Active sonar imaging method based on complex approximate message passing

Abstract

The invention provides an active sonar imaging method based on complex approximate message passing, and the method comprises the steps: carrying out the matched filtering of a multi-channel echo signal received by an active sonar, and obtaining a pulse pressure signal; segmenting the pulse pressure signal according to a preset distance, and performing Fourier transform processing on each segment of pulse pressure signal; carrying out azimuth estimation on the signal after Fourier transform processing by using a complex approximate message passing algorithm to obtain azimuth information at a corresponding distance; and all azimuth information is arranged together according to the distance to obtain a high-resolution active sonar imaging picture. According to the method, the real value is improved through the complex approximate message passing algorithm so as to be better applied to the field of underwater active sonar target detection, and high calculation complexity caused by matrix inversion in a traditional method is avoided; compared with a traditional imaging method, the target resolution is higher, reverberation is suppressed to a certain extent, and robustness and detection precision are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of sonar imaging, and in particular to an active sonar imaging method based on complex approximate message passing. Background Art

[0002] Active sonar is mainly used for the detection, positioning, tracking and identification of small targets at long distances underwater. The main problems faced by small target active detection sonar in shallow water environments are the strong clutter and false alarms caused by a large amount of reverberation interference (seabed, sea surface, water body scattering, aquatic organisms, etc.). This is also the focus and difficulty of research in this field. In order to suppress reverberation interference to a certain extent from the perspective of underwater acoustic signal processing, the imaging algorithm of the sonar array is an important technical means.

[0003] The beamforming algorithm is the core of the active sonar imaging method, which is used to enhance the echo signal of weak underwater targets and estimate the azimuth of the targets. Due to the dual constraints of the detection distance and transducer technology on the sonar array, the operating frequency of active sonar cannot be too high and the size of the sonar array is small. Therefore, if the conventional beamforming method is used, the spatial gain of the array is limited and the target azimuth resolution is low. Increasing the signal transmission frequency and the physical aperture of the receiving array can both improve the spatial resolution of the target, but this is also unrealistic in actual engineering. Therefore, establishing a high-resolution active sonar imaging method is of great significance in the field of sonar detection. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides an active sonar imaging method based on complex approximate message passing, which can improve the target resolution and reduce the reverberation intensity.

[0005] The technical solution of the present invention is as follows: An active sonar imaging method based on complex approximate message passing, comprising the following steps:

[0006] S1) Performing matched filtering processing on the multi-channel echo signals received by the active sonar to obtain pulse compression signals;

[0007] S2) Segmenting the pulse compression signals according to a preset distance, and performing Fourier transform processing on each segment of the pulse compression signals;

[0008] S3) Performing azimuth estimation on the signals after Fourier transform processing by using the complex approximate message passing algorithm to obtain azimuth information at corresponding distances;

[0009] S4) Arranging all the azimuth information together according to the distance to obtain a high-resolution active sonar imaging diagram.

[0010] Preferably, in step S1), a matched filter is used to perform matched filtering processing on the multi-channel echo signals, specifically:

[0011]

[0012] Wherein, p(t) represents the pulse pressure signal; y(t) represents the multi-channel echo signal; s(t) represents the transmitted signal; - represents the conjugate operator; * represents the convolution operator; t represents the time index.

[0013] Preferably, in step S2), the pulse pressure signal is segmented according to a preset distance r, and Fourier transform processing is performed on each segmented signal, specifically:

[0014] P(r,f) = fft(p(r,t)); (2)

[0015] Wherein, p(r,f) represents the frequency-domain pulse pressure signal at distance r; P(r,t) represents the time-domain pulse pressure signal at distance r; fft(t) represents the fast Fourier transform; f represents the frequency index.

[0016] Preferably, in step S3), the azimuth estimation is performed on the signal after Fourier transform processing by using the complex approximate message passing algorithm, which specifically includes the following steps:

[0017] S31), Model the frequency-domain pulse pressure signal at distance r as:

[0018] p(r,f) = Hq(r,f) + n(f); (3)

[0019] Wherein, p(r,f) represents the frequency-domain multi-channel pulse pressure signal with dimension M×1; H represents the measurement matrix with dimension M×N; q(r,f) represents the sparse target signal to be reconstructed with dimension N×1; n(f) represents the additive Gaussian white noise with dimension N×1. M represents the number of receiving array channels, and N represents the number of spatial azimuth angle divisions.

[0020] S32), Initialize the target signal q (0) = 0 and the residual term z (0) = p(r,f), the iteration step t = 0, the stop iteration error ε, and the regularization parameter λ;

[0021] S33), Calculate the threshold parameter θ of the soft threshold function (t) :

[0022]

[0023] Wherein, med() is the operator for taking the median of a vector; H H represents the conjugate transpose of the measurement matrix; z (t-1) represents the residual term at time t-1; q (t-1) represents the sparse target signal to be reconstructed at time t-1;

[0024] S34), update the target signal q (t) :

[0025] q (t) = μ(H H z (t-1) + q (t-1) ; θ (t) ); (5)

[0026] where μ is the complex soft threshold function;

[0027] S35), update the residual term z (t) :

[0028]

[0029] where δ = M / N represents the measurement rate; <> is the mean operator; is the real part μ of the complex soft threshold function R for the real part x of the input R take the partial derivative; is the imaginary part μ of the complex soft threshold function I for the imaginary part x of the input I take the partial derivative;

[0030] S36), repeat steps S33) - S35) until the error of the reconstructed target signal is less than the stop iteration error ε, and then output the target signal q (t) .

[0031] Preferably, in step S4), after reconstructing the target for each segmented signal by the complex approximate message passing algorithm, they are arranged in distance order to obtain the active sonar imaging map.

[0032] The beneficial effects of the present invention are as follows:

[0033] 1. The present invention improves the real value through the complex approximate message passing algorithm to better apply in the field of underwater active sonar target detection, avoiding the high computational complexity brought by matrix inversion in traditional methods;

[0034] 2. Compared with traditional imaging methods, the present invention has higher target resolution, suppresses reverberation to a certain extent, and improves robustness and detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a schematic flow chart of the method of the present invention;

[0036] Figure 2 is the imaging map of the shallow sea environment in the embodiment of the present invention; where (a) is the imaging schematic diagram of the traditional method; (b) is the imaging map of the method of this embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0037] The specific embodiments of the present invention will be further described below in conjunction with the accompanying drawings:

[0038] As Figure 1 shown, this embodiment provides an active sonar imaging method based on complex approximate message passing, including the following steps:

[0039] S1), perform matched filtering on the multi-channel echo signals received by the active sonar to obtain the pulse compression signal; specifically:

[0040]

[0041] where p(t) represents the pulse compression signal; y(t) represents the multi-channel echo signal; s(t) represents the transmitted signal; - represents the conjugate operator; * represents the convolution operator; t represents the time index.

[0042] S2), segment the pulse compression signal according to a preset distance, and perform Fourier transform processing on each segment of the pulse compression signal;

[0043] In this embodiment, the specific Fourier transform processing of each segment of the pulse compression signal is as follows:

[0044] p(r,f) = fft(p(r,t)); (2)

[0045] where p(r,f) represents the frequency-domain pulse compression signal at distance r; P(r,t) represents the time-domain pulse compression signal at distance r; fft(t) represents the fast Fourier transform; f represents the frequency index.

[0046] S3), perform azimuth estimation on the signal after Fourier transform processing using the complex approximate message passing algorithm to obtain the azimuth information at the corresponding distance; specifically including the following steps:

[0047] S31), model the frequency-domain pulse compression signal at distance r as:

[0048] p(r,f) = Hq(r,f) + n(f); (3)

[0049] where p(r,f) represents the frequency-domain multi-channel pulse compression signal with dimension M×1; H represents the measurement matrix with dimension M×N; q(r,f) represents the sparse target signal to be reconstructed with dimension N×1; n(f) represents the additive white Gaussian noise with dimension N×1. M represents the number of receiving array channels, and N represents the number of spatial azimuth angle divisions.

[0050] S32), initialize the target signal q (0) = 0 and the residual term z (0) = p(r,f), the iteration step t = 0, the stop iteration error ε, and the regularization parameter λ;

[0051] S33), calculate the threshold parameter θ of the soft threshold function (t) :

[0052]

[0053] where med() is the median operator for the vector; H H denotes the conjugate transpose of the measurement matrix; z (t-1) denotes the residual term at time t - 1; q (t-1) denotes the sparse target signal to be reconstructed at time t - 1;

[0054] S34), update the target signal q (t) :

[0055] q (t) = μ(H H z (t-1) + q (t-1) ; θ (t) ); (5)

[0056] where μ is the complex soft threshold function;

[0057] S35), update the residual term z (t) :

[0058]

[0059] where δ = M / N represents the measurement rate; <> is the mean operator; is the real part μ of the complex soft threshold function R partial derivative with respect to the real part x of the input R ; is the imaginary part μ of the complex soft threshold function I partial derivative with respect to the imaginary part x of the input I ;

[0060] S36), repeat steps S33) - S35) until the error of the reconstructed target signal is less than the stop iteration error ε, then output the target signal q (t) .

[0061] S4), arrange all the azimuth information in distance order to obtain a high-resolution active sonar imaging map.

[0062] In this embodiment, after reconstructing the target for each segmented signal using the complex approximate message passing algorithm, they are arranged in distance order to obtain the active sonar imaging map.

[0063] As Figure 2 shown, this embodiment uses the method of this embodiment and the traditional method to image a certain shallow sea environment, and the target is circled with a red frame. AsFigure 2 As shown in (a) of Figure 2 As shown in (b) of , the imaging method of this embodiment can suppress most of the reverberation, the imaging diagram is cleaner, and the main lobe width is small, and the target resolution is high, which is beneficial to subsequent target detection and tracking based on image processing.

[0064] The above embodiments and descriptions in the specification only illustrate the principles and the best embodiments of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. An active sonar imaging method based on complex approximate message passing, characterized in that Including the following steps: S1), perform matched filtering on the multi-channel echo signals received by the active sonar to obtain a pulse compression signal; S2), segment the pulse compression signal according to a preset distance, and perform Fourier transform processing on each segment of the pulse compression signal; S3), perform azimuth estimation on the signal after Fourier transform processing using the complex approximate message passing algorithm to obtain the azimuth information at the corresponding distance; S4), arrange all the azimuth information in order of distance to obtain a high-resolution active sonar imaging map.

2. The active sonar imaging method based on complex approximate message passing according to claim 1, wherein: In step S1), a matched filter is used to perform matched filtering on the multi-channel echo signals.

3. The active sonar imaging method based on complex approximate message passing according to claim 1, wherein: In step S1), the matched filtering of the multi-channel echo signals is specifically as follows: Wherein, p(t) represents the pulse pressure signal; y(t) represents the multi-channel echo signal; s(t) represents the transmitted signal; - represents the conjugate operator; * represents the convolution operator; t represents the time index.

4. A method for active sonar imaging based on complex approximate message passing according to claim 1, characterized in that: In step S2), the Fourier transform processing of each segmented signal is specifically as follows: p(r,f) = fft(p(r,t)); (2) In the formula, p(r,f) represents the frequency-domain pulse compression signal at distance r; p(r,t) represents the time-domain pulse compression signal at distance r; fft(t) represents the fast Fourier transform; f represents the frequency index.

5. A method for active sonar imaging based on complex approximate message passing according to claim 1, characterized in that: In step S3), the azimuth estimation of the signal after Fourier transform processing using the complex approximate message passing algorithm specifically includes the following steps: S31), model the frequency-domain pulse compression signal at distance r as: p(r,f) = Hq(r,f) + n(f); (3) In the formula, p(r,f) represents the frequency-domain multi-channel pulse compression signal with dimension M×1; H represents the measurement matrix with dimension M×N; q(r,f) represents the sparse target signal to be reconstructed with dimension N×1; n(f) represents the additive Gaussian white noise with dimension N×1. M represents the number of receiving array channels, and N represents the number of spatial azimuth angle divisions. S32), Initialize the target signal q (0) = 0 and the residual term z (0) = p(r, f), iteration step t = 0, stopping iteration error ε, regularization parameter λ; S33), calculating the threshold parameter θ of the soft threshold function (t) : where med() is the median operator for the vector; H H represents the conjugate transpose of the measurement matrix; z (t-1) represents the residual term at time t-1; q (t-1) represents the sparse target signal to be reconstructed at time t-1; S34), update the target signal q (t) : q (t) = μ(H H z (t-1) + q (t-1) ; θ (t) )); (5) In the formula, μ is the complex soft threshold function; S35), update the residual term z (t) : where δ = M / N represents the measurement rate; <> is the mean operator; is the real part μ of the complex soft threshold function R for the real part x of the input R take the partial derivative; is the imaginary part μ of the complex soft threshold function I for the imaginary part x of the input I take the partial derivative; S36), repeat steps S33) - S35) until the error of the reconstructed target signal is less than the stop iteration error ε, and then output the target signal q (t) .

6. A method for active sonar imaging based on complex approximate message passing according to claim 1, characterized in that: Step S4), arrange all the azimuth information in order of distance to obtain a high-resolution active sonar imaging map.

Images