A high-precision multi-beam acoustic three-dimensional imaging method based on big data

Abstract

This invention relates to the field of three-dimensional imaging technology, and in particular to a high-precision multi-beam acoustic three-dimensional imaging method based on big data. The method includes determining imaging parameters; arranging a transducer array according to the imaging parameters; controlling the transducer array to output multiple beams to scan a target area and receiving echo signals; preprocessing the echo signals by noise reduction and filtering; mapping the preprocessed echo data to a three-dimensional spatial coordinate system to generate an initial three-dimensional imaging model; matching and estimating the features and volume of missing areas in the initial three-dimensional imaging model based on big data, and calculating the imaging missing rate; determining whether the current imaging is qualified based on the imaging missing rate; and if the current imaging is unqualified, determining the reason for the unqualified imaging based on the complexity of the initial three-dimensional imaging model. This invention achieves intelligent diagnosis and optimization of the acoustic three-dimensional imaging process by combining big data missing analysis and imaging model complexity assessment.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional imaging technology, and in particular to a high-precision multi-beam acoustic three-dimensional imaging method based on big data. Background Technology

[0002] Multibeam sonar systems are core equipment for seabed topographic mapping and underwater target detection. They construct three-dimensional seabed models by transmitting multiple narrow beams through an array and receiving echo signals. However, obtaining high-precision and high-completeness three-dimensional images in practical applications still faces many challenges. First, due to the characteristics of sound wave propagation, topographic masking effects, and limitations of transducer beamwidth, blind spots or areas with weak signals are inevitable in the scanned data. This results in holes or geometric distortions in the generated three-dimensional model, affecting the integrity and reliability of the imaging. Traditional interpolation or extrapolation methods are insufficient to accurately estimate the true shape and volume of these missing areas.

[0003] Chinese Patent Application No. CN105259557A discloses a multi-frequency transmit beamforming method, comprising: (1) dividing the beam direction of the transmit array in a cross-shaped array into multiple sectors, and sequentially transmitting a series of sector-shaped sonar beam signals of different frequencies in each sector, wherein each frequency of the sector-shaped sonar beam signal points to a beam direction within the corresponding sector; (2) after the transmission of all frequencies of the sector-shaped sonar beam signals in each sector is completed, receiving the sonar echo signal using the receive array in the cross-shaped array, extracting the frequency information corresponding to all sector-shaped sonar beam signals in each sector through discrete Fourier transform, and performing beamforming calculation in the frequency domain corresponding to the frequency information to obtain the beam intensity result. This invention can reduce the transmission time of the cross-shaped array, obtain beam performance similar to that of a two-dimensional triangular facet receive array, and reduce the complexity of the underwater real-time three-dimensional acoustic imaging system.

[0004] However, existing technologies still have the following problems:

[0005] The lack of automated and quantitative assessment capabilities for image quality makes it impossible to intelligently diagnose the root causes of imaging defects and perform adaptive parameter optimization accordingly, thus making it difficult to ensure the integrity and accuracy of 3D imaging in complex scenarios. Summary of the Invention

[0006] To address this, the present invention provides a high-precision multi-beam acoustic 3D imaging method based on big data, which overcomes the lack of automated and quantitative assessment capabilities for imaging quality in existing technologies, the inability to intelligently diagnose the root causes of imaging defects, and the difficulty in ensuring the integrity and accuracy of 3D imaging in complex scenarios by adaptively optimizing parameters accordingly.

[0007] To achieve the above objectives, this invention provides a high-precision multi-beam acoustic three-dimensional imaging method based on big data. It includes:

[0008] Step S1: Determine the imaging parameters, which include the spacing between adjacent transducers in the transducer array, the beam parameters of a single beam output by each transducer, and data processing standard parameters.

[0009] Step S2: Arrange the transducer array according to the imaging parameters, control the transducer array to output multiple beams to scan the target area, and receive echo signals;

[0010] Step S3: Perform noise reduction and filtering preprocessing on the echo signal, and map the preprocessed echo data to a three-dimensional spatial coordinate system to generate an initial three-dimensional imaging model.

[0011] Step S4: Based on big data, match and estimate the features of the missing regions in the initial three-dimensional imaging model to determine the estimated missing volume, and calculate the imaging missing rate based on the estimated missing volume.

[0012] Step S5: Determine whether the current imaging is qualified based on the imaging missing rate. If the current imaging is unqualified, determine the reason for the unqualified imaging based on the complexity of the initial three-dimensional imaging model, and generate corresponding imaging parameter correction instructions according to the reason for the unqualified imaging.

[0013] Further, in step S4, the imaging missing rate is the ratio of the estimated missing volume to the scanned terrain volume.

[0014] Further, in step S5, determining whether the current imaging is qualified based on the imaging missing rate includes:

[0015] If the imaging missing rate is less than the preset imaging missing rate, the current imaging is determined to be qualified, and the final three-dimensional imaging model is output.

[0016] If the imaging missing rate is greater than or equal to the preset imaging missing rate, the current imaging is determined to be unqualified, and the reason for the imaging failure is determined based on the complexity of the initial three-dimensional imaging model.

[0017] Further, in step S5, the reasons for imaging failure are determined based on the complexity of the initial three-dimensional imaging model, including:

[0018] Step S51: Calculate the complexity based on the total number of triangular facets in the initial three-dimensional imaging model and the area regularity of each triangular facet.

[0019] Step S52: Compare the complexity with a preset complexity threshold;

[0020] Step S53: If the complexity is greater than or equal to the preset complexity threshold, then the reason for the imaging failure is determined to be insufficient beam coverage due to the complexity of the terrain to be tested.

[0021] Step S54: If the complexity is less than the preset complexity threshold, then the reason for the imaging failure is determined to be environmental interference.

[0022] Further, in step S51, the complexity is calculated based on the total number of triangular facets in the initial three-dimensional imaging model and the area regularity of each triangular facet, including:

[0023] Step S511: Count the total number of all discrete triangular patches in the initial three-dimensional imaging model, and use it as the first calculation factor;

[0024] Step S512: Calculate the ratio of the actual area of ​​each triangular facet in the model to a preset standard area, and calculate the arithmetic mean of all ratios as the second calculation factor;

[0025] Step S513: Calculate the intermediate complexity value and adjust the intermediate complexity value using a correction coefficient determined based on historical data to obtain the final complexity.

[0026] Furthermore, in step S53, when it is determined that the reason for the imaging failure is insufficient beam coverage due to the complex terrain to be measured, the vertical angle width in the beam parameters of a single beam is adjusted.

[0027] Specifically, the adjustment range of the vertical angle width is determined based on the complexity of the initial three-dimensional imaging model and the average depth gradient of the edge points of each missing region. The higher the complexity and the larger the average depth gradient, the greater the increase in the vertical angle width.

[0028] Furthermore, in step S53, during the process of adjusting the vertical angle width, the maximum spacing between adjacent transducers in the transducer array is simultaneously corrected based on the total estimated volume of all missing regions; wherein, the larger the total estimated volume, the smaller the reduction in the maximum spacing.

[0029] Further, in step S53, steps S2 to S5 are re-executed according to the adjusted parameters. If the imaging missing rate is still greater than or equal to the preset imaging missing rate, and the determined reason for the imaging failure is still insufficient beam coverage due to the complex terrain to be measured, the normal angle of the output beam of each transducer is adjusted.

[0030] The initial adjustment angle is determined based on the angle between the original normal angle of each beam and the line connecting the center point of the corresponding missing region. The initial adjustment angle is then corrected based on the current adjusted vertical angle width to obtain the final adjustment angle.

[0031] The larger the width of the currently adjusted vertical angle, the greater the reduction in the initial adjustment angle.

[0032] Further, in step S54, when it is determined that the cause of the imaging failure is environmental interference, the following steps are included:

[0033] Detect the resolution of the initial three-dimensional imaging model;

[0034] If the resolution is less than a preset resolution threshold, the preprocessing parameters are adjusted.

[0035] If the resolution is greater than or equal to the preset resolution threshold, then the azimuth width in the beam parameters of a single beam is increased.

[0036] Further, in step S54, the increase in the lateral angle width in the beam parameters of a single beam is determined based on the ratio of the current model resolution to the preset resolution threshold and the imaging missing rate.

[0037] Compared with existing technologies, the beneficial effects of this invention lie in its transformation of the traditional imaging quality assessment and parameter adjustment process, which relies on operator experience, into an automated and intelligent decision-making process based on historical big data and quantitative analysis of model features. The dynamic determination method of preset thresholds enables the system to possess self-learning and adaptive capabilities, allowing for the setting of reasonable evaluation criteria for different detection environments and mission objectives. This mechanism not only significantly improves the accuracy and efficiency of imaging defect diagnosis, avoiding blind trial and error, but also ensures that subsequent parameter optimization directly targets the root cause of the problem through precise cause localization. Therefore, it systematically improves the integrity, accuracy, and overall operational efficiency of multibeam acoustic 3D imaging in complex underwater terrain or harsh acoustic environments.

[0038] Furthermore, this invention transforms the traditional reliance on operator subjective experience for vague judgments of imaging defect causes into objective and automated diagnosis based on quantitative indicators (complexity) of model geometric features and statistical patterns of historical big data. The dynamic determination method of preset thresholds enables the system to possess self-learning and scene adaptation capabilities. This mechanism can quickly and accurately pinpoint whether the root cause of imaging problems is terrain or environment, thus providing precise decision-making basis for subsequent highly targeted parameter optimization (such as adjusting beam angle or improving filtering), greatly improving the efficiency and reliability of troubleshooting and performance optimization of 3D acoustic imaging in complex underwater scenes.

[0039] Furthermore, this invention transforms beam parameter adjustment from an experience-based qualitative operation into a precise calculation process driven by model quantification features and historical data. It accurately assesses complex spatial patterns in the terrain using the average distribution distance as an indicator, and intelligently determines the adjustment amount using mapping relationships learned from big data. This allows the beam shape adjustment to adapt to the specific characteristics of current terrain defects. This effectively avoids the problems of under-adjustment (failure to cover missing areas) or over-adjustment (leading to decreased resolution and increased sidelobe interference), significantly improving the acoustic system's first-scan coverage of complex underwater terrain and the completeness of the final imaging.

[0040] Furthermore, this invention achieves intelligent linkage adjustment between transducer array parameters and terrain loss features. When the total volume of the loss is large, it indicates extremely complex terrain and extensive blind spots. In this case, excessively reducing the spacing, while increasing beam density, would significantly increase data volume and processing time, and have limited effect on improving the extensive blind spots. This solution achieves an optimal balance between coverage, imaging resolution, and system processing load by making the reduction magnitude decrease as the volume increases. This collaborative adjustment mechanism systematically and economically improves the integrity and accuracy of the final 3D imaging for complex terrain.

[0041] Furthermore, when the initial parameter optimization is insufficient, this invention initiates a secondary fine-tuning based on model feedback. This method does not simply point the beam towards the center of the missing region; instead, it uses the increment of the vertical angle width as adjustment damping to dynamically calculate a more robust and less prone-to-overshoot correction angle. The preset value λ is learned from historical data, ensuring the reliability and adaptability of the adjustment strategy. This step-by-step, progressive intelligent adjustment mechanism, where each step depends on the result of the previous step, significantly improves the system's adaptive imaging capability for extremely complex terrain, approximating the optimal beam pointing configuration with higher efficiency, thereby minimizing detection blind spots and enhancing the integrity of 3D imaging.

[0042] Furthermore, this invention transforms the adjustment of preprocessing parameters from experience-based trial and error to data-driven intelligent decision-making. By establishing and utilizing a historical big data case library, the system can "recall" and integrate historically proven effective processing strategies based on the specific interference characteristics encountered, thereby achieving precise and efficient parameter optimization. This not only significantly improves the success rate and efficiency of parameter adjustment but also reduces reliance on the professional experience of operators, enhancing the imaging system's adaptability and overall robustness in complex and variable underwater acoustic environments. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the workflow of the high-precision multi-beam acoustic three-dimensional imaging method based on big data of this invention.

[0044] Figure 2This is a flowchart illustrating the process of determining the cause of imaging failure based on the complexity of the initial three-dimensional imaging model in the high-precision multi-beam acoustic three-dimensional imaging method based on big data of the present invention. Detailed Implementation

[0045] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0046] It should be noted that the data in this embodiment are all derived from a comprehensive analysis and evaluation of historical data from the six months prior to this determination and the corresponding historical determination results by the system described in this invention. Those skilled in the art will understand that the system described in this invention can determine the above-mentioned parameters for a single item by selecting the value with the highest proportion based on the data distribution as the preset standard parameter, using weighted summation to obtain the value as the preset standard parameter, substituting each historical data point into a specific formula and using the value obtained by that formula as the preset standard parameter, or other selection methods, as long as the system described in this invention can clearly define different specific situations in the single-item determination process through the obtained values.

[0047] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0048] Please see Figures 1-2 As shown, Figure 1 This is a flowchart illustrating the workflow of the high-precision multi-beam acoustic three-dimensional imaging method based on big data of this invention. Figure 2 This is a flowchart illustrating the process of determining the cause of imaging failure based on the complexity of the initial three-dimensional imaging model in the high-precision multi-beam acoustic three-dimensional imaging method based on big data of the present invention.

[0049] This invention relates to a high-precision multi-beam acoustic three-dimensional imaging method based on big data, comprising:

[0050] Step S1: Determine the imaging parameters, which include the spacing between adjacent transducers in the transducer array, the beam parameters of a single beam output by each transducer, and data processing standard parameters.

[0051] Step S2: Arrange the transducer array according to the imaging parameters, control the transducer array to output multiple beams to scan the target area, and receive echo signals;

[0052] Step S3: Perform noise reduction and filtering preprocessing on the echo signal, and map the preprocessed echo data to a three-dimensional spatial coordinate system to generate an initial three-dimensional imaging model.

[0053] Step S4: Based on big data, match and estimate the features of the missing regions in the initial three-dimensional imaging model to determine the estimated missing volume, and calculate the imaging missing rate based on the estimated missing volume.

[0054] Step S5: Determine whether the current imaging is qualified based on the imaging missing rate. If the current imaging is unqualified, determine the reason for the unqualified imaging based on the complexity of the initial three-dimensional imaging model, and generate corresponding imaging parameter correction instructions according to the reason for the unqualified imaging.

[0055] In this embodiment of the invention, the beam parameters of a single beam output by each transducer include, but are not limited to, normal angle, vertical angular width perpendicular to the heading (beam opening angle perpendicular to the heading), lateral angular width parallel to the heading (beam opening angle parallel to the heading), transmission frequency, pulse length, and beam opening angle; the data processing standard parameters include, but are not limited to, signal sampling rate, type and threshold of noise reduction filtering algorithm, definition of the three-dimensional spatial coordinate system on which data mapping is based, and rule parameters of the point cloud mesh generation model.

[0056] Specifically, in this embodiment, the big data-based decision-making relies on a pre-built and continuously updated deep-sea acoustic imaging case library. This library contains: original environmental parameters (depth, temperature, salinity), equipment parameters (element spacing, beamwidth), processing parameters, generated 3D models, and their quality evaluation indicators (missing rate, complexity, resolution) from historical missions. For new missions, the system matches similar scenes from the case library based on the target sea area type (e.g., continental slope, hydrothermal vent area) and recommends initial imaging parameters (e.g., element spacing, normal angle distribution). The "preset imaging missing rate," "preset complexity threshold," "correction coefficient k," "coefficients in the adjustment amplitude mapping relationship," and "sensitivity coefficient δ" are all determined by analyzing the case studies. The data from successful imaging cases in the database are statistically analyzed, regression fitted, or trained using machine learning to enable the system to adapt and learn. A neural network model is trained using the correspondence between complete terrain models in the database and artificially simulated or real missing models. This model takes the boundary geometric features and contextual terrain information of the missing area as input and outputs the estimated range of its volume. The high-precision 3D acoustic model output after optimization by this invention is not only the final result itself, but its high-precision geographic coordinate framework and terrain details can provide accurate navigation paths and region of interest (ROI) positioning for subsequent optical payloads (such as laser scanning and high-definition cameras), achieving an efficient collaborative operation mode of "wide-area acoustic survey and detailed local optical survey".

[0057] Specifically, in step S4, the imaging missing rate is the ratio of the estimated missing volume to the scanned terrain volume.

[0058] Specifically, in step S5, determining whether the current imaging is qualified based on the imaging missing rate includes:

[0059] If the imaging missing rate is less than the preset imaging missing rate, the current imaging is determined to be qualified, and the final three-dimensional imaging model is output.

[0060] If the imaging missing rate is greater than or equal to the preset imaging missing rate, the current imaging is determined to be unqualified, and the reason for the imaging failure is determined based on the complexity of the initial three-dimensional imaging model.

[0061] In this embodiment of the invention, the preset imaging missing rate is not a fixed value, but rather a dynamic optimization result based on big data. The system maintains a historical imaging task database, which records successful imaging cases of different sea area types (such as flat continental shelves, complex seamounts, and continental slopes) and different equipment operating modes. For a new imaging task, the system first matches similar cases from the database according to task parameters (such as target sea area type and expected resolution). The preset imaging missing rate is set as the statistical average of the imaging missing rates among these similar historical cases. If no matching cases are found, an empirical safety value (e.g., 15%) is used as the initial threshold. This threshold is fine-tuned during task execution based on the actual data quality of the current scan.

[0062] In this embodiment of the invention, after determining that the imaging is unqualified, the cause of the unqualification is diagnosed. The preset complexity threshold is also determined based on historical big data analysis: the system calculates the complexity of all case models in the database that need parameter adjustment due to insufficient coverage caused by terrain complexity, and performs statistical analysis on this set of complexity (for example, calculating its 85th percentile). This statistical value is set as the preset complexity threshold to distinguish between the two main causes of terrain complexity and environmental interference. If the calculated current model complexity is greater than or equal to the preset complexity threshold, the cause is diagnosed as insufficient beam coverage due to the complexity of the terrain to be tested. The system will then generate targeted parameter correction instructions, such as prioritizing increasing the vertical angle width of the beam to cover more terrain undulations. Conversely, if the complexity is less than the threshold, the main problem is determined to be caused by environmental interference. The system will then check the model resolution and decide whether to enhance data preprocessing or adjust the beam's azimuth angle width based on the resolution.

[0063] Through the above-described scheme, this invention transforms the traditional imaging quality assessment and parameter adjustment process, which relies on operator experience, into an automated and intelligent decision-making process based on historical big data and quantitative analysis of model features. The dynamic determination method for preset thresholds enables the system to possess self-learning and adaptive capabilities, allowing for the setting of reasonable evaluation criteria for different detection environments and mission objectives. This mechanism not only significantly improves the accuracy and efficiency of imaging defect diagnosis, avoiding blind trial and error, but also ensures that subsequent parameter optimization directly targets the root cause of the problem through precise cause localization. Therefore, it systematically improves the integrity, accuracy, and overall operational efficiency of multibeam acoustic 3D imaging in complex underwater terrain or harsh acoustic environments.

[0064] Specifically, in step S5, the reasons for imaging failure are determined based on the complexity of the initial three-dimensional imaging model, including:

[0065] Step S51: Calculate the complexity based on the total number of triangular facets in the initial three-dimensional imaging model and the area regularity of each triangular facet.

[0066] Step S52: Compare the complexity with a preset complexity threshold;

[0067] Step S53: If the complexity is greater than or equal to the preset complexity threshold, then the reason for the imaging failure is determined to be insufficient beam coverage due to the complexity of the terrain to be tested.

[0068] Step S54: If the complexity is less than the preset complexity threshold, then the reason for the imaging failure is determined to be environmental interference.

[0069] Specifically, in step S51, the complexity is calculated based on the total number of triangular facets in the initial three-dimensional imaging model and the area regularity of each triangular facet, including:

[0070] Step S511: Count the total number of all discrete triangular patches in the initial three-dimensional imaging model, and use it as the first calculation factor N;

[0071] Step S512: Calculate the ratio ri=Si / Sstd of the actual area Si of each triangular facet in the model to a preset standard area S_std, and calculate the arithmetic mean r_avg of all ratios ri as the second calculation factor; wherein, the second calculation factor r_avg characterizes the average regularity of the size of the triangular facets in the model, and the smaller its value, the more fragmented the terrain surface.

[0072] Step S513: Calculate the intermediate complexity value C_temp=N / r_avg, and adjust the intermediate complexity value using a correction coefficient k determined based on historical data to obtain the final complexity C=k×C_temp.

[0073] In an embodiment of the present invention, all discrete triangular patches constituting the three-dimensional surface are extracted from the grid data of the initial three-dimensional imaging model; the total number of the discrete triangular patches is counted and used as the first calculation factor (N) characterizing the richness of the geometric details of the model; for each triangular patch extracted from the model, the ratio (ri = Si / S_std) of its actual area (Si) to a standard reference area (S_std) preset according to the task resolution and device performance is calculated, and the arithmetic mean of this ratio for all patches is calculated to obtain the second calculation factor (r_avg) characterizing the regularity degree of the patches of the model; the first calculation factor is divided by the second calculation factor to obtain an initial complexity index, and then multiplied by a preset correction coefficient (k, such as 0.85) less than 1 to obtain the final model complexity (C), that is, C = k×(N / r_avg); the calculated complexity (C) is compared with a preset complexity threshold (C_th); if the complexity (C) is greater than or equal to the preset complexity threshold (C_th), it is intelligently diagnosed and determined that the reason for the unqualified imaging is insufficient beam coverage caused by the complexity of the terrain to be measured; if the complexity (C) is less than the preset complexity threshold (C_th), it is intelligently diagnosed and determined that the reason for the unqualified imaging is environmental interference, where the correction coefficient k (0 < k ≤ 1) is a normalization or calibration factor obtained based on historical big data analysis and is used to adjust the calculated intermediate complexity value to a numerical range matching the historical diagnosis experience.

[0074] Through the above solution, the present invention transforms the fuzzy judgment of the cause of imaging defects that traditionally relies on the subjective experience of operators into an objective and automatic diagnosis based on the quantitative index (complexity) of the geometric features of the model and the statistical law of historical big data. The dynamic determination method of the preset threshold enables the system to have the ability of self-learning and scene adaptation. This mechanism can quickly and accurately lock the root cause of the imaging problem as the terrain or the environment, thereby providing an accurate decision-making basis for subsequent highly targeted parameter optimization (such as adjusting the beam angle or improving filtering), and greatly improving the efficiency and reliability of troubleshooting and performance optimization of three-dimensional acoustic imaging in complex underwater scenes.

[0075] Specifically, in step S53, when it is determined that the reason for the unqualified imaging is insufficient beam coverage caused by the complexity of the terrain to be measured, the vertical angle width in the beam parameters of a single beam is adjusted;

[0076] Among them, based on the complexity of the initial three-dimensional imaging model and the average depth gradient of the edge points of each missing area, the adjustment amplitude of the vertical angle width is determined, and the higher the complexity and the greater the average depth gradient, the greater the increase amplitude of the vertical angle width.

[0077] In this embodiment of the invention, firstly, the system identifies the centroid (center point) coordinates of all missing regions. Then, it calculates the Euclidean distance between each pair of centroids of all missing regions, obtaining a distance set. Finally, it calculates the arithmetic mean of this distance set as the average distribution distance characterizing the dispersion of the missing regions. The larger this value, the more dispersed the undetected terrain features are in space, suggesting that the terrain may have large-scale undulations or isolated complex structures. The increase in vertical angle width is determined based on the average distribution distance through a preset mapping function. The core parameters of this mapping function (such as the scaling factor and segmented interval threshold) are derived from machine learning using historical big data. The system maintains a database containing a large number of complex terrain cases, each case recording its average distribution distance and the effective increase in vertical angle width ultimately confirmed through manual or iterative optimization. By performing regression analysis on these historical data, the system learns and establishes a predictive model from the average distribution distance to the suggested increase. In practical applications, the system inputs the calculated average distribution distance of the current task into this predictive model to directly output a quantified suggested increase value.

[0078] For example, based on historical models, a simplified implementation rule could be: set a baseline distance threshold D0. If the average distribution distance is less than D0, the terrain is considered locally complex, and the vertical angle width is recommended to be increased by a base value Δθ1 (e.g., 5°). If the average distribution distance is greater than or equal to D0, it is considered a large-scale complex terrain. In this case, the increase is positively correlated with the distance, for example, by the formula: Increase = Δθ1 + h × (Average distribution distance - D0), where h is a positive coefficient (e.g., 0.1° / meter) obtained by fitting based on historical successful adjustment cases.

[0079] Through the above-described scheme, this invention transforms beam parameter adjustment from an experience-based qualitative operation into a precise calculation process driven by model quantification features and historical data. It accurately assesses complex spatial patterns in the terrain using the average distribution distance as an indicator, and intelligently determines the adjustment amount using mapping relationships learned from big data. This allows the beam shape adjustment to adapt to the specific characteristics of current terrain defects. This effectively avoids the problems of under-adjustment (failure to cover missing areas) or over-adjustment (leading to decreased resolution and increased sidelobe interference), significantly improving the acoustic system's first-scan coverage of complex underwater terrain and the completeness of the final imaging.

[0080] Specifically, in step S53, during the process of adjusting the vertical angle width, the maximum spacing between adjacent transducers in the transducer array is simultaneously corrected based on the total estimated volume of all missing regions; wherein, the larger the total estimated volume, the smaller the reduction in the maximum spacing.

[0081] In this embodiment of the invention, firstly, all volumes estimated for each missing region in step S4 are summarized to obtain the total estimated volume (V_missing). Then, this volume value is substituted into a preset correction function to calculate the spacing correction coefficient (α). One possible implementation of this function is: α = max(0.6, 1 - β × V_missing). Here, the max() function ensures that the correction coefficient has a lower limit (e.g., 0.6) to prevent the spacing from being reduced too much, leading to a surge in the number of beams and excessive processing burden; β is the aforementioned volume influence weighting coefficient. According to this formula, the larger the total estimated volume V_missing, the larger the β × V_missing term, resulting in a smaller (1 - β × V_missing). However, due to the existence of the max function, α will eventually approach the lower limit value of 0.6, meaning that the maximum spacing reduction is stable at 40%, rather than shrinking indefinitely. Conversely, if V_missing is very small, the α value is close to 1, and the spacing hardly needs to be reduced. Finally, the system will multiply the maximum spacing D0 by the calculated correction coefficient α to obtain the maximum spacing limit that the array should follow after this adjustment (D_new=D0×α). Based on this new limit, the transducer array is re-optimized within the possible range. For example, if D0 is 0.5 meters, V_missing is 200 cubic meters, and β=0.0005 / cubic meter, then α=max(0.6,1-0.0005×200)=max(0.6,0.9)=0.9, therefore the new D_new=0.5×0.9=0.45 meters.

[0082] Through the above scheme, this invention achieves intelligent linkage adjustment between transducer array parameters and terrain loss features. When the total volume of the loss is large, it indicates extremely complex terrain and extensive blind spots. In this case, excessively reducing the spacing, while increasing beam density, would significantly increase data volume and processing time, and have limited effect on improving the extensive blind spots. This scheme achieves an optimal balance between coverage, imaging resolution, and system processing load by making the reduction magnitude decrease as the volume increases. This collaborative adjustment mechanism systematically and economically improves the integrity and accuracy of the final 3D imaging for complex terrain.

[0083] Specifically, in step S53, steps S2 to S5 are re-executed according to the adjusted parameters. If the imaging missing rate is still greater than or equal to the preset imaging missing rate, and the determined reason for the imaging failure is still insufficient beam coverage due to the complex terrain to be measured, the normal angle of the output beam of each transducer is adjusted.

[0084] The initial adjustment angle is determined based on the angle between the original normal angle of each beam and the line connecting the center point of the corresponding missing region. The initial adjustment angle is then corrected based on the current adjusted vertical angle width to obtain the final adjustment angle.

[0085] The larger the width of the currently adjusted vertical angle, the greater the reduction in the initial adjustment angle.

[0086] In this embodiment of the invention, firstly, for each beam requiring adjustment, the system locates the geometric center point of one or more missing regions within its coverage area based on the initial three-dimensional imaging model. The absolute angle between the beam's current original normal angle (i.e., the direction of the beam's central axis) and the target direction pointing from the transducer to the center point of the missing region is calculated and denoted as the initial adjustment angle θ_initial. This angle represents the theoretically required adjustment amount to directly turn the beam center towards the missing region; direct adjustment according to θ_initial may lead to over-adjustment because the increased vertical angular width itself has already expanded the beam's coverage. Therefore, an angle correction coefficient μ (0 < μ ≤ 1) is introduced to reduce θ_initial. The correction coefficient μ is determined by the current adjusted vertical angular width φ_current of the beam, and its functional relationship is: μ = max(0.1, λ / φ_current). Here, λ is a key preset angle attenuation benchmark value (in degrees). The method for determining the preset angle attenuation benchmark value λ is based on historical big data: the system analyzes successful optimization cases in history where insufficient coverage was caused by complex terrain, and extracts from these cases the "actual adjustment amount (θ_final) of the effective normal angle that ultimately makes the imaging acceptable under a specific vertical angle width (φ_current)". By analyzing a large number of (θ_final, φ_current) data pairs, an empirical relationship of θ_final≈(λ / φ_current)×θ_initial is fitted, and then the optimal value of λ is statistically determined (e.g., λ=5.0 degrees). This value reflects the average influence level of vertical angle width on pointing adjustment sensitivity; finally, the formula for calculating the final adjustment angle θ_final is: θ_final=μ×θ_initial=(λ / φ_current)×θ_initial. This formula reflects the core principle that "the larger the current adjusted vertical angle width φ_current, the greater the reduction in the initial adjustment angle θ_initial". For example, if the current φ of a beam increases from an initial 10° to 15°, and λ is 5.0 degrees, the calculated μ will be 0.33. If its initial θ is 9°, then only about 3° of adjustment is needed in the end.

[0087] Through the above scheme, when the initial parameter optimization effect is insufficient, this invention initiates a secondary fine-tuning based on model feedback. This method does not simply point the beam towards the center of the missing region, but rather uses the increment of the vertical angle width as adjustment damping to dynamically calculate a more robust and less prone-to-over-adjustment correction angle. The preset value λ is learned through historical data, ensuring the reliability and adaptability of the adjustment strategy. This step-by-step, progressive, and each-step-dependent-on-the-previous-step intelligent adjustment mechanism significantly improves the system's adaptive imaging capability for extremely complex terrain, approximating the optimal beam pointing configuration with higher efficiency, thereby minimizing detection blind spots and improving the integrity of 3D imaging.

[0088] Specifically, in step S54, when it is determined that the cause of the imaging failure is environmental interference, the following steps are included:

[0089] Detect the resolution of the initial three-dimensional imaging model;

[0090] If the resolution is less than a preset resolution threshold, the preprocessing parameters are adjusted.

[0091] If the resolution is greater than or equal to the preset resolution threshold, then the azimuth width in the beam parameters of a single beam is increased.

[0092] Specifically, in step S54, the increase in the lateral angle width in the beam parameters of a single beam is determined based on the ratio of the current model resolution to the preset resolution threshold and the imaging missing rate.

[0093] In this embodiment of the invention, the preset resolution threshold R_threshold is a key preset value used to distinguish the type of interference. Its determination method is based on historical big data: the system maintains a case library of known cases affected by environmental interference (such as strong ocean currents or high suspended matter concentrations) but which ultimately achieved successful imaging after parameter optimization. The model resolution data of these cases before optimization, when the imaging was unsatisfactory, are extracted to form a sample set. R_threshold is then set as the statistical lower quartile (i.e., the 25th percentile) of the resolution of this sample set.

[0094] In this embodiment of the invention, a historical case library is maintained, storing a large number of cases where low resolution was caused by environmental interference and was successfully optimized by adjusting preprocessing parameters. Each case includes at least: the original echo signal segment or key interference feature vectors extracted from it (e.g., energy distribution, signal-to-noise ratio statistics, time-domain pulse broadening coefficient, etc. in a specific frequency band). The case ultimately adopts and verifies an effective combination of preprocessing parameters (e.g., a specific set of filter bandwidths, thresholds, and algorithm types), and the final resolution improvement achieved after adjustment. When it is determined that preprocessing parameters need to be adjusted, firstly, interference feature vectors consistent with the definition in the historical case library are extracted from the original echo signal acquired this time. The similarity between the interference feature vector of the current signal and all feature vectors in the historical case library is calculated (e.g., using Euclidean distance or cosine similarity). The top K historical cases with the highest similarity (e.g., K=5) are selected as references. The selected reference cases are analyzed, and adjustment instructions are generated using a weighted fusion method. The instruction generation logic is as follows: Filtering algorithm selection: If more than a certain proportion (e.g., 60%) of the reference cases use the same advanced algorithm (e.g., wavelet threshold denoising replacing conventional bandpass filtering), the instruction will preferentially switch to that algorithm; Parameter adjustment calculation: For continuously adjustable parameters (e.g., bandwidth B, threshold T), the suggested adjustment value (B_new, T_new) is calculated based on the parameter values ​​of the reference cases and their similarity weight with the current signal.

[0095] In this embodiment of the invention, the amplification factor F of the lateral angle width is calculated using a preset adjustment formula based on the ratio of the current model resolution R_current to the preset resolution threshold R_threshold (resolution factor) and the current imaging missing rate M. The formula is designed as: Amplification factor F = L × (R_current / R_threshold) × (1 / (1+δ×M));

[0096] Wherein, L is the base increase factor, a preset constant (e.g., 1.2) determined based on historical optimization case statistics, representing the recommended baseline adjustment range under ideal conditions (resolution meets the standard and there are no missing areas); (R_current / R_threshold) is the resolution factor, which is greater than or equal to 1 (because the condition for triggering this branch is R_current≥R_threshold). The higher the resolution, the greater the sacrifice of angular resolution is allowed in exchange for heading coverage; (1+δ×M) is the missing rate excitation factor, where δ is a preset positive sensitivity coefficient (e.g., 0.3). This design is based on deep-sea operation experience: when environmental interference (such as suspended matter, turbulence) leads to a high imaging missing rate M, it indicates that the single beam signal-to-noise ratio is insufficient; moderately increasing the lateral angle width F is equivalent to performing beamforming along the heading, which can effectively improve the echo signal strength and suppress random interference, thereby potentially compensating for missing areas, although it will slightly lose lateral resolution. The sensitivity coefficient δ is obtained by fitting the "effect of lateral angle width adjustment on the reduction of missing rate under different interference intensities" in historical cases.

[0097] Specifically, in this embodiment, taking the engineering prototype of the 4500-meter-class deep-sea multibeam three-dimensional acoustic detection payload as an example, its initial parameters are set as follows: transducer element spacing 0.05m, beamwidth 0.8° (along) × 1.5° (vertical), number of beams 512, frame rate 50Hz; when scanning a complex hydrothermal vent area, the calculated imaging missing rate after the first round of imaging is 25% (greater than the preset threshold of 15%), and the model complexity is calculated to be 85 (greater than the preset threshold of 60); the system diagnoses the cause as insufficient beam coverage due to extremely complex terrain. Based on the complexity and the steep edge gradient of the missing area, the vertical angle width was calculated to need to be increased to 2.0°. At the same time, based on the total estimated missing volume, the maximum spacing of the array elements was fine-tuned to 0.048m. After rescanning with the new parameters, the missing rate was reduced to 18%, but it was still unacceptable. The system diagnosed the cause again as complex terrain. A second optimization was triggered, calculating the deviation angle between the original direction of each beam and the center of the remaining missing area, and using the increased vertical angle width (2.0°) as damping for correction, generating a new beam normal angle command. After the third scan, the missing rate was reduced to 10%, the system determined that the imaging was qualified, and output a high-precision three-dimensional model of the hydrothermal area. This model then guided the onboard high-definition camera on the same platform to perform close-range optical imaging verification of the identified suspected chimney, achieving efficient acoustic-optical collaborative detection.

[0098] Through the above-described approach, this invention transforms the adjustment of preprocessing parameters from experience-based trial and error to data-driven intelligent decision-making. By establishing and utilizing a historical big data case library, the system can "recall" and integrate historically proven effective processing strategies based on the specific interference characteristics encountered, thereby achieving precise and efficient parameter optimization. This not only significantly improves the success rate and efficiency of parameter adjustment but also reduces reliance on the professional experience of operators, enhancing the imaging system's adaptability and overall robustness in complex and variable underwater acoustic environments.

[0099] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0100] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A high-precision multi-beam acoustic three-dimensional imaging method based on big data, characterized in that, include: Step S1: Determine the imaging parameters, which include the spacing between adjacent transducers in the transducer array, the beam parameters of a single beam output by each transducer, and data processing standard parameters. Step S2: Arrange the transducer array according to the imaging parameters, control the transducer array to output multiple beams to scan the target area, and receive echo signals; Step S3: Perform noise reduction and filtering preprocessing on the echo signal, and map the preprocessed echo data to a three-dimensional spatial coordinate system to generate an initial three-dimensional imaging model. Step S4: Based on big data, match and estimate the features of the missing regions in the initial three-dimensional imaging model to determine the estimated missing volume, and calculate the imaging missing rate based on the estimated missing volume. Step S5: Determine whether the current imaging is qualified based on the imaging missing rate, and if the current imaging is unqualified, determine the reason for the imaging failure based on the complexity of the initial three-dimensional imaging model, and generate corresponding imaging parameter correction instructions according to the reason for the imaging failure. In step S5, determining whether the current imaging is qualified based on the imaging missing rate includes: If the imaging missing rate is less than the preset imaging missing rate, the current imaging is determined to be qualified, and the final three-dimensional imaging model is output. If the imaging missing rate is greater than or equal to the preset imaging missing rate, the current imaging is determined to be unqualified, and the reason for the imaging failure is determined based on the complexity of the initial three-dimensional imaging model. In step S5, the reasons for imaging failure are determined based on the complexity of the initial three-dimensional imaging model, including: Step S51: Calculate the complexity based on the total number of triangular facets in the initial three-dimensional imaging model and the area regularity of each triangular facet. Step S52: Compare the complexity with a preset complexity threshold; Step S53: If the complexity is greater than or equal to the preset complexity threshold, then the reason for the imaging failure is determined to be insufficient beam coverage due to the complexity of the terrain to be tested. Step S54: If the complexity is less than the preset complexity threshold, then the reason for the imaging failure is determined to be environmental interference.

2. The high-precision multi-beam acoustic three-dimensional imaging method based on big data according to claim 1, characterized in that, In step S4, the imaging missing rate is the ratio of the estimated missing volume to the scanned terrain volume.

3. The high-precision multi-beam acoustic three-dimensional imaging method based on big data according to claim 1, characterized in that, In step S51, the complexity is calculated based on the total number of triangular facets in the initial three-dimensional imaging model and the area regularity of each triangular facet, including: Step S511: Count the total number of all discrete triangular patches in the initial three-dimensional imaging model, and use it as the first calculation factor; Step S512: Calculate the ratio of the actual area of ​​each triangular facet in the model to a preset standard area, and calculate the arithmetic mean of all ratios as the second calculation factor; Step S513: Calculate the intermediate complexity value and adjust the intermediate complexity value using a correction coefficient determined based on historical data to obtain the final complexity.

4. The high-precision multi-beam acoustic three-dimensional imaging method based on big data according to claim 1, characterized in that, In step S53, when it is determined that the imaging failure is due to insufficient beam coverage caused by complex terrain, the vertical angle width in the beam parameters of a single beam is adjusted. Specifically, the adjustment range of the vertical angle width is determined based on the complexity of the initial three-dimensional imaging model and the average depth gradient of the edge points of each missing region. The higher the complexity and the larger the average depth gradient, the greater the increase in the vertical angle width.

5. The high-precision multi-beam acoustic three-dimensional imaging method based on big data according to claim 4, characterized in that, In step S53, during the adjustment of the vertical angle width, the maximum spacing between adjacent transducers in the transducer array is simultaneously corrected based on the total estimated volume of all missing regions; wherein, the larger the total estimated volume, the smaller the reduction in the maximum spacing.

6. The high-precision multi-beam acoustic three-dimensional imaging method based on big data according to claim 4, characterized in that, In step S53, steps S2 to S5 are re-executed according to the adjusted parameters. If the imaging missing rate is still greater than or equal to the preset imaging missing rate, and the determined reason for the imaging failure is still insufficient beam coverage due to the complex terrain to be measured, the normal angle of the output beam of each transducer is adjusted. The initial adjustment angle is determined based on the angle between the original normal angle of each beam and the line connecting the center point of the corresponding missing region. The initial adjustment angle is then corrected based on the current adjusted vertical angle width to obtain the final adjustment angle. The larger the width of the currently adjusted vertical angle, the greater the reduction in the initial adjustment angle.

7. The high-precision multi-beam acoustic three-dimensional imaging method based on big data according to claim 1, characterized in that, In step S54, when it is determined that the cause of the imaging failure is environmental interference, the following steps are included: Detect the resolution of the initial three-dimensional imaging model; If the resolution is less than a preset resolution threshold, the preprocessing parameters are adjusted. If the resolution is greater than or equal to the preset resolution threshold, then the azimuth width in the beam parameters of a single beam is increased.

8. The high-precision multi-beam acoustic three-dimensional imaging method based on big data according to claim 7, characterized in that, In step S54, the increase in the lateral angle width in the beam parameters of a single beam is determined based on the ratio of the current model resolution to the preset resolution threshold and the imaging missing rate.

Images