A three-dimensional distribution monitoring and feeding management method for factory farming

Abstract

This invention discloses a three-dimensional distribution monitoring and feeding management method for factory farming, which utilizes the effective sound wave intensity three-dimensional point cloud P v Spatial partitioning statistics can be performed to analyze the three-dimensional distribution characteristics of fish populations within rectangular aquaculture ponds, including: Depth Intensity Concentration Index (DCI), which quantifies the concentration of fish populations along the depth direction within the rectangular aquaculture pond. I The Edge Strength Aggregation Index (ECI) is used to quantify the concentration of fish at the edge of a rectangular aquaculture pond. I And the stratification intensity index (LSI) used to quantify the distribution of fish populations at the surface and bottom of rectangular culture ponds. I Therefore, this invention can quantitatively and comprehensively describe the depth-direction and horizontal-direction distribution information of fish in a rectangular aquaculture pond using three characteristic parameters. Furthermore, based on the depth intensity concentration (DCI)... I The monitoring system can detect when fish are highly concentrated in a certain depth layer and feed them in layers, thus achieving precise feeding.

Description

Technical Field

[0001] This invention relates to the field of aquaculture monitoring technology, specifically a three-dimensional distribution monitoring and feeding management method for factory farming. Background Technology

[0002] Factory-style aquaculture is an important development direction for modern fisheries, and rectangular aquaculture ponds are widely used due to their ease of management and standardization. Accurate monitoring of the distribution within the aquaculture ponds is crucial for scientific feeding, density control, and environmental management.

[0003] Traditional methods for monitoring fish populations mainly include:

[0004] Method 1: Manual counting: This method involves manually counting fish through fishing or underwater observation. It is labor-intensive, inefficient, causes stress to the fish population, and is difficult to monitor continuously.

[0005] Method 2, underwater photography: This method uses a camera to capture images of fish schools. However, it is greatly affected by water turbidity. In high-density aquaculture environments, the fish bodies severely obscure the view, making accurate counting difficult. Furthermore, it can only observe a localized area.

[0006] Method 3: Acoustic detection method: Sonar equipment is used to detect fish schools, but the existing technology is mainly for open water or deep-water cages, and has poor adaptability to shallow rectangular aquaculture ponds with a depth of 2-5 meters.

[0007] Existing acoustic monitoring technologies mainly suffer from the following problems:

[0008] First, there is the issue of distribution analysis: existing methods lack quantitative parameters for analyzing the three-dimensional distribution of fish populations, and most can only provide qualitative descriptions, making it impossible to make scientific decisions on feeding, density control, and environmental management based on accurate distribution information.

[0009] Secondly, there is the issue of interference from the pond walls: the monitoring environment of rectangular aquaculture pond walls differs fundamentally from that of deep-sea cages. Deep-sea cages use nylon or polyethylene netting with acoustic impedance close to seawater and a reflection coefficient of only 0.1-0.3. In contrast, rectangular aquaculture pond walls are made of concrete or brick, with acoustic impedance far exceeding that of seawater (approximately 4-6 times higher) and a sound wave reflection coefficient of 0.7-0.9. This results in strong acoustic wave reflections with echo intensities of 80-110 dB, significantly higher than the 40-60 dB echo intensity of flexible netting and the 60-90 dB echo intensity of fish bodies. More seriously, the corner reflectors at the four corners of the rectangular pond walls create an echo intensity reaching 120 dB. Therefore, the echoes from the rectangular pond walls severely interfere with sonar identification of fish schools within the pond. If these echoes are not distinguished and eliminated, they may be misinterpreted as large or dense schools of fish, severely impacting the accuracy of fish school identification.

[0010] Thirdly, the problem of scanning mode: traditional sonar adopts horizontal scanning mode, which is simultaneously interfered by strong echoes from four pool walls in a rectangular culture pond, and the echoes are superimposed on each other, resulting in extremely serious interference; furthermore, the equipment needs to move vertically to cover the entire depth range, which leads to complex scanning paths and long data acquisition time of 10-20 minutes, and changes in fish school positions result in poor data consistency.

[0011] Fourthly, the problem of equipment deployment: ROV (Remotely Operated Vehicle) has high cost and complex operation, which is not suitable for daily continuous monitoring; the fixed vertical lifting rod occupies the central position, interferes with the activities of fish schools, and has a complex mechanical structure. SUMMARY OF THE INVENTION

[0012] The technical problem to be solved by the present invention is: to provide a three-dimensional distribution monitoring and feeding management method for industrial aquaculture, so as to solve the problem in the prior art that accurate feeding control cannot be performed due to the lack of a quantitative description method for the three-dimensional distribution of fish schools in a rectangular culture pond.

[0013] To solve the above technical problem, the technical solution adopted by the present invention is as follows:

[0014] A three-dimensional distribution monitoring method for industrial aquaculture, comprising:

[0015] Step S1, performing sonar scanning on the interior of the rectangular culture pond to obtain a three-dimensional point cloud of acoustic intensity P={(x i ,y i , z i , I i )丨i=1,2,…,N}, wherein N represents the total number of measurement points obtained by the sonar scanning, which is usually hundreds of thousands to millions of measurement points, (x i , y i , z i ) and I i respectively represent the three-dimensional spatial coordinate and acoustic intensity of the i-th measurement point;

[0016] Step S2, performing filtering processing on the acoustic intensity three-dimensional point cloud P, and recording the set of measurement points remaining after filtering as effective acoustic intensity three-dimensional point cloud P v ={(x j , y j , z j , I j )丨j=1,2,…,N v}, wherein N v represents the total number of effective measurement points, (x j , y j , z j ) and I jLet represent the three-dimensional spatial coordinates and sound wave intensity of the j-th valid measurement point, respectively;

[0017] Step S3: For the effective acoustic wave intensity three-dimensional point cloud P... v Spatial partitioning statistics were performed to analyze the three-dimensional distribution characteristics of fish populations within rectangular aquaculture ponds, including: Depth Intensity Concentration Index (DCI), used to quantify the concentration of fish populations along the depth direction within the rectangular aquaculture ponds. I The Edge Strength Aggregation Index (ECI) is used to quantify the concentration of fish at the edge of a rectangular aquaculture pond. I And the stratification intensity index (LSI) used to quantify the distribution of fish populations at the surface and bottom of rectangular culture ponds. I .

[0018] The method for performing spatial partition statistics in step S3 is as follows:

[0019] Step S3-1: Divide the rectangular aquaculture pond into n equal depth layers along its depth direction, and number the depth layers from bottom to top as layer 1 to layer n. Then, count the number of effective measurement points N in the kth depth layer. zk Total sound wave intensity I zk Volume V zk Sound wave intensity density ρ zk =I zk / V zk Where k = 1, 2, ..., n, and N is the number of valid measurement points. zk That is, the total number of all valid measurement points in the k-th depth layer, and the sum of the sound wave intensities I. zk That is, the sum of the sound wave intensities at all valid measurement points located in the k-th depth layer;

[0020] Step S3-2: Divide the rectangular aquaculture pond into m equal intervals along both the length and width directions. L column and m W The row is divided into m in the horizontal direction. L ×m W A rectangular horizontal grid, and the 2m area located at the edge of the rectangular aquaculture pond. L +2m W - The four horizontal grids are designated as edge grids; and the number of valid measurement points N for the horizontal grid in column l and row h is statistically obtained. lh Total sound wave intensity I lh Area A lh Sound wave intensity density ρ lh =I lh / A lh Where l = 1, 2, ..., m L h=1,2,...,m W, the number of effective measurement points N lh that is, the number of all effective measurement points within the horizontal grid range of the l-th column and h-th row where the rectangular aquaculture pond is located, the sum of acoustic intensity I lh that is, the sum of the acoustic intensity of all effective measurement points within the horizontal grid range of the l-th column and h-th row where the rectangular aquaculture pond is located; and, the horizontal area A of the rectangular aquaculture pond, the sum of the acoustic intensity of all measurement points of the rectangular aquaculture pond ΣI i , the average intensity density ρ in the plane of the rectangular aquaculture pond mean =ΣI i / A, the average intensity density ρ of the edge grid edge =ΣI ic / A c , where ΣI ic is the sum of the acoustic intensity of all edge grids, and A c is the sum of the area of all edge grids;

[0021] Step S3-3, calculating three-dimensional distribution characteristics of fish schools in said rectangular aquaculture pond:

[0022] Depth Concentration Intensity DCI I =(ΣΣ|ρ zk1 - ρ zk2 |) / (2n 2 ×ρ z );

[0023] where, |ρ zk1 - ρ zk2 | represents the absolute difference between the acoustic intensity density of the k1-th layer and the k2-th depth layer, ΣΣ|ρ zk1 - ρ zk2 | represents the sum of |ρ zk1 - ρ zk2 | corresponding to all combinations of k1=1,2,...,n and k2=1,2,...,n; n is the number of depth layers; ρ z is the average value of the acoustic intensity density ρ zk of all depth layers;

[0024] Edge Concentration Index ECI I =ρ edge / ρ mean ;

[0025] Layered Strength Index LSI I =(ρ upper – ρ lower ) / (ρ upper + ρ lower );

[0026] where, ρupper and ρ lower The acoustic intensity density ρ represents the depth stratification of the upper and lower halves of the rectangular aquaculture pond, respectively. zk The average value, and when the number of layers n is odd, the upper half and the lower half respectively contain the upper half and the lower half of the (n+1) / 2th layer depth layer.

[0027] Therefore, this invention utilizes the three-dimensional point cloud P of effective sound wave intensity. v Spatial partitioning statistics can be performed to analyze the three-dimensional distribution characteristics of fish populations within rectangular aquaculture ponds, including: Depth Intensity Concentration Index (DCI), which quantifies the concentration of fish populations along the depth direction within the rectangular aquaculture pond. I The Edge Strength Aggregation Index (ECI) is used to quantify the concentration of fish at the edge of a rectangular aquaculture pond. I And the stratification intensity index (LSI) used to quantify the distribution of fish populations at the surface and bottom of rectangular culture ponds. I Therefore, the present invention can quantitatively and comprehensively describe the depth-direction and horizontal-direction distribution information of fish in a rectangular breeding pond using three characteristic parameters.

[0028] Among them, the depth intensity concentration (DCI) I The criteria for judgment are: DCI I A concentration < 0.3 indicates low concentration, meaning the fish population is not highly concentrated in the depth direction of the rectangular culture pond and is evenly distributed; 0.3 ≤ DCI I A concentration level <0.6 indicates a moderate concentration; DCI I A value ≥0.6 indicates high concentration, meaning the fish are concentrated in a certain depth layer.

[0029] The edge intensity aggregation index (ECI) I The criteria for determination are: ECI I When ECI < 0.8, it indicates central aggregation, meaning the fish population tends to concentrate in the center of the rectangular rearing pond, with lower density at the edges; 0.8 ≤ ECI I When <1.3, it indicates that the fish population is evenly distributed; 1.3≤ECI I A value <1.8 indicates an edge-gathering warning, meaning the fish begin to gather at the edge of the rectangular rearing pond; ECI I A value ≥1.8 indicates a severe edge aggregation alarm, meaning that a large number of fish are gathering at the edge of the rectangular aquaculture pond, which may indicate an environmental anomaly.

[0030] The layering strength index LSI I The criterion for determination is: LSI I When LSI is less than -0.3, it indicates that the fish are mainly distributed at the bottom of the rectangular aquaculture pond, with the fish tending towards the bottom; -0.3 ≤ LSI IWhen the value is <0.3, the fish population is evenly distributed on both the surface and bottom layers of the rectangular culture pond; LSI I When the value is ≥0.3, it indicates that the fish are mainly distributed on the surface of the rectangular aquaculture pond, and the fish tend to be closer to the water surface.

[0031] Preferably, in step S3-1, the stratification interval Δz of the rectangular aquaculture pond, i.e., the height of the stratification depth, ranges from 0.8 meters to 2.5 meters. This is to avoid the stratification interval Δz being too small, which could cause fish in adjacent stratification depths to overlap and lose the significance of stratification statistics, and to avoid the stratification interval Δz being too large, which would fail to reflect the details of the vertical distribution of fish. This ensures that each stratification depth has sufficient resolution without having too few samples. Furthermore, within the range of the stratification interval Δz, the number of stratifications n is preferably between 2 and 4 stratifications.

[0032] In step S3-2, the edge size of the horizontal grid ranges from 8 meters to 12 meters to balance spatial resolution and sample size, avoiding: if the horizontal grid is too small, the number of measurement points in each horizontal grid will be too small, resulting in statistical instability; if the measurement points are too large, the spatial resolution will be insufficient, and the distribution differences will not be reflected.

[0033] In addition, the three-dimensional distribution monitoring method can generate a plot of data based on the statistical data in step S3 to visually display the distribution of fish schools, including: firstly, generating a depth bar chart to display the intensity distribution of each layer: the horizontal axis is the depth layer number (1,2,3,...,n, from bottom to top), and the vertical axis is the sum of the acoustic wave intensity I of that layer. zk Each layer is represented by a pillar, and the pillar height reflects the fish density at that layer. Secondly, a horizontal heatmap is generated to display m. L ×m W Mesh intensity distribution: Rectangular mesh layout (m) L line × m W (columns), each grid displays the acoustic intensity density ρ of that region. lh A color gradient is used to represent the intensity (e.g., blue represents low density and red represents high density).

[0034] As a preferred embodiment of the present invention: such as Figure 4 and Figure 5 As shown, in step S1, the sonar scanning method is as follows: the sonar moves horizontally along the length direction of the rectangular aquaculture pond, and the sonar performs a 360° scan in the vertical plane of the length direction, and the 360° scanning range of the sonar covers the cross section of the rectangular aquaculture pond perpendicular to the length direction.

[0035] In step S2, the three-dimensional point cloud P of the sound wave intensity is subjected to multi-level filtering processing, which includes at least removing the measurement points generated by the strong reflected echo interference of the rectangular aquaculture pond wall on the sonar scan.

[0036] Therefore, in step S1, the present invention uses a sonar scanning method in which the sonar moves horizontally along the length direction and performs a 360° scan in the vertical plane. It is only affected by strong reflected echoes from the two side walls and bottom of the rectangular aquaculture pond. There is a release gap for the echo in the direction of the water surface, that is, the water surface is a free surface. Although the reflection coefficient is high (about 0.95), the echo can be scattered upward and escape from the rectangular aquaculture pond. Compared with the sonar using a horizontal scanning method, it significantly reduces the number of times the echo reflects and interferes with the sonar in the rectangular aquaculture pond, so that it is easier to eliminate the strong reflected echo interference caused by the pond walls of the rectangular aquaculture pond.

[0037] Furthermore, by eliminating measurement points caused by strong reflected echo interference from the rectangular aquaculture pond walls in step S2, an effective three-dimensional point cloud Pv of acoustic intensity can be obtained without being affected by strong reflected echo interference from the rectangular aquaculture pond walls. This ensures the accuracy of estimating the number of fish in the rectangular aquaculture pond. Through experiments, the estimation error has been reduced from 20-40% to 5-10%, and the accuracy has been improved by 2-4 times.

[0038] Preferably, the multi-level filtering process in step S2 includes:

[0039] Filtering Strategy 1: Intensity Dual Threshold Filtering: Eliminating sound wave intensity I i Less than the preset lower limit of strength I min Or greater than the preset intensity upper limit value I max The measurement points, among which, the preset lower limit value I min The value range is 50dB to 70dB, used for filtering seawater noise (including plankton, suspended particles, equipment noise, etc., with an intensity typically <50dB); the preset upper limit of intensity is I. max The value range is 90dB to 100dB, used for strong reflections in the filter tank wall;

[0040] Filtering Strategy 2: Spatial Range Filtering: Eliminating 3D spatial coordinates (x... i , y i , z i Measurement points extending beyond the boundary of the rectangular aquaculture pond, wherein the boundary of the rectangular aquaculture pond, with length, width, and height L, W, and H respectively, includes: the virtual top surface of the aquaculture pond (i.e., excluding z...). i ≥H measurement points), bottom surface of the aquaculture pond (i.e., excluding z) i ≤0 measurement points), two aquaculture pond walls along the length direction (i.e., excluding x) i <0 or x i The measurement point of >L), the two aquaculture pond walls in the width direction are respectively recessed inward by a preset boundary margin δ to form two virtual boundary end faces (i.e., removing |y i|≤W / 2-δ measurement point), the preset boundary margin δ ranges from 0.4 meters to 0.6 meters; among them, because the sonar moves horizontally along the length direction and performs a 360° scan in the vertical plane, the two aquaculture pond walls in the length direction do not need to be set with a preset boundary margin δ.

[0041] Filtration Strategy 3: Strong Echo Filtration from Pool Walls

[0042] For sound wave intensity I i Greater than the preset suspicious echo intensity threshold I th The measurement point, if its distance from any wall of the rectangular aquaculture pond is less than the preset echo distance threshold δ, is considered valid. wall If the measurement point is not found, then the measurement point is discarded; wherein, the preset suspicious echo intensity threshold I th The value range is 80dB to 90dB, and the preset echo distance threshold δ wall The value ranges from 0.5 meters to 0.8 meters;

[0043] For the measurement points in the four corner reflection areas of a rectangular aquaculture pond, if their sound wave intensity I i If the value is greater than 80dB, the measurement point is discarded; wherein, the corner reflection area is the area within 1 meter of the corner of the bottom of the aquaculture pond, and the 1-meter threshold is an empirical value that covers the main influence range of corner reflection.

[0044] Therefore, this invention, through the combined use of filtration strategies one through three, can distinguish between pond wall echoes and fish body echoes, completely eliminating measurement points caused by strong reflected echo interference from the pond wall of rectangular aquaculture ponds to sonar scans. This reduces the misjudgment rate of sonar scans due to strong reflections from the pond wall from 15-25% to below 2%, and increases the purity of the effective acoustic intensity three-dimensional point cloud from 75-80% to 95-98%. Specifically:

[0045] Regarding intensity dual threshold filtering: When the preset intensity lower limit value I is passed... min In addition to filtering seawater noise including plankton, suspended particles, and equipment noise, the system also addresses the strong reflective environment of rectangular aquaculture pond walls by setting a preset upper limit for intensity I. max The filter tank wall has strong reflectivity; relatively speaking, for the netted environment of deep-sea cages, since the net echo (40-60dB) is much lower than the fish body echo (60-90dB), there is no need to use the preset upper limit value I. max Perform filtering;

[0046] Regarding spatial range filtering: For the strong reflection environment of rectangular aquaculture pond walls and the scanning method of sonar moving horizontally along the length direction and scanning 360° in the vertical plane, a preset boundary margin δ is used to eliminate measurement points between the aquaculture pond wall and the virtual boundary end face. This eliminates chaotic echo areas near the aquaculture pond wall, and avoids near-field reverberation echoes generated by multiple reflections between the pond wall and the water in chaotic echo areas, which cannot be accurately located and have large intensity fluctuations, thus causing misjudgment of sonar scanning.

[0047] Regarding strong echo filtration from pond walls: On the one hand, for the strong reflection environment of rectangular aquaculture pond walls, a preset threshold I for suspected echo intensity is used. th and preset echo distance threshold δ wall By combining the intensity and distance, measurement points that are likely to be reflected by the pond wall rather than fish body echoes can be more accurately eliminated. On the other hand, for rectangular aquaculture ponds, the corner reflection areas are ideal corner reflectors due to the perpendicular bottom surfaces of the ponds and the two pond walls, which can generate extremely strong echoes of up to 120dB. The echo intensity far exceeds the upper limit of 90dB for fish body echoes. Moreover, the corner reflection areas are usually not the main activity areas of fish (fish tend to avoid corners). Therefore, high-intensity measurement points in the corner reflection areas can be directly eliminated without losing effective fish information.

[0048] Preferred: The preset intensity lower limit value I of the filtering strategy one described in step S2 min and preset intensity upper limit value I max The preset boundary margin δ of the second filtering strategy is set to 64dB and 95dB respectively; the preset suspicious echo intensity threshold I of the third filtering strategy is set to 0.5 meters. th and preset echo distance threshold δ wall The values ​​were taken as 85dB and 0.6 meters, respectively.

[0049] in:

[0050] Preset lower limit of strength I min The basis for setting it to 64dB is that the swim bladder echo intensity of farmed fish (30-50cm in length) is usually in the range of 60-90dB, and the noise of seawater is usually <50dB. 64dB is located in the upper middle of the effective signal range, which can filter out most of the noise (<50dB) and retain most of the fish body signal (60-90dB).

[0051] Preset intensity upper limit value I maxThe 95dB threshold is based on the following: the upper limit of fish body echo is approximately 90dB (corresponding to the strongest cases of large fish, close-range, and frontal reflection), while the lower limit of pool wall echo is approximately 80dB (corresponding to the weakest cases of long-range and lateral reflection). 95dB is the dividing threshold between the two, filtering out most pool wall echoes (80-110dB) while retaining almost all fish body echoes (60-90dB). In practical applications, approximately 90% of points with I>95dB originate from the pool wall, and <10% originate from fish in special cases (such as when the sonar is directly aimed at the swim bladder of a large fish at very close range), which will be further filtered in subsequent steps.

[0052] The basis for setting the preset boundary margin δ to 0.5 meters is as follows: For the cage environment, δ=0.4 meters is determined based on the sound wave wavelength λ=c / f≈2.2mm (c=1500m / s, f=677kHz) and the 2λ near-field theory. However, in the pool wall environment, the near-field reverberation area is larger due to the high acoustic impedance of the concrete, requiring an increase to 0.5 meters (25% more than the cage). Actual tests show that when δ<0.4 meters, the interference from the pool wall is significant, and when δ>0.6 meters, the loss in the effective monitoring area is too large. δ=0.5 meters is the optimal balance point.

[0053] Preset suspicious echo intensity threshold I th The basis for taking a value of 85dB is based on the overlap range of fish body echo (60-90dB) and pool wall echo (80-110dB). Points below 85dB are basically fish bodies, while points above 85dB require further judgment.

[0054] Preset echo distance threshold δ wall The basis for choosing a value of 0.6 meters is that 0.6 meters is slightly larger than the boundary margin δ=0.5 meters, in order to establish a stricter filtration standard near the pool wall and ensure that the pool wall echo is fully removed.

[0055] Preferably, the multi-level filtering process in step S2 further includes:

[0056] Filtering strategy four: Statistical outlier filtering, including:

[0057] First, the remaining measurement points after filtering by filtering strategies one to three are recorded as initial filter measurement points.

[0058] Then, the distance distribution from all initial filter measurement points to the sonar and the sound wave intensity distribution of all initial filter measurement points are calculated to obtain the distance P98 percentile and the sound wave intensity P98 percentile, respectively.

[0059] Finally, the initial filter measurement points whose distance to the sonar is greater than the distance P98 percentile are removed, as are the initial filter measurement points whose sound wave intensity is greater than the sound wave intensity P98 percentile. These points are usually clutter or false detections at very long distances or with high sound wave intensity. The P98 percentile method is a robust statistical method that can automatically adapt to different data distributions and remove extreme outliers.

[0060] Preferably, the multi-level filtering process in step S2 further includes:

[0061] Filtering strategy five, voxel downsampling, includes:

[0062] First, for the remaining measurement points after filtering by filtering strategies one to three, multiple voxels are used for downsampling, wherein the diameter of the voxels ranges from 5 cm to 15 cm.

[0063] Then, for each voxel, its representative point is calculated ( , , , ): , , , In the formula, This represents the sum of the products of the sound wave intensity and the x-coordinate of all measurement points contained in the voxel. This represents the sum of the products of the acoustic wave intensity and the y-coordinate of all measurement points contained within a voxel. This represents the sum of the products of the acoustic wave intensity and the z-coordinate of all measurement points contained in the voxel. This represents the sum of the sound wave intensities of all remaining measurement points after filtering by filtering strategies one through three; thus, the representative point calculated in this way ( , , , It is not a geometric centroid, but an intensity-weighted centroid, biased towards points with higher intensity, which is closer to the actual location of the fish's body. This is because the fish's swim bladder (the main echo source) is usually located slightly above the center of the fish's body, and the point with the highest intensity is closest to the swim bladder. Therefore, the intensity-weighted average is more accurate than a simple geometric average.

[0064] Finally, the representative points of each voxel ( , , , ) is used as an effective measurement point to obtain the three-dimensional point cloud P of the effective sound wave intensity. v ={(x j , y j , z j , I j) where j=1, 2, ..., N v}Therefore, the number of said effective measurement points can be reduced and redundant data can be removed while preserving the geometric features of the original point cloud as much as possible.

[0065] Accordingly, based on filtering strategies 1 to 3, the present invention removes extreme outliers by a statistical method through filtering strategy four, and performs voxel down-sampling through filtering strategy five, so as to reduce the number of effective measurement points and remove redundant data while preserving the geometric features of the original point cloud as much as possible. Therefore, when the original acoustic intensity three-dimensional point cloud P contains a total of 1 to 2 million measurement points, effective measurement points of about 50,000 to 100,000 points can be obtained after multi-level filtering processing, and the data volume is compressed to 5-10% of the original. Meanwhile, noise, pond wall interference and redundant data are removed, and the key features of fish schools are retained.

[0066] Preferably: in said filtering strategy five, the value of the voxel diameter is 8 cm, which is based on the following: the body width of the cultured fish species is 4-10 cm (average 8 cm), and the spatial resolution of the 677 kHz sonar is about 10 cm. V=8 cm both matches the characteristic size of the fish body (can characterize the position of a single fish) and adapts to the resolution of the sonar, achieving a balance between data compression and feature retention. Within the range of 5-15 cm, the down-sampling effect is not obvious when the value is lower than 5 cm (the data volume is still large), and when the value exceeds 15 cm, adjacent fish bodies may be merged (spatial resolution is lost).

[0067] Preferably: in said step S2, a pond wall echo reference map of said rectangular culture pond is established based on data obtained by the first sonar scanning, which comprises the three-dimensional space coordinates and acoustic intensity of fixed structures of the rectangular culture pond (such as pond wall, feeding machine, aeration pipe, etc.); and for data obtained by subsequent sonar scanning, if within the error range of the three-dimensional space coordinates of the fixed structures of said rectangular culture pond (for example, the position change is less than 10 cm), a measurement point whose obtained acoustic intensity falls within the error range of the acoustic intensity of the fixed structures of said rectangular culture pond, it is automatically determined that the measurement point belongs to the fixed structures of the rectangular culture pond and is eliminated, so as to further improve the filtering accuracy of pond wall echo, and increase the purity of effective point cloud from 95-98% to 98-99%.

[0068] A feeding management method for industrial aquaculture, comprising:

[0069] Step 1: when a preset monitoring time is reached or a monitoring instruction is received, calculating the three-dimensional distribution characteristics of fish schools in said rectangular culture pond based on said three-dimensional distribution monitoring method;

[0070] Step 2: when it is necessary to feed the fish schools in the rectangular culture pond, if the depth intensity concentration obtained from the latest monitoring satisfies DCI IThe criterion of ≥0.6 indicates that the fish are highly concentrated in a certain depth layer, and stratified feeding should be carried out. Specifically:

[0071] First, a corresponding density of feed is configured for each depth layer, so that each type of feed can be suspended in the corresponding depth layer after being put into the rectangular breeding pond.

[0072] Then, the first stage of feeding is carried out: 30% of the total feed is fed into the rectangular breeding pond, and in the first stage, the proportion of various feeds is inversely proportional to the sound wave intensity density ρ of each corresponding layer depth. zk The proportion observed in the most recent monitoring resulted in: a greater amount of feed being suspended at the sound wave intensity density ρ. zk In low-density fish populations, fewer feed particles are suspended at the sound wave intensity density ρ. zk In the high-density fish strata, the fish in the high-density fish strata actively disperse to the low-density fish strata to feed.

[0073] Finally, after the first stage of feeding, the depth intensity set (DCI) was continuously monitored through step one. I until DCI is met I When the value is less than 0.4, it indicates that the fish are evenly distributed along the depth direction of the rectangular culture pond. The second stage of feeding then begins: 70% of the total feed is fed into the rectangular culture pond. Furthermore, in this second stage, the proportion of each type of feed is proportional to the sonic intensity density ρ of its corresponding depth layer. zk The proportions used in this monitoring ensured that the fish population remained evenly distributed across the depth, avoiding excessive concentration that could lead to localized oxygen depletion, while also ensuring that fish in each depth layer could feed adequately, preventing hunger and feed waste.

[0074] Therefore, this invention is based on the depth intensity concentration (DCI) I The monitoring system can perform stratified feeding when fish are found to be highly concentrated in a certain depth layer. The first stage of feeding induces fish in the high-density depth layer to actively disperse to the low-density depth layer to feed. The second stage of feeding ensures that the fish are evenly distributed in the depth direction, avoiding excessive concentration that could lead to local hypoxia. At the same time, fish in each depth layer can feed fully, avoiding hunger and feed waste, thus achieving precise feeding.

[0075] Preferably, the feeding management method further includes:

[0076] Step 3: If the monitored edge intensity aggregation index satisfies ECI IA score of ≥1.8 indicates that a large number of fish are gathering at the edge of the rectangular culture pond, which suggests that there may be an environmental anomaly in the rectangular culture pond (such as low oxygen at the bottom, excessively high water temperature in the central area, or excessively low dissolved oxygen). It is recommended to check the dissolved oxygen and temperature distribution immediately, and turn on the aeration equipment or adjust the water flow if necessary.

[0077] Step 4: If the monitored stratification intensity index satisfies LSI I The condition that the value is >0.3 and continues to increase in multiple monitoring tests indicates that the fish are continuously gathering at the surface of the rectangular aquaculture pond, which suggests that there may be discomfort in the bottom environment (such as bottom hypoxia or bottom sediment deterioration). It is recommended to increase bottom aeration or check the bottom sediment condition.

[0078] In addition, a historical comparison database can be established to calculate the historical mean μ and standard deviation σ of each parameter. When a parameter exceeds the range of μ±2σ, an anomaly alarm is triggered, prompting the aquaculture personnel to check in time.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] First, this invention utilizes the three-dimensional point cloud P of effective sound wave intensity. v Spatial partitioning statistics can be performed to analyze the three-dimensional distribution characteristics of fish within rectangular culture pond 1, including: [data to quantify the fish population's distribution along the depth direction D of rectangular culture pond 1]. H Depth of Concentration, Intensity, Concentration (DCI) I The Edge Strength Aggregation Index (ECI) is used to quantify the concentration of fish at the edge of a rectangular aquaculture pond. I And the stratification intensity index LSI used to quantify the distribution ratio of fish in the surface and bottom layers of a rectangular culture pond. I Therefore, the present invention can quantitatively and comprehensively describe the depth-direction distribution information and horizontal-direction distribution information of fish in a rectangular breeding pond 1 using three characteristic parameters.

[0081] Second, this invention is based on depth intensity concentration (DCI). I The monitoring system allows for stratified feeding when fish are highly concentrated in a specific depth layer (1A). The first stage of feeding induces fish in the high-density depth layer (1A) to actively disperse to the lower-density depth layer (1A) for feeding. The second stage of feeding then further disperses the fish population in the depth direction (D). H While maintaining a uniform distribution and avoiding excessive concentration that could lead to localized hypoxia, fish in each depth layer 1A can feed fully, avoiding hunger and feed waste, thus achieving precise feeding.

[0082] Third, in step S1 of this invention, the sonar is used along the length direction D LThe sonar scanning method, which moves horizontally and performs a 360° scan in the vertical plane, is only affected by strong reflected echoes from the two side walls and bottom of the rectangular aquaculture pond 1. There is a release gap for the echo in the direction of the water surface, that is, the water surface is a free surface. Although the reflection coefficient is high (about 0.95), the echo can be scattered upward and escape from the rectangular aquaculture pond 1. Compared with the horizontal scanning method of sonar, it significantly reduces the number of times the echo reflects and interferes with the sonar in the rectangular aquaculture pond 1, so that it is easier to eliminate the strong reflected echo interference caused by the pond walls of the rectangular aquaculture pond 1.

[0083] Furthermore, by eliminating measurement points caused by strong reflected echo interference from the walls of the rectangular aquaculture pond 1 in step S2, an effective three-dimensional point cloud Pv of acoustic intensity can be obtained without being affected by strong reflected echo interference from the walls of the rectangular aquaculture pond 1. This ensures the accuracy of the estimation of the number of fish in the rectangular aquaculture pond 1. Through experiments, the estimation error has been reduced from 20-40% to 5-10%, and the accuracy has been improved by 2-4 times.

[0084] Fourth, this invention, through the combined use of filtration strategies one through three, can distinguish between pond wall echoes and fish body echoes, completely eliminating measurement points caused by strong reflected echo interference from the pond wall of the rectangular aquaculture pond 1. This reduces the misjudgment rate of strong reflections from the pond wall by sonar scanning from 15-25% to below 2%, and increases the purity of the effective acoustic intensity three-dimensional point cloud from 75-80% to 95-98%. Specifically:

[0085] Regarding intensity dual threshold filtering: When the preset intensity lower limit value I is passed... min In addition to filtering seawater noise including plankton, suspended particles, and equipment noise, the system also addresses the strong reflective environment of the rectangular aquaculture pond 1 by setting a preset upper limit value I for the intensity. max The filter tank wall has strong reflectivity; relatively speaking, for the netted environment of deep-sea cages, since the net echo (40-60dB) is much lower than the fish body echo (60-90dB), there is no need to use the preset upper limit value I. max Perform filtering;

[0086] Regarding spatial range filtering: This is for the strongly reflective environment of the rectangular aquaculture pond 1, and the sonar along the length direction D... L The scanning method, which moves horizontally and performs a 360° scan in the vertical plane, uses a preset boundary margin δ to eliminate measurement points between the aquaculture pond wall 1-4 and the virtual boundary end face 1-4a. This eliminates chaotic echo areas near the aquaculture pond wall 1-4, and avoids near-field reverberation echoes generated by multiple reflections between the pond wall and the water in chaotic echo areas, which cannot be accurately located and have large intensity fluctuations, thus causing misjudgments in sonar scanning.

[0087] Regarding strong echo filtration of the pond walls: On the one hand, for the strong reflection environment of the rectangular aquaculture pond 1, a preset suspicious echo intensity threshold I is used. th and preset echo distance threshold δ wall The intensity and distance are combined to make a more accurate determination of the measurement points that are likely to be reflected by the pond wall rather than fish body echoes. On the other hand, for rectangular breeding pond 1, the corner reflection area 1a is an ideal corner reflector formed by the mutually perpendicular bottom surfaces 1-2 of the breeding pond and the two breeding pond walls. It can generate extremely strong echoes of up to 120dB, which far exceeds the upper limit of 90dB for fish body echoes. Moreover, the corner reflection area 1a is usually not the main activity area of ​​the fish (fish tend to avoid corners). Therefore, the high sound intensity measurement points in the corner reflection area 1a can be directly eliminated without losing effective fish information.

[0088] Fifth, based on filtering strategies one through three, this invention uses filtering strategy four to remove extreme outliers using statistical methods and filtering strategy five to perform voxel downsampling. This reduces the number of effective measurement points and removes redundant data while preserving the geometric features of the original point cloud as much as possible. Therefore, the original three-dimensional point cloud P containing 1-2 million measurement points can be processed through multi-level filtering to obtain approximately 50,000-100,000 effective measurement points, reducing the data volume to 5-10% of the original. At the same time, noise, pool wall interference, and redundant data are removed, while retaining the key characteristics of the fish school. Attached Figure Description

[0089] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments:

[0090] Figure 1 This is a flowchart of the present invention;

[0091] Figure 2 This is a schematic diagram of the depth layering in step S3-1 of the present invention;

[0092] Figure 3 This is a schematic diagram of the horizontal grid in step S3-2 of the present invention;

[0093] Figure 4 This is a side view of the installation structure of the scanning device in a rectangular aquaculture pond according to the present invention;

[0094] Figure 5 This is a top view of the scanning device in a rectangular aquaculture pond according to the present invention. Detailed Implementation

[0095] The present invention will be described in detail below with reference to the examples and the accompanying drawings to help those skilled in the art better understand the inventive concept of the present invention. However, the protection scope of the claims of the present invention is not limited to the following examples. For those skilled in the art, all other embodiments obtained without creative work on the premise of not departing from the inventive concept of the present invention shall fall within the protection scope of the present invention.

[0096] Example 1

[0097] As shown in Figures 1 to 3 , the present invention discloses a three-dimensional distribution monitoring method for industrial culture, comprising:

[0098] Step S1: performing sonar scanning on the interior of the rectangular culture pond 1 to obtain a sound intensity three-dimensional point cloud P={(x i , y i , z i , I i )丨i=1,2,…,N}, wherein N represents the total number of measurement points obtained by the sonar scanning, which is usually hundreds of thousands to millions of measurement points, (x i , y i , z i ) and I i respectively represent the three-dimensional spatial coordinate and sound intensity of the i-th measurement point;

[0099] Step S2: performing filtering processing on the sound intensity three-dimensional point cloud P, and recording the set of measurement points remaining after filtering as an effective sound intensity three-dimensional point cloud P v ={(x j , y j , z j , I j )丨j=1,2,…,N v}, wherein N v represents the total number of effective measurement points, (x j , y j , z j ) and I j respectively represent the three-dimensional spatial coordinate and sound intensity of the j-th effective measurement point;

[0100] Step S3: performing spatial partition statistics on the effective sound intensity three-dimensional point cloud P v to analyze and obtain the three-dimensional distribution characteristics of fish schools in the rectangular culture pond 1, which includes: depth concentration index DCI H for quantifying the concentration degree of fish schools in the depth direction D I of the rectangular culture pond 1, and edge concentration index ECI IAnd the stratification intensity index LSI used to quantify the distribution ratio of fish in the surface and bottom layers of a rectangular culture pond. I .

[0101] The method for performing spatial partition statistics in step S3 is as follows:

[0102] Step S3-1: Move the rectangular aquaculture pond 1 along its depth direction D H The depth is divided into n equally spaced layers 1A, and these layers are numbered from bottom to top as layer 1 to layer n. The number of valid measurement points N in the kth layer 1A is then calculated. zk Total sound wave intensity I zk Volume V zk Sound wave intensity density ρ zk =I zk / V zk Where k = 1, 2, ..., n, and N is the number of valid measurement points. zk That is, the total number of all valid measurement points in the k-th depth layer 1A, and the sum of the sound wave intensities I. zk That is, the sum of the sound wave intensities at all valid measurement points in the k-th depth layer 1A;

[0103] Step S3-2: The rectangular aquaculture pond 1 is positioned along its length D... L and width direction D W The upper equal interval is divided into m L column and m W The row is divided into m in the horizontal direction. L ×m W A rectangular horizontal grid 1B is formed, and the 2m area located at the edge of the rectangular aquaculture pond 1 is... L +2m W - Four horizontal grids 1B are denoted as edge grids 1C; and the number of valid measurement points N of the horizontal grid 1B in the l-th column and h-th row is statistically obtained. lh Total sound wave intensity I lh Area A lh Sound wave intensity density ρ lh =I lh / A lh Where l = 1, 2, ..., m L h=1,2,...,m W Number of valid measurement points N lh That is, the number of all valid measurement points within the horizontal grid 1B of the l-th column and h-th row of the rectangular aquaculture pond 1, and the total sound wave intensity I. lhthat is, the sum of the sound wave intensities of all effective measurement points located within the horizontal grid 1B in the l-th column and the h-th row of the rectangular culture pond 1; and, the horizontal area A of the rectangular culture pond 1, the sum of the sound wave intensities ΣI of all measurement points of the rectangular culture pond 1 i , the average intensity density ρ in the plane of the rectangular culture pond 1 mean =ΣI i / A, the average intensity density ρ of the edge grid 1C edge =ΣI ic / A c , wherein ΣI ic is the sum of the sound wave intensities of all edge grids 1C, and A c is the sum of the areas of all edge grids 1C;

[0104] Step S3-3: Calculate the three-dimensional distribution characteristics of fish schools in the rectangular culture pond 1:

[0105] Depth Intensity Concentration DCI I =(ΣΣ|ρ zk1 - ρ zk2 |) / (2n 2 ×ρ z );

[0106] wherein, |ρ zk1 - ρ zk2 | represents the absolute value of the difference between the sound wave intensity densities of the depth layer 1A at the k1-th layer and the k2-th layer, and ΣΣ|ρ zk1 - ρ zk2 | represents the sum of |ρ zk1 - ρ zk2 | corresponding to traversing all combinations of k1=1,2,...,n and k2=1,2,...,n; n is the number of layers of the depth layers 1A; ρ z is the average value of the sound wave intensity density ρ of all depth layers 1A zk ;

[0107] Edge Intensity Aggregation Index ECI I =ρ edge / ρ mean ;

[0108] Layered Strength Index LSI I =(ρ upper – ρ lower ) / (ρ upper + ρ lower );

[0109] wherein, ρ upper and ρ lowerThe acoustic intensity density ρ represents the depth stratification 1A contained in the upper and lower halves of the rectangular aquaculture pond 1, respectively. zk The average value, and when the number of layers n is odd, the upper half and the lower half respectively contain the upper half and the lower half of the (n+1) / 2th layer depth layer 1A.

[0110] Therefore, this invention utilizes the three-dimensional point cloud P of effective sound wave intensity. v Spatial partitioning statistics can be performed to analyze the three-dimensional distribution characteristics of fish within rectangular culture pond 1, including: [data to quantify the fish population's distribution along the depth direction D of rectangular culture pond 1]. H Depth of Concentration, Intensity, Concentration (DCI) I The Edge Strength Aggregation Index (ECI) is used to quantify the concentration of fish at the edge of a rectangular aquaculture pond. I And the stratification intensity index LSI used to quantify the distribution ratio of fish in the surface and bottom layers of a rectangular culture pond. I Therefore, the present invention can quantitatively and comprehensively describe the depth-direction distribution information and horizontal-direction distribution information of fish in a rectangular breeding pond 1 using three characteristic parameters.

[0111] Among them, the depth intensity concentration (DCI) I The criteria for judgment are: DCI I When the value is less than 0.3, it indicates low concentration, meaning the fish population is concentrated in the depth direction D of the rectangular culture pond 1. H The concentration is low and the distribution is uniform; 0.3≤DCI I A concentration level <0.6 indicates a moderate concentration; DCI I A value ≥0.6 indicates high concentration, meaning the fish are concentrated in a certain depth layer 1A.

[0112] The edge intensity aggregation index (ECI) I The criteria for determination are: ECI I When the density is <0.8, it indicates central aggregation, meaning the fish population tends to concentrate in the center of the rectangular rearing pond 1, with lower density at the edges; 0.8 ≤ ECI I When <1.3, it indicates that the fish population is evenly distributed; 1.3≤ECI I When the value is less than 1.8, it indicates an edge-gathering warning, meaning the fish begin to gather towards the edge of the rectangular aquaculture pond 1; ECI I When the value is ≥1.8, it is a serious edge aggregation alarm, indicating that a large number of fish are gathering at the edge of the rectangular aquaculture pond 1, which may indicate an environmental anomaly.

[0113] The layering strength index LSI I The criterion for determination is: LSI I When LSI is less than -0.3, it indicates that the fish are mainly distributed at the bottom of the rectangular aquaculture pond 1, with the fish tending towards the bottom; -0.3 ≤ LSII When the value is <0.3, the fish population is evenly distributed on both the surface and bottom layers of rectangular culture pond 1; LSI I When the value is ≥0.3, it indicates that the fish are mainly distributed on the surface of the rectangular aquaculture pond 1, and the fish tend to be closer to the water surface.

[0114] The above is the basic implementation method of this embodiment one, and further optimizations, improvements and limitations can be made based on this basic implementation method:

[0115] Preferably, in step S3-1, the stratification interval Δz of the rectangular aquaculture pond 1, that is, the height of the depth stratification 1A, ranges from 0.8 meters to 2.5 meters. This is to avoid the stratification interval Δz being too small, which may cause fish populations in adjacent depth stratification 1A to overlap and lose the significance of stratification statistics, and to avoid the stratification interval Δz being too large, which may fail to reflect the details of the vertical distribution of fish populations. This ensures that each depth stratification 1A has sufficient resolution without having too few samples. Furthermore, within the range of the stratification interval Δz, the number of stratifications n is preferably between 2 and 4.

[0116] In step S3-2, the edge size of the horizontal grid 1B ranges from 8 meters to 12 meters to balance spatial resolution and sample size, avoiding the following: if the horizontal grid 1B is too small, the number of measurement points in each horizontal grid 1B will be too small, resulting in statistical instability; if the measurement points are too large, the spatial resolution will be insufficient, and the distribution differences will not be reflected.

[0117] In addition, the three-dimensional distribution monitoring method can generate a plot of data based on the statistical data in step S3 to visually display the distribution of fish schools, including: firstly, generating a depth bar chart to display the intensity distribution of each layer: the horizontal axis is the depth layer number (1,2,3,...,n, from bottom to top), and the vertical axis is the sum of the acoustic wave intensity I of that layer. zk Each layer is represented by a pillar, and the pillar height reflects the fish density at that layer. Secondly, a horizontal heatmap is generated to display m. L ×m W Mesh intensity distribution: Rectangular mesh layout (m) L line × m W (columns), each grid displays the acoustic intensity density ρ of that region. lh A color gradient is used to represent the intensity (e.g., blue represents low density and red represents high density).

[0118] Example 2

[0119] Based on the above embodiment one, this embodiment two also adopts the following preferred implementation method:

[0120] like Figure 4 and Figure 5 As shown, in step S1, the sonar scanning method is as follows: the sonar scans along the length direction D of the rectangular aquaculture pond 1.L Horizontal movement, and the sonar in the length direction D L A 360° scan is performed within the vertical plane, and the 360° scan range of the sonar covers the rectangular aquaculture pond 1 perpendicular to the length direction D. L The cross-section;

[0121] In step S2, the three-dimensional point cloud P of the sound wave intensity is subjected to multi-level filtering processing, which at least includes removing the measurement points generated by the strong reflected echo interference of the pool wall of the rectangular aquaculture pond 1 on the sonar scan.

[0122] Therefore, in step S1, the present invention employs sonar along the length direction D L The sonar scanning method, which moves horizontally and performs a 360° scan in the vertical plane, is only affected by strong reflected echoes from the two side walls and bottom of the rectangular aquaculture pond 1. There is a release gap for the echo in the direction of the water surface, that is, the water surface is a free surface. Although the reflection coefficient is high (about 0.95), the echo can be scattered upward and escape from the rectangular aquaculture pond 1. Compared with the horizontal scanning method of sonar, it significantly reduces the number of times the echo reflects and interferes with the sonar in the rectangular aquaculture pond 1, so that it is easier to eliminate the strong reflected echo interference caused by the pond walls of the rectangular aquaculture pond 1.

[0123] Furthermore, by eliminating measurement points caused by strong reflected echo interference from the walls of the rectangular aquaculture pond 1 in step S2, an effective three-dimensional point cloud Pv of acoustic intensity can be obtained without being affected by strong reflected echo interference from the walls of the rectangular aquaculture pond 1. This ensures the accuracy of the estimation of the number of fish in the rectangular aquaculture pond 1. Through experiments, the estimation error has been reduced from 20-40% to 5-10%, and the accuracy has been improved by 2-4 times.

[0124] The above is the basic implementation method of this embodiment two, and further optimizations, improvements and limitations can be made based on this basic implementation method:

[0125] Preferably, the multi-level filtering process in step S2 includes:

[0126] Filtering Strategy 1: Intensity Dual Threshold Filtering: Eliminating sound wave intensity I i Less than the preset lower limit of strength I min Or greater than the preset intensity upper limit value I max The measurement points, among which, the preset lower limit value I min The value range is 50dB to 70dB, used for filtering seawater noise (including plankton, suspended particles, equipment noise, etc., with an intensity typically <50dB); the preset upper limit of intensity is I. max The value range is 90dB to 100dB, used for strong reflections in the filter tank wall;

[0127] Filtering Strategy 2: Spatial Range Filtering: Eliminating 3D spatial coordinates (x... i , y i , z i Measurement points extending beyond the boundary of the rectangular aquaculture pond 1, wherein the boundary of the rectangular aquaculture pond 1, with length, width, and height L, W, and H respectively, includes: the virtual top surface 1-1 of the aquaculture pond, i.e., excluding z. i Measurement points ≥H, 1-2 on the bottom of the aquaculture pond (i.e., excluding z) i (Measurement points ≤ 0), length direction D L The walls of the two aquaculture ponds above, 1-3 (i.e., removing x) i <0 or x i >L (measurement point), width direction D W The two aquaculture pond walls 1-4 are recessed inward by a preset boundary margin δ to form two virtual boundary end faces 1-4a (i.e., removing |y). i |≤W / 2-δ measurement point), the preset boundary margin δ ranges from 0.4 meters to 0.6 meters; among them, because the sonar moves horizontally along the length direction DL and performs a 360° scan in the vertical plane, the two aquaculture pond walls 1-3 on the length direction DL do not need to be set with a preset boundary margin δ.

[0128] Filtration Strategy 3: Strong Echo Filtration from Pool Walls

[0129] For sound wave intensity I i Greater than the preset suspicious echo intensity threshold I th If the distance between the measurement point and any wall of the rectangular aquaculture pond 1 is less than the preset echo distance threshold δ, then... wall If the measurement point is not found, then the measurement point is discarded; wherein, the preset suspicious echo intensity threshold I th The value range is 80dB to 90dB, and the preset echo distance threshold δ wall The value ranges from 0.5 meters to 0.8 meters;

[0130] For the measurement points in the four corner reflection areas 1a of the rectangular aquaculture pond 1, if their sound wave intensity I i If the value is greater than 80dB, the measurement point is discarded; wherein, the corner reflection area 1a is the area within 1 meter of the corner point 1-2 of the bottom surface of the aquaculture pond, and the 1-meter threshold is an empirical value that covers the main influence range of corner reflection.

[0131] Therefore, by combining filtration strategies one through three, this invention can distinguish between pond wall echoes and fish body echoes, completely eliminating measurement points caused by strong reflected echo interference from the pond wall of the rectangular aquaculture pond 1. This reduces the misjudgment rate of strong reflections from the pond wall by sonar scanning from 15-25% to below 2%, and increases the purity of the effective acoustic intensity three-dimensional point cloud from 75-80% to 95-98%. Specifically:

[0132] Regarding intensity dual threshold filtering: When the preset intensity lower limit value I is passed... min In addition to filtering seawater noise including plankton, suspended particles, and equipment noise, the system also addresses the strong reflective environment of the rectangular aquaculture pond 1 by setting a preset upper limit value I for the intensity. max The filter tank wall has strong reflectivity; relatively speaking, for the netted environment of deep-sea cages, since the net echo (40-60dB) is much lower than the fish body echo (60-90dB), there is no need to use the preset upper limit value I. max Perform filtering;

[0133] Regarding spatial range filtering: This is for the strongly reflective environment of the rectangular aquaculture pond 1, and the sonar along the length direction D... L The scanning method, which moves horizontally and performs a 360° scan in the vertical plane, uses a preset boundary margin δ to eliminate measurement points between the aquaculture pond wall 1-4 and the virtual boundary end face 1-4a. This eliminates chaotic echo areas near the aquaculture pond wall 1-4, and avoids near-field reverberation echoes generated by multiple reflections between the pond wall and the water in chaotic echo areas, which cannot be accurately located and have large intensity fluctuations, thus causing misjudgments in sonar scanning.

[0134] Regarding strong echo filtration of the pond walls: On the one hand, for the strong reflection environment of the rectangular aquaculture pond 1, a preset suspicious echo intensity threshold I is used. th and preset echo distance threshold δ wall The intensity and distance are combined to make a more accurate determination of the measurement points that are likely to be reflected by the pond wall rather than fish body echoes. On the other hand, for rectangular breeding pond 1, the corner reflection area 1a is an ideal corner reflector formed by the mutually perpendicular bottom surfaces 1-2 of the breeding pond and the two breeding pond walls. It can generate extremely strong echoes of up to 120dB, which far exceeds the upper limit of 90dB for fish body echoes. Moreover, the corner reflection area 1a is usually not the main activity area of ​​the fish (fish tend to avoid corners). Therefore, the high sound intensity measurement points in the corner reflection area 1a can be directly eliminated without losing effective fish information.

[0135] Preferred: The preset intensity lower limit value I of the filtering strategy one described in step S2 min and preset intensity upper limit value I maxThe preset boundary margin δ of the second filtering strategy is set to 64dB and 95dB respectively; the preset suspicious echo intensity threshold I of the third filtering strategy is set to 0.5 meters. th and preset echo distance threshold δ wall The values ​​were taken as 85dB and 0.6 meters, respectively.

[0136] in:

[0137] Preset lower limit of strength I min The basis for setting it to 64dB is that the swim bladder echo intensity of farmed fish (30-50cm in length) is usually in the range of 60-90dB, and the noise of seawater is usually <50dB. 64dB is located in the upper middle of the effective signal range, which can filter out most of the noise (<50dB) and retain most of the fish body signal (60-90dB).

[0138] Preset intensity upper limit value I max The 95dB threshold is based on the following: the upper limit of fish body echo is approximately 90dB (corresponding to the strongest cases of large fish, close-range, and frontal reflection), while the lower limit of pool wall echo is approximately 80dB (corresponding to the weakest cases of long-range and lateral reflection). 95dB is the dividing threshold between the two, filtering out most pool wall echoes (80-110dB) while retaining almost all fish body echoes (60-90dB). In practical applications, approximately 90% of points with I>95dB originate from the pool wall, and <10% originate from fish in special cases (such as when the sonar is directly aimed at the swim bladder of a large fish at very close range), which will be further filtered in subsequent steps.

[0139] The basis for setting the preset boundary margin δ to 0.5 meters is as follows: For the cage environment, δ=0.4 meters is determined based on the sound wave wavelength λ=c / f≈2.2mm (c=1500m / s, f=677kHz) and the 2λ near-field theory. However, in the pool wall environment, the near-field reverberation area is larger due to the high acoustic impedance of the concrete, requiring an increase to 0.5 meters (25% more than the cage). Actual tests show that when δ<0.4 meters, the interference from the pool wall is significant, and when δ>0.6 meters, the loss in the effective monitoring area is too large. δ=0.5 meters is the optimal balance point.

[0140] Preset suspicious echo intensity threshold I th The basis for taking a value of 85dB is based on the overlap range of fish body echo (60-90dB) and pool wall echo (80-110dB). Points below 85dB are basically fish bodies, while points above 85dB require further judgment.

[0141] Preset echo distance threshold δ wall The basis for choosing a value of 0.6 meters is that 0.6 meters is slightly larger than the boundary margin δ=0.5 meters, in order to establish a stricter filtration standard near the pool wall and ensure that the pool wall echo is fully removed.

[0142] Preferably, the multi-level filtering process in step S2 further includes:

[0143] Filtering strategy four: Statistical outlier filtering, including:

[0144] First, the remaining measurement points after filtering by filtering strategies one to three are recorded as initial filter measurement points.

[0145] Then, the distance distribution from all initial filter measurement points to the sonar and the sound wave intensity distribution of all initial filter measurement points are calculated to obtain the distance P98 percentile and the sound wave intensity P98 percentile, respectively.

[0146] Finally, the initial filter measurement points whose distance to the sonar is greater than the distance P98 percentile are removed, as are the initial filter measurement points whose sound wave intensity is greater than the sound wave intensity P98 percentile. These points are usually clutter or false detections at very long distances or with high sound wave intensity. The P98 percentile method is a robust statistical method that can automatically adapt to different data distributions and remove extreme outliers.

[0147] Preferably, the multi-level filtering process in step S2 further includes:

[0148] Filtering strategy five, voxel downsampling, includes:

[0149] First, for the remaining measurement points after filtering by filtering strategies one to three, multiple voxels are used for downsampling, wherein the diameter of the voxels ranges from 5 cm to 15 cm.

[0150] Then, for each voxel, its representative point is calculated ( , , , ): , , , In the formula, This represents the sum of the products of the sound wave intensity and the x-coordinate of all measurement points contained in the voxel. This represents the sum of the products of the acoustic wave intensity and the y-coordinate of all measurement points contained within a voxel. This represents the sum of the products of the acoustic wave intensity and the z-coordinate of all measurement points contained in the voxel. This represents the sum of the sound wave intensities of all remaining measurement points after filtering by filtering strategies one through three; thus, the representative point calculated in this way ( , , , ) It is not a geometric centroid, but an intensity-weighted centroid, which is biased towards points with higher intensity and closer to the actual position of the fish body. This is because the swim bladder of the fish body (the main echo source) is usually located at the upper-central position of the fish body, and the point with the highest intensity is closest to the position of the swim bladder. Therefore, intensity-weighted averaging is more accurate than simple geometric averaging.

[0151] Finally, the representative point of each voxel ( , , , ) is taken as one said effective measurement point, so as to obtain said effective acoustic intensity three-dimensional point cloud P v ={(x j , y j , z j , I j )丨j=1,2,…,N v} Therefore, the number of said effective measurement points can be reduced and redundant data can be removed on the premise of retaining the geometric features of the original point cloud as much as possible.

[0152] Accordingly, based on the first filtering strategy to the third filtering strategy, the present invention removes extreme outliers by a statistical method through the fourth filtering strategy, and performs voxel downsampling through the fifth filtering strategy, so as to reduce the number of effective measurement points and remove redundant data on the premise of retaining the geometric features of the original point cloud as much as possible. Therefore, when the original acoustic intensity three-dimensional point cloud P contains a total of 1-2 million measurement points, after multi-level filtering processing, about 50,000-100,000 effective measurement points can be obtained, the data volume is compressed to 5-10% of the original, while noise, pond wall interference and redundant data are removed, and the key features of the fish school are retained.

[0153] Preferably: in said fifth filtering strategy, the value range of the voxel diameter is 8 cm, and the basis is: the body width of cultured fish is 4-10 cm (average 8 cm), the spatial resolution of 677 kHz sonar is about 10 cm, V=8 cm matches the characteristic size of the fish body (can represent the position of a single fish) and adapts to the sonar resolution, achieving a balance between data compression and feature retention. In the range of 5-15 cm, the downsampling effect is not obvious when the value is lower than 5 cm (the data volume is still large), and adjacent fish bodies may be merged when the value exceeds 15 cm (resulting in loss of spatial resolution).

[0154] Preferably, in step S2, the data obtained from the initial sonar scan is used to establish a reference map of the echo of the rectangular aquaculture pond 1, which includes the three-dimensional spatial coordinates and acoustic intensity of the fixed structures of the rectangular aquaculture pond (such as the pond wall, feeder, aerator, etc.). Furthermore, for the data obtained from subsequent sonar scans, if the three-dimensional spatial coordinates of the fixed structures of the rectangular aquaculture pond are within the error range (e.g., the position change is less than 10 cm), and the measurement points obtained from the scan are within the error range of the acoustic intensity of the fixed structures of the rectangular aquaculture pond, the measurement points are automatically determined to belong to the fixed structures of the rectangular aquaculture pond and are removed, so as to further improve the accuracy of the pond wall echo filtration and increase the effective point cloud purity from 95-98% to 98-99%.

[0155] Example 3

[0156] Based on the above embodiment one or embodiment two, this embodiment three also adopts the following preferred implementation method:

[0157] The sonar scanning in step S1 is performed using the following scanning device:

[0158] The rectangular aquaculture pond 1 is internally fixedly equipped with a device D along its length direction. L An extended horizontal track 2 is provided, and a slider 3 is slidably mounted on the horizontal track 2 via rollers or linear bearings. A drive control system can drive the slider 3 to move along the length direction D of the horizontal track 2. L The sonar 4 moves at a constant speed; and the sonar 4 is mounted laterally on the slider 3, so that the sonar 4 is located in the depth direction D of the rectangular aquaculture pond 1. H The sonar 4 is located in the middle position and is capable of operating in the length direction D. L A 360° scan is performed within the vertical plane of sonar 4, covering the rectangular aquaculture pond 1 perpendicular to the length direction D. L The cross-section of the sonar 4; the slider 3 is equipped with a position sensor for measuring the distance from the sonar 4 to the length direction D. L The distance between the aquaculture pond wall and the position sensor can be 1-3 cm. The position sensor can be an encoder, a rope displacement sensor or a laser rangefinder, etc., with an accuracy of ≤2 cm.

[0159] The horizontal track 2 is preferably made of stainless steel or aluminum alloy, which has corrosion resistance and high strength; the slider 3 moves at a speed of 0.05-0.3 m / s on the horizontal track 2, preferably 0.1 m / s; the sonar 4 has a scanning angle range of 120 degrees.

[0160] The intensity matrix I of the measurement point is obtained by the sonar 4 scan. The distances (x, r, t) and synchronously acquired by the position sensors can be converted into a three-dimensional point cloud of sound wave intensity, P={(x, r, t)}. i , y i, z i , I i ) | i=1,2,…,N}, the specific conversion method is as follows:

[0161] First, obtain the intensity matrix I( ,r,t) and the corresponding distance collected by the position sensor, wherein t is the scanning time of the measurement point, is the vertical scanning angle of the measurement point, that is, the rotation angle of the direction from the sonar 4 to the measurement point relative to the horizontal reference direction, and r is the distance between the sonar 4 and the measurement point;

[0162] Then, establish a three-dimensional coordinate system, with the length direction D of the rectangular aquaculture pond 1 L , width direction D W and depth direction D H as the X-axis, Y-axis and Z-axis of the three-dimensional coordinate system respectively;

[0163] Finally, according to the following coordinate conversion formula, convert the intensity matrix I( ,r,t) into the three-dimensional space coordinates and acoustic intensity of the measurement point, so as to obtain the three-dimensional acoustic intensity point cloud P={(x i ,y i , z i , I i ) | i=1,2,…,N}:

[0164] x i = x rov ;

[0165] y i = r×cos ;

[0166] z i =H / 2 + r×sin ;

[0167] In the formula, x rov is converted from the distance collected by the position sensor under the three-dimensional coordinate system; H is the height of the rectangular aquaculture pond 1.

[0168] I( ,r,t) is converted into the three-dimensional space coordinates and acoustic intensity of the measurement point, so as to obtain the three-dimensional acoustic intensity point cloud P={(x i , y i , z i , I i ) | i=1,2,…,N}.

[0169] Therefore, step S1 of the present invention uses a scanning device consisting of a horizontal track 2, a slider 3, a drive control system, a sonar 4, and a position sensor to achieve sonar scanning. It has the advantages of simple structure, low cost (saving 50-70%), convenient maintenance, automatic reciprocating scanning, and high degree of automation.

[0170] Furthermore, the scanning device achieves full three-dimensional spatial coverage of the pool through a combination of horizontal sonar movement and vertical fan-shaped scanning. Compared to ROV or boom lift solutions, the horizontal track system has the following advantages: First, the track is fixed, ensuring precise horizontal movement of the sonar without drift; second, it is installed at the middle depth position, providing uniform coverage of both the upper and lower parts; third, the horizontal installation of the sonar enables vertical scanning, covering the entire depth range with a single horizontal movement, eliminating the need for vertical lifting; fourth, it has a simple structure, lower cost than ROV (saving 50-70%), and is easy to maintain; fifth, it can achieve automatic reciprocating scanning, making it suitable for continuous monitoring.

[0171] The above is the basic implementation method of this embodiment three, and further optimizations, improvements and limitations can be made based on this basic implementation method:

[0172] Preferably, the scanning device operates according to a preset automatic reciprocating scanning mode, including: the sonar 4 moves unidirectionally from the starting point to the ending point on the horizontal track 2 with the slider 3 to complete one single-pass scan, and then automatically returns to the starting point to perform the next single-pass scan; the single-pass scan is automatically performed at set times through a timer or remote control, for example, once at 10:00 AM and once at 4:00 PM every day; the scanning data of the sonar 4 and the sensing data of the position sensor are automatically uploaded to the cloud or local server, and corresponding analysis reports are generated, and alarm information is automatically sent in abnormal situations.

[0173] Example 4

[0174] This invention also discloses a feeding management method for factory farming, comprising:

[0175] Step 1: Calculate the three-dimensional distribution characteristics of the fish population in the rectangular aquaculture pond 1 according to the preset monitoring time or upon receiving a monitoring instruction, based on the three-dimensional distribution monitoring method described in any one of Examples 1 to 3.

[0176] Step 2: When feeding the fish in rectangular culture pond 1, if the depth intensity concentration detected most recently meets the DCI (Distribution Criteria) standard... I The criterion of ≥0.6 indicates that the fish are highly concentrated in a certain depth layer 1A, in which case stratified feeding shall be carried out, specifically as follows:

[0177] First, a feed of a corresponding density is configured for each depth layer 1A, so that each feed can be suspended in the corresponding depth layer 1A after being put into the rectangular breeding pond 1.

[0178] Then, the first stage of feeding is carried out: 30% of the total feed is fed into the rectangular aquaculture pond 1, and in the first stage, the proportion of various feeds is inversely proportional to the sound wave intensity density ρ of each corresponding layer depth 1A. zk The proportion observed in the most recent monitoring resulted in: a greater amount of feed being suspended at the sound wave intensity density ρ. zk In low-density fish strata (1A), fewer feed particles are suspended at the sound intensity density ρ. zk In the high-density fish stratum 1A, the fish population in the high-density fish stratum 1A is induced to actively disperse to the low-density fish stratum 1A to feed.

[0179] Finally, after the first stage of feeding, the depth intensity set (DCI) was continuously monitored through step one. I until DCI is met I When <0.4, it indicates that the fish population is in the depth direction D of the rectangular culture pond 1. H The feed is evenly distributed, and the second stage of feeding is carried out: 70% of the total feed is fed into the rectangular aquaculture pond 1. In the second stage, the proportion of various feeds fed is proportional to the sound wave intensity density ρ of the corresponding layer depth 1A. zk The proportions observed during this monitoring resulted in: the fish population being in the depth direction D... H While maintaining a uniform distribution and avoiding excessive concentration that could lead to localized hypoxia, fish in each depth layer 1A can feed fully, avoiding hunger and feed waste.

[0180] Therefore, this invention is based on the depth intensity concentration (DCI) I The monitoring system allows for stratified feeding when fish are highly concentrated in a specific depth layer (1A). The first stage of feeding induces fish in the high-density depth layer (1A) to actively disperse to the lower-density depth layer (1A) for feeding. The second stage of feeding then further disperses the fish population in the depth direction (D). H While maintaining a uniform distribution and avoiding excessive concentration that could lead to localized hypoxia, fish in each depth layer 1A can feed fully, avoiding hunger and feed waste, thus achieving precise feeding.

[0181] The above is the basic implementation method of this embodiment four, and further optimizations, improvements and limitations can be made based on this basic implementation method:

[0182] Preferably, the feeding management method further includes:

[0183] Step 3: If the monitored edge intensity aggregation index satisfies ECI IThe judgment condition of ≥1.8 indicates that a large number of fish are gathering at the edge of the rectangular culture pond 1, which gives a prompt that there may be an environmental abnormality in the rectangular culture pond 1 (such as low oxygen at the bottom, excessively high water temperature in the central area, or excessively low dissolved oxygen). It is recommended to check the dissolved oxygen and temperature distribution immediately, and turn on the aeration equipment or adjust the water flow if necessary.

[0184] Step 4: If the monitored stratification intensity index satisfies LSI I The condition that the value is >0.3 and continues to increase in multiple monitoring tests indicates that the fish are continuously gathering at the surface of the rectangular culture pond 1, which suggests that there may be discomfort in the bottom environment (such as bottom hypoxia or bottom sediment deterioration). It is recommended to increase bottom oxygenation or check the bottom sediment condition.

[0185] In addition, a historical comparison database can be established to calculate the historical mean μ and standard deviation σ of each parameter. When a parameter exceeds the range of μ±2σ, an anomaly alarm is triggered, prompting the aquaculture personnel to check in time.

[0186] This invention is not limited to the specific embodiments described above. Based on the above content and in accordance with common technical knowledge and conventional methods in the field, without departing from the basic technical concept of this invention, this invention can also make other equivalent modifications, substitutions or alterations, all of which fall within the protection scope of this invention.

Claims

1. A three-dimensional distribution monitoring method for factory farming, characterized in that, include: Step S1: performing sonar scanning on the interior of the rectangular culture pond (1) to obtain a three-dimensional point cloud of sound wave intensity P={(x i , y i , z i , I i )丨i=1,2,…,N}, wherein N represents the total number of measurement points obtained by the sonar scanning, (x i , y i , z i ) and I i represent the three-dimensional spatial coordinate and sound wave intensity of the i-th measurement point respectively; Step S2: filtering the three-dimensional point cloud P of sound wave intensity, and recording the set of measurement points remaining after filtering as the effective three-dimensional point cloud P of sound wave intensity v ={(x j , y j , z j , I j )|j=1,2,…,N v}, wherein, N v represents the total number of effective measurement points, (x j , y j , z j ) and I j represent the three-dimensional spatial coordinate and sound wave intensity of the j-th effective measurement point respectively; Step S3: For the effective acoustic wave intensity three-dimensional point cloud P... v Spatial partitioning statistics were performed to analyze the three-dimensional distribution characteristics of fish in the rectangular aquaculture pond (1), including: quantifying the depth direction (D) of the fish population in the rectangular aquaculture pond (1). H Depth of concentration intensity concentration (DCI) I The Edge Strength Aggregation Index (ECI) is used to quantify the concentration of fish at the edge of a rectangular aquaculture pond (1). I And the stratification intensity index LSI used to quantify the distribution ratio of fish in the surface and bottom layers of a rectangular culture pond (1). I .

2. The three-dimensional distribution monitoring method for factory farming according to claim 1, characterized in that: The method for performing spatial partition statistics in step S3 is as follows: Step S3-1: Move the rectangular aquaculture pond (1) along its depth direction (D) H The depth is divided into n equal-interval layers (1A), and the depth layers (1A) are numbered from bottom to top as layer 1 to layer n. The total acoustic intensity of the kth depth layer (1A) is statistically obtained as I. zk Volume V zk Sound wave intensity density ρ zk =I zk / V zk Where k = 1, 2, ..., n; Step S3-2: The rectangular aquaculture pond (1) is divided into sections along its length (D) L ) and width direction (D W The upper part is divided into m equal intervals. L column and m W The row is divided into m in the horizontal direction. L ×m W A rectangular horizontal grid (1B) is formed, and 2m of the area located at the edge of the rectangular aquaculture pond (1) is... L +2m W - Four horizontal grids (1B) are designated as edge grids (1C); and the total acoustic intensity of the horizontal grids (1B) in the l-th column and h-th row is statistically obtained as I. lh Area A lh Sound wave intensity density ρ lh =I lh / A lh Where l = 1, 2, ..., m L h=1,2,...,m W ; and, the horizontal area A of the rectangular aquaculture pond (1), the sum of the sound wave intensities ΣI at all measurement points of the rectangular aquaculture pond (1). i The planar average strength density ρ of the rectangular aquaculture pond (1) mean =ΣI i / A, Average intensity density ρ of the edge mesh (1C) edge =ΣI ic / A c , where ΣI ic A is the sum of the acoustic wave intensities of all edge grids (1C). c This represents the total area of ​​all edge grids (1C). Step S3-3: Calculate the three-dimensional distribution characteristics of the fish population within the rectangular aquaculture pond (1): Depth Intensity Concentration DCI I =(ΣΣ|ρ zk1 - ρ zk2 |) / (2n 2 ×ρ z ); wherein |ρ zk1 - ρ zk2 | represents the absolute value of the difference between the sound intensity densities of the k1-th layer and the k2-th depth layer (1A), ΣΣ|ρ zk1 - ρ zk2 | represents |ρ corresponding to all combinations of k1=1,2,...,n and k2=1,2,...,n zk1 - ρ zk2 | sum; n is the number of layers of the depth layer (1A); ρ z is the average value of the sound intensity density ρ of all depth layers (1A) zk ; Edge Intensity Aggregation Index (ECI) I =ρ edge / ρ mean ; Stratified Intensity Index (LSI) I =(ρ upper – ρ lower ) / (ρ upper + ρ lower ); Where, ρ upper and ρ lower The acoustic intensity density ρ represents the depth stratification (1A) contained in the upper and lower halves of the rectangular aquaculture pond (1), respectively. zk The average value, and when the number of layers n is odd, the upper half and the lower half respectively contain the upper half and the lower half of the (n+1) / 2th depth layer (1A).

3. The three-dimensional distribution monitoring method for factory farming according to claim 2, characterized in that: In step S3-1, the layering interval Δz of the rectangular aquaculture pond (1), that is, the height of the depth layer (1A), ranges from 0.8 meters to 2.5 meters. In step S3-2, the edge size of the horizontal grid (1B) ranges from 8 meters to 12 meters.

4. The three-dimensional distribution monitoring method for factory farming according to claim 1, characterized in that: In step S1, the sonar scanning method is as follows: the sonar is scanned along the length direction (D) of the rectangular aquaculture pond (1). L ) moves horizontally, and the sonar moves along the length direction (D) L A 360° scan is performed in the vertical plane of the sonar, and the 360° scan range of the sonar covers the rectangular aquaculture pond (1) perpendicular to the length direction (D). L The cross-section of ) In step S2, the three-dimensional point cloud P of the sound wave intensity is subjected to multi-level filtering processing, which includes at least the removal of measurement points generated by the strong reflected echo interference of the pool wall of the rectangular aquaculture pond (1) on the sonar scan.

5. The three-dimensional distribution monitoring method for factory farming according to claim 4, characterized in that: The multi-level filtering process described in step S2 includes: Filtering Strategy 1: Intensity Dual Threshold Filtering: Eliminating sound wave intensity I i Less than the preset lower limit of strength I min Or greater than the preset intensity upper limit value I max The measurement points, among which, the preset lower limit value I min The value range is 50dB to 70dB, with a preset upper limit of intensity I. max The value range is from 90dB to 100dB; Filtering Strategy 2: Spatial Range Filtering: Eliminating 3D spatial coordinates (x... i , y i , z i ) Measurement points extending beyond the boundary of the rectangular aquaculture pond (1), wherein the boundary of the rectangular aquaculture pond (1) includes: the virtual top surface (1-1) of the aquaculture pond, the bottom surface (1-2) of the aquaculture pond, and the length direction (D) L The two aquaculture pond walls (1-3) on the surface, in the width direction (D) W The two aquaculture pond walls (1-4) on the ) are respectively recessed inward by a preset boundary margin δ to form two virtual boundary end faces (1-4a). The preset boundary margin δ ranges from 0.4 meters to 0.6 meters. Filtration Strategy 3: Strong Echo Filtration from Pool Walls For sound wave intensity I i Greater than the preset suspicious echo intensity threshold I th If the distance between the measurement point and any one of the walls of the rectangular aquaculture pond (1) is less than the preset echo distance threshold δ, then... wall If the measurement point is not found, then the measurement point is discarded; wherein, the preset suspicious echo intensity threshold I th The value range is 80dB to 90dB, and the preset echo distance threshold δ wall The value ranges from 0.5 meters to 0.8 meters; For the measurement points in the four corner reflection areas (1a) of the rectangular aquaculture pond (1), if its sound wave intensity I i If the value is greater than 80dB, the measurement point is discarded; wherein, the corner reflection area (1a) is the area within 1 meter of the corner point of the bottom surface (1-2) of the aquaculture pond.

6. The three-dimensional distribution monitoring method for factory farming according to claim 5, characterized in that: The preset intensity lower limit value I of the filtering strategy one described in step S2 min and preset intensity upper limit value I max The preset boundary margin δ for filtering strategy two is set to 64dB and 95dB respectively; the preset suspicious echo intensity threshold I for filtering strategy three is set to 0.5 meters. th and preset echo distance threshold δ wall The values ​​were taken as 85dB and 0.6 meters, respectively.

7. The three-dimensional distribution monitoring method for factory farming according to claim 5, characterized in that: The multi-level filtering process described in step S2 further includes: Filtering strategy four: Statistical outlier filtering, including: First, the measurement points remaining after filtering by filtering strategies one to three are denoted as initial filter measurement points. Then, the distance distribution from all initial filter measurement points to the sonar and the sound wave intensity distribution of all initial filter measurement points are calculated to obtain the distance P98 percentile and the sound wave intensity P98 percentile, respectively. Finally, the initial filter measurement points whose distance to the sonar is greater than the distance to the P98 percentile are eliminated, as are the initial filter measurement points whose sound wave intensity is greater than the P98 percentile of the sound wave intensity.

8. The three-dimensional distribution monitoring method for factory farming according to claim 5, characterized in that: The multi-level filtering process described in step S2 further includes: Filtering strategy five, voxel downsampling, includes: First, for the remaining measurement points after filtering by filtering strategies one to three, multiple voxels are used for downsampling, wherein the diameter of the voxels ranges from 5 cm to 15 cm. Then, for each voxel, its representative point is calculated ( , , , ): , , , In the formula, This represents the sum of the products of the sound wave intensity and the x-coordinate of all measurement points contained in the voxel. This represents the sum of the products of the acoustic wave intensity and the y-coordinate of all measurement points contained within a voxel. This represents the sum of the products of the acoustic wave intensity and the z-coordinate of all measurement points contained in the voxel. This represents the sum of the sound wave intensities of all remaining measurement points after filtering by filtering strategies one through three. Finally, the representative point of each voxel ( , , , ) is used as one said effective measurement point, so as to obtain said three-dimensional point cloud P of effective sound wave intensity v ={(x j , y j , z j , I j )|j=1,2,…,N v}.

9. A feeding management method for factory farming, characterized in that, include: Step 1: Calculate the three-dimensional distribution characteristics of the fish population in the rectangular aquaculture pond (1) according to the preset monitoring time or when a monitoring instruction is received, based on the three-dimensional distribution monitoring method described in any one of claims 1 to 8. Step 2: When it is necessary to feed the fish in the rectangular aquaculture pond (1), if the depth intensity concentration detected most recently meets the DCI standard... I If the criterion is ≥0.6, then stratified feeding will be implemented, specifically as follows: First, a feed of a corresponding density is configured for each depth layer (1A) so that each feed can be suspended in the corresponding depth layer (1A) after being put into the rectangular breeding pond (1); Then, the first stage of feeding is carried out: 30% of the total feed is fed into the rectangular aquaculture pond (1), and in the first stage, the feeding ratio of various feeds is inversely proportional to the sound intensity density ρ of each corresponding layer depth (1A). zk The proportion during the most recent monitoring period; Finally, after the first stage of feeding, the depth intensity set (DCI) was continuously monitored through step one. I until DCI is met I When <0.4, the second stage of feeding is carried out: 70% of the total feed is fed into the rectangular aquaculture pond (1), and in the second stage, the feeding ratio of various feeds is proportional to the sound intensity density ρ of each corresponding layer depth (1A). zk The proportion during this monitoring.

10. The feeding management method for factory farming according to claim 9, characterized in that: The feeding management method further includes: Step 3: If the monitored edge intensity aggregation index satisfies ECI I If the judgment condition is ≥1.8, then a prompt will be given that there may be an environmental anomaly in the rectangular aquaculture pond (1); Step 4: If the monitored stratification intensity index satisfies LSI I If the threshold is >0.3 and continues to increase across multiple monitoring tests, a warning will be given that there may be an unsuitable underlying environment.

Images