A sonar imaging-based marine organism distribution detection system and method

Abstract

This invention relates to the field of sonar imaging technology and discloses a marine biological distribution detection system and method based on sonar imaging. The method includes: setting up multiple depth layers and acquiring acoustic echo signals from each depth layer; performing sparse observation of marine biomass blocks based on the three-dimensional acoustic echo signals to obtain an observation matrix and calculate a set of effective biological signal blocks; calculating the spatial coordinates and reflection intensity of each biological target based on the set of effective biological signal blocks to obtain biological positioning data for each depth layer and obtaining a three-dimensional acoustic image of each depth layer through three-dimensional imaging; calculating the biological density index, Shannon-Wiener diversity index, and vertical migration velocity of the scattering coefficient within each ecological layer, and outputting the biological detection results for each depth layer. This method acquires data at multiple depth layers, improving detection reliability in low signal-to-noise ratio marine environments and accurately identifying marine biological populations and distribution characteristics at different depth layers.

Description

Technical Field

[0001] This invention relates to the field of sonar imaging technology, and in particular to a sonar imaging-based system and method for detecting the distribution of marine organisms. Background Technology

[0002] Traditional methods for detecting marine organism distribution rely on single-frequency or multi-beam sonar for two-dimensional planar scanning. These methods have significant limitations when dealing with complex marine organism distributions. Traditional sonar systems typically employ a horizontal scanning mode at a single depth layer, making it difficult to obtain complete vertical distribution information of marine organisms. Furthermore, the lack of accurate modeling of marine acoustic propagation characteristics during signal processing leads to insufficient target localization accuracy. In densely populated marine environments with schools of fish and plankton, the echo signals from multiple biological targets can severely overlap and interfere, resulting in inaccurate counting of individual organisms, misidentification of species, and spatial positioning errors. Summary of the Invention

[0003] This invention provides a marine organism distribution detection system and method based on sonar imaging. This invention collects data at multiple depth layers, which improves the detection reliability in marine environments with low signal-to-noise ratios and can accurately identify marine organism populations and distribution characteristics at different depth layers.

[0004] In a first aspect, the present invention provides a method for detecting the distribution of marine organisms based on sonar imaging, the method comprising:

[0005] Set up multiple depth layers and acquire acoustic echo signals from each depth layer;

[0006] Based on the three-dimensional acoustic echo signal, sparse observation of marine biomass blocks is performed to obtain an observation matrix, and the set of effective biomass signal blocks is calculated according to each signal block in the observation matrix.

[0007] The spatial coordinates and reflection intensity of each biological target are calculated based on the set of effective biological signal blocks to obtain biological positioning data for each depth layer. Three-dimensional imaging is then performed based on the biological positioning data to obtain a three-dimensional acoustic image for each depth layer.

[0008] Based on the three-dimensional acoustic images of each depth layer, the biological density index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer are calculated, and the biological detection results of each depth layer are output.

[0009] In conjunction with the first aspect, in a first implementation of the first aspect of the present invention, the step of setting multiple depth layers and acquiring acoustic echo signals from each depth layer includes:

[0010] Multiple depth layers are set up, and the autonomous underwater vehicle is controlled to emit linear frequency modulated signals using preset sonar at multiple depth layers according to a preset spiral trajectory and receive the raw echo data of each depth layer.

[0011] The original echo data is digitized in the time domain while the spatial position information of each azimuth angle is recorded to obtain the acoustic echo signal of each depth layer.

[0012] In conjunction with the first aspect, in a second implementation of the first aspect of the present invention, the step of performing sparse observation of marine biomass blocks based on the three-dimensional acoustic echo signal to obtain an observation matrix, and calculating a set of effective biosignal blocks according to each signal block in the observation matrix, includes:

[0013] The acoustic echo signal is expanded into a received signal matrix according to the time sampling points and azimuth angle, and the spatial grid points of the target detection area in each depth layer are set as the signal observation data combination.

[0014] The propagation compensation coefficient is calculated based on the propagation path from each spatial grid point to the receiving array element in the combined signal observation data.

[0015] Based on the propagation compensation coefficient, the target detection area is divided into multiple sparse blocks, and the acoustic transfer parameters of the spatial grid points in each sparse block are calculated.

[0016] An observation matrix for each depth layer is generated based on the acoustic transfer parameters.

[0017] The set of effective biosignal blocks for each depth layer is calculated based on the observation matrix.

[0018] In conjunction with the first aspect, in a third implementation of the first aspect of the present invention, the step of dividing the target detection region into multiple sparse blocks according to the propagation compensation coefficient and calculating the acoustic transfer parameters of the spatial grid points within each sparse block includes:

[0019] The acoustic intensity distribution characteristics of each spatial grid point in the target detection area are analyzed using the propagation compensation coefficient, and the block size allocation strategy is determined by combining the marine biological community aggregation distribution pattern and the acoustic intensity distribution characteristics.

[0020] The fish aggregation area and the plankton aggregation area are divided into multiple sparse blocks according to the block size allocation strategy described above;

[0021] The propagation delay and phase information of each receiving array element are calculated for each spatial grid point in each sparse block, and the acoustic attenuation effect is corrected by the propagation compensation coefficient to obtain the acoustic transmission parameters of each spatial grid point in each sparse block.

[0022] In conjunction with the first aspect, in a fourth implementation of the first aspect of the present invention, the step of calculating the set of effective biosignal blocks for each depth layer based on the observation matrix includes:

[0023] Based on the observation matrix, calculate the residual ratio function sequence for each sparse block;

[0024] The residual ratio statistical mean and standard deviation are calculated by combining the shape and scale parameters of the gamma distribution under pure ocean noise environment, and the detection threshold of marine biological signals is set based on the residual ratio statistical mean and standard deviation;

[0025] The effective biological signal block set for each depth layer is selected based on the residual ratio function sequence of each sparse block and the marine biological signal detection threshold.

[0026] In conjunction with the first aspect, in a fifth implementation of the first aspect of the present invention, the step of calculating the spatial coordinates and reflection intensity of each biological target based on the set of effective biosignal blocks to obtain biolocation data for each depth layer, and performing three-dimensional imaging based on the biolocation data to obtain a three-dimensional acoustic image for each depth layer, includes:

[0027] A prior distribution parameter for the reflection intensity of each biological target in the set of effective biological signal blocks is set, and a combination of hyperparameters is determined according to the marine species to construct a prior probability model of biological targets.

[0028] Based on the prior probability model of the biological target, the expected step is executed to calculate the posterior distribution parameters of each biological signal block;

[0029] The updated model parameters are obtained by performing a maximization step using the posterior distribution parameters.

[0030] Calculate the rate of change of the updated model parameters, and when the rate of change of the parameters is lower than the convergence threshold, output the spatial coordinates and reflection intensity of each biological target to obtain the biological positioning data of each depth layer;

[0031] Three-dimensional imaging is performed based on the biolocation data to obtain three-dimensional acoustic images of each depth layer.

[0032] In conjunction with the first aspect, in a sixth implementation of the first aspect of the present invention, the step of performing a maximization step update using the posterior distribution parameters to obtain the updated model parameters includes:

[0033] The squared value of the second norm is calculated based on the posterior mean vector of each biological signal block in the posterior distribution parameters. At the same time, the trace value of the posterior covariance matrix is ​​calculated and added to the shape parameter of the corresponding marine species to obtain the numerator and denominator statistics.

[0034] The updated biosignal block precision parameters are obtained by performing fractional calculations using the numerator and denominator statistics combined with the scale parameters of each marine organism species.

[0035] The squared residual vector norms of the observed signal and the reconstructed signal are calculated based on the updated biosignal block precision parameters and then added to the residual norms to obtain the total residual statistics.

[0036] The updated noise accuracy parameters are generated based on the total residual statistics, and the updated noise accuracy parameters are combined with the updated biosignal block accuracy parameters to obtain the updated model parameters.

[0037] In conjunction with the first aspect, in the seventh implementation of the first aspect of the present invention, the step of performing three-dimensional imaging based on the bio-location data to obtain three-dimensional acoustic images of each depth layer includes:

[0038] Based on the spatial coordinates and reflection intensity of each biological target in the biological positioning data, the linear frequency modulated signal is subjected to frequency domain conjugate matched filtering to obtain range-compressed echo data.

[0039] The azimuth-focused imaging data is calculated by combining the range-compressed echo data with the spiral trajectory of the autonomous underwater vehicle.

[0040] Fractional Fourier transform compensation is performed on the azimuth-focused imaging data to obtain Doppler-corrected signal data, and the Doppler-corrected signal data is reconstructed to obtain a three-dimensional acoustic image of each depth layer.

[0041] In conjunction with the first aspect, in the eighth implementation of the first aspect of the present invention, the step of calculating the biological density index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer based on the three-dimensional acoustic image of each depth layer, and outputting the biological detection results of each depth layer, includes:

[0042] Marine organisms were classified and identified from the three-dimensional acoustic images of each depth layer to obtain the biological identification results for each depth layer.

[0043] Based on the biometric results, the biological density index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer are calculated, and the biological detection results of each depth layer are output.

[0044] Secondly, the present invention provides a sonar imaging-based marine organism distribution detection system, the sonar imaging-based marine organism distribution detection system comprising:

[0045] The acquisition module is used to set up multiple depth layers and acquire the acoustic echo signal of each depth layer;

[0046] The calculation module is used to perform sparse observation of marine biomass based on the three-dimensional acoustic echo signal, obtain the observation matrix, and calculate the set of effective biosignal blocks according to each signal block in the observation matrix.

[0047] The construction module is used to calculate the spatial coordinates and reflection intensity of each biological target based on the set of effective biological signal blocks, obtain biological positioning data for each depth layer, and perform three-dimensional imaging based on the biological positioning data to obtain a three-dimensional acoustic image for each depth layer.

[0048] The biodetection module is used to calculate the biodensity index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer based on the three-dimensional acoustic image of each depth layer, and output the biodetection results of each depth layer.

[0049] The technical solution provided by this invention employs an autonomous underwater vehicle (AUV) to collect data at multiple depth layers along a preset spiral trajectory, breaking through the limitations of traditional single-layer horizontal scanning. This allows for the acquisition of complete three-dimensional acoustic information. A block sparse signal model, established based on the aggregation and distribution characteristics of marine biological populations, fully considers marine acoustic propagation characteristics such as spherical diffusion loss and frequency absorption attenuation, making it more consistent with the actual distribution patterns of marine organisms compared to traditional point sparse models. Adaptive block partitioning is performed according to the aggregation scale of different marine biological species, setting different block size ranges for large fish and plankton, effectively improving the accuracy and adaptability of sparse reconstruction. A gamma distribution model is established based on statistical analysis of a pure marine noise environment, and an adaptive detection threshold is set to screen effective biological signal blocks, improving detection reliability in low signal-to-noise ratio marine environments. A hierarchical prior distribution and expectation-maximization iterative Bayesian framework are used to achieve automatic separation and localization of multiple overlapping biological sound sources without requiring prior target quantity information, demonstrating excellent blind source separation capabilities. By combining range compression, azimuth focusing, and fractional Fourier transform Doppler compensation in a multi-domain processing technique, the impact of marine life movement on imaging quality is effectively compensated, generating high-quality 3D acoustic images. In the process of marine species identification, ecological parameter calculations are incorporated to achieve automatic layered detection at different depths, accurately identifying marine biological populations and distribution characteristics at different depths.

[0050] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of an embodiment of the marine organism distribution detection method based on sonar imaging in this invention.

[0053] Figure 2 This is a schematic diagram of one embodiment of the marine organism distribution detection system based on sonar imaging in this invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] The terms "comprising" and "having," and any variations thereof, used in the embodiments of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0056] To facilitate understanding of this embodiment, a detailed description of a sonar imaging-based method for detecting the distribution of marine organisms disclosed in this invention will be provided first. For example... Figure 1 As shown, this method includes the following steps:

[0057] 101. Set up multiple depth layers and acquire the acoustic echo signal of each depth layer;

[0058] Specifically, multiple vertically distributed depth layers are established within the target marine area, with a vertical spacing of 10 meters between each layer to ensure sufficient acoustic resolution and spatial coverage continuity in the vertical direction. An autonomous underwater vehicle (AUV) carries a 256-element circular transducer array. The array elements are uniformly arranged in a ring with a λ / 2 spacing, providing omnidirectional transmission and reception capabilities. A circular array synthetic aperture sonar system implements a three-dimensional spiral trajectory at each depth layer, with a fixed spiral descent angle of 15 degrees and a fixed circular scanning radius of 200 meters. Approximately 30% overlap in acoustic coverage areas is ensured between adjacent layers, achieving multi-layer continuous three-dimensional sampling. During operation at each depth layer, the control system drives the sonar transmitter to periodically emit a preset linear frequency modulated signal, with a frequency range around 120 kHz, a pulse width of 1 ms, and a bandwidth of 30 kHz. After propagating in the water, the emitted signal is scattered by biological targets and the reverse echo signal is captured by the array receiver. The raw echo data is sampled at high frequency, and the echo response of each array element is recorded synchronously using a time-domain digital sampling rate of 500kHz. At the same time, the corresponding azimuth angle θ and spatial position coordinates are recorded in real time based on the AUV's inertial navigation system and trajectory planning data. Each set of data collected contains time-domain response and three-dimensional spatial position information. Three-dimensional acoustic echo matrices containing time sampling points, azimuth angles, and depth index dimensions are constructed at each depth layer.

[0059] 102. Based on three-dimensional acoustic echo signals, sparse observation of marine biomass was carried out to obtain the observation matrix, and the set of effective biomass signal blocks was calculated according to each signal block in the observation matrix.

[0060] Specifically, the three-dimensional acoustic echo signal is expanded according to the time sampling points and azimuth dimensions to construct a two-dimensional received signal matrix Y. Each row corresponds to the echo data of an array element at a certain time snapshot, and each column represents a time sampling sequence. Simultaneously, the position and attitude information of the AUV is combined to associate each set of data in the matrix with specific three-dimensional spatial coordinates. A three-dimensional spatial grid point set of the target detection area is constructed for each depth layer. Equidistantly distributed three-dimensional coordinate points are set in the water space as the signal inversion location domain, forming a grid structure of the observation data combination X. Each grid point represents the possible location of a potential biological target. Based on the propagation path distance from each grid point to each receiving element in the array, a composite propagation compensation coefficient including spherical diffusion loss and frequency absorption attenuation is calculated, specifically the spherical term α(r) = 1 / r. 2 With the absorption term β(f,r)=exp(-αf) 2The propagation distance (r) is calculated, where r is the propagation distance, f is the operating frequency, and α is the absorption coefficient. This establishes an acoustic transfer relationship between each grid point and the receiving array element. Based on the propagation compensation coefficient, the three-dimensional target detection area is divided into several block-shaped sparse units. Each sparse block contains multiple spatially adjacent grid points. The acoustic transfer parameters, i.e., the transfer function from grid point n to receiving array element m, are calculated for each grid point within each sparse block, forming an observation matrix with a block-sparse structure. Based on the observation matrix, a statistical algorithm based on the residual ratio detection principle, such as the marine biological block residual ratio detection method, is used to solve the acoustic echo signal matrix and the observation matrix simultaneously. This dynamically evaluates the rate of change of residual signal energy in each sparse block and uses a threshold strategy set by the statistical model to select effective sparse blocks with actual scattering characteristics, forming a set of effective biological signal blocks corresponding to each depth layer.

[0061] 103. Calculate the spatial coordinates and reflection intensity of each biological target based on the set of effective biological signal blocks to obtain the biological positioning data of each depth layer, and perform three-dimensional imaging based on the biological positioning data to obtain the three-dimensional acoustic image of each depth layer.

[0062] Specifically, for potential biological targets within the effective biosignal block set, prior distribution parameters for the reflectance intensity of biological targets are set based on their spatial scale, reflectance characteristics, and ecological attributes. Corresponding hyperparameter combinations are determined according to marine organism categories such as large fish, small fish, or plankton to construct an adaptive prior probability model for biological targets. The model employs a hierarchical Bayesian structure, where the reflectance intensity variable is set to a Gaussian distribution and the accuracy parameter to a Gamma distribution, forming an automatic correlation determination mechanism with inherent sparsity control capabilities. Based on the prior model, the expectation step of the expectation-maximization algorithm is executed. By inputting the observed signal and the constructed observation matrix, combined with the prior structure, the posterior distribution parameters of each biosignal block are calculated, obtaining the posterior mean and posterior covariance matrix of each biosignal block, describing the estimated reflectance intensity and uncertainty of the current biosignal block in the spatial grid. Using the posterior distribution parameters, the hyperparameter combinations in the model are iteratively updated in the maximization step, including the accuracy coefficients of each sparse block and the system noise accuracy term, achieving stepwise optimization of the model in three dimensions: spatial, intensity, and uncertainty. In each iteration, calculate the relative rate of change of the model parameters before and after the update, and compare the relative rate of change with a set convergence threshold (e.g., 10). -4The model is considered converged when all parameter change rates are below the convergence threshold. Spatial coordinates with significant posterior intensity values ​​and their corresponding reflection intensities are extracted from each effective signal block to obtain all locatable marine biological targets in the current depth layer. Based on the biological positioning data, spatial inversion and image restoration are performed using multi-domain fusion techniques such as combined time-delay compression, azimuth focusing, and depth-direction sparse reconstruction. Biological targets are mapped to a three-dimensional grid space according to the three dimensions of distance, azimuth, and depth, forming a three-dimensional acoustic image with actual physical size and acoustic response characteristics, reflecting the distribution pattern of marine organisms in each depth layer.

[0063] 104. Calculate the biological density index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer based on the three-dimensional acoustic images of each depth layer, and output the biological detection results of each depth layer.

[0064] Specifically, the 3D acoustic images at each depth layer are spatially segmented, and marine organism classification and identification are performed using machine learning strategies. The identification module employs a deep convolutional neural network (BioCNN) architecture, using the 3D local voxel regions in each image as input units. Multi-scale feature extraction is performed using a multi-layer convolutional structure, and the trained classification model outputs the probability distribution of typical marine organisms corresponding to each input block, yielding the bio-identification results for each depth layer. Based on the bio-identification results, the scattering coefficient distribution of the identified organisms is statistically analyzed in each ecological layer, and a bio-density index reflecting the target population density is calculated to measure the concentration of acoustic reflection energy per unit volume. Simultaneously, to assess the ecological diversity of species distribution, the Shannon-Wiener diversity index is used to quantify the relative abundance of each organism, effectively reflecting the balance and complexity of the biological community in terms of species composition. To capture the dynamic changes of organisms in the vertical direction, the positional changes of the same biological category in the depth dimension are extracted based on acoustic image data across time frames. The vertical migration speed is obtained by calculating the rate of change of biological position at continuous moments, used to identify the upstream and downstream movement behavior of organisms during the diurnal cycle. Based on ecological characteristics such as biological classification labels, biological density, diversity index, and migration speed, the biological detection results for each depth layer are output separately.

[0065] In one specific embodiment, the process of performing step 101 may specifically include the following steps:

[0066] Multiple depth layers are set up, and the autonomous underwater vehicle is controlled to emit linear frequency modulated signals using preset sonar at multiple depth layers according to a preset spiral trajectory and receive the raw echo data of each depth layer.

[0067] The original echo data is digitized in the time domain while the spatial location information of each azimuth angle is recorded to obtain the acoustic echo signal of each depth layer.

[0068] Specifically, a vertical layering scheme is determined based on the spatial extent and ecological characteristics of the target marine area. The entire water column is vertically divided into multiple depth layers, with equidistant spacing between each layer, resulting in a three-dimensional acoustic coverage structure with strong continuity between different water layers. An autonomous underwater vehicle (AUV) with high-precision attitude control and navigation positioning capabilities is used as a platform, equipped with a circular transducer array system with omnidirectional receiving capabilities. The circular transducer array consists of 256 acoustic elements evenly distributed around the circumference at half the wavelength spacing to ensure the same receiving sensitivity in any azimuth direction. Following a preset helical scanning trajectory, the AUV completes a full circular motion at each depth layer. The helical trajectory settings include a fixed scanning radius, a constant descent angle, and layered speed control parameters, enabling the AUV to continuously complete multiple circular scans at different depth layers while maintaining acoustic coverage overlap between layers, thus constructing a dense and continuous sampling structure in space. During each circular motion, the sonar transmission module periodically emits a linear frequency modulated (LFM) signal. The LFM signal frequency is set around 120kHz, with a specific bandwidth and pulse duration. The emitted acoustic signal propagates to the surface of marine life or other sound scatterers in the water and is reflected. The reflected signal is synchronously received by a circular array receiver, forming the original echo response sequence. At the receiving end, the original analog echo signal is digitized in the time domain at a sampling rate of 500 kHz, transforming it into discrete data with high time resolution. Simultaneously, the inertial navigation system and attitude sensors of the vehicle record the spatial position and azimuth angle corresponding to each sampling time point, so that each time-domain sampled data is accompanied by its corresponding spatial information tag, including metadata such as sampling time, array element number, azimuth angle information, and depth layer number. During the entire acquisition process, the scanning time of each depth layer is approximately 6 minutes, covering 360 degrees and achieving acoustic information sampling in all directions at the corresponding depth layer. Each depth layer forms a multi-dimensional acoustic echo signal dataset, which contains time series information, azimuth angle data, and spatial positioning information, organized in matrix form.

[0069] In one specific embodiment, the process of performing step 102 may specifically include the following steps:

[0070] The acoustic echo signal is expanded into a received signal matrix according to the time sampling points and azimuth angle, and the spatial grid points of the target detection area in each depth layer are set as the signal observation data combination.

[0071] The propagation compensation coefficient is calculated based on the propagation path from each spatial grid point to the receiving array element in the signal observation data combination;

[0072] Based on the propagation compensation coefficient, the target detection area is divided into multiple sparse blocks, and the acoustic transfer parameters of the spatial grid points in each sparse block are calculated.

[0073] The observation matrix for each depth layer is generated based on the acoustic transfer parameters.

[0074] The set of effective biosignal blocks for each depth layer is calculated based on the observation matrix.

[0075] Specifically, data structures in the time and azimuth dimensions are extracted from the acoustic echo signals acquired at each depth layer and reconstructed into a two-dimensional received signal matrix. The rows of this matrix represent the responses of different receiving array elements, and the columns represent time sampling points or equivalent distance indices. Each matrix element carries amplitude and phase information of the acoustic signal along the transmission and reception path. The three-dimensional spatial region covered by each depth layer is divided into a set of equidistant voxel grid points, and these spatial grid points serve as a candidate set of potential target echo reflection points, forming a combination of signal data to be observed. The spatial location of the grid points and their geometric relationship with the array are used to construct the echo propagation path model. For each spatial grid point in the combination, the propagation path length is calculated based on the geometric distance from the current grid point to each receiving array element. Considering the physical mechanism of acoustic wave propagation in the marine environment, propagation loss and frequency absorption terms are comprehensively introduced to compensate for path effects. Specifically, this includes an amplitude attenuation factor generated by spherical diffusion and a frequency-squared related absorption factor, generating a set of propagation compensation coefficients to characterize the energy attenuation and phase delay characteristics of each path. Based on the propagation compensation coefficient, the entire target detection area is divided into multiple spatially continuous sparse block structures. Each sparse block contains several adjacent grid points and represents an aggregation region of a type of biological community. Based on the propagation compensation relationship between each grid point in each sparse block and all receiving array elements, a set of acoustic transfer parameters is calculated to describe the energy attenuation, phase rotation, and path delay experienced by the echo signal emitted from a specific spatial point during propagation to the receiving array. The transfer parameters of all sparse blocks are combined, and an observation matrix is ​​constructed according to the correspondence between receiving array elements and spatial grid points. The matrix exhibits multidimensional sparsity in structure and shows a structural mapping relationship with the actual biological distribution within the target area. By combining the observation matrix with the acquired echo signal data, the sparse block set with significant energy residual changes in acoustic inversion is identified through the block sparse reconstruction method. Combined with statistical modeling methods, such as the residual ratio statistical distribution model, the signal intensity change trend of each sparse block is judged and a dynamic threshold is set for filtering. The output is the sparse block set in which biological target echo responses actually exist at the current depth layer, which is the effective biological signal block set at the current depth layer.

[0076] In one specific embodiment, the process of dividing the target detection area into multiple sparse blocks according to the propagation compensation coefficient and calculating the acoustic transfer parameters of the spatial grid points within each sparse block can specifically include the following steps:

[0077] The acoustic intensity distribution characteristics of each spatial grid point in the target detection area are analyzed using the propagation compensation coefficient, and the block size allocation strategy is determined by combining the aggregation and distribution patterns of marine biological populations and the acoustic intensity distribution characteristics.

[0078] The fish aggregation area and the plankton aggregation area are divided into multiple sparse blocks according to the block size allocation strategy;

[0079] The propagation delay and phase information of each receiving array element are calculated for each spatial grid point in each sparse block, and the acoustic attenuation effect is corrected by the propagation compensation coefficient to obtain the acoustic transmission parameters of each spatial grid point in each sparse block.

[0080] Specifically, within the three-dimensional spatial grid constructed at each depth layer, the propagation path from the sonar transmitter to the receiver element at each grid point is analyzed using a propagation compensation model. The composite propagation loss characteristics reflected by the propagation path length and operating frequency are used as propagation compensation coefficients, forming a set of acoustic intensity estimation data at the grid point level. This acoustic intensity data constitutes a preliminary acoustic scattering energy spectrum in the spatial domain, reflecting the distribution differences between potential biological aggregation areas and background noise areas in the entire target region. Ecological understanding of typical marine biological population aggregation patterns is introduced. For example, fish exhibit localized dense aggregation, with small scales and concentrated distribution, while plankton appear in large-scale, low-intensity, and high-diffusion community forms. Therefore, based on indicators such as the continuity of acoustic intensity spatial distribution characteristics, local peak density, and gradient of change, a block size allocation strategy conforming to ecological attributes is established. This block size allocation strategy allows for adaptive adjustment of the spatial scale and boundary morphology of sparse blocks. According to the block size allocation strategy, fish aggregation areas are divided into multiple sparse block structures with smaller block scales and clear boundaries, while plankton aggregation areas are divided into a set of sparse blocks with larger block sizes, blurred boundaries, but extensive coverage. After sparse block partitioning, propagation relationships between all spatial grid points within each sparse block and the receiving array elements are established. This includes propagation delay information calculated based on sound velocity and propagation path length, and phase rotation information derived from the propagation path and frequency, forming a propagation channel response model for each point. Based on the propagation path analysis results, and incorporating propagation compensation coefficients, numerical corrections are made to the energy loss and frequency absorption effects of each path, eliminating intensity attenuation shifts caused by the water propagation environment, ensuring that the acoustic response of each spatial grid point reflects its true scattering capability. The propagation delay, phase response, and compensated energy weights are then integrated to form an acoustic transfer parameter cluster between all spatial grid points and the receiving array elements within each sparse block.

[0081] In one specific embodiment, the process of calculating the set of effective biosignal blocks for each depth layer based on the observation matrix may specifically include the following steps:

[0082] Based on the observation matrix, calculate the residual ratio function sequence for each sparse block;

[0083] The residual ratio statistical mean and standard deviation were calculated by combining the shape and scale parameters of the gamma distribution under pure ocean noise environment, and the detection threshold of marine biological signals was set based on the residual ratio statistical mean and standard deviation.

[0084] The effective biological signal block set for each depth layer is selected based on the residual ratio function sequence of each sparse block and the marine biological signal detection threshold.

[0085] Specifically, the observation matrix of each depth layer is simultaneously processed with the corresponding echo signal matrix, and initial zero-value coefficients are assigned to all sparse blocks in the initialization state. An iterative solution method is adopted, gradually updating the estimated coefficient values ​​within the sparse blocks by minimizing the residual energy. In each iteration, the error signal between the predicted echo and the actual observation data is recalculated, and the error signal is projected onto the structure of each sparse block to evaluate the residual changes of each block in the current iteration round. Based on the error structure, for the residual signal of each sparse block at different iteration steps, the ratio of the residual energy between two iterations is calculated, thus constructing the residual ratio function sequence of the current sparse block, effectively reflecting the signal contribution and error convergence rate of each sparse block during the iterative convergence process. To determine whether the changes in the residual ratio function originate from real biological signal characteristics, statistical modeling results under pure ocean background noise are introduced as a reference benchmark. Long-term sampling of sonar echo data from pure waters without any targets is performed, and statistical distribution fitting analysis is conducted on the residual ratio function sequence. This type of background residual ratio follows a gamma distribution characteristic, and its shape and scale parameters are obtained by fitting historical data. The statistical mean and standard deviation of the residual ratio are calculated by combining the shape and scale parameters of the gamma distribution under pure ocean noise environment, and a detection threshold for marine biological signals is set based on this. This detection threshold is applied to the residual ratio function sequence of each sparse block. By setting coherent judgment criteria, such as the residual ratio consistently falling below the detection threshold across multiple iterations, the presence of genuine marine biological scattering signals in the current sparse block is determined. Furthermore, a redundant judgment mechanism is introduced to effectively avoid misjudgments caused by occasional disturbances or local fluctuations, improving the stability and reliability of the screening results. The set of sparse blocks that meet the detection conditions is considered the set of valid biological signal blocks at the current depth layer.

[0086] In one specific embodiment, the process of performing step 103 may specifically include the following steps:

[0087] Prior distribution parameters of the reflection intensity of each biological target in the effective biological signal block set are set, and hyperparameter combinations are determined according to marine biological species to construct a prior probability model of biological targets;

[0088] Based on the prior probability model of biological targets, the expected steps are executed to calculate the posterior distribution parameters of each biological signal block;

[0089] The updated model parameters are obtained by performing a maximization step using the posterior distribution parameters;

[0090] Calculate the rate of change of the updated model parameters. When the rate of change of the parameters is lower than the convergence threshold, output the spatial coordinates and reflection intensity of each biological target to obtain the biological localization data for each depth layer.

[0091] Three-dimensional imaging is performed based on biolocation data to obtain three-dimensional acoustic images of each depth layer.

[0092] Specifically, within the set of valid biological signal blocks selected at each depth layer, spatial regions with acoustic reflection characteristics are identified, and a prior distribution model of biological target reflection intensity is established for each spatial grid point. This prior distribution model is constructed based on a Bayesian learning framework, describing the scattering intensity of biological targets with a zero-mean, Gaussian distribution. A hierarchical prior structure is constructed by introducing precision control parameters associated with biological species. Simultaneously, hyperparameter combinations are set according to the acoustic reflection characteristics of different biological types. For example, large fish, due to their strong reflection intensity, are assigned smaller hyperparameter scale factors, while plankton, due to their weak reflection, are assigned larger scale factors, allowing the model to reflect adaptability to biological ecological characteristics at the prior level. Based on the biological target prior probability model, the expectation step is executed. Based on the previous round of estimation results and observation data, the posterior distribution of reflection intensity within each sparse block is inferred using the current prior probability model, calculating its posterior mean and covariance. The posterior parameters reflect the uncertainty distribution of the possible location and intensity of biological targets in the current state, and demonstrate the effect of observational information on correcting prior beliefs. The model employs a maximization step update using posterior distribution parameters, and uses posterior statistics to back-update the prior precision hyperparameters and noise control factors, making the model more closely resemble the statistical structure of the actual observation data in the next iteration. After each iteration, the differences in model parameters before and after the update are evaluated. By normalizing the rate of change of the posterior mean and the rate of change of the prior precision, it is determined whether the update has reached convergence. If the rate of change is lower than a preset convergence threshold, the iteration is considered complete, and the 3D coordinates and corresponding reflection values ​​of spatial grid points with significant reflection intensity within each effective sparse block are output as the bio-localization result. Based on the bio-localization data, acoustic scattering images for each depth layer are constructed using spatial coordinates and intensity values. By fusing time compression, azimuth focusing, and sparse inversion techniques along the depth direction, 3D mapping and imaging rendering are performed on the target spatial points, generating 3D acoustic images with physical consistency and spatial resolution.

[0093] In one specific embodiment, the process of performing a maximization step to update the model parameters using the posterior distribution parameters and obtaining the updated model parameters can specifically include the following steps:

[0094] The squared value of the second norm is calculated based on the posterior mean vector of each biological signal block in the posterior distribution parameters. At the same time, the trace value of the posterior covariance matrix is ​​calculated and added to the shape parameter of the corresponding marine species to obtain the numerator and denominator statistics.

[0095] The updated biosignal block precision parameters are obtained by performing fractional operations using numerator and denominator statistics combined with scale parameters of each marine species.

[0096] The squared residual vector norms of the observed and reconstructed signals are calculated based on the updated biosignal block precision parameters and then added to the residual norms to obtain the total residual statistics.

[0097] The updated noise accuracy parameters are generated based on the total residual statistics, and then combined with the updated biosignal block accuracy parameters to obtain the updated model parameters.

[0098] Specifically, within the Bayesian inference framework, the posterior distribution of identified valid biosignal blocks in each depth layer is processed. Each sparse block corresponds to a posterior mean vector and a posterior covariance matrix, which together reflect the current optimal estimate of the biosignal intensity and its uncertainty characteristics. The components of each posterior mean vector are squared and summed to obtain the overall energy distribution information of the signal block estimate. Simultaneously, the trace of the posterior covariance matrix corresponding to the sparse block is calculated to extract the overall variance sum contained in the sparse block estimate, representing the model's current tolerance for errors in that signal block. The trace of the posterior covariance matrix is ​​superimposed with the shape parameter corresponding to the biosignal block to form the numerator and denominator statistical basis required for hyperparameter updates. Based on the numerator and denominator statistics, combined with the pre-defined scale parameter, a fractional update operation is performed by combining the scale parameter as a divisor with the statistics to calculate the updated accuracy parameter of the current signal block in the current iteration. This reflects the model's re-evaluation of the sparsity of the sparse block's signal and dynamically adjusts its importance weights. After updating the accuracy parameters of all sparse blocks, the overall residual of the entire model reconstruction is evaluated. The norm squared value is calculated for the difference vector between all observed signals and their corresponding reconstructed signals, and all residual terms are summed to obtain the total residual statistic representing the overall reconstruction error level. By introducing a predefined noise prior structure and combining it with the total residual value, the noise accuracy parameter for the current iteration is calculated, reflecting the system's estimation level of background noise energy. The updated noise accuracy parameters are then combined with the updated accuracy parameters corresponding to each sparse block to form the updated model parameter set after this iteration.

[0099] In one specific embodiment, the process of performing three-dimensional imaging based on biolocation data to obtain three-dimensional acoustic images of each depth layer can specifically include the following steps:

[0100] Based on the spatial coordinates and reflection intensity of each biological target in the biological positioning data, the linear frequency modulated signal is subjected to frequency domain conjugate matched filtering to obtain range-compressed echo data.

[0101] By combining the range-compressed echo data with the spiral trajectory of the autonomous underwater vehicle, the azimuth-focused imaging data is calculated.

[0102] Fractional Fourier transform compensation is performed on the imaging data focused in the azimuth direction to obtain Doppler-corrected signal data, and the Doppler-corrected signal data is reconstructed to obtain a three-dimensional acoustic image of each depth layer.

[0103] Specifically, based on the three-dimensional spatial coordinates and reflection intensity of each biological target in the bio-location data, this is used as an excitation source to back-match the acquired acoustic echo signal. Simultaneously, based on the structural characteristics of the linear frequency modulated signal transmitted by the system in the frequency domain, a conjugate matched filter is constructed. This filter employs a frequency-domain back-frequency modulation strategy, causing the spectral structure of the transmitted signal to be convolved with the received echo signal in the frequency domain, achieving energy compression in the target range dimension. All echo data are synchronously input into the matched filter, and the main lobe of the output signal in the range dimension is concentrated at the actual target's location, while the edge side lobes are effectively suppressed, resulting in high-resolution range-compressed echo data. The range-compressed echo data is spatially aligned with the spiral motion trajectory executed by the autonomous underwater vehicle during data acquisition. Using the azimuth, position coordinates, and depth parameters at each moment in the trajectory information, combined with the receiving channel encoding under a circular array structure, the compressed echoes acquired from the same target at different azimuth angles are rearranged into a complete azimuth signal sequence, and an azimuth focusing model is constructed accordingly. The azimuth-focusing model calculates the relative positional differences between the same target point and the sonar platform at different azimuth angles and compensates for the corresponding phase errors. This eliminates the geometric distortion effects introduced by the motion trajectory in the azimuth dimension, outputting imaging data with continuous azimuth information. Since targets exhibit certain motion behaviors in water, especially when plankton or fish migrate vertically and radially, frequency shifts are introduced into the echo signals. Therefore, Doppler effect compensation is performed on the azimuth-focused imaging data. Based on the fractional Fourier transform method, the signal energy is refocused in the time-frequency plane by adjusting the rotation angle parameter. The optimal rotation angle is automatically selected using the minimum entropy criterion, concentrating the main signal energy within the expected frequency bandwidth, thus correcting the Doppler frequency shift error caused by biological targets with different radial velocities. The Doppler-compensated signal data is then remapped back to a three-dimensional spatial grid. Based on the positioning coordinates of each biological target and the corrected reflection intensity, a voxel-level acoustic reflection field is reconstructed. Combined with the spatial dimension boundary corresponding to each depth layer, all voxels are mapped to the corresponding three-dimensional volume space, forming a three-dimensional acoustic image.

[0104] In one specific embodiment, the process of performing step 104 may specifically include the following steps:

[0105] Marine organisms were classified and identified from the three-dimensional acoustic images of each depth layer to obtain the biological identification results for each depth layer.

[0106] Based on the biometric results, the biological density index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer are calculated, and the biological detection results of each depth layer are output.

[0107] Specifically, spatial segmentation and feature extraction are performed on the 3D acoustic image. Within each depth layer, the 3D voxel data is divided into several fixed-size local voxel blocks. A trained deep convolutional neural network is then used to analyze each local region block by block. The deep convolutional neural network has the ability to recognize the scattering structure features of typical organisms in the marine environment, and can automatically extract multi-dimensional features such as local texture, edge structure, and reflection intensity distribution in 3D space. Through the combined action of multiple convolutional layers, activation functions, pooling structures, and fully connected layers, the input voxel blocks are mapped into probability distribution vectors. Each dimension of the probability vector corresponds to a specific marine organism category, enabling the identification of the dominant organism species in each local region. After all local voxel blocks have been identified, the identification results are back-mapped to the 3D image space through a spatial fusion strategy, resulting in a biological classification label distribution map covering the entire region on each depth layer, which constitutes the biological identification result of the current depth layer. Based on biometric results, the spatial density of different biological groups within the current layer is statistically analyzed. The biological density index is calculated by statistically analyzing the proportion of the total volume occupied by all voxel blocks belonging to the same category at the current depth layer, combined with the sum of scattering coefficients within the voxel blocks to reflect the energy distribution per unit volume, thus reflecting the local aggregation characteristics of the biological community. Simultaneously, to measure biodiversity levels, the Shannon-Wiener diversity index is used. Based on the proportion of each species in the layer, the complexity of population distribution is calculated, reflecting the balance of community structure and ecological stability. A high index value indicates the presence of diverse and relatively evenly distributed biological groups within the current layer, while a low value indicates that the current layer is dominated by a few dominant species. To monitor the dynamic behavior of marine organisms at different depth layers, the positional changes of the same species at different depth layers over continuous time periods are analyzed. By comparing the classification results at two different times, the vertical positional change trend of the center of gravity of the same type of organism is tracked, and its vertical migration speed is calculated. This is used to identify the diurnal vertical migration behavior of zooplankton or the movement patterns of fish between different thermoclines. The distribution map of biological species, biological density parameters, diversity index and vertical migration speed in each depth layer are fused and output to form a comprehensive biological detection result for each depth layer.

[0108] The above describes the sonar imaging-based marine organism distribution detection method in the embodiments of the present invention. The following describes the sonar imaging-based marine organism distribution detection system in the embodiments of the present invention. Please refer to [link to relevant documentation]. Figure 2 One embodiment of the marine organism distribution detection system based on sonar imaging in this invention includes:

[0109] Acquisition module 201 is used to set multiple depth layers and acquire the acoustic echo signal of each depth layer;

[0110] The calculation module 202 is used to perform sparse observation of marine biomass based on three-dimensional acoustic echo signals, obtain an observation matrix, and calculate the set of effective biosignal blocks based on each signal block in the observation matrix.

[0111] The construction module 203 is used to calculate the spatial coordinates and reflection intensity of each biological target based on the set of effective biological signal blocks, obtain the biological positioning data of each depth layer, and perform three-dimensional imaging based on the biological positioning data to obtain the three-dimensional acoustic image of each depth layer.

[0112] The biological detection module 204 is used to calculate the biological density index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer based on the three-dimensional acoustic image of each depth layer, and output the biological detection results of each depth layer.

[0113] Through the collaborative efforts of the aforementioned components, an autonomous underwater vehicle (AUV) collects data at multiple depths along a pre-set spiral trajectory, overcoming the limitations of traditional single-layer horizontal scanning. This allows for the acquisition of complete three-dimensional acoustic information. A block-sparse signal model, established based on the aggregation and distribution characteristics of marine biological populations, fully considers marine acoustic propagation characteristics such as spherical diffusion loss and frequency absorption attenuation, making it more consistent with the actual distribution patterns of marine organisms compared to traditional point-sparse models. Adaptive block partitioning is performed according to the aggregation scale of different marine biological species, setting different block size ranges for large fish and plankton, effectively improving the accuracy and adaptability of sparse reconstruction. A gamma distribution model is established based on statistical analysis of a pure marine noise environment, and an adaptive detection threshold is set to screen effective biological signal blocks, improving detection reliability in low signal-to-noise ratio marine environments. Employing a hierarchical prior distribution and expectation-maximization iterative Bayesian framework, automatic separation and localization of multiple overlapping biological sound sources can be achieved without prior target quantity information, demonstrating excellent blind source separation capabilities. By combining range compression, azimuth focusing, and fractional Fourier transform Doppler compensation in a multi-domain processing technique, the impact of marine life movement on imaging quality is effectively compensated, generating high-quality 3D acoustic images. In the process of marine species identification, ecological parameter calculations are incorporated to achieve automatic layered detection at different depths, accurately identifying marine biological populations and distribution characteristics at different depths.

[0114] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0115] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0116] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting the distribution of marine organisms based on sonar imaging, characterized in that, include: Set up multiple depth layers and acquire the three-dimensional acoustic echo signal of each depth layer; Based on the three-dimensional acoustic echo signal, sparse observation of marine biomass blocks is performed to obtain an observation matrix, and the set of effective biomass signal blocks is calculated according to each signal block in the observation matrix. The spatial coordinates and reflection intensity of each biological target are calculated based on the set of effective biological signal blocks to obtain biological positioning data for each depth layer. Three-dimensional imaging is then performed based on the biological positioning data to obtain a three-dimensional acoustic image for each depth layer. Based on the three-dimensional acoustic images of each depth layer, the biological density index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer are calculated, and the biological detection results of each depth layer are output.

2. The method for detecting the distribution of marine organisms based on sonar imaging according to claim 1, characterized in that, The process of setting multiple depth layers and acquiring the three-dimensional acoustic echo signal of each depth layer includes: Multiple depth layers are set up, and the autonomous underwater vehicle is controlled to emit linear frequency modulated signals using preset sonar at multiple depth layers according to a preset spiral trajectory and receive the raw echo data of each depth layer. The original echo data is digitized in the time domain, and the spatial position information of each azimuth angle is recorded to obtain the three-dimensional acoustic echo signal of each depth layer.

3. The method for detecting the distribution of marine organisms based on sonar imaging according to claim 1, characterized in that, The process involves sparse observation of marine biomass blocks based on the three-dimensional acoustic echo signal to obtain an observation matrix, and calculating a set of effective biosignal blocks based on each signal block in the observation matrix, including: The three-dimensional acoustic echo signal is expanded into a received signal matrix according to time sampling points and azimuth angle, and at the same time, the spatial grid points of the target detection area in each depth layer are set as the signal observation data combination. The propagation compensation coefficient is calculated based on the propagation path from each spatial grid point to the receiving array element in the combined signal observation data. Based on the propagation compensation coefficient, the target detection area is divided into multiple sparse blocks, and the acoustic transfer parameters of the spatial grid points in each sparse block are calculated. An observation matrix for each depth layer is generated based on the acoustic transfer parameters. The set of effective biosignal blocks for each depth layer is calculated based on the observation matrix.

4. The method for detecting the distribution of marine organisms based on sonar imaging according to claim 3, characterized in that, The step of dividing the target detection area into multiple sparse blocks according to the propagation compensation coefficient and calculating the acoustic transfer parameters of the spatial grid points within each sparse block includes: The acoustic intensity distribution characteristics of each spatial grid point in the target detection area are analyzed using the propagation compensation coefficient, and the block size allocation strategy is determined by combining the marine biological community aggregation distribution pattern and the acoustic intensity distribution characteristics. The fish aggregation area and the plankton aggregation area are divided into multiple sparse blocks according to the block size allocation strategy described above; The propagation delay and phase information of each receiving array element are calculated for each spatial grid point in each sparse block, and the acoustic attenuation effect is corrected by the propagation compensation coefficient to obtain the acoustic transmission parameters of each spatial grid point in each sparse block.

5. The method for detecting the distribution of marine organisms based on sonar imaging according to claim 4, characterized in that, The calculation of the effective biosignal block set for each depth layer based on the observation matrix includes: Based on the observation matrix, calculate the residual ratio function sequence for each sparse block; The residual ratio statistical mean and standard deviation are calculated by combining the shape and scale parameters of the gamma distribution under pure ocean noise environment, and the detection threshold of marine biological signals is set based on the residual ratio statistical mean and standard deviation; The effective biological signal block set for each depth layer is selected based on the residual ratio function sequence of each sparse block and the marine biological signal detection threshold.

6. The method for detecting the distribution of marine organisms based on sonar imaging according to claim 1, characterized in that, The process of calculating the spatial coordinates and reflection intensity of each biological target based on the set of effective biosignal blocks to obtain biolocation data for each depth layer, and performing three-dimensional imaging based on the biolocation data to obtain a three-dimensional acoustic image for each depth layer, includes: A prior distribution parameter for the reflection intensity of each biological target in the set of effective biological signal blocks is set, and a combination of hyperparameters is determined according to the marine species to construct a prior probability model of biological targets. Based on the prior probability model of the biological target, the expected step is executed to calculate the posterior distribution parameters of each biological signal block; The updated model parameters are obtained by performing a maximization step using the posterior distribution parameters. Calculate the rate of change of the updated model parameters, and when the rate of change of the parameters is lower than the convergence threshold, output the spatial coordinates and reflection intensity of each biological target to obtain the biological positioning data of each depth layer; Three-dimensional imaging is performed based on the biolocation data to obtain three-dimensional acoustic images of each depth layer.

7. The method for detecting the distribution of marine organisms based on sonar imaging according to claim 6, characterized in that, The step of performing a maximization step using the posterior distribution parameters to obtain the updated model parameters includes: The squared value of the second norm is calculated based on the posterior mean vector of each biological signal block in the posterior distribution parameters. At the same time, the trace value of the posterior covariance matrix is ​​calculated and added to the shape parameter of the corresponding marine species to obtain the numerator and denominator statistics. The updated biosignal block precision parameters are obtained by performing fractional calculations using the numerator and denominator statistics combined with the scale parameters of each marine organism species. The squared residual vector norms of the observed signal and the reconstructed signal are calculated based on the updated biosignal block precision parameters and then added to the residual norms to obtain the total residual statistics. The updated noise accuracy parameters are generated based on the total residual statistics, and the updated noise accuracy parameters are combined with the updated biosignal block accuracy parameters to obtain the updated model parameters.

8. The method for detecting the distribution of marine organisms based on sonar imaging according to claim 7, characterized in that, The process of performing three-dimensional imaging based on the biolocation data to obtain three-dimensional acoustic images of each depth layer includes: Based on the spatial coordinates and reflection intensity of each biological target in the biological positioning data, the linear frequency modulated signal is subjected to frequency domain conjugate matched filtering to obtain range-compressed echo data. The azimuth-focused imaging data is calculated by combining the range-compressed echo data with the spiral trajectory of the autonomous underwater vehicle. Fractional Fourier transform compensation is performed on the azimuth-focused imaging data to obtain Doppler-corrected signal data, and the Doppler-corrected signal data is reconstructed to obtain a three-dimensional acoustic image of each depth layer.

9. The method for detecting the distribution of marine organisms based on sonar imaging according to claim 1, characterized in that, The biological density index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer are calculated based on the three-dimensional acoustic images of each depth layer, and the biological detection results of each depth layer are output, including: Marine organisms were classified and identified from the three-dimensional acoustic images of each depth layer to obtain the biological identification results for each depth layer. Based on the biometric results, the biological density index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer are calculated, and the biological detection results of each depth layer are output.

10. A marine organism distribution detection system based on sonar imaging, characterized in that, A method for performing the sonar imaging-based marine organism distribution detection method as described in any one of claims 1-9, comprising: The acquisition module is used to set multiple depth layers and acquire the three-dimensional acoustic echo signal of each depth layer; The calculation module is used to perform sparse observation of marine biomass based on the three-dimensional acoustic echo signal, obtain the observation matrix, and calculate the set of effective biosignal blocks according to each signal block in the observation matrix. The construction module is used to calculate the spatial coordinates and reflection intensity of each biological target based on the set of effective biological signal blocks, obtain biological positioning data for each depth layer, and perform three-dimensional imaging based on the biological positioning data to obtain a three-dimensional acoustic image for each depth layer. The biodetection module is used to calculate the biodensity index, Shannon-Wiener diversity index, and vertical migration velocity of each ecological layer based on the three-dimensional acoustic image of each depth layer, and output the biodetection results of each depth layer.

Images