Method for monitoring three-dimensional distribution of fish in net cage based on spatial statistics of sound wave intensity

Abstract

This invention discloses a method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of acoustic intensity, comprising the following steps: S1. Positioning and attitude adjustment of an ROV carrying sonar; S2. Horizontal movement of the ROV and vertical scanning of the sonar; S3. Generation of a three-dimensional point cloud of acoustic intensity; S4. Preprocessing of the point cloud data; S5. Spatial partitioning statistics of acoustic intensity, specifically including depth layer statistics and horizontal region statistics; S6. Calculation of quantitative distribution parameters. This invention, by performing spatial partitioning statistics on intensity-controlled depth layers and horizontal regions, forms quantifiable three-dimensional distribution parameters of fish schools, which can bring significant improvements in feeding optimization, environmental monitoring, and density management.

Description

Technical Field

[0001] This invention relates to the field of marine aquaculture monitoring technology, specifically to a method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity. Background Technology

[0002] In deep-sea cage aquaculture, the three-dimensional spatial distribution (vertical and horizontal) of fish schools directly reflects their health status, feeding behavior, and environmental adaptability. Understanding fish school distribution information is crucial for precise feed delivery, stocking density assessment, behavioral pattern recognition, and early warning of stress responses.

[0003] Existing methods for monitoring fish distribution mainly include manual observation, single-beam vertical scanning, multi-layer horizontal scanning, and image recognition. Manual observation has a small field of view (less than 5 meters) and is limited by water turbidity, making it unable to obtain complete information about the entire fish cage. Single-beam scanning cannot simultaneously acquire complete vertical and horizontal distribution in a single operation; for example, vertical scanning only obtains one-dimensional depth distribution from fish echoes, missing horizontal echo information. Multi-layer horizontal scanning requires ROVs to scan at multiple depth layers, which is cumbersome and time-consuming (10-20 minutes), with varying data acquisition times at different layers and poor data consistency due to changes in fish location. Image recognition heavily relies on water transparency, requires multiple cameras, and demands high computing power.

[0004] The existing fish monitoring methods mentioned above lack quantitative parameters for fish distribution, and most can only provide qualitative descriptions of fish distribution (e.g., monitoring results indicate that the fish are mainly in the middle layer of the water). Therefore, they cannot be used for further analysis and decision-making based on precise fish distribution information. In addition, the monitoring results of fish distribution are not intuitive enough and are difficult for non-experts to understand. Summary of the Invention

[0005] One of the objectives of this invention is to provide a method for monitoring the three-dimensional distribution of fish in net cages based on spatial statistics of acoustic wave intensity. A single rapid scan using an ROV equipped with sonar can acquire three-dimensional intensity distribution data of the entire net cage. By performing spatial partitioning statistics on the intensity-controlled depth layer and horizontal region, quantifiable three-dimensional distribution parameters of the fish population can be formed.

[0006] The technical solution adopted to achieve the above objectives of this invention is as follows:

[0007] A method for monitoring the three-dimensional distribution of fish in net cages based on spatial statistics of sound wave intensity includes the following steps:

[0008] S1. Positioning and attitude adjustment of ROV carrying sonar: The ROV equipped with omnidirectional scanning sonar is submerged along the edge of the net cage under test to the middle layer of the net cage. The submersion depth is 40%-60% of the total depth of the net cage. The ROV is an unmanned remotely operated vehicle and existing products can be used. The attitude of the ROV is controlled so that the scanning plane of the sonar is vertical and the scanning area covers the entire depth range from the top to the bottom of the net cage. The ROV is an unmanned remotely operated vehicle and existing products can be used.

[0009] S2. ROV horizontal movement and sonar vertical scanning: The ROV maintains the middle layer depth and moves in a uniform linear motion along the horizontal plane inside the cage. The sonar continuously scans 360 degrees in the vertical plane. The sonar measures the sound wave intensity I at each time t, scanning angle θ, and detection distance r to form an intensity matrix I(θ,r,t).

[0010] S3. Generation of 3D point cloud of acoustic wave intensity: Establish a 3D rectangular coordinate system with the center of the bottom of the cage as the origin O, the X-axis as the ROV movement direction, the Z-axis as the vertical direction, and the Y-axis perpendicular to the XZ plane, combined with the real-time position information of the ROV ( , , The sonar's scanning angle θ and detection distance r are used to transform the acoustic intensity matrix I(θ,r,t) into a three-dimensional point cloud. Each detection point contains spatial coordinates (x,y,z) and an intensity value I. The coordinate transformation formula is:

[0011] ,

[0012] ,

[0013] ,

[0014] The generated 3D intensity point cloud is represented as a set P={( , , , )|i=1,2,3,…, },in Total points;

[0015] S4. Point cloud data preprocessing: Perform intensity threshold filtering, spatial range filtering, statistical outlier filtering, and voxel downsampling on the 3D spatial point cloud to obtain effective point cloud;

[0016] S5. Spatial Partition Statistics of Sound Wave Intensity: Spatial partition statistics are performed on the effective point cloud, including the following steps:

[0017] S51. Depth layer statistics: Divide the depth of the cage into n equal layers, and count the number of points, total intensity, average intensity and intensity density of the effective point cloud in each layer, where n≥4;

[0018] S52. Horizontal area statistics: Taking the center of the cage in the horizontal direction as the point, divide the horizontal plane of the cage into m annular areas along the radial direction. Count the number of points, the total intensity, and the intensity density of the effective point cloud in each annular area, where m≥3;

[0019] S6. Calculation of Quantitative Distribution Parameters: The following quantitative distribution parameters are calculated based on the spatial zoning statistical results of acoustic wave intensity:

[0020] Depth intensity concentration is used to assess the degree of concentration of fish in the vertical direction;

[0021] Edge intensity aggregation index is used to assess whether fish gather at the edge of the net cage;

[0022] The stratification strength index is used to assess the strength of vertical stratification in a fish population.

[0023] This invention provides quantified three-dimensional distribution parameters of fish populations. All three parameters are dimensionless normalized indices with clearly defined thresholds, supporting precise analysis, historical comparison, and cross-cage comparison. Numerical range analysis is performed on each parameter. For historical comparison analysis, a time-series database is established for trend analysis, periodic analysis, and anomaly detection, enabling early detection of environmental problems, identification of normal behavioral patterns, and optimization of aquaculture management. For cross-cage comparison analysis, since the parameters are normalized indices, direct comparisons across cages are possible without considering cage size differences, allowing for the identification of abnormal individuals, evaluation of aquaculture program effectiveness, and analysis of environmental impact. Therefore, the precise three-dimensional distribution results of fish populations monitored by this invention can significantly improve feeding optimization, environmental monitoring, and density management, demonstrating broad application value.

[0024] The present invention has the following preferred design:

[0025] In step S6 of the present invention, the formula for calculating the depth intensity concentration is: ,in For depth intensity concentration, Let be the intensity density of the i-th layer. Let the intensity density be the intensity density of the j-th layer. The average intensity density across all depth layers;

[0026] The criteria for judgment are: <0.3 indicates low concentration, 0.3≤ <0.6 indicates moderate concentration. ≥0.6 indicates high concentration. The larger the value, the more concentrated the fish are in the depth direction; The smaller the value, the more evenly the fish are distributed across different layers. This is a method based on the Gini coefficient to quantify the degree of unevenness in depth distribution.

[0027] In step S6 of the present invention, the formula for calculating the edge intensity aggregation index is: = ,in The edge intensity aggregation index, The intensity density refers to the ring-shaped region at the edge. The average intensity density of all annular regions;

[0028] The criteria for judgment are: A warning is issued when the value is greater than (1 + 0.5 / m). A critical alarm is issued if the value is ≥(1+1.0 / m).

[0029] In step S6 of the present invention, the formula for calculating the delamination strength index is:

[0030]

[0031] in , These represent the maximum and minimum intensity densities across all depth layers, respectively.

[0032] The criteria for judgment are: <0.5 indicates weak stratification, 0.5≤ <1.0 indicates a medium stratification. A value of ≥1.0 indicates strong stratification.

[0033] In step S1 of the present invention, the ROV descends to 50% of the total depth of the cage, and the sonar scans at this depth, which can completely cover the depth range above and below the cage.

[0034] In step S2 of the present invention, the ROV moves at a constant linear speed of 0.1-0.5 m / s, the sonar scanning frequency is 5-15 Hz, and the scanning duration is 40-60 seconds. The three-dimensional intensity distribution data of the entire cage can be obtained through a single rapid scan.

[0035] In step S4 of this invention, the intensity threshold T for intensity threshold filtering of the three-dimensional spatial point cloud is 50-80dB. The swim bladder echo intensity of farmed fish (30-50cm in length) is typically in the range of 60-90dB, and seawater noise (plankton, suspended particles) is typically <50dB. The lower limit of 50dB is a typical dividing point between fish body echo and noise, and the upper limit of 80dB avoids missing small fish, sideways fish, or distant targets. 64dB is preferred, as 64dB is located in the upper-middle position of the effective signal range, which can filter out most of the noise while retaining most of the fish body signal. This achieves an optimal balance between noise filtering and signal integrity, resolving noise interference caused by excessively low T and signal loss caused by excessively high T. The spatial range filtering removes points exceeding the physical boundary of the wire mesh cage, with a 0.3-0.5 meter margin inward of the cage's physical boundary to eliminate mesh echoes. This margin is referred to as the boundary margin, determined based on acoustic principles: sound wave wavelength λ = c / f ≈ 0.21 meters (c = 1500 m / s, f = 700 kHz), and the theoretical minimum distance to avoid mesh reverberation ≈ 2λ ≈ 0.4 meters, where c is the distance at a temperature of 15°C. The speed of sound in standard seawater with a salinity of 35‰ and a temperature of ℃, where f is the operating frequency of the sonar; the statistical anomaly filtering involves calculating the distance distribution from all points to the ROV, removing outliers with distances exceeding the 98th percentile, and calculating the intensity gradation of all points, removing outliers with intensity exceeding the 98th percentile; the voxel downsampling uses a voxel grid with a side length of 5-15 cm. Within this range, downsampling is less effective below 5 cm, and may merge adjacent fish bodies above 15 cm. The three-dimensional space is divided into a cubic grid with a side length of V, and multiple points within each voxel are replaced by its centroid. The coordinates are the weighted average of the coordinates of all points within the voxel (weighted by intensity), and the centroid intensity is the sum of the intensities of all points within the voxel. Downsampling reduces the amount of data and improves computational efficiency while preserving spatial distribution characteristics and intensity information. The preferred value for the voxel grid side length V is 8 cm. This value is determined based on the fish body feature size and sonar resolution: the body width of farmed fish is 4-10 cm (average 8 cm), and the spatial resolution of a 677 kHz sonar is about 10 cm. V=8 cm matches both the fish body feature size (which can characterize the position of a single fish) and the sonar resolution, achieving a balance between data compression and feature preservation.

[0036] In step S51 of the present invention, the depth of the cage is divided into 4 to 8 layers, and the layer interval is generally 2-3 meters; in step S52, the horizontal plane of the cage is divided into 3 to 6 annular areas.

[0037] The present invention also includes the following steps:

[0038] S7. Generate a three-dimensional distribution visualization chart based on the calculation results of the quantitative distribution parameters;

[0039] S8. Generate aquaculture management suggestions based on the calculation results of quantitative distribution parameters.

[0040] Furthermore, the three-dimensional distribution visualization chart in step S7 includes a three-dimensional intensity scatter plot, a depth intensity bar chart, and a horizontal intensity heatmap, wherein,

[0041] The three-dimensional intensity scatter plot draws all valid points in a three-dimensional coordinate system, and the intensity value is represented by the size or color of the points.

[0042] The horizontal axis of the depth intensity bar chart represents the depth layers in sequential numbering, and the vertical axis represents the sum of the intensities of the depth layers.

[0043] The horizontal intensity heat map is divided into grids on a horizontal plane, and different grid colors represent the intensity density of each annular region.

[0044] Three visualization charts show the distribution from different perspectives, making them intuitive and easy to understand. Even non-professionals can quickly understand the state of the fish school.

[0045] The present invention has the following beneficial effects:

[0046] 1. Full coverage of all cages in all dimensions: Complete three-dimensional distribution information (vertical + horizontal) can be obtained simultaneously through a single scan, eliminating the need for multiple scans or multi-layer operations, improving efficiency by 10-20 times, and ensuring good data consistency over time.

[0047] 2. Intensity-based direct statistics: No complex clustering calculations are required. Spatial partitioning statistics of intensity are performed directly. The calculation speed is fast (1-2 seconds), which is suitable for real-time monitoring and frequent monitoring (daily or even hourly).

[0048] 3. Intensity-weighted quantitative parameters: Depth intensity concentration, edge intensity aggregation index, and stratification intensity index are all calculated based on intensity weighting. They have clear physical meanings and reflect the actual distribution density of fish schools, which is more accurate than simple point counting.

[0049] 4. Standardized parameters and judgment criteria: The three parameters, depth intensity concentration, edge intensity aggregation index and layer intensity index, are all dimensionless normalized indicators (0-1 or ratio), with clear judgment thresholds, supporting accurate analysis, historical comparison and cross-cage comparison.

[0050] 5. Multi-dimensional visualization: Three types of charts display the distribution from different angles, making it intuitive and easy to understand. Even non-professionals can quickly understand the state of the fish school.

[0051] 6. Based on the monitoring results of this invention, precise feeding can be carried out, reducing feed waste rate by 18-25%; the detection rate of environmental anomalies (such as hypoxia at the bottom or stress response) can be increased by more than 40%, and the detection time can be advanced by 2-6 hours; the accuracy of stocking density assessment can be improved by more than 30%. Attached Figure Description

[0052] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0053] Figure 1 This is a flowchart of a method for monitoring the three-dimensional distribution of fish in net cages based on spatial statistics of sound wave intensity, according to the present invention.

[0054] Figure 2 This is a schematic diagram of the ROV scanning path perpendicular to the sonar scanning of the present invention;

[0055] Figure 3 This is a schematic diagram of depth-level hierarchical statistics in an embodiment of the present invention;

[0056] Figure 4 This is a schematic diagram of horizontal region statistics in an embodiment of the present invention;

[0057] Figure 5 This is a three-dimensional intensity scatter plot of an embodiment of the present invention;

[0058] Figure 6 This is a depth intensity histogram of an embodiment of the present invention;

[0059] Figure 7 This is a horizontal intensity thermogram of an embodiment of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of the present invention.

[0061] It should be understood that the invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0062] Example:

[0063] A method for monitoring the three-dimensional distribution of fish in net cages based on spatial statistics of sound wave intensity, such as Figure 1 and Figure 2 As shown, the steps include:

[0064] S1. Positioning and attitude adjustment of ROV carrying sonar: The ROV equipped with omnidirectional scanning sonar is lowered along the edge of the net cage under test to the middle layer of the net cage. The lowering depth is 40%-60% of the total depth of the net cage. The ROV is an unmanned remotely operated vehicle and existing products can be used. The attitude of the ROV is controlled so that the scanning plane of the sonar is vertical and the scanning area covers the entire depth range from the top to the bottom of the net cage. The ROV is an unmanned remotely operated vehicle and existing products can be used. In this embodiment, the preferred lowering depth is 50% of the total depth of the net cage. After the lowering is completed, the ROV is controlled to rotate 90 degrees around its longitudinal axis so that the scanning plane of the omnidirectional sonar changes from the initial horizontal direction to the vertical direction.

[0065] S2. ROV Horizontal Movement and Sonar Vertical Scan: The ROV maintains a mid-depth and moves at a constant speed in a straight line along the horizontal plane inside the cage. The sonar continuously scans 360 degrees in the vertical plane. The sonar measures the sound wave intensity I at each time t, scanning angle θ, and detection distance r, forming an intensity matrix I(θ,r,t). Preferably, the ROV's constant speed in a straight line is 0.1-0.5 m / s, and in this embodiment, it is preferably 0.3 m / s. The sonar scanning frequency is 5-15 Hz, and in this embodiment, it is preferably 10 Hz. The scanning duration is 40-60 seconds. A single rapid scan can obtain the three-dimensional intensity distribution data of the entire cage.

[0066] S3. Generation of 3D point cloud of acoustic wave intensity: Establish a 3D rectangular coordinate system with the center of the bottom of the cage as the origin O, the X-axis as the ROV movement direction, the Z-axis as the vertical direction, and the Y-axis perpendicular to the XZ plane, combined with the real-time position information of the ROV ( , , The sonar's scanning angle θ and detection distance r are used to transform the acoustic intensity matrix I(θ,r,t) into a three-dimensional point cloud. Each detection point contains spatial coordinates (x,y,z) and an intensity value I. The coordinate transformation formula is:

[0067] ,

[0068] ,

[0069] ,

[0070] The generated 3D intensity point cloud is represented as a set P={( , , , )|i=1,2,3,…, },in Total points;

[0071] S4. Point Cloud Data Preprocessing: The 3D point cloud is subjected to intensity threshold filtering, spatial range filtering, statistical outlier filtering, and voxel downsampling to obtain a valid point cloud. Among these steps...

[0072] The intensity threshold T for intensity threshold filtering of the three-dimensional spatial point cloud is 50-80dB. The swim bladder echo intensity of farmed fish (30-50cm in length) is usually in the range of 60-90dB, and the seawater noise (plankton, suspended particles) is usually <50dB. The lower limit of 50dB is the typical dividing point between fish body echo and noise, and the upper limit of 80dB avoids missing small fish, sideways fish or distant targets. In this embodiment, 64dB is preferred. 64dB is located in the upper middle of the effective signal range, which can filter most of the noise and retain most of the fish body signal. It achieves the optimal balance between noise filtering and signal integrity, and solves the noise interference problem caused by too low T and the signal omission problem caused by too high T.

[0073] The spatial range filtering is used to remove points that exceed the physical boundary of the net cage, and the physical boundary of the net cage is recessed by 0.3-0.5 meters to eliminate net echo. The recessed margin of the physical boundary of the net cage is referred to as the boundary margin. This boundary margin is determined based on acoustic principles: the sound wave wavelength λ=c / f≈0.21 meters (c=1500m / s, f=700kHz), and the theoretical minimum distance to avoid net reverberation ≈2λ≈0.4 meters. c is the speed of sound in standard seawater at a temperature of 15℃ and a salinity of 35‰, and f is the operating frequency of the sonar.

[0074] The statistical anomaly filtering involves calculating the distance distribution from all points to the ROV, removing outliers whose distance exceeds the 98th percentile, and calculating the intensity gradation of all points, removing outliers whose intensity exceeds the 98th percentile.

[0075] The voxel downsampling uses a voxel grid with a side length of 5-15 cm. Within this range, downsampling is not effective below 5 cm, and may merge adjacent fish bodies if the length exceeds 15 cm. The three-dimensional space is divided into a cubic grid with a side length of V. Multiple points within each voxel are replaced by its centroid. The centroid coordinates are the weighted average of the coordinates of all points within the voxel (weighted by intensity), and the centroid intensity is the sum of the intensities of all points within the voxel. Downsampling reduces the amount of data and improves computational efficiency while preserving spatial distribution characteristics and intensity information. In this embodiment, the preferred value of the voxel grid side length V is 8 cm. This value is determined based on the fish body feature size and sonar resolution: the width of farmed fish is 4-10 cm (average 8 cm), and the spatial resolution of a 677 kHz sonar is about 10 cm. V=8 cm matches both the fish body feature size (which can characterize the position of a single fish) and the sonar resolution, achieving a balance between data compression and feature preservation.

[0076] After preprocessing, effective point clouds are obtained. , ={( , , , )|i=1,2,3,…, },in Valid points;

[0077] S5. Spatial Partition Statistics of Sound Wave Intensity: Spatial partition statistics are performed on the effective point cloud, including the following steps:

[0078] S51. Depth-layered statistics: Divide the cage depth into n equal layers, and count the number of points, total intensity, average intensity, and intensity density of effective point clouds in each layer. Divide the cage depth H into n equal layers, with each layer having a thickness Δz = H / n. For the i-th layer (i = 1, 2, ..., n, numbered from top to bottom), count the information of all points in that layer:

[0079] Number of points: ,

[0080] Total intensity: = ,

[0081] Average strength: ,

[0082] Strength and density: = ,

[0083] in Let n be the volume of the i-th layer. Preferably, n ≥ 4. In practical applications, the layer interval is generally 2-3 meters. The number of layers n in the depth of the cage can be set to 4 to 8. For cylindrical cages: =π×R²×Δz, In this embodiment, a cylindrical wire mesh cage is used, preferably with n=6. Figure 3 This is a schematic diagram of the depth-level hierarchical statistics in this embodiment;

[0084] S52. Horizontal Region Statistics: Taking the center of the cage in the horizontal direction as the center point, divide the horizontal plane of the cage into m annular regions along the radial direction. Statistically count the number of effective point clouds, the total intensity, and the intensity density within each annular region, where m ≥ 3. The statistics for the k-th region can be expressed as follows:

[0085] Number of points: =count{points are in the k-th region},

[0086] Total intensity: =Σ{I|point is in the k-th region},

[0087] Strength and density: = ,

[0088] in, Let be the volume of the k-th region, where k is a positive integer from 1 to m. Taking a cylindrical wire mesh cage with m=3 as an example... Figure 4 This is a schematic diagram of the horizontal region statistics in this embodiment, based on radial distance d= Divide the horizontal plane of the cage into 3 annular areas:

[0089] Central region (k=1): 0≤d<R / 3,

[0090] Intermediate region (k=2): R / 3 ≤ d < 2R / 3,

[0091] Edge region (k=3): 2R / 3≤d<R,

[0092] (Central area, solid cylinder)

[0093] (Central area, circular cylinder)

[0094] Edge area, circular cylinder);

[0095] S6. Calculation of Quantitative Distribution Parameters: The following quantitative distribution parameters are calculated based on the spatial zoning statistical results of acoustic wave intensity:

[0096] Depth intensity concentration is used to assess the degree of concentration of fish in the vertical direction. In this embodiment, the formula for calculating depth intensity concentration is: ,in For depth intensity concentration, Let be the intensity density of the i-th layer. Let the intensity density be the intensity density of the j-th layer. Given the average intensity density of all depth layers, the sum of the absolute values ​​of the intensity density differences between each pair of depth layers is calculated by doubly summing through all combinations of depth layer pairs (i,j) using the above formula.

[0097] The criteria for judgment are: <0.3 indicates low concentration, 0.3≤ <0.6 indicates moderate concentration. ≥0.6 indicates high concentration. The larger the value, the more concentrated the fish are in the depth direction; The smaller the value, the more evenly the fish are distributed across all layers. This is a method based on the Gini coefficient to quantify the degree of unevenness in depth distribution.

[0098] The edge intensity aggregation index is used to assess whether fish congregate at the edge of the net cage. The formula for calculating the edge intensity aggregation index is: = ,in The edge intensity aggregation index, The intensity density refers to the ring-shaped region at the edge. The average intensity density of all annular regions is used to divide the radial region into m concentric annular regions: the radial range of the k-th region. ≤ < ,in Intensity density of region k , It is the area of ​​the k-th annular region, and the formula is:

[0099] = ,

[0100] Average intensity density of all annular regions Edge intensity clustering index = The m-th region is the edge region, and the determination criteria correspond to the value of m: >(1+0.5 / m) is a warning value. >(1+1.0 / m) indicates a severe alarm. For m=6, the warning threshold is 1.08 and the alarm threshold is 1.17.

[0101] The stratification strength index is used to assess the intensity of vertical stratification in a fish population. The formula for calculating the stratification strength index is:

[0102]

[0103] in , These represent the maximum and minimum intensity densities across all depth layers, respectively.

[0104] The criteria for judgment are: <0.5 indicates weak stratification, meaning the fish population is relatively evenly distributed at different depths; 0.5≤ <1.0 indicates a medium stratification. A value of ≥1.0 indicates strong stratification, suggesting that there may be significant vertical stratification in the water temperature or dissolved oxygen.

[0105] The monitoring method of this invention yields three parameters—depth intensity concentration, edge intensity aggregation index, and stratification intensity index—all dimensionless normalized indicators with clearly defined thresholds. These parameters support precise analysis, historical comparison, and cross-cage comparison. Numerical range analysis is performed on each parameter. For historical comparison analysis, a time-series database is established for trend analysis, periodic analysis, and anomaly detection, enabling early detection of environmental problems, identification of normal behavior patterns, and optimization of aquaculture management. For cross-cage comparison analysis, since the parameters are all normalized indicators, direct cross-cage comparisons are possible without considering differences in cage size, allowing for the identification of abnormal individuals, evaluation of aquaculture program effectiveness, and analysis of environmental impact.

[0106] S7. Generate a three-dimensional distribution visualization chart based on the calculation results of the quantitative distribution parameters;

[0107] S8. Generate aquaculture management suggestions based on the calculation results of quantitative distribution parameters.

[0108] The 3D distribution visualization charts include a 3D intensity scatter plot, a depth intensity bar chart, and a horizontal intensity heatmap.

[0109] A 3D intensity scatter plot draws all valid points in a 3D coordinate system, such as... Figure 5 As shown, the intensity value is represented by the size or color of the dots, which can intuitively show the distribution of the fish in three-dimensional space;

[0110] like Figure 6 As shown, the horizontal axis of the depth intensity bar chart is numbered sequentially to represent depth layers, and the vertical axis represents the total intensity of that depth layer, which can intuitively show the distribution of fish in each depth layer;

[0111] like Figure 7 As shown, the horizontal intensity heatmap divides the horizontal plane into grids (such as 20×20), and uses different grid colors to represent the intensity density of each annular area, with a gradient from blue (low intensity) to red (high intensity), which visually shows the distribution pattern of fish on the horizontal plane.

[0112] Three visualization charts show the distribution from different perspectives, making them intuitive and easy to understand. Even non-professionals can quickly understand the state of the fish school.

[0113] This invention automates the entire process from data collection and analysis to decision-making recommendations, reducing manual judgment and reliance on specialized knowledge. Practical application results show that using this method for precise feeding reduces feed waste by 18-25%; increases the detection rate of environmental anomalies (such as bottom hypoxia and stress response) by over 40%, advancing the detection time by 2-6 hours; improves the accuracy of stocking density assessment by over 30%; and intelligently generates aquaculture management recommendations based on monitoring results.

[0114] The above embodiments of the present invention are not intended to limit the scope of protection of the present invention. The implementation of the present invention is not limited thereto. All other modifications, substitutions or alterations made to the above structure of the present invention based on the above content of the present invention, in accordance with ordinary technical knowledge and common practice in the field, without departing from the basic technical idea of ​​the present invention, shall fall within the scope of protection of the present invention.

Claims

1. A method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity, characterized in that, Including the following steps: S1. Positioning and attitude adjustment of ROV carrying sonar: The ROV equipped with omnidirectional scanning sonar is lowered along the edge of the net cage under test to the middle layer of the net cage. The lowering depth is 40%-60% of the total depth of the net cage. The attitude of the ROV is controlled so that the scanning plane of the sonar is vertical and the scanning area covers the entire depth range from the top to the bottom of the net cage. S2. ROV horizontal movement and sonar vertical scanning: The ROV maintains the middle layer depth and moves in a uniform linear motion along the horizontal plane inside the cage. The sonar continuously scans 360 degrees in the vertical plane. The sonar measures the sound wave intensity I at each time t, scanning angle θ, and detection distance r to form an intensity matrix I(θ,r,t). S3. Generation of 3D point cloud of acoustic wave intensity: Establish a 3D rectangular coordinate system with the center of the bottom of the cage as the origin O, the X-axis as the ROV movement direction, the Z-axis as the vertical direction, and the Y-axis perpendicular to the XZ plane, combined with the real-time position information of the ROV ( , , The sonar's scanning angle θ and detection distance r are used to transform the acoustic intensity matrix I(θ,r,t) into a three-dimensional point cloud. Each detection point contains spatial coordinates (x,y,z) and an intensity value I. The coordinate transformation formula is: ， ， ， The generated 3D intensity point cloud is represented as a set P={( , , , )|i=1,2,3,…, },in Total points; S4. Point cloud data preprocessing: Perform intensity threshold filtering, spatial range filtering, statistical outlier filtering, and voxel downsampling on the 3D spatial point cloud to obtain effective point cloud; S5. Spatial Partition Statistics of Sound Wave Intensity: Spatial partition statistics are performed on the effective point cloud, including the following steps: S51. Depth layer statistics: Divide the depth of the cage into n equal layers, and count the number of points, total intensity, average intensity and intensity density of the effective point cloud in each layer, where n≥4; S52. Horizontal area statistics: Taking the center of the cage in the horizontal direction as the point, divide the horizontal plane of the cage into m annular areas along the radial direction. Count the number of points, the total intensity, and the intensity density of the effective point cloud in each annular area, where m≥3; S6. Calculation of Quantitative Distribution Parameters: The following quantitative distribution parameters are calculated based on the spatial zoning statistical results of acoustic wave intensity: Depth intensity concentration is used to assess the degree of concentration of fish in the vertical direction; The edge intensity aggregation index is used to assess whether fish gather at the edge of the net cage; The stratification strength index is used to assess the strength of vertical stratification in a fish population.

2. The method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity according to claim 1, characterized in that, In step S6, the formula for calculating the depth intensity concentration is: ,in For depth intensity concentration, Let be the intensity density of the i-th layer. Let the intensity density be the intensity density of the j-th layer. The average intensity density across all depth layers; The criteria for judgment are: <0.3 indicates low concentration, 0.3≤ <0.6 indicates moderate concentration. A concentration of ≥0.6 indicates a high concentration level.

3. The method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity according to claim 1, characterized in that, In step S6, the formula for calculating the edge intensity aggregation index is: = ,in The edge intensity aggregation index, The intensity density refers to the ring-shaped region at the edge. The average intensity density of all annular regions; The criteria for judgment are: A warning is issued when the value is greater than (1 + 0.5 / m). A critical alarm is issued if the value is ≥(1+1.0 / m).

4. The method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity according to claim 1, characterized in that, In step S6, the formula for calculating the delamination strength index is: in , These represent the maximum and minimum intensity densities across all depth layers, respectively. The criteria for judgment are: <0.5 indicates weak stratification, 0.5≤ <1.0 indicates a medium stratification. A value of ≥1.0 indicates strong stratification.

5. The method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity according to claim 1, characterized in that: In step S1, the ROV descends to 50% of the total depth of the cage.

6. The method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity according to claim 1, characterized in that: In step S2, the ROV moves at a constant linear speed of 0.1-0.5 m / s, the sonar scanning frequency is 5-15 Hz, and the scanning duration is 40-60 seconds.

7. The method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity according to claim 1, characterized in that: In step S4, the intensity threshold T for intensity threshold filtering of the three-dimensional spatial point cloud is 50-80 dB; the spatial range filtering removes points that exceed the physical boundary of the net cage, and the physical boundary of the net cage is recessed by 0.3-0.5 meters to exclude net echo; the statistical anomaly filtering calculates the distance distribution from all points to the ROV, removes anomaly points whose distance exceeds the 98th percentile, calculates the intensity division of all points, and removes anomaly intensity points whose intensity exceeds the 98th percentile; the voxel downsampling voxel grid side length is 5-15 cm.

8. The method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity according to claim 1, characterized in that: In step S51, the depth of the cage is divided into 4 to 8 layers (n); in step S52, the horizontal plane of the cage is divided into 3 to 6 annular areas.

9. The method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity according to claim 1, characterized in that, It also includes the following steps: S7. Generate a three-dimensional distribution visualization chart based on the calculation results of the quantitative distribution parameters; S8. Generate aquaculture management suggestions based on the calculation results of quantitative distribution parameters.

10. The method for monitoring the three-dimensional distribution of fish schools in net cages based on spatial statistics of sound wave intensity according to claim 9, characterized in that: The three-dimensional distribution visualization chart in step S7 includes a three-dimensional intensity scatter plot, a depth intensity bar chart, and a horizontal intensity heatmap, wherein... The three-dimensional intensity scatter plot draws all valid points in a three-dimensional coordinate system, and the intensity value is represented by the size or color of the points. The horizontal axis of the depth intensity bar chart represents the depth layers in sequential numbering, and the vertical axis represents the sum of the intensities of the depth layers. The horizontal intensity heat map is divided into grids on a horizontal plane, and different grid colors represent the intensity density of each annular region.

Images