Underwater sound image super-resolution enhancement method based on space-time two-dimensional deconvolution

Abstract

The invention relates to the technical field of underwater acoustic image processing, in particular to an underwater acoustic image super-resolution enhancement method based on space-time two-dimensional deconvolution, which comprises the following steps of: performing matched filtering on a simulation signal and a received signal, converting the simulation signal and the received signal into a frequency domain through FFT (Fast Fourier Transform), searching in an azimuth-frequency domain to generate a two-dimensional CBF (Customer Broadband Function) result, obtaining a time delay-azimuth two-dimensional spectrum through iFFT, and obtaining a time delay-azimuth two-dimensional spectrum through iFFT; and calculating two-dimensional deconvolution by using an R-L algorithm to obtain a high-resolution spectrum. The method can be combined with a circumferential synthetic aperture technology to be used for underwater target high-resolution imaging, and can also be applied to multi-beam squint sonar imaging. Lake tests and pool experiments verify that time delay and azimuth resolution can be effectively improved, side lobes are inhibited, high-quality image information is provided for underwater target detection and recognition, and the method has important application value.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater acoustic image processing, and specifically to an underwater acoustic image super-resolution enhancement method based on spatiotemporal two-dimensional deconvolution, which can be used to improve the resolution and clarity of underwater target imaging and provide more accurate image information for underwater target detection, identification and analysis. Background Art

[0002] In the field of active imaging of underwater acoustic targets, traditional matched filtering (MF) and conventional beamforming (CBF) methods have certain limitations. The temporal resolution of MF depends primarily on the signal bandwidth. When the signal bandwidth is limited, its temporal resolution cannot meet the requirements for high-precision detection of underwater targets. The spatial spectrum obtained by the CBF algorithm has a wide main lobe and high sidelobes, which leads to blurred target imaging, affecting accurate target identification and positioning, and making it difficult to meet the high-resolution imaging requirements of modern underwater detection. Therefore, a new method is needed to improve the resolution and quality of underwater acoustic images. Summary of the Invention

[0003] The present invention aims to provide a method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution. By breaking through the limitations of traditional imaging algorithms in terms of time delay and azimuth resolution, the method effectively suppresses side lobes, significantly improves the clarity and accuracy of underwater acoustic images, and provides more reliable image data support for the detection, identification and analysis of underwater targets.

[0004] The technical solution adopted by the present invention to solve the technical problem is: a method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution, comprising the following steps:

[0005] Perform matched filtering on the simulation signal and the received signal respectively;

[0006] Perform fast Fourier transform (FFT) operation on the result after matched filtering to convert it into frequency domain;

[0007] Search is performed in the azimuth-frequency domain, and beamforming technology is used to generate two-dimensional conventional beamforming (CBF) results;

[0008] Perform inverse fast Fourier transform (iFFT) on the power spectrum of a certain azimuth on the two-dimensional CBF to obtain a delay-azimuth two-dimensional spectrum. The delay-azimuth two-dimensional spectrum obtained by simulation signal processing is used as the point spread function (PSF), and the delay-azimuth two-dimensional spectrum obtained from the actual received signal is used as the output energy distribution.

[0009] The PSF and the two-dimensional deconvolution of the output energy distribution are calculated using the Richardson-Lucas (RL) algorithm to obtain a high-resolution time-delay-azimuth two-dimensional spectrum.

[0010] Specifically, when combining with the circular synthetic aperture technology to achieve high-resolution imaging of underwater targets, the following steps are also included:

[0011] The echoes received at 360 directions in the circular synthetic aperture are equivalently regarded as signals received by 360 primitives, and the element spacing is calculated based on the rotation radius and rotation angle;

[0012] The received signal is divided into several sub-arrays, each sub-array contains 20 directions, and adjacent sub-arrays overlap by 10 directions;

[0013] Construct a two-dimensional PSF based on the transmission signal parameters, array motion parameters and environmental parameters;

[0014] The delay-azimuth two-dimensional CBF of each subarray signal is estimated, and the two-dimensional CBF result is used to perform deconvolution operation with the two-dimensional PSF to obtain a high-resolution delay-azimuth spectrum.

[0015] Specifically, when applied to multi-beam squint sonar high-resolution imaging, the method further includes the following steps:

[0016] According to the operating frequency, number of primitives and primitive spacing parameters of the multi-beam squint sonar, the two-dimensional PSF is obtained by using the two-dimensional deconvolution of time and space.

[0017] The signals received by the multi-beam squint sonar are processed, and the delay-azimuth spectrum estimation results are obtained through two-dimensional CBF and two-dimensional space-time deconvolution.

[0018] Specifically, the transmission signal is a linear frequency modulation (LFM) signal, and its parameters include a starting frequency, a cutoff frequency, a pulse width, a repetition period, a starting phase, and a sampling rate.

[0019] Specifically, in the circular synthetic aperture technology, the sonar array rotates 360° around the target at a specific radius, transmits a signal every 1° of rotation, and receives an echo signal.

[0020] Specifically, the multi-beam squint sonar has an operating frequency of 1.2 MHz, contains 96 primitives, and has a primitive spacing of 1 mm.

[0021] Specifically, when calculating the two-dimensional deconvolution, the RL algorithm performs an iterative solution based on the two-dimensional convolution relationship of signal energy distribution in the deconvolution imaging principle.

[0022] Specifically, by combining the spatiotemporal two-dimensional deconvolution imaging algorithm with the circular synthetic aperture technology, the pulse width can be compressed by about 40 times and the azimuth resolution can be increased by 5 times.

[0023] Specifically, when the spatiotemporal two-dimensional deconvolution imaging algorithm is applied to multi-beam squint sonar imaging, it can simultaneously improve the resolution of time delay and azimuth, and suppress side lobes.

[0024] Beneficial effects of the present invention:

[0025] The present invention's underwater acoustic image super-resolution enhancement method based on spatiotemporal two-dimensional deconvolution has significant beneficial effects. Through spatiotemporal two-dimensional deconvolution processing, it effectively breaks through the resolution bottleneck of traditional imaging algorithms and significantly improves the quality of underwater acoustic images.

[0026] In the application scenario combined with circular synthetic aperture technology, compared with traditional imaging methods, the present invention can compress the pulse width by about 40 times and increase the azimuth resolution by 5 times, making the imaging of underwater targets clearer and more accurate, and being able to more accurately locate the position and outline of the target, providing more powerful support for the detection and identification of underwater targets.

[0027] This method also performs well in multi-beam squint sonar imaging applications, simultaneously improving both latency and azimuth resolution while effectively suppressing sidelobes. This enables multi-beam squint sonar to more clearly capture target information in complex underwater environments, reducing the impact of interference and noise, and improving underwater target detection and identification accuracy.

[0028] In summary, the method of the present invention provides clearer and more accurate image information for the detection, identification and analysis of underwater targets, and has important practical application value and broad market prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention will be further described below with reference to the accompanying drawings and examples.

[0030] Figure 1 It is the operation process of spatiotemporal two-dimensional deconvolution imaging;

[0031] The figure shows the complete processing process from the input of the simulated signal and the received signal, through matched filtering, FFT operation, azimuth-frequency domain search to generate the two-dimensional CBF result, then to obtain the delay-azimuth two-dimensional spectrum through iFFT, and finally using the RL algorithm to perform two-dimensional deconvolution to obtain the high-resolution delay-azimuth two-dimensional spectrum. It clearly presents the operating steps and data flow of the method of the present invention.

[0032] Figure 2 Schematic diagram of the division of circular synthetic aperture subarray;

[0033] This figure illustrates how, in a circular synthetic aperture, echoes received in 360 directions are converted into signals received by 360 primitives and divided into several subarrays, each containing 20 directions and with adjacent subarrays overlapping by 10 directions. This specific division helps us understand how data is processed and organized in circular synthetic aperture technology.

[0034] Figure 3 is the spatiotemporal two-dimensional deconvolution PSF;

[0035] This figure presents the two-dimensional PSF results constructed based on the transmission signal parameters, array motion parameters and environmental parameters. It intuitively shows the shape and characteristics of the PSF, provides a basic basis for the subsequent deconvolution calculation using the RL algorithm, and also reflects the imaging characteristics of the system.

[0036] Figure 4 Comparison of the delay-azimuth spectrum estimation results between 2D-CBF and dCv2;

[0037] This figure compares the results of the traditional 2D-CBF method and the spatiotemporal two-dimensional deconvolution method (dCv2) of the present invention in terms of delay-azimuth spectrum estimation, clearly showing the advantages of the method of the present invention in improving resolution and suppressing sidelobes. Through comparison, we can more intuitively see the improvement effect of the method of the present invention on the quality of underwater acoustic images.

[0038] Figure 5 Comparison of the estimation results of 2D-CBF and dCv2 in the time delay and azimuth domains;

[0039] The figure further compares and analyzes the estimation results of the 2D-CBF method and the dCv2 method from the two dimensions of time delay and azimuth, showing in more detail the improvement of resolution and suppression of sidelobes in different dimensions by the method of the present invention, which helps to deepen the understanding of the performance characteristics of the method of the present invention.

[0040] Figure 6 This is the PSF for the multi-beam squint sonar tank experiment;

[0041] This figure shows the two-dimensional PSF obtained by using spatiotemporal two-dimensional deconvolution in a multi-beam squint sonar pool experimental scenario, reflecting the imaging characteristics of the system in this experimental scenario and providing key data for deconvolution calculation and image enhancement in this scenario.

[0042] Figure 7 Comparison of the time delay-azimuth spectrum estimation results between 2D-CBF and dCv2 of squint sonar;

[0043] This figure compares the delay-azimuth spectrum estimation results of the 2D-CBF method and the dCv2 method in squint sonar imaging, once again proving that the method of the present invention can effectively improve the resolution and suppress sidelobes in multi-beam squint sonar imaging, providing more reliable image information for underwater target detection in this application scenario. DETAILED DESCRIPTION

[0044] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0045] like Figure 1-Figure 7 As shown, the present invention includes the following technical solutions:

[0046] Among them, the principle of spatiotemporal two-dimensional deconvolution imaging

[0047] The most commonly used algorithm for active target imaging first estimates the target's delay and azimuth parameters, then uses this information to locate the target, and finally image the target. Among algorithms for estimating target delay and azimuth, the most commonly used are matched filtering (MF) and conventional beamforming (CBF), but their resolution is often limited by the signal bandwidth and array aperture. To improve delay and azimuth resolution, a two-dimensional deconvolution algorithm, denoted as dCv2, can be used.

[0048] In an ideal noise-free environment, when the signal x(t) is input to the shift-invariant system h(t), the output y(t) can be expressed as The “*” represents a one-dimensional convolution operation. If the signal is described in terms of energy, then Here I y =|y| 2 , I 2 =|x| 2 , I h =|h| 2 , while I h It is the point spread function (PSF). With the help of Richardson-Lucas (RL) deconvolution algorithm, I can be solved by iterative method. x , and the solution is unique. The solution of the mth iteration is

[0049] For the CBF of an N-element uniform linear array, at frequency f, its beam response H(sin φ, f) has a convolution relationship with the sound source distribution X(sin φ, f);

[0050] Specifically, it can be expressed as H(si nφ,f)=N sin c(2πNdf sinφ / c)

[0051] and The convolution form of .

[0052] Where d represents the array element spacing, φ m and a m represent the direction of arrival and complex amplitude of the mth signal, respectively, and δ(·) is the one-dimensional Dirac function.

[0053] When the sound sources are uncorrelated or coherent, the CBF beam output energy can be expressed as the convolution of the beam pattern and the signal strength, which is: in is the PSF, and the M uncorrelated signals in the beam domain can be expressed as

[0054] In the actual scenario of active sensing, MF and CBF are usually used to estimate and display the sinφ-τ plane of the echo signal. However, the time resolution of MF is limited by the signal bandwidth, and the spatial spectrum obtained by the CBF algorithm has the problem of wide main lobe and high side lobe. The deconvolution algorithm (dCv) can effectively improve the beam resolution and suppress the side lobes. For each data segment (i-th), the dCv beam distribution of all frequency beam outputs is I i (sinφ)=∑ f S i (sinφ,f).

[0055] In order to achieve the improvement of both temporal and spatial resolution, the outputs of MF and CBF are re-represented in the spatiotemporal domain, i.e., through a specific two-dimensional convolution operation;

[0056] Denote it as I MF-CBF (sinφ,t)=|h MF-CBF (sinφ,t)**a m δ(sinφ-sinφ m ,t-τ m )| 2 , which involves the two-dimensional Dirac function δ(·,·) and the two-dimensional convolution operation "**". At the same time, for the two-dimensional h MF-CBF (sinφ,t) is clearly defined as: Where e(t) is the transmitted signal waveform, H represents the complex transposition, and its frequency domain expression is Here E(f) is the signal spectrum.

[0057] In summary, the output in the time and space domain is: I MF-CBF (sinφ,t)=Bp2(sinφ,t)**S2(sinφ,t), and use the RL algorithm to solve the signal energy distribution obtained from the above formula, that is, Among them, I MF-CBF (sinφ,t) is the beam time series intensity, Bp2(sinφ,t) is the two-dimensional PSF. The deconvolution result of S2(sinφ,t) is I dCv2 (sinφ,t) is the high-resolution representation we desire.

[0058] Among them, the spatiotemporal two-dimensional deconvolution imaging operation process

[0059] The operation process of using the spatiotemporal two-dimensional deconvolution algorithm to obtain high-resolution time delay-azimuth results is shown in the attached figure. Figure 1 As shown in the figure, throughout the entire imaging process, the simulated and received signals follow the same processing steps. The simulated signals received by each element of the sonar array are generated based on the parameters of the transmitted signal, and the receive delay is uniformly set to 0. First, matched filtering is performed on both the simulated and actual received signals. This process enhances the match between the signal and the target characteristics, improving the accuracy of subsequent processing. Next, the matched filtering results are subjected to frequency domain analysis. This involves performing a fast Fourier transform (FFT) operation to convert the time domain signal into the frequency domain, enabling more in-depth analysis and processing in the frequency dimension.

[0060] Next, a search is performed in the azimuth-frequency domain, using beamforming techniques to generate a two-dimensional conventional beamforming (CBF) result. The beamforming process weights the received signal based on the array geometry and the signal's direction of arrival, thereby enhancing signal strength in a specific direction and providing a preliminary estimate of the target's azimuth. Finally, an inverse fast Fourier transform (iFFT) is performed on the power spectrum at a specific azimuth on the two-dimensional CBF, converting the frequency domain information back to the time domain to obtain a two-dimensional delay-azimuth spectrum. In this process, the two-dimensional delay-azimuth spectrum obtained through simulated signal processing serves as the PSF, reflecting the imaging characteristics of the system; the two-dimensional delay-azimuth spectrum obtained from the actual received signal serves as the output energy distribution in the delay-azimuth domain. Using a RL algorithm to calculate the two-dimensional deconvolution of these two results in a high-resolution two-dimensional delay-azimuth spectrum, providing accurate data support for subsequent target positioning and imaging.

[0061] Among them, the application of spatiotemporal two-dimensional deconvolution imaging

[0062] The spatiotemporal two-dimensional deconvolution imaging algorithm of the present invention has broad application prospects in multiple fields.

[0063] 1) Combined with circumferential synthetic aperture technology

[0064] Combined with circular synthetic aperture technology, high-resolution imaging of underwater targets can be achieved. During the lake test, based on experimental requirements and underwater acoustic propagation characteristics, the transmission signal was set to a linear frequency modulation (LFM) signal with a starting frequency of 70kHz and a cutoff frequency of 100kHz. This frequency range can effectively cover the target detection frequency band and adapt to the underwater acoustic channel transmission characteristics; the pulse width is 4ms and the repetition period is 50ms, which is optimized to balance signal energy and detection accuracy; the starting phase is 90° to facilitate signal synchronization and processing; the sampling rate is 2MHz to meet the signal sampling theorem and accurately collect echo signals. The sonar array rotates 360° around the target with a radius of 7.9920m, emitting a signal every 1° of rotation, and beamforming the echo signal to output a time domain beam sequence. In this way, target echo signals can be obtained in 360 directions after spatial filtering.

[0065] In the processing of circular synthetic aperture, the echoes received in 360 directions are equivalently regarded as signals received by 360 primitives, and the element spacing is accurately calculated based on the rotation radius and rotation angle. In order to facilitate subsequent processing, the 360 ​​"primitive" received signals are divided into several sub-arrays. In actual operation, 20 directions are usually divided into one sub-array, and each sub-array overlaps 10 directions, as shown in the attached figure. Figure 2 This sub-array division method can make full use of the redundant information of the data while reducing the computational complexity.

[0066] For each subarray, a time-space two-dimensional deconvolution algorithm is used to accurately calculate the time delay and direction of the target echo. First, a two-dimensional PSF is constructed based on the transmission signal parameters, array motion parameters, and environmental parameters, as shown in the attached figure. Figure 3 As shown in the figure. From the PSF result, it can be observed that it satisfies the shift invariance property, which provides the necessary conditions for the subsequent calculation using the RL algorithm. Then, the delay-azimuth two-dimensional CBF of each sub-array signal is estimated, and the two-dimensional CBF result and the two-dimensional PSF are used for deconvolution operation, and finally a high-resolution delay-azimuth spectrum is successfully obtained. As shown in the attached figure Figure 4 and attached Figure 5 As shown in the figure, by comparing with the traditional 2D-CBF method, it is clearly found that the two-dimensional deconvolution algorithm has significant advantages in resolution improvement and sidelobe suppression. It can compress the pulse width by about 40 times and improve the azimuth resolution by 5 times, greatly improving the quality of underwater target imaging.

[0067] 2) Application in multi-beam squint sonar

[0068] The spatiotemporal two-dimensional deconvolution imaging algorithm can also be applied to the high-resolution imaging of multi-beam squint sonar. Taking the pool experiment as an example, the squint sonar uses an operating frequency of 1.2MHz, which is the optimized operating frequency of the equipment for underwater target detection, and can achieve better resolution and detection distance; the number of primitives is 96, which is determined according to the equipment design specifications; the primitive spacing is 1mm, which is precisely measured and calibrated to ensure the effective acquisition and processing of signals by the sonar array. In this experimental scenario, spatiotemporal two-dimensional deconvolution is used to obtain a specific two-dimensional PSF. By processing the experimental data, 2D-CBF and spatiotemporal two-dimensional deconvolution are used to obtain the delay-azimuth spectrum estimation results, as shown in the attached figure. Figure 6 The comparison results show that Figure 7 As shown in the figure, in squint sonar imaging, the spatiotemporal two-dimensional deconvolution algorithm can also effectively improve the resolution of time delay and azimuth at the same time, and can effectively suppress side lobes, providing more powerful technical support for the application of multi-beam squint sonar in underwater target detection.

[0069] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above-described embodiments. The above-described embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution, characterized in that: The following steps are involved: Perform matched filtering on the simulation signal and the received signal respectively; Perform fast Fourier transform on the result of matched filtering to convert it into frequency domain; Searching in the azimuth-frequency domain, using beamforming technology to generate two-dimensional conventional beamforming results; Perform an inverse fast Fourier transform on the power spectrum of a certain azimuth on the two-dimensional conventional beamforming to obtain a delay-azimuth two-dimensional spectrum. The delay-azimuth two-dimensional spectrum obtained by simulation signal processing is used as the point spread function, and the delay-azimuth two-dimensional spectrum obtained from the actual received signal is used as the output energy distribution. The Richardson-Lucas algorithm is used to calculate the two-dimensional deconvolution of the PSF and the output energy distribution to obtain a high-resolution time-delay-azimuth two-dimensional spectrum.

2. The method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution according to claim 1, characterized in that: When combined with the circular synthetic aperture technology to achieve high-resolution imaging of underwater targets, the following steps are also included: The echoes received at 360 directions in the circular synthetic aperture are equivalently regarded as signals received by 360 primitives, and the element spacing is calculated based on the rotation radius and rotation angle; The received signal is divided into several sub-arrays, each sub-array contains 20 directions, and adjacent sub-arrays overlap by 10 directions; A two-dimensional spectrum is constructed as a point spread function according to the transmission signal parameters, array motion parameters and environmental parameters; A two-dimensional conventional beamforming estimation of the delay and azimuth is performed on each subarray signal, and a deconvolution operation is performed using the two-dimensional conventional beamforming result and the two-dimensional spectrum as the point spread function to obtain a high-resolution delay-azimuth spectrum.

3. The method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution according to claim 1, characterized in that: When applied to multi-beam squint sonar high-resolution imaging, the following steps are also included: According to the operating frequency, number of primitives and spacing parameters of the multi-beam squint sonar, a two-dimensional spectrum is obtained as a point spread function by using time-space two-dimensional deconvolution. The signals received by the multi-beam squint sonar are processed and the delay-azimuth spectrum estimation results are obtained through two-dimensional conventional beamforming and two-dimensional spatiotemporal deconvolution.

4. The method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution according to claim 1, characterized in that: The transmission signal is a linear frequency modulation signal, and its parameters include a starting frequency, a cut-off frequency, a pulse width, a repetition period, a starting phase and a sampling rate.

5. The method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution according to claim 2, characterized in that: In the circular synthetic aperture technology, the sonar array rotates 360° around the target at a specific radius, transmits a signal every 1° of rotation, and receives an echo signal.

6. The method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution according to claim 3, characterized in that: The multi-beam squint sonar has an operating frequency of 1.2 MHz, contains 96 primitives, and has a primitive spacing of 1 mm.

7. The method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution according to claim 1, characterized in that: The Richardson-Lucas algorithm performs iterative solution based on the two-dimensional convolution relationship of signal energy distribution in the deconvolution imaging principle when calculating the two-dimensional deconvolution.

8. The method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution according to claim 2, characterized in that: By combining the spatiotemporal two-dimensional deconvolution imaging algorithm with the circular synthetic aperture technology, the pulse width can be compressed by about 40 times and the azimuth resolution can be increased by 5 times.

9. The method for super-resolution enhancement of underwater acoustic images based on spatiotemporal two-dimensional deconvolution according to claim 3, characterized in that: When the spatiotemporal two-dimensional deconvolution imaging algorithm is applied to multi-beam squint sonar imaging, it can simultaneously improve the resolution of time delay and azimuth, and suppress side lobes.

Images