100 seconds and contains three targets: Target 1's azimuth varies uniformly from -40° to -30°, with a signal strength of 1; Target 2's azimuth varies uniformly from -10° to 15°, with a signal strength of 20; and Target 3's azimuth varies from 20° to 15°, with a signal strength of 1. The receiving array is a 100-element uniform linear array with an element spacing of 1.5 meters. The simulation frequency is 500 Hz.

[0040] The simulation scenario is processed according to this technical solution as follows:

[0041] Execute step 1, divide the receiving array into two sub-arrays, perform conventional beamforming of 256 beams on each sub-array, and perform correlation processing on the outputs of the two sub-arrays and perform cross-spectral whitening to obtain the target's azimuth history as follows: Figure 2shown.

[0042] Execute step 2, first perform Fourier transform along the azimuth dimension to the azimuth frequency dimension, then quickly implement Keystone transform by CZT-IFFT algorithm in the frequency band (0,250], and inversely transform back to the azimuth dimension to obtain the azimuth change corrected azimuth history spectrum as shown in the figure below: Figure 3 shown.

[0043] Execute step 3, perform Fourier transform along the slow time dimension, complete the coherent accumulation of energy, and obtain the focusing azimuth spectrum Ψ(θ,f) as follows: Figure 4 As shown in Figure 2, due to the greater signal strength of target 2, the response of target 3 in the azimuth focused energy spectrum will be affected by the side lobe interference of target 2. The influence of the side lobe can be suppressed by adding a window. For example, the response result obtained by adding a Hanning window is as follows: Figure 6 Compared with the original focused energy spectrum ( Figure 5 ), the sidelobe effect is indeed improved after windowing, but the response amplitude is significantly reduced and the response width is also widened, which will reduce the target detection and resolution capabilities in applications.

[0044] Execute step 4 to initialize the focus response storage matrix T(θ, f) = 0 and the track storage matrix Φ(θ, τ) = 0.

[0045] Set the maximum number of loops K = 3 and iterate the following steps:

[0046] Execute step 5 and traverse the coherent cumulative spectrum Ψ k (θ,f), search for the current maximum response value and store T(θ k ,f k )=a k .

[0047] Execute step 6 to generate the azimuth history spectrum according to the response azimuth and angular frequency, and record the azimuth history. Then perform the same Keystone transformation on the generated azimuth history spectrum to obtain the reference response.

[0048] Execute step 7 and convert the coherent cumulative spectrum Ψ k (θ,f) is inversely transformed to the slow time dimension and expressed as To measure the atom, the MP algorithm is used to offset the energy in each direction to obtain ψ Key,k+1 (θ,τ).

[0049] Execute step 8 and set ψ Key,k+1 (θ,τ) is then transformed along the slow time dimension to the angular frequency dimension to obtain Ψ k+1 (θ,f), return to step 5.

[0050] After the iteration, the projection of the result of T(θ,f) on the slow frequency f axis is as follows: Figure 7 As shown, the response focusing matrix processed by this scheme is a two-dimensional δ response, which can basically obtain an ideal focusing response matrix without sidelobe interference, and compared with the original response intensity Figure 5 It can be seen that the target coherent cumulative response amplitude remains basically unchanged. The result of the stored track Φ(θ,τ) is as follows Figure 8 As shown, this method can achieve rapid backtracking of the target track. Compared with the detection-association-state filtering processing framework, this technical solution has better continuity and can avoid problems such as track misassociation during navigation and intersection.

[0051] Based on the CLEAN algorithm, the present invention applies the same transformation to the reference response to maintain amplitude consistency with the energy accumulation spectrum. The focused response spectrum is then transformed into an azimuth history spectrum. The corrected azimuth history spectrum is offset azimuthally using a matching pursuit algorithm. When the iteration termination threshold is reached, a response intensity matrix is output and the corresponding trajectory is calculated, enabling target detection and track tracing.

[0052] In the above description, many specific details are set forth in order to fully understand the present invention. However, the above description is only a preferred embodiment of the present invention. The present invention can be implemented in many other ways different from those described herein, so the present invention is not limited to the specific implementation disclosed above. At the same time, any person skilled in the art can make many possible changes and modifications to the technical solution of the present invention using the methods and technical contents disclosed above without departing from the scope of the technical solution of the present invention, or modify it into an equivalent embodiment of equivalent changes. Any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of protection of the technical solution of the present invention.

Claims

1. A passive sonar target detection and track tracing method based on the CLEAN algorithm, characterized by: include: S1, first whiten the output result of the SAKT algorithm, and generate a reference azimuth history map in real time according to the position of the maximum response value; S1, then perform the same scale transformation on the reference azimuth history diagram so that the amplitude and phase between the real-time generated reference response and the observation result maintain a linear relationship; S3, inverse Fourier transform of the slow frequency of the observation result image into the slow time dimension, and use the reference response as the measurement atom in the corrected azimuth history spectrum to cancel each other azimuthally through the matching pursuit algorithm; Repeat steps S2 to S3 for multiple iterations to obtain the impulse response distribution matrix, and detect the target and trace the track based on the response parameters; The SAKT algorithm processes the array and outputs a coherent cumulative energy spectrum; then it goes through the following steps: initialization , used to store the response sequence; , used to store target tracks; Execute Iteration; The coherent accumulated residual energy spectrum in the iteration is: ; traverse Retrieve the direction corresponding to the maximum response amplitude and angular frequency , and record ; According to the response direction and angular frequency , combined with the array's steering vector to regenerate the split array cross-correlation azimuth process spectrum , and record the track corresponding to the target energy response ; The azimuth frequency range is ,in ,right Perform Keystone transformation to obtain the reference azimuth history spectrum ; Will along Inverse transform from dimension to slow time dimension , and As a reference, the matching pursuit algorithm is used to offset the orientation dimension one by one The directional process energy in ,in , is the energy normalization factor; Will Along the slow time Fourier transform, we get ; Repeat the above steps until the number of cycles or The maximum residual value in is less than the given threshold. After the cycle ends, is the two-dimensional impulse response matrix after target coherent accumulation, It is a clean sonar target position history map.

2. The passive sonar target detection and track tracing method based on the CLEAN algorithm according to claim 1, characterized in that: SAKT algorithm processes arrays, including: First of all, The frequency of the array at the moment is The received signal is split into two sub-arrays, each of which is aligned with the azimuth. Perform conventional beamforming, then perform cross-correlation of the two sub-array beam outputs, output the phase of the cross-correlation, and obtain Cross-correlation azimuth spectrum at time ; Will Along the azimuth dimension Fourier transform gives , the transformation pair is ; and in the frequency band Perform Keystone transformation within the range to correct the azimuth change, where , is the array aperture, is the reference sound speed during beamforming; after correction right Inverse transform to obtain the azimuth process spectrum ; Will get Perform Fourier transform along the observation time dimension to obtain the coherent cumulative energy spectrum , the transformation pair is .

Images