Multi-beam sonar image real-time generation method based on time delay beam forming

Abstract

The invention discloses a multi-beam sonar image real-time generation method based on time delay beam forming, a real-time generation method of a multi-beam sonar original image is designed based on accurate time delay beam forming, and a sonar original image with high distance resolution is obtained by performing matched filtering operation on each path of beam data.

Description

Technical Field

[0001] The invention relates to a method for real-time generation of multi-beam sonar images based on time-delay beamforming, and belongs to the technical field of signal processing. Background Art

[0002] Multi-beam sonar uses a series of signal processing technologies to achieve clear imaging in turbid underwater environments. However, due to the interference of the electronic circuits of the device itself and the noise and reverberation in the underwater environment, the acquired acoustic images are still interfered by complex noise. Compared with optical images, acoustic images have lower resolution, and the separation between the target and the background is not obvious, making it difficult to intuitively obtain the characteristic information of the target. Summary of the invention

[0003] Purpose of the invention: In view of the problems and shortcomings in the prior art, the present invention provides a method for real-time generation of multi-beam sonar images based on time-delay beamforming. Based on precise time-delay beamforming, a method for real-time generation of multi-beam sonar original images is designed. By performing matched filtering operation on each beam data, a sonar original image with high distance resolution is obtained.

[0004] Technical solution: A real-time generation method of multi-beam sonar images based on time-delay beamforming, including three steps: high-speed sampling, time-domain beamforming, and matched filtering.

[0005] Multi-beam sonar systems usually use beamforming and matched filtering to obtain the azimuth and distance information of the target object. After obtaining the azimuth and distance information, the signal of each array element is delayed and compensated to align them in the time domain so that they can be coherently superimposed. Therefore, each receiving channel must open a data buffer to store the previous sampling data.

[0006] In time domain beamforming, the beamforming angle is related to the delay of the echo signal reaching each array element. In order to achieve beamforming, it is necessary to compensate the delay of the signal of each array element to align them in the time domain so that they can be coherently superimposed. Therefore, each receiving channel must open a data buffer to store the previous sampling data.

[0007] Since the delay of each channel varies according to the beam pointing angle, the number of sampling points required to be stored also varies according to the beam angle. Under far-field conditions, for a linear array with N spacing d, the maximum number of sampling data points required to be stored is:

[0008] M max =f s ·τ max =f s (N-1)dsinθ max(1)

[0009] Where f s is the sampling frequency, τ max is the maximum delay, θ max is the maximum beam opening angle. In order to ensure that all targets within the field of view can be observed, the sampling data buffer of all channels must have at least M max size.

[0010] The frame length of the matched filter is directly related to the length of the matched signal. When the pulse width of the transmitted signal is T and the sampling frequency is f s When the frame length of the matched filter is Tf s The frame interval directly affects the distance resolution of the system. The relationship between distance resolution and frame interval is:

[0011]

[0012] In the formula, c is the speed of sound, Δs is the distance resolution, and Δn is the frame interval. The meaning of the frame interval is how many sampling points are used to perform a matched filter operation. The shorter the frame interval, the higher the distance resolution of the system, and the higher the speed requirement for the matched filter operation.

[0013] Matching means that the filter can best correlate with the input signal. The input or echo signal of the filter can be expressed as:

[0014] s i (t) = s(t) + n i (t) (3)

[0015] Where s(t) is the filter input signal, n i (t) is the input noise. The output of the filter can be expressed as:

[0016] y(t)=s o (t)+n o (t) (4)

[0017] According to the linear time-invariant theory, we have

[0018] s o (t)=s(t)*h(t) (5)

[0019] n o (t) = n i (t)*h(t) (6)

[0020] In the formula, * represents the convolution operation, and h(t) is the impulse response of the optimal filter. According to the inverse Fourier transform formula, formula (5) can be written as follows at t = t0:

[0021]

[0022] Where S(w) and s(t), H(w) and h(t) are Fourier transform pairs. The instantaneous power E of the signal at t = t0 is s For o The square of the modulus of (t0):

[0023]

[0024] Assuming that the input noise is a stationary random process, the noise power spectral density of the filter output is

[0025]

[0026] Where N i (w) is the filter input noise n i The Fourier transform of (t), the instantaneous signal-to-noise ratio at time t0 is:

[0027]

[0028] By Schwarz inequality

[0029] |∫a(ω)b(ω)dω| 2 ≤∫|a(ω)| 2 dω∫|b(ω)| 2 dω (11)

[0030] If and only if a(w)=Kb * (w), equation (2.34) can be equal, and K is an arbitrary constant. Substituting Schwarz inequality into equation (2.33), and assuming

[0031]

[0032] It can be obtained that when H(w) satisfies

[0033]

[0034] The signal-to-noise ratio has a maximum value

[0035]

[0036] When the input noise is Gaussian white noise, N i (ω) is a constant N0 / 2, SNR max =2E / N0, where E is the energy of the input signal

[0037]

[0038] The above proves that under Gaussian white noise, the frequency response of the linear filter that maximizes the output signal-to-noise ratio is

[0039]

[0040] The corresponding filter impulse response is

[0041] h(t)=Ks * (t0-t) (18)

[0042] Substituting equation (18) into equation (5), we can obtain that the input or echo signal of the filter is:

[0043]

[0044] The result of matched filtering is the correlation function of the input signal. The correlation peak is obtained at t=t0. For each frame, the real-time processing of multi-beam matched filtering requires the calculation of the multiplication and accumulation of the data points in the frame and the matching signal data points, and uses it as the correlation value of the current point. In order to ensure real-time performance, it should be at least Δn / f s within the time limit, complete b n Tf s times multiplication and b n (Tf s -1) additions, where b n is the number of beams.

[0045] Therefore, in the design of the frame interval of the matched filter, the multi-beam multiplication and accumulation operations are parallelized in the FPGA to improve the calculation speed and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is the basic flow chart of original image generation;

[0047] Figure 2 is a schematic diagram of the time domain beamforming pipeline buffer;

[0048] Figure 3 It is a schematic diagram of matched filter frame length and frame step;

[0049] Figure 4 The beam history diagram obtained by single-frequency pulse time-domain beamforming, (a) beam history diagram, (b) history diagram after matched filtering;

[0050] Figure 5 Beam history diagram obtained by linear frequency modulation pulse time-domain beamforming, (a) beam history diagram, (b) history diagram after matched filtering. DETAILED DESCRIPTION

[0051] The present invention is further explained below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0052] A real-time generation method of multi-beam sonar images based on time-delay beamforming. The basic process of generating original images by multi-beam sonar is as follows: Figure 1 As shown. Multi-beam sonar systems usually use beamforming and matched filtering to obtain the azimuth and distance information of target objects. The azimuth information is reflected by the beam strength of beams at different angles, and the distance information is obtained by converting the time difference between the echo signal and the transmitted signal into distance. By matching the sound beams at multiple angles, the resolution of the distance dimension can be improved, while suppressing noise and enhancing the accuracy and reliability of the image. After obtaining the azimuth and distance information, the multi-beam sonar system can process and integrate these data to generate the original sonar image.

[0053] In time domain beamforming, the beamforming angle is related to the delay of the echo signal reaching each array element. In order to achieve beamforming, it is necessary to compensate the delay of the signal of each array element to align them in the time domain so that they can be coherently superimposed. Therefore, each receiving channel must open a data buffer to store the previous sampling data, such as Figure 2 shown.

[0054] exist Figure 2 In the figure, the shaded area represents the buffer size required by each channel during the beamforming operation. Since the delay of each channel varies according to the beam pointing angle, the number of sampling points required to be stored also varies according to the beam angle. Under far-field conditions, for a linear array with N spacing d, the maximum number of sampling data points required to be stored is:

[0055] M max =f s ·τ max =f s (N-1)dsinθ max (1)

[0056] Where f s is the sampling frequency, τ max is the maximum delay, θ max is the maximum beam opening angle. In order to ensure that all targets within the field of view can be observed, the sampling data buffer of all channels must have at least M max size.

[0057] In order to ensure the real-time performance of matched filter calculation, it is very important to reasonably design the frame length and frame interval. The design of frame length and frame interval involves a trade-off in the spatial and temporal dimensions, especially for the spatial resolution and real-time requirements in the distance dimension. The schematic diagram of frame length and frame interval is shown in the figure below. Figure 3 shown.

[0058] Generally speaking, the frame length of matched filtering is directly related to the length of the matched signal. For example, when the pulse width of the transmitted signal is T and the sampling frequency is f s When the frame length of the matched filter is Tf s The frame interval directly affects the distance resolution of the system. The relationship between distance resolution and frame interval is:

[0059]

[0060] In the formula, c is the speed of sound, Δs is the distance resolution, and Δn is the frame interval. The meaning of the frame interval is how many sampling points are used to perform a matched filter operation. The shorter the frame interval, the higher the distance resolution of the system, and the higher the speed requirement for the matched filter operation.

[0061] The input or echo signal to the filter is:

[0062]

[0063] It can be seen that the result of matched filtering is the correlation function of the input signal, and the correlation peak is obtained at t=t0. Therefore, under the background of white noise, the essence of matched filtering is a cross-correlator that can perform cross-correlation operations. For each frame, the real-time processing of multi-beam matched filtering requires calculating the multiplication and accumulation of the data points in the frame and the matching signal data points, and using it as the correlation value of the current point. In order to ensure real-time performance, it should be at least Δn / f s within the time limit, complete b n Tf s multiplications and bn(Tfs-1) additions, where b n is the number of beams.

[0064] Therefore, in the design of the frame interval of the matched filter, it is necessary to consider the real-time requirements and weigh the performance and time overhead of signal processing. Through reasonable algorithm design, the multi-beam multiplication and accumulation operations can be parallelized in the FPGA to improve the calculation speed and efficiency.

[0065] The original two-dimensional sonar image is simulated by using the precise time-delay beamforming method to pre-form multiple beams. The simulation conditions are as follows: the transmission signal is in the form of a single-frequency rectangular pulse with a pulse width T of 1ms, a carrier frequency f0 of 400kHz, and a sampling frequency f sThe simulation uses a 96-element linear array, and the target is set at 10°, 3m away from the array and meets the far-field condition. The signal-to-noise ratio is -6dB. The beam-related parameters are:

[0066] Number of beams: 200

[0067] Field of view: ±50°

[0068] Main lobe width: 1.0°

[0069] Array element spacing: 2.2mm

[0070] Fractional delay filter order: 16

[0071] The beam history diagram obtained by simulation is as follows Figure 4 shown.

[0072] according to Figure 4 (a), we can see that there is a bright line perpendicular to the azimuth axis at 10°, indicating that the target is located at 10° horizontally. However, in the noise background, there are a lot of noise points in the history map, and the quality of the original image is not high. Figure 4 (b) is the two-dimensional acoustic image obtained after the matched filter. It can be seen that the signal energy is concentrated at about 3m, thereby improving the distance resolution of the target. At the same time, the signal-to-noise ratio is also significantly improved.

[0073] The single-frequency signal is changed to a linear frequency modulation pulse with a signal bandwidth of 200kHz. The other parameters are the same as above. The obtained two-dimensional acoustic image is as follows: Figure 5 shown.

[0074] contrast Figure 5 (a) and Figure 5 (b) It can be found that matched filtering not only improves the signal-to-noise ratio for linear frequency modulation pulses, but also compresses the pulse width to a great extent, thereby improving the distance resolution.

Claims

1. A method for real-time generation of multi-beam sonar images based on time-delay beamforming, characterized in that: Includes high-speed sampling, time-domain beamforming, and matched filtering; Beamforming and matched filtering are used to obtain the azimuth and distance information of the target object. After obtaining the azimuth and distance information, the signal of each array element is delayed compensated to align them in the time domain so that they can be coherently superimposed; each receiving channel must open a data buffer to store the previous sampling data.

2. The method for real-time generation of multi-beam sonar images based on time-delay beamforming according to claim 1, characterized in that: In time-domain beamforming, since the delay of each channel varies according to the beam pointing angle, the number of sampling points required to be stored also varies according to the beam angle; under far-field conditions, for a linear array with N spacing d, the maximum number of sampling data points required to be stored is: M max =f s ·t max =f s (N-1)dsinθ max Where f s is the sampling frequency, τ max is the maximum delay, θ max is the maximum beam angle; in order to ensure that all targets within the field of view can be observed, the sampling data buffer of all channels must have at least M max size; The frame length of the matched filter is directly related to the length of the matched signal. When the pulse width of the transmitted signal is T and the sampling frequency is f s When the frame length of the matched filter is Tf s ; The frame interval directly affects the distance resolution of the system; the relationship between distance resolution and frame interval is: Where c is the speed of sound, Δs is the range resolution, and Δn is the frame interval.

3. The method for real-time generation of multi-beam sonar images based on time-delay beamforming according to claim 1, characterized in that: The input or echo signal to the filter is: The result of matched filtering is the correlation function of the input signal. The correlation peak is obtained at t=t0. For each frame, the real-time processing of multi-beam matched filtering requires the calculation of the multiplication and accumulation of the data points in the frame and the matching signal data points, and uses it as the correlation value of the current point. In order to ensure real-time performance, it should be at least Δn / f s within the time limit, complete b n Tf s multiplications and bn(Tfs-1) additions, where b n is the number of beams; In the design of the frame interval of the matched filter, the multi-beam multiplication and accumulation operations are parallelized in the FPGA.

Images