A method for achieving underwater acoustic imaging with the lowest resolution on the order of wavelength.

Abstract

This invention relates to the field of underwater acoustic imaging technology, specifically a method for achieving underwater acoustic imaging with the lowest possible resolution on the wavelength order. It involves acquiring the target scattering spatial spectrum using a circular synthetic aperture experimental system and combining it with super-resolution image processing technology to overcome the traditional λ / 4 resolution limit. Here, λ is 17.41 mm (center frequency 85 kHz), and λ / 4 = 4.4 mm. After super-resolution processing, the minimum 3dB bandwidth of the image point reaches 0.1 mm, a 54-fold improvement over traditional methods. Through three-dimensional coordinate calibration (error < 0.01 m), pulse compression (range accuracy improved to below 0.5 mm), and a two-dimensional deconvolution super-resolution algorithm (breaking the λ / 4 = 4.4 mm limit after 10 iterations, achieving a final resolution of 0.1 mm), wavelength-level resolution is achieved. An imaging system is constructed using a 64-element high-frequency hydrophone vertical array and a directional sound source. Through target rotation sampling, three-dimensional coordinate calibration, pulse compression, and two-dimensional deconvolution super-resolution processing, a resolution better than λ / 4 is achieved. Experimental results show that this method can improve the resolution to 0.1 mm.

Description

Technical Field

[0001] This invention relates to the field of underwater acoustic imaging technology, specifically to an underwater acoustic imaging method that achieves the lowest resolution on the wavelength scale, enabling high-precision target detection with the lowest resolution on the wavelength scale. Background Technology

[0002] Underwater acoustic imaging technology has significant application value in fields such as ocean exploration, underwater target identification, and environmental monitoring. Imaging resolution, as a core indicator, directly determines the system's ability to identify minute targets. Traditional underwater acoustic imaging theory states that the theoretical physical limit of the imaging system's minimum resolution is λ / 4 (where λ is the wavelength of the sound wave), a limit jointly determined by the wavelength of the sound wave and the spatial spectral coverage capability of the acquisition system. When the target size is smaller than λ / 4, conventional imaging methods struggle to resolve it, thus limiting the ability to detect fine structures.

[0003] Current underwater acoustic imaging technologies primarily improve resolution by increasing the acoustic wave frequency (shortening the wavelength) or increasing the array aperture. However, high-frequency signals attenuate significantly in water, limiting the detection range. Large-aperture arrays, on the other hand, face challenges such as difficult installation and high costs. Furthermore, traditional imaging algorithms like focused beamforming (CBF) are limited by incomplete spatial spectrum acquisition of scattered echoes, resulting in actual resolutions that are often close to λ / 4 but difficult to surpass. For example, conventional CBF methods achieve a resolution of only 5.4 mm at a center frequency of 85 kHz, close to λ / 4 (4.4 mm) but not a breakthrough, and are sensitive to multipath interference and array attitude deviations, leading to blurred target edges. With the development of super-resolution imaging theory, improving resolution through signal processing algorithms has become a research hotspot. However, existing methods suffer from problems such as sensitivity to noise, high computational complexity, and poor three-dimensional imaging continuity, especially in complex multipath environments where resolution improvement is limited.

[0004] Traditional methods suffer from high-frequency information loss due to the inability to acquire the entire spatial spectrum (e.g., angular sampling intervals > 0.5°). This invention, however, achieves complete acquisition of the target scattering spatial spectrum through circular rotation sampling at 0.1° intervals, providing ample data support for super-resolution processing. To address these issues, this invention proposes an underwater acoustic imaging method combining circular synthetic aperture technology and a two-dimensional deconvolution super-resolution algorithm. By constructing an experimental system with full-spatial spectrum acquisition capabilities, coupled with high-precision three-dimensional coordinate calibration and pulse compression preprocessing, the quality of the original data is effectively improved. Post-processing of the reconstructed image using a super-resolution algorithm breaks through the λ / 4 resolution limit, achieving wavelength-level fine imaging. This method significantly improves imaging performance through technological innovation while maintaining existing hardware conditions, possessing significant practical application value.

[0005] Traditional methods suffer from spatial spectrum loss due to large angle sampling intervals (e.g., >0.5°), while this invention achieves full spectrum acquisition through precise 0.1° rotation sampling. Existing super-resolution algorithms are limited by vertical array sampling in 3D imaging, resulting in discontinuous energy along the z-axis. This invention improves this defect through high-precision calibration and spectral energy preservation strategies. Summary of the Invention

[0006] This paper proposes a method for underwater acoustic imaging that can achieve a resolution better than λ / 4 relative minimum resolution, thus solving the problems of limited resolution and poor adaptability to complex environments in traditional technologies.

[0007] The technical solution adopted by this invention to solve its technical problem is: a method for achieving underwater acoustic imaging with the lowest relative resolution on the order of wavelength, comprising the following steps:

[0008] A circular synthetic aperture test system was constructed, including a directional sound source, a 64-element high-frequency hydrophone vertical array, a rotatable target, and a stepper motor control unit. The directional sound source emits a linear frequency modulated signal with a center frequency of 85kHz and a bandwidth of 30kHz. The vertical array has an element spacing of 7.5cm and an aperture of 4.8m.

[0009] Three-dimensional coordinate calibration of the target: The direct signal from the sound source is received by a standard hydrophone at the target to determine the relative position between the sound source and the target, and obtain the three-dimensional absolute coordinates of the sound source; The signal emitted by the sound source at the target is collected by a cross-shaped high-frequency target listener on the vertical array to calibrate the relative position between the array and the target, and obtain the three-dimensional absolute coordinates of the array; The direct signal from the target sound source is collected by the vertical array, the time delay residual of each channel is calculated and compensated, and the three-dimensional coordinate calibration error is controlled within 0.01m; The time difference of the received signal of each array element is measured by the direct wave signal emitted by the target sound source, and the deviation between the actual distance and the theoretical distance is calculated by combining the sound speed of 1480m / s, and the time delay residual Δt_i=Δd_i / c is obtained;

[0010] The stepper motor is controlled to drive the target to rotate around the vertical axis, and the sample is taken once every 0.1° to cover the 360° full-angle scattered echo acquisition. The backsound scattered echo signals at different angles are acquired simultaneously to form the raw imaging data.

[0011] Preprocessing of raw imaging data includes secondary cropping and stitching to form a full-angle long file, shift correction based on direct wave signal, and improvement of range accuracy by pulse compression algorithm, which is implemented by performing correlation operation between the raw signal and an ideal linear frequency modulated signal.

[0012] Image reconstruction: Input correction parameters and preprocessed data, reconstruct the target image through tomographic sonar imaging algorithm, and obtain three-dimensional imaging results containing the target feature size;

[0013] A two-dimensional deconvolution super-resolution algorithm is used to process the reconstructed image. By iterative operation, side lobes are suppressed and the pixel bandwidth is reduced by 3dB. When the pixel bandwidth is less than 4.4mm, a relative minimum resolution better than λ / 4 is finally achieved.

[0014] Specifically, the signal parameters of the directional sound source include: a sound source level of 190dB, a pulse width of 1.5ms (lake test) or 0.2ms (pool test), a signal period of 100ms or 50ms, and a sampling rate of 2MHz.

[0015] Specifically, the target includes a square cage target and an elliptical ring target. The square cage target is used for resolution detection and target recognition rate verification, and the elliptical ring target is used for extreme resolution testing, with a diameter of 1.5 cm.

[0016] Specifically, during the three-dimensional coordinate calibration process, the precise position of the sound source and the array is determined through phase comparison and phase residual analysis, with the error controlled within 0.01m.

[0017] Specifically, the pulse compression algorithm improves the distance accuracy to below 0.5 mm by performing correlation operations between the original signal and the ideal linear frequency modulated signal.

[0018] Specifically, the two-dimensional deconvolution super-resolution algorithm performs two-dimensional deconvolution iterative operations on the beam-delay map based on the system point spread function (PSF) for 10-90 iterations. When the number of iterations is ≥10, the 3dB bandwidth of the image point is ≤4.2mm (<λ / 4=4.4mm), and finally reaches 0.1mm.

[0019] Specifically, during the image reconstruction process, the center position, unit length, and threshold of the reconstruction region are set to generate a 3D imaging map containing the xy and yz planes, which clearly displays the size of the target features.

[0020] Specifically, the test environment included the Huangcai Reservoir (lake test) with a water depth of 16m and the Dingjialing Laboratory Pool (pool test) with specific dimensions, and the measured sound velocity was 1480m / s.

[0021] Specifically, the vertical array is positioned 0.6m above the water surface at its top and 3m above the water surface at its center. Attitude correction is achieved through four cross-shaped high-frequency receivers, with an equivalent noise pressure of less than 40dB.

[0022] Specifically, the super-resolution processed imaging result is used to determine whether the resolution index is met by comparing the numerical relationship between the 3dB width of the image point and λ / 4 (4.4mm). Here, λ is the wavelength of the sound wave, which is calculated from the center frequency of 85kHz. λ=17.41mm is calculated based on the speed of sound of 1480m / s and the formula λ=c / f, where c is the speed of sound and f is the center frequency.

[0023] The beneficial effects of this invention are:

[0024] 1. Breaking through the resolution limit: By combining circumferential synthetic aperture full-space spectrum acquisition with super-resolution algorithm, a resolution better than λ / 4 (minimum 0.1mm) is achieved, which is 54 times higher than the traditional CBF method. This significantly enhances the ability to identify small targets and can clearly distinguish steel columns with a spacing of 5cm (design value 0.05m, actual measurement 0.0519m) and elliptical rings with a diameter of 1.5cm (actual measurement 2.05cm).

[0025] 2. High-precision calibration technology: The three-dimensional coordinate calibration error is less than 0.01m. Time delay residual compensation eliminates the influence of array attitude and position deviation, ensuring the geometric accuracy of the reconstructed image. The measured side length of the square cage is 0.1866m (design value 0.2m), with a measurement error of 3.3%.

[0026] 3. Multi-environment adaptability: Suitable for different scenarios such as lakes and laboratory pools. By adjusting the signal pulse width (1.5ms / 0.2ms) and sampling rate (2MHz), it can meet the needs of both far-field detection and near-field fine imaging.

[0027] 4. High-efficiency data processing: The pulse compression algorithm improves the signal-to-noise ratio by more than 20dB and shortens the preprocessing time by 50%; the super-resolution algorithm has a fast iterative convergence speed, and the sidelobe suppression ratio is improved from 10dB to 40dB, effectively suppressing noise interference (see Table 1). It can break through the λ / 4 limit in 10 iterations to meet the requirements of real-time imaging.

[0028] 5. 3D imaging capability: Generates 3D images containing xy and yz planes, clearly displaying the spatial structure of the target. The measured diameter of the elliptical ring is 2.05cm (design value 1.5cm), and the geometric feature restoration accuracy is improved by 43%, providing support for the analysis of complex target morphology. Attached Figure Description

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

[0030] Figure 1 Schematic diagram of the lake test setup for principle verification (linear array tomography);

[0031] This diagram illustrates the system layout of the Huangcai Reservoir lake test, including the test platform, a loudspeaker sound source, a vertical array of 64-element high-frequency hydrophones, and a rotatable target. The sound source and array are fixed at the same underwater depth (center 3m above the water surface), and the target is approximately 7.1m from the array, rotating around the vertical axis via a stepper motor (sampling once every 0.1°). The diagram marks the location parameters of key equipment: sound source 3.6m above the water surface, top of the array 0.6m above the water surface, array aperture 4.8m (64 elements, 7.5cm spacing). A calibration point sound source (Rst) and a target area calibration sound source (Rta) are installed at the target end to accurately calibrate the relative position of the sound source and array, ensuring full-spectrum acquisition capability and laying the foundation for subsequent resolution testing.

[0032] Figure 2 This is a real photo of the square cage used as a target.

[0033] This image shows a square cage target used for resolution testing. The number and spacing of the pillars on its four sides vary (designed spacing 0.05m, side length 0.2m), which verifies the imaging system's ability to resolve target details. The gap structure on the target surface serves as feature points to evaluate whether the imaging method can accurately reproduce the target's geometry. The actual image shows that the target is constructed of a metal frame, with a rotating shaft at the bottom connected to a stepper motor, supporting 360° rotation sampling. Combined with a calibration sound source, it enables multi-angle scattered echo acquisition.

[0034] Figure 3 A schematic diagram of the test environment and test system (Dingjialing Laboratory water tank);

[0035] This image shows the laboratory water tank setup used for extreme resolution testing of an elliptical ring target. The directional sound source is 1m above the water surface, with emission parameters identical to the lake test (center frequency 85kHz, bandwidth 30kHz). The target (elliptical ring, 1.5cm in diameter) is fixed to a rotating motor, 0.75m from the array (at the same depth as the array center). While the tank dimensions are not fully displayed, the core layout is similar to the lake test. Reducing the test distance improves near-field imaging accuracy and verifies the super-resolution algorithm's ability to handle small targets.

[0036] Figure 4 The image shows the water tank control device and the elliptical ring target.

[0037] The left image shows the rotation control device for the water tank test, with a stepper motor accuracy of 0.1° to ensure accurate target angle sampling. The right image is a close-up of the elliptical ring target, a 1.5cm diameter metal ring structure used to test the system's imaging capability for curved targets. The actual photograph shows the target mounted on a transparent bracket to reduce hydrodynamic interference, working in conjunction with a high-frequency hydrophone array to acquire high signal-to-noise ratio echo signals, providing a high-quality data source for super-resolution processing.

[0038] Figure 5The results of sound source calibration (phase comparison and phase residual).

[0039] Figure 6 This is a schematic diagram of the sound source calibration location.

[0040] Figure 7 The results of array calibration (phase comparison and phase residual).

[0041] Figure 8 This is a schematic diagram of the array calibration location.

[0042] Figure 9 This is a schematic diagram of time delay residual calibration;

[0043] This figure shows the distribution of theoretical and actual time delay residuals for each channel of the vertical array. By acquiring the direct signal from the target sound source, the time delay deviation of each array element (Δt = actual time delay - theoretical time delay) is calculated. During reconstruction, the position residual effect is eliminated by the phase compensation formula, which improves the imaging focusing accuracy and avoids image blurring caused by array attitude deviation.

[0044] Figure 10 This is the sine curve after shift correction;

[0045] The horizontal axis of the graph represents the target rotation angle (in 0.1° intervals), and the vertical axis represents the channel number. The bright areas represent direct wave signals. After shift correction, the signal positions are aligned, eliminating time axis misalignment caused by array tilt.

[0046] Figure 11 For comparison of pulse pressure data and raw data;

[0047] By performing correlation operations with an ideal LFM signal, the range resolution of this image was improved from 24.67 mm (theoretical value) to 5.4 mm (measured CBF result), and noise was effectively suppressed, providing high-contrast data for subsequent super-resolution processing.

[0048] Figure 12 This is a 3D reconstruction image.

[0049] Figure 13 For 3D images (angle 1).

[0050] Figure 14 For 3D images (angle 2).

[0051] Figure 15 For 3D images (xy plane).

[0052] Figure 16 For 3D images (yz plane).

[0053] Figure 17 This is a schematic diagram of the feature dimensions.

[0054] Figures 18-19This is an image of an elliptical ring and a schematic diagram of its feature dimensions.

[0055] The measured width of the elliptical ring in the figure is 0.1228m (design value 0.14m), and the diameter is 2.05cm (design value 1.5cm), demonstrating the system's ability to geometrically reconstruct curved targets and providing support for the identification of complex targets.

[0056] Figure 20 The results are from a super-resolution algorithm imaging method.

[0057] The image shows a comparison of CBF and dCv algorithms after 10 iterations. In the right super-resolution image (dCv), the edges of adjacent steel columns are sharper, and the 3dB bandwidth is reduced from 5.4mm to 4.2mm, breaking the λ / 4 (4.4mm) limit for the first time. The 3dB bandwidth is defined as the horizontal width when the peak intensity drops to 70.7%. The CBF method is 5.4mm, while the dCv algorithm reduces it to 4.2mm after 10 iterations.

[0058] Figure 21 For comparison of super-resolution algorithm imaging results and 3dB bandwidth;

[0059] The image shows the ultimate resolution result after 90 iterations. The 3dB bandwidth of the image point reaches 0.1mm, and the side lobes are basically eliminated. This verifies the efficient processing capability of the two-dimensional deconvolution algorithm for the continuous spectrum. Although the z-axis energy fluctuations cause slightly poor continuity, the lateral resolution is significantly improved, providing a possibility for underwater sub-millimeter level detection. Detailed Implementation

[0060] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0061] like Figures 1-21 As shown, the underwater acoustic imaging method of the present invention, which achieves the lowest resolution on the order of wavelength, includes the following core steps:

[0062] The experimental system was constructed using a directional sound source (emitting a high-frequency linear frequency modulated signal) with a center frequency of 85 kHz and a vertical array of 64 elementary high-frequency hydrophones to form the imaging system. The sound source and array were fixed to the experimental platform, and the target was rotated around the vertical axis by a stepper motor to achieve full-angle sampling at 0.1° intervals. In the lake test, the target was 7.1 m away from the array, and in the pool test, the distance was 0.75 m, ensuring coverage of the target's scattering spatial spectrum.

[0063] Three-dimensional coordinate calibration: Sound source calibration involves receiving the direct sound signal from the target using a standard hydrophone, analyzing the signal amplitude and phase to determine the three-dimensional coordinates of the sound source, with an error controlled within 0.01m. Array calibration utilizes the target sound source's emitted signal; data is collected by crosshairs on the array to calculate the relative position of the array and the target, obtaining the array's three-dimensional absolute coordinates. Time delay residual compensation involves acquiring the direct sound signal from the target, calculating the residual between the actual and theoretical time delays of each channel, and performing phase correction during reconstruction to improve positioning accuracy.

[0064] Data acquisition and preprocessing: Echo signals are simultaneously acquired during target rotation to form a raw data file containing 64 channels and covering all angles. Secondary cropping, splicing, and shift correction are used to eliminate multipath interference. A pulse compression algorithm (correlation calculation) is employed to improve range resolution, achieving a range accuracy of less than 0.5 mm.

[0065] Image reconstruction, based on tomographic sonar imaging algorithm, inputs calibration parameters and preprocessed data, sets the reconstruction area (such as xy plane, yz plane), and generates a three-dimensional image through frequency domain back projection or time domain focusing algorithm, clearly showing the target feature size (such as the side length of square cage, the diameter of elliptical ring).

[0066] Super-resolution processing employs a two-dimensional deconvolution (dCv) algorithm to iteratively process the reconstructed image, suppressing sidelobe noise. As the number of iterations increases (e.g., 10-90 times), the 3dB bandwidth of the image point gradually decreases from 5.4mm (CBF result) to 0.1mm, breaking through the λ / 4 (4.4mm) limit and achieving wavelength-level resolution.

[0067] The test results of the resolution improvement of the two-dimensional deconvolution imaging algorithm are shown in Table 1 below:

[0068]

[0069] As shown in the table above, the focused beamforming (CBF) resolution obtained by the tomographic method is 5.4 mm, close to the quarter-wavelength limit resolution size (the imaging system's center frequency is 85 kHz, corresponding to a wavelength of 17.41 mm, and λ / 4 is approximately 4.4 mm). The resolution can be improved by using the super-resolution method—two-dimensional deconvolution (dCv), and the 3 dB bandwidth of the image point is continuously reduced with the increase of the number of iterations. The relationship between the number of iterations and resolution is shown in Table 1. When the number of iterations is ≥10, the resolution exceeds λ / 4, and after 90 iterations, it reaches 0.1 mm. It can be seen that after 10 iterations, the resolution obtained by super-resolution processing is 4.2 mm, which is better than the quarter-wavelength limit resolution size (λ / 4 is approximately 4.4 mm). A comparison of the results after 10 iterations is attached. Figure 20 (Results of CBF method on the left, results of dCv method on the right).

[0070] Example 1: Experimental System Setup and Parameter Configuration

[0071] Hardware deployment

[0072] Lake test environment: Huangcai Reservoir, water depth 16m, sound speed 1480m / s. The directional sound source is fixed 3.6m underwater, can rotate horizontally, and emits an LFM signal with a center frequency of 85kHz and a bandwidth of 30kHz (sound source level 190dB, pulse width 1.5ms, period 100ms).

[0073] Vertical array: 64-element high-frequency hydrophone, element spacing 7.5cm, aperture 4.8m, top 0.6m above water surface, center 3m above water surface. The array integrates four cross-shaped high-frequency receivers for attitude correction, with an equivalent noise level <40dB.

[0074] Target: Square cage target, designed with a side length of 0.2m and a steel column spacing of 0.05m. The target calibration point sound source and the target area listening device are installed and rotated around the vertical axis by a stepper motor with a sampling interval of 0.1°.

[0075] Water tank test configuration

[0076] The Dingjialing laboratory water tank is fitted with an elliptical ring target (1.5cm in diameter). The sound source is 1m above the water surface, and the transmission parameters are adjusted to a pulse width of 0.2ms, a period of 50ms, and a sampling rate of 2MHz. The target is 0.75m from the array and at the same depth as the array center.

[0077] System synchronization

[0078] The acoustic emission and reception are synchronized using a 48MHz clock source to ensure signal phase consistency; the stepper motor and data acquisition card (NI) are triggered and synchronized to achieve strict angle-signal correspondence.

[0079] Example 2: Three-dimensional coordinate calibration method

[0080] Sound source absolute coordinate calibration

[0081] With the target fixed, the sound source is rotated, and the signal amplitude is monitored using a standard hydrophone. The angle of the sound source is fixed at the point of maximum amplitude (the direction with the strongest directivity).

[0082] Move the target up and down until the signal amplitude is maximized again to determine the target depth.

[0083] Acquire direct wave signals from different azimuth angles and compare them by phase (e.g.) Figure 5 The coordinates (x, y, z) of the sound source were calculated, and the error was corrected by phase residual analysis, with the final coordinate accuracy being ±0.01m.

[0084] Array absolute coordinate calibration

[0085] The target sound source emits LFM signals, and the crosshair receiver (4 channels) collects the signals. The relative distance between the array and the target is calculated by the time difference of arrival (TDOA).

[0086] Establish a coordinate system with the target center as the origin, and derive the coordinates of each array element (e.g., ...). Figure 8 Combined with attitude correction data, a three-dimensional coordinate array is generated.

[0087] Time Delay Residual Compensation

[0088] Collect the direct signal from the target sound source and calculate the difference between the theoretical time delay (t = distance / c) and the actual time delay of each array element (e.g., Figure 9 ).

[0089] During reconstruction, time delay compensation is performed on the signals of each channel. The formula is as follows:

[0090] Where Δt_i is the difference between the actual delay and the theoretical delay of the i-th channel, which is calculated by the time difference of arrival (TDOA) of the direct wave signal and is used to correct the phase error caused by the array attitude deviation.

[0091] Example 3: Data Acquisition and Preprocessing Process

[0092] Raw data collection

[0093] The motor drives the target to rotate 360°, triggering a data acquisition every 0.1°, generating a 64-channel .mat file (naming rule: CF-1_4_probe number_angle.mat).

[0094] The lake test took about 10 minutes to collect 3,600 angle samples at a time; the pool test, due to its proximity, reduced the sampling time to 2 minutes.

[0095] Secondary cutting and splicing

[0096] The host computer initially removes invalid data (such as transient signals during motor startup) and retains 10 seconds of valid data.

[0097] A program was written to stitch together the data from each angle by channel, forming a long file containing all angles, which facilitates the subsequent generation of a sine curve.

[0098] Shift correction and pulse compression

[0099] Based on the position of the direct wave signal, the sine curve is laterally shifted to eliminate time axis misalignment caused by array attitude deviation (e.g., Figure 10 ).

[0100] Pulse compression formula:

[0101] Where h(t) is the ideal LFM signal, the range resolution is improved to c / (2B)=1480 / (2×30×10³)=24.67mm after correlation operation, and the actual test reached 5.4mm (affected by noise).

[0102] Example 4: Image Reconstruction and Feature Extraction

[0103] Chromatography Reconstruction Parameter Settings

[0104] Input the calibrated sound source coordinates, array coordinates, and time delay compensation parameters, and set the reconstruction area: x range [-0.3m, 0.3m], y range [-8m, -6m], z range [-4.5m, -3.5m], unit length 0.001m.

[0105] A frequency domain back-projection algorithm is used to map echo signals from each angle onto the reconstruction grid, and the accumulated energy value generates a grayscale image.

[0106] 3D image generation

[0107] Generate xy-plane projection diagram (e.g.) Figure 15 The yz plane displays the target's lateral dimensions; the yz plane displays the depth distribution (e.g., ...). Figure 16 ).

[0108] The display threshold was optimized to highlight the target edge. The measured spacing between the square cage steel columns was 0.0519m (design value 0.05m), with an error of 3.8%.

[0109] Feature size calculation

[0110] Extract the orthogonal profile intensity curve of the peak point and calculate the 3dB bandwidth (full width at half maximum). The formula is as follows:

[0111] in , The coordinates are 0.707 times the peak intensity. The traditional CBF method measured 5.4 mm, which gradually decreased after super-resolution processing.

[0112] Example 5: Super-resolution processing and resolution verification

[0113] Implementation of 2D deconvolution algorithm

[0114] Define the system point spread function (PSF) and construct the PSF matrix based on the array response model.

[0115] Iteration formula:

[0116] in For image estimation, For observation data, For PSF, Represents convolution. This indicates point division, which suppresses sidelobe noise through iteration.

[0117] Resolution improvement effect

[0118] After 10 iterations, the resolution decreased from 5.4 mm to 4.2 mm (better than λ / 4=4.4 mm), as shown in Table 1 of the resolution improvement test results of the two-dimensional deconvolution imaging algorithm; after 90 iterations, it reached 0.1 mm, and the sidelobe suppression ratio was improved by 30 dB.

[0119] Compare CBF and dCv imaging results (e.g.) Figure 20 After super-resolution processing, the target edges are sharper, and adjacent steel columns can be clearly separated.

[0120] Indicator Judgment

[0121] The ratio of the actual image point's 3dB bandwidth to λ / 4 is calculated, and a bandwidth < 4.4mm is considered acceptable. Both lake and pool tests passed this test, validating the method's effectiveness.

[0122] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of protection claimed by the present invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method of underwater acoustic imaging with a relative minimum resolution of the order of a wavelength, characterized in that, The method comprises the following steps: A circumferential synthetic aperture test system is built, which comprises a directional sound source, a 64-element high-frequency vertical array, a rotatable target and a stepping motor control unit, the directional sound source emits a linear frequency modulation signal with a center frequency of 85 kHz and a bandwidth of 30 kHz, the vertical array has an element spacing of 7.5 cm and an aperture of 4.8 m; Three-dimensional coordinate calibration of the target: the relative position relationship between the sound source and the target is determined by receiving the direct signal of the sound source at the standard hydrophone on the target, and the three-dimensional absolute coordinates of the sound source are obtained; the relative position between the array and the target is calibrated by emitting signals from the sound source on the target, and the signals are collected by the cross high-frequency hydrophone on the vertical array, and the three-dimensional absolute coordinates of the array are obtained; the time delay residual of each channel is calculated by using the vertical array to collect the direct signal of the target sound source, and compensation is performed, and the three-dimensional coordinate calibration error is controlled within 0.01 m; The stepping motor is controlled to drive the target to rotate around the vertical axis, and the target is sampled every 0.1°, covering 360° full-angle scattering echo collection, and the backward acoustic scattering echo signals at different angles are collected synchronously to form the original imaging data; The original imaging data is preprocessed: including secondary clipping, splicing to form a full-angle long file, shift correction based on the direct wave signal, and pulse compression algorithm to improve the distance precision, the pulse compression algorithm is realized by correlating the original signal with the ideal linear frequency modulation signal; Image reconstruction: input correction parameters and preprocessed data, and reconstruct the target image by tomographic sonar imaging algorithm to obtain three-dimensional imaging results containing target feature size; A two-dimensional deconvolution super-resolution algorithm is used to process the reconstructed image, the sidelobe is suppressed by iterative operation, and the 3dB bandwidth of the image point is reduced, when the 3dB bandwidth of the image point is less than 4.4mm, the relative minimum resolution better than λ / 4 is finally realized, and λ is the wavelength of the sound wave.

2. The method according to claim 1, wherein the relative minimum resolution is of the order of the wavelength. The signal parameters of the directional sound source include: sound source level 190 dB, lake test pulse width 1.5 ms, pool test pulse width 0.2 ms, signal period 100 ms or 50 ms, and sampling rate 2 MHz.

3. The method of claim 1, wherein the relative minimum resolution is on the order of a wavelength. The target includes a square cage target and an elliptical ring target, the square cage target is used for resolution detection and target recognition rate verification, and the elliptical ring target is used for limit resolution test, and the diameter is 1.5 cm.

4. The method of claim 1, wherein the relative minimum resolution is on the order of a wavelength. In the three-dimensional coordinate calibration process, the accurate positions of the sound source and the array are determined by phase contrast and phase residual analysis, and the error is controlled within 0.01 m.

5. The method of claim 1, wherein the relative minimum resolution is on the order of a wavelength. The pulse compression algorithm improves the distance precision to below 0.5 mm by correlating the original signal with the ideal linear frequency modulation signal.

6. The method of claim 1, wherein the relative minimum resolution is on the order of a wavelength. The two-dimensional deconvolution super-resolution algorithm performs two-dimensional deconvolution iterative operation on the beam-time delay graph based on the system point spread function, and the iteration number is 10-90 times, when the iteration number is greater than or equal to 10 times, the 3dB bandwidth of the image point is less than or equal to 4.2 mm, and finally reaches 0.1 mm.

7. The method of claim 1, wherein the relative minimum resolution is on the order of a wavelength. In the image reconstruction process, the center position, unit length and threshold of the reconstruction region are set, a 3D imaging graph containing x-y and y-z planes is generated, and the target feature size is clearly displayed.

8. The method of claim 1, wherein the relative minimum resolution is on the order of a wavelength. The vertical array is 0.6 m from the water surface at the top and 3 m from the water surface at the center, and is corrected in posture by 4 cross-shaped high-frequency receivers, with equivalent noise pressure less than 40 dB.

9. The method of claim 1, wherein the relative minimum resolution is on the order of a wavelength. The imaging result after super-resolution processing is judged whether the resolution index is reached by comparing the 3dB width of the image point with the value of λ / 4, and λ is calculated by the formula λ=c / f, wherein c is the sound speed 1480 m / s, and f is the center frequency 85 kHz.

Images