Method for evaluating fish biomass in a recirculating aquaculture tank

Abstract

This invention discloses a method for assessing fish biomass in recirculating aquaculture ponds. First, sonar images corresponding to each rotational scan of the recirculating aquaculture pond are acquired through steps S1 and S2. The spatial three-dimensional coordinates of each identified farmed fish in each sonar image are extracted based on image recognition. Then, step S3 filters out noise points caused by repeated identification of the farmed fish coordinates. The number of fish corresponding to each sonar image is counted based on the actual coordinates of the farmed fish obtained by point cloud merging. Finally, step S4 obtains the effective area S of the pond wall. i With sampling depth h i The linear correlation curve S i (h i ), and fish density ρ i With sampling depth h i The nonlinear correlation curve ρ i (h i Finally, in step S5, the two curves are integrated by interval, the number of fish in each interval is calculated in layers, and the sum is obtained to obtain the biomass assessment value P of the fish in the recirculating aquaculture pond. This invention can improve the measurement accuracy of fish biomass in recirculating aquaculture ponds.

Description

TECHNICAL FIELD

[0001] The present application relates to the evaluation measurement of aquaculture biomass, in particular to a method for evaluating the biomass of fish population in a recirculating aquaculture pond. BACKGROUND

[0002] In modern aquaculture, the closed system can artificially regulate the water environment indicators in the system, and provide a suitable living environment and space for the cultured organisms. However, how to efficiently and accurately monitor the number of fish population in the recirculating aquaculture pond is still a key problem in the management of aquaculture. Real-time monitoring of the number of fish population not only directly affects the amount of feed feeding, but also promotes the rapid and healthy growth of fish population. The traditional manual observation or monitoring method is often unable to provide accurate fish population information due to the limitations of water, light conditions and labor costs, which affects the evaluation of aquaculture yield and the optimization of feeding strategy. At present, the biomass statistics method of cultured fish population mainly relies on optical cameras and acoustic sonar technology. As an advanced underwater detection technology, sonar technology can penetrate water and provide real-time fish population distribution information, and is widely used in aquaculture monitoring. However, the existing sonar monitoring technology is mostly limited to two-dimensional distribution identification, and cannot accurately estimate the overall number of fish population in the recirculating aquaculture pond, especially in complex environments where the fish population density distribution is uneven. The data of cultured fish population obtained by sonar contains many noise points, and the measurement error is large. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a method for evaluating the biomass of fish population in a recirculating aquaculture pond.

[0004] To solve the above technical problems, the technical scheme adopted by the present application is as follows:

[0005] A method for evaluating the biomass of fish population in a recirculating aquaculture pond, characterized in that it comprises:

[0006] Step S1, referring to Figure 2 The sonar is arranged at the center of the recirculating aquaculture pond, and is lowered from the water surface to the bottom of the recirculating aquaculture pond or raised from the bottom of the recirculating aquaculture pond to the water surface at a fixed speed v in the vertical direction, while the sonar rotates and scans in the horizontal direction at a constant angular velocity ω. Thus, the vertical movement of the sonar is synchronized with the horizontal rotation and scanning, and the entire recirculating aquaculture pond is scanned to obtain complete sonar image data of the aquaculture pond. The fixed speed v at which the sonar is raised and lowered needs to be moderate, that is, the overlap of each scan should not be too large, and there should be no large area that is not scanned.

[0007] Step S2, referring to Figure 3From the complete sonar image data of the aquaculture pond, based on the sonar completing one full rotation scan, the sonar image corresponding to each full rotation scan is extracted; wherein, the complete sonar image data of the aquaculture pond can be saved as a video format, and the sonar image is extracted based on the time T required for one full rotation scan of the sonar, or other parameters can be used as the basis for extraction.

[0008] And see also Figure 4 Each extracted sonar image is identified by an image recognition model to obtain the spatial three-dimensional coordinates of each identified farmed fish in the sonar image, denoted as the coordinate point P(x,y,h) of the identified farmed fish. Here, x and y represent the horizontal horizontal coordinate and the horizontal vertical coordinate of the identified farmed fish at the time of identification (i.e., the time when the sonar scan obtains the corresponding sonar image), respectively, and h represents the sampling depth of the identified farmed fish at the time of identification, which can be calculated based on the uniform movement speed of the sonar and the angular velocity of the sonar to complete one rotation scan.

[0009] The conversion of pixel coordinates in sonar images to x, y coordinates in three-dimensional space can be achieved using existing conversion methods.

[0010] Step S3, see Figure 5 The noise points generated by repeated identification in the identified farmed fish coordinate points obtained in filtering step S2 include:

[0011] Step S3-1: Calculate the coordinates P of any two identified farmed fish. i (x q ,y q ,h q ) and P j (x j ,y j ,h j The Euclidean distance D between them:

[0012]

[0013] Step S3-2: Divide the identified farmed fish coordinate points obtained in step S2 into multiple point clouds, and each point cloud meets the following requirements: the Euclidean distance D between each identified farmed fish coordinate point in the point cloud and another identified farmed fish coordinate point in the same point cloud satisfies the condition D<△h.

[0014] △h=T·v;

[0015] In the formula, Δh is the depth of a single rotation in the vertical direction corresponding to the completion of one rotational scan by the sonar in step S1; T is the time corresponding to each rotational scan of the sonar.

[0016] Step S3-3, merging each point cloud into one actual coordinate point of the farmed fish, and the merging manner is: calculating the average value of the horizontal plane horizontal coordinate x, the average value of the horizontal plane vertical coordinate y, and the average value of the sampling depth h of all the identified coordinate points of the farmed fish in the point cloud, and taking them as the horizontal plane horizontal coordinate, the horizontal plane vertical coordinate and the sampling depth of the actual coordinate point of the farmed fish corresponding to the point cloud;

[0017] Step S3-4, referring to Figure 6 , according to the sampling depth of the actual coordinate point of the farmed fish, the number of the actual coordinate points of the farmed fish contained in each sonar image is counted, which is recorded as the fish population number of the sonar image; wherein the fish population number and the sampling depth of the sonar image corresponding to the i-th rotation scanning are recorded as N i and h i , respectively, h i+1 -h i =△h, i is a positive integer, and 1≤i≤k, k is the total number of rotation scanning in the scanning process of the sonar in step S1, and △h is the single-circle depth corresponding to the vertical direction movement of the sonar in step S1 for completing one rotation scanning;

[0018] Step S4, obtaining linear correlation curve S i (h i ) and nonlinear correlation curve p i (h i ), and the specific steps include:

[0019] Step S4-1, referring to Figure 7 , according to the following formula, the effective area of the pool wall of the recirculating aquaculture pond corresponding to the part of the pool wall in the time period of the i-th rotation scanning is taken to approximately represent the irregular shape of the pool wall of the recirculating aquaculture pond and / or the deformation change in the water body when the sonar is in each rotation scanning:

[0020]

[0021] In the formula, S i represents the effective area of the pool wall of the recirculating aquaculture pond corresponding to the i-th rotation scanning, M i represents the total number of pixel points of the internal area of the pool wall of the recirculating aquaculture pond in the sonar image corresponding to the i-th rotation scanning, which is the square area in Figure 6 ; T represents the total number of pixel points of the sonar image corresponding to the i-th rotation scanning; S represents the rotation scanning area of the sonar, which is determined by the scanning radius r of the sonar and calculated according to πr 2 ;

[0022] Step S4-2, referring to Figure 8 , based on the sampling depth h i and the effective area Si , the effective area of the pool wall S i is obtained by linear fitting i and the linear correlation curve S i (h i ) of the sampling depth h i ;

[0023] Step S4-3, the fish population quantity obtained in step S3-4 is converted into fish population density according to the following formula, so as to obtain a fish population density data point based on each sonar image, and the fish population density data point of the sonar image corresponding to the i-th rotating scanning is recorded as (h i , p i ):

[0024]

[0025] In the formula, p i represents the fish population density of the sonar image corresponding to the i-th rotating scanning;

[0026] And, for the adjacent two fish population density data points (h i+1 , p i ) and (h i , p i+1 ) corresponding to each sampling depth interval [h i+1 , h i ], interpolation is performed by using the same interpolation method, so as to obtain a batch of interpolated fish population density data points corresponding to each sampling depth interval;

[0027] Therefore, based on all the fish population density data points and the interpolated fish population density data points, the nonlinear correlation curve p i (h i ) of the sampling depth h i and the fish population density p i is fitted;

[0028] Step S5, the fish population biomass evaluation value P in the recirculating aquaculture pool is calculated by integral accumulation according to the following formula:

[0029]

[0030] Therefore, the sonar corresponding to each rotating scanning of the recirculating aquaculture pool is obtained by steps S1 and S2, and the spatial three-dimensional coordinates of each recognized farmed fish in each sonar image are extracted based on image recognition, then the noise points in the recognized farmed fish coordinate points due to repeated recognition are filtered by step S3, the fish population quantity corresponding to each sonar image is counted according to the actual coordinate points of the farmed fish obtained by point cloud merging, then the effective area of the pool wall S i and the sampling depth h i is obtained by linear fittingi (h i ), and fish population density p i with the nonlinear correlation curve p i (h i (h i ), and finally the two curves are integrated by step S5 to calculate the fish population in each interval and accumulate to obtain the fish biomass evaluation value P in the recirculating aquaculture pond. Therefore, the noise points generated by repeated identification in the identified breeding fish coordinate points are filtered to adapt to the complex environment of uneven fish density distribution in the recirculating aquaculture pond, and the non-uniformity of fish density and the area of the recirculating aquaculture pond wall with depth is fully considered, which can improve the measurement accuracy of the fish biomass in the recirculating aquaculture pond, and has the advantages of high accuracy, efficient and fast and convenient implementation.

[0031] In step S3 of the present application, the identified breeding fish coordinate points with close distance are regarded as the echo reflection of the same fish according to the spatial position relationship, and the actual coordinate points of the breeding fish are merged, which can adapt to the situation of rapid acceleration or abnormal movement of the fish population in the recirculating aquaculture pond, and avoid the problem of existing technology using clustering to realize merging, that is, when the fish population moves too fast or abnormally, the local area may not meet the minimum point parameter required for clustering, resulting in isolated clusters, which makes the data too high.

[0032] Preferably, in step S1, the fixed speed v is calculated according to the following formula:

[0033]

[0034] In the formula, H is the total depth of the recirculating aquaculture pond, T is the time required for the sonar to rotate and scan a circle at the angular velocity, and L is the scannable length of the sonar to the recirculating aquaculture pond, as shown in Figure 2 The scannable length L is the length of the sonar scanning range projected onto the recirculating aquaculture pond, which is calculated according to the opening angle a of the sonar.

[0035] Therefore, it can be ensured that the water area scanned by the sonar when moving in the vertical direction is as little as possible. At the same time, it can avoid the situation of not being scanned.

[0036] Preferably, in step S1, the sonar is driven by a motor-driven mechanical lifting device or by manpower to move in the vertical direction; and the sonar is an omnidirectional sonar, which can rotate and scan 360° in the horizontal direction by driving the scanning beam.

[0037] Preferably, in step S2, the image recognition model uses a YOLOv8s model, and the training method is as follows:

[0038] Step S2-1, 500 sonar images are randomly extracted from existing sonar images, divided into a training set and a test set according to a 7:3 ratio, and the sonar images are labeled for fish position by using YoloLabel software;

[0039] Step S2-2, based on ultralytics, the labeled sonar images are used for model training, and the training parameters are: 300 training rounds, an initial learning rate of 0.001, and a picture size of 800x600, and finally the image recognition model is obtained, so that the fish position in the sonar image can be recognized by the image recognition model, and the corresponding fish quantity is obtained by counting the fish position. Figure 4

[0040] Therefore, in step S2, the YOLOv8s model is used to efficiently and accurately count the fish quantity, which greatly improves the data acquisition efficiency and accuracy compared with manual observation or traditional two-dimensional sonar analysis.

[0041] Preferably, in step S4-2, the linear fitting uses a linear regression analysis, and the regression model is optimized by minimizing the mean square error.

[0042] The formula is:

[0043]

[0044] Where m is the sample size, is the predicted value of the model, y i is the true value.

[0045] Preferably, in step S4-3, the interpolation method uses cubic spline interpolation to avoid overfitting and underfitting, and obtain the optimal fitting effect of the nonlinear correlation curve ρ i (h i ).

[0046] That is, a cubic polynomial ρ i (h i ) is constructed for each sampling depth interval [h i , h i+1 ], and the cubic polynomial function ρ i (h i ) is used to approximate the relationship between the data points to obtain the final nonlinear correlation curve ρ i (h i ). Each segment of the cubic polynomial has the following form:

[0047] ρ i (h i ) = a i +b i (h-h i ​)+c i (h-h i ) 2 +d i (h-h i ) 3

[0048] wherein a i ,b i ,c i ,d i are pending coefficients.

[0049] Compared with the prior art, the present application has the following beneficial effects:

[0050] Firstly, the present application firstly obtains the sonar image corresponding to each rotation scanning of the circular water breeding pond by the steps S1 and S2, and extracts the spatial three-dimensional coordinates of each identified breeding fish in each sonar image based on image recognition, and then filters the noise points in the identified breeding fish coordinate points caused by repeated identification by the step S3, and counts the fish population corresponding to each sonar image according to the actual coordinate points of the breeding fish obtained by point cloud merging, and then obtains the effective area S i of the pool wall by the step S4, and the linear correlation curve S i (h i ) of the sampling depth h i , and the nonlinear correlation curve p i (h i ) of the fish population density p i and the sampling depth h i , and finally integrates the two curves by the step S5, calculates the fish population in each interval and accumulates to obtain the fish population biomass evaluation value P in the circular water breeding pond, so that the present application filters the noise points in the identified breeding fish coordinate points caused by repeated identification to adapt to the complex environment of uneven fish population density distribution in the circular water breeding pond, and fully considers the non-uniformity of the fish population density and the pool wall area of the circular water breeding pond with the depth, which can improve the measurement accuracy of the fish population biomass in the circular water breeding pond, and has the advantages of high accuracy, efficient and convenient implementation.

[0051] In the present application, the step S3 is based on the spatial position relationship to regard the identified breeding fish coordinate points with close distance as the echo reflection of the same fish, so as to merge the actual coordinate points of the breeding fish, which can adapt to the situation of rapid acceleration or abnormal movement of the fish population in the circular water breeding pond, and avoid the problem of the prior art using clustering cluster to realize merging, that is, when the fish population moves too fast or abnormally, the local area may not meet the minimum point parameter required by clustering, resulting in isolated clusters, so that the data is high.

[0052] Secondly, the fixed speed v of driving the sonar to move up and down in the vertical direction in step S1 is calculated according to a formula based on the total depth H, the time T and the scannable length L, which can ensure that the water body scanned by the sonar when moving in the vertical direction is as little as possible to be repeated, and at the same time, the situation of not being scanned is avoided.

[0053] Thirdly, in step S2, the YOLOv8s model is used to efficiently and accurately count the number of fish groups, which greatly improves the data acquisition efficiency and accuracy compared with artificial observation or traditional two-dimensional sonar analysis.

[0054] Fourthly, in step S4-3, the cubic spline interpolation is used to avoid overfitting and underfitting, and the optimal fitting effect of the nonlinear correlation curve p i (h i ) is obtained. BRIEF DESCRIPTION OF DRAWINGS

[0055] The present application will be further described in detail below in combination with the drawings and specific embodiments:

[0056] Figure 1 is the flow chart of the present application;

[0057] Figure 2 is the schematic diagram of the sonar scanning in the water body in step S1 of the present application;

[0058] Figure 3 is the sonar image corresponding to the i-th rotation scanning in step S2 of the present application;

[0059] Figure 4 is the image data obtained by the image recognition model recognizing the sonar image in step S2 of the present application;

[0060] Figure 5 is the schematic diagram of the filtered noise points in step S3 of the present application;

[0061] Figure 6 is the relationship curve diagram of the sampling depth h i and the number of fish groups N i obtained in step S3-4 of the present application;

[0062] Figure 7 is the schematic diagram of the total sum M i of the pixel points of the internal area of the pool wall of the recirculating aquaculture pond in step S4-1 of the present application;

[0063] Figure 8 is the relationship curve diagram of the sampling depth h i , the effective area S i of the pool wall and the linear correlation curve S i (hi ) is a schematic view. DETAILED DESCRIPTION

[0064] The application will be described in greater detail by way of reference only to the examples and accompanying drawings in which:

[0065] As shown in Figure 1 , the application discloses a method for evaluating fish biomass in a recirculating aquaculture pond, comprising:

[0066] Step S1, referring to Figure 2 , the recirculating aquaculture pond is scanned by a sonar in the following manner: the sonar is arranged at the center of the recirculating aquaculture pond and is lowered from the water surface to the bottom of the recirculating aquaculture pond or raised from the bottom of the recirculating aquaculture pond to the water surface at a constant speed v in the vertical direction, while the sonar is rotated and scanned in the horizontal direction at a constant angular velocity ω; thus, the vertical movement of the sonar is synchronized with the horizontal rotation and scanning, and the recirculating aquaculture pond is scanned as a whole to obtain complete sonar image data of the recirculating aquaculture pond; wherein the constant speed v at which the sonar is raised and lowered needs to be moderate, that is, the overlap of each scan should not be too large, and a large area that is not scanned should not occur.

[0067] Preferably, in the step S1, the constant speed v is calculated according to the following formula:

[0068]

[0069] In the formula, H is the total depth of the recirculating aquaculture pond, T is the time required for the sonar to rotate and scan one circle at the angular velocity, and L is the scannable length of the recirculating aquaculture pond by the sonar, as shown in Figure 2 The scannable length L is the length of the scanning range of the sonar projected onto the recirculating aquaculture pond, which is calculated according to the opening angle α of the sonar.

[0070] Thus, the water area scanned by the sonar when moving in the vertical direction can be ensured to be as little repeated as possible, while avoiding the occurrence of unscanned areas.

[0071] Preferably, in the step S1, the sonar is moved in the vertical direction by a mechanical lifting device driven by a motor or by manpower; and the sonar is an omnidirectional sonar, which can be rotated and scanned in the horizontal direction by 360° through driving the scanning beam.

[0072] Step S2, referring to Figure 3From the complete sonar image data of the culture pond, the sonar image corresponding to each rotation scanning is extracted based on each completion of a rotation scanning of the sonar; wherein the complete sonar image data of the culture pond can be saved in a video format, and the sonar image is extracted based on the time T required for one rotation scanning of the sonar, or can be extracted based on other parameters.

[0073] And referring to Figure 4 , each sonar image extracted is identified by an image recognition model to obtain the spatial three-dimensional coordinates of each identified cultured fish in the sonar image, denoted as the identified cultured fish coordinate point P(x, y, h), wherein x and y respectively represent the horizontal plane horizontal coordinate and horizontal plane vertical coordinate of the identified cultured fish at the identified time (i.e. the time when the corresponding sonar image is scanned by the sonar), and h represents the sampling depth of the identified cultured fish at the identified time, which can be calculated according to the speed of uniform movement of the sonar and the angular velocity of one rotation scanning of the sonar.

[0074] Wherein the conversion of the pixel coordinates in the sonar image to x, y in the spatial three-dimensional coordinates can adopt the conversion method in the prior art.

[0075] Preferably: in the step S2, the image recognition model adopts a YOLOv8s model, and the training method is as follows:

[0076] Step S2-1, 500 sonar images are randomly extracted from the existing sonar images, divided into a training set and a test set according to a ratio of 7:3, and the fish school position of the sonar images is labeled by using YoloLabel software;

[0077] Step S2-2, based on ultralytics, the labeled sonar images are used for model training, and the training parameters are: 300 training rounds, an initial learning rate of 0.001, and an image size of 800x600, and finally the image recognition model is obtained, which can be used to identify the fish school position in the sonar image, referring to Figure 4 , the fish school position is counted to obtain the corresponding fish school quantity.

[0078] Thus, in the step S2, the YOLOv8s model is adopted, which can efficiently and accurately count the fish school quantity, and compared with artificial observation or traditional two-dimensional sonar analysis, the data acquisition efficiency and accuracy are greatly improved.

[0079] Step S3, referring to Figure 5 , noise points generated due to repeated identification are filtered from the identified cultured fish coordinate points obtained in the step S2, including:

[0080] Step S3-1, the distance between any two identified cultured fish coordinate points P i (xq , y q , h q ) and P j (x j , y j , h j ) between the Euclidean distance D:

[0081]

[0082] Step S3-2, the identified aquaculture fish coordinate points obtained in step S2 are divided into a plurality of point clouds, and each point cloud meets the following requirements: the Euclidean distance D between each identified aquaculture fish coordinate point in the point cloud and another identified aquaculture fish coordinate point in the point cloud satisfies the condition of D < △h;

[0083] △h = T · v;

[0084] In the formula, △h is the single-circle depth corresponding to the movement in the vertical direction of the sonar in completing a circle of rotational scanning in step S1; T is the time corresponding to each circle of rotational scanning of the sonar;

[0085] Step S3-3, each point cloud is combined into an actual aquaculture fish coordinate point, and the combination manner is: the average value of the horizontal plane abscissa x, the average value of the horizontal plane ordinate y, and the average value of the sampling depth h of all the identified aquaculture fish coordinate points in the point cloud are calculated respectively, and are taken as the horizontal plane abscissa, the horizontal plane ordinate, and the sampling depth of the actual aquaculture fish coordinate point corresponding to the point cloud;

[0086] Step S3-4, referring to Figure 6 , the number of actual aquaculture fish coordinate points contained in each sonar image is counted according to the sampling depth of the actual aquaculture fish coordinate point, and is recorded as the fish population number of the sonar image; wherein the fish population number and the sampling depth of the sonar image corresponding to the i-th circle of rotational scanning are recorded as N i and h i , respectively, h i+1 -h i = △h, i is a positive integer, and 1 ≤ i ≤ k, k is the total number of rotational circles in the scanning process of the sonar in step S1, and △h is the single-circle depth corresponding to the movement in the vertical direction of the sonar in completing a circle of rotational scanning in step S1;

[0087] Step S4, linear correlation curve S i (h i ) and nonlinear correlation curve p i (h i ) are obtained, and the specific steps include:

[0088] Step S4-1, referring to Figure 7The following formula is used to approximate the irregular shape and / or deformation of the recirculating aquaculture pond wall during each sonar rotation scan, corresponding to the area scanned in that rotation scan period:

[0089]

[0090] In the formula, S i M represents the effective area of ​​the recirculating aquaculture pond wall corresponding to the i-th rotation scan. i This represents the total number of pixels in the internal region of the recirculating aquaculture pond wall in the sonar image corresponding to the i-th rotational scan, see... Figure 6 The square region in the diagram; T represents the total number of pixels in the sonar image corresponding to the i-th rotational scan; S represents the rotational scanning area of ​​the sonar, which is determined by the scanning radius r of the sonar, according to πr 2 calculate;

[0091] Step S4-2, see Figure 8 Based on the sampling depth h corresponding to each rotation scan i and the effective area S of the pool wall i The effective area S of the pool wall was obtained through linear fitting. i With sampling depth h i The linear correlation curve S i (h i );

[0092] Preferably, in step S4-2, the linear fitting adopts univariate linear regression analysis, and the regression model is optimized by minimizing the mean square error.

[0093] The formula is:

[0094]

[0095] Where m is the number of samples. It is the model's predicted value, y i It is the actual value.

[0096] Step S4-3: Convert the fish population obtained in step S3-4 into fish density using the following formula, so as to obtain a fish density data point based on each sonar image, and record the fish density data point of the sonar image corresponding to the i-th rotation scan as (h i ,ρ i ):

[0097]

[0098] In the formula, ρ i This represents the fish density in the sonar image corresponding to the i-th rotational scan.

[0099] And, for each sampling depth interval [h i ,h i+1 ], the adjacent two fish population density data points (h i ,ρ i ) and (h i+1 ,ρ i+1 ) are interpolated by the same interpolation method to obtain a batch of interpolated fish population density data points corresponding to each sampling depth interval;

[0100] Thus, based on all fish population density data points and interpolated fish population density data points, a nonlinear correlation curve ρ i (h i ) of fish population density ρ i and sampling depth h i is fitted.

[0101] Preferably, in the step S4-3, the interpolation method adopts cubic spline interpolation to avoid overfitting and underfitting and obtain optimal fitting effect of the nonlinear correlation curve ρ i (h i ).

[0102] That is, a cubic polynomial ρ i (h i ) is constructed for each sampling depth interval [h i ,h i+1 ], and the cubic polynomial function ρ i (h i ) of the interval segment is used to approximate the relationship between the data points to obtain the final nonlinear correlation curve ρ i (h i ). Each cubic polynomial of the segment has the following form:

[0103] ρ i (h i )=a i +b i (h-h i )+c i (h-h i ) 2 +d i (h-h i ) 3

[0104] where a i , b i , c i , and d i are undetermined coefficients.

[0105] Step S5: Calculate the biomass assessment value P of the fish population in the recirculating aquaculture pond by integrating and summing the results using the following formula:

[0106]

[0107] Therefore, this invention first obtains sonar images corresponding to each rotational scan of the recirculating aquaculture pond through steps S1 and S2, and extracts the spatial three-dimensional coordinates of each identified farmed fish in each sonar image based on image recognition. Then, in step S3, it filters out noise points caused by repeated identification in the coordinate points of the identified farmed fish, and counts the number of fish corresponding to each sonar image based on the actual coordinate points of the farmed fish obtained by point cloud merging. Finally, in step S4, it obtains the effective area S of the pond wall. i With sampling depth h i The linear correlation curve S i (h i ), and fish density ρ i With sampling depth h i The nonlinear correlation curve ρ i (h i Finally, in step S5, the two curves are integrated by interval, the number of fish in each interval is calculated in layers, and the sum is accumulated to obtain the biomass assessment value P of the fish in the recirculating aquaculture pond. Therefore, this invention filters out noise points caused by repeated identification in the coordinate points of the identified farmed fish, so as to adapt to the complex environment of uneven fish density distribution in the recirculating aquaculture pond. It also fully considers the non-uniformity of fish density and the pond wall area of ​​the recirculating aquaculture pond with depth, which can improve the measurement accuracy of fish biomass in the recirculating aquaculture pond. It has the advantages of high accuracy, high efficiency, speed and convenience in implementation.

[0108] In this invention, step S3 is based on spatial positional relationships to treat the coordinates of the identified farmed fish that are close to each other as echo reflections of the same fish, thereby merging the actual coordinates of the farmed fish. This can adapt to situations where the fish in the recirculating aquaculture pond accelerates rapidly or moves abnormally, avoiding the problems of existing technologies that use clustering to merge the fish. That is, when the fish move too fast or abnormally, local areas may not meet the minimum number of points required for clustering, resulting in the formation of isolated clusters and thus higher data.

[0109] 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 method for assessing fish biomass in a recirculating aquaculture pond, characterized in that, include: Step S1: Scan the recirculating aquaculture pond with sonar in the following manner: The sonar is positioned at the center of the recirculating aquaculture pond and descends vertically at a fixed speed v from the water surface to the bottom of the recirculating aquaculture pond or rises from the bottom of the recirculating aquaculture pond to the water surface. At the same time, the sonar rotates horizontally at a constant angular velocity ω. Thus, complete sonar image data of the aquaculture pond is obtained. Step S2: From the complete sonar image data of the aquaculture pond, based on the sonar completing one rotational scan each time, extract the sonar image corresponding to each rotational scan. Furthermore, each extracted sonar image is identified using an image recognition model to obtain the spatial three-dimensional coordinates of each identified farmed fish in the sonar image, denoted as the coordinate point P(x,y,h) of the identified farmed fish, where x and y represent the horizontal horizontal coordinate and the horizontal vertical coordinate of the identified farmed fish at the time of identification, respectively, and h represents the sampling depth of the identified farmed fish at the time of identification. Step S3: Filter out noise points caused by repeated identification in the identified farmed fish coordinate points obtained in step S2, including: Step S3-1: Calculate the coordinates P of any two identified farmed fish. i (x q ,y q ,h q ) and P j (x j ,y j ,h j The Euclidean distance D between them: Step S3-2: Divide the identified farmed fish coordinate points obtained in step S2 into multiple point clouds, and each point cloud meets the following requirements: the Euclidean distance D between each identified farmed fish coordinate point in the point cloud and another identified farmed fish coordinate point in the same point cloud satisfies the condition D<△h. △h=T·v; In the formula, Δh is the depth of a single rotation in the vertical direction corresponding to the completion of one rotational scan by the sonar in step S1; T is the time corresponding to each rotational scan of the sonar. Step S3-3: Merge each point cloud into a single actual coordinate point of a farmed fish. The merging method is as follows: calculate the average value of the horizontal plane x-coordinate, the average value of the horizontal plane y-coordinate, and the average value of the sampling depth h of all identified farmed fish coordinate points in the point cloud, and use these values ​​as the horizontal plane x-coordinate, horizontal plane y-coordinate, and sampling depth of the corresponding actual coordinate point of the farmed fish in the point cloud. Step S3-4: Based on the sampling depth of the actual coordinates of the farmed fish, count the number of actual coordinates of the farmed fish contained in each sonar image, and record this as the number of fish in that sonar image; wherein, the number of fish and the sampling depth of the sonar image corresponding to the i-th rotation scan are respectively denoted as N. i and h i h i+1 -h i = △h, i is a positive integer, and 1≤i≤k, k is the total number of rotations of the sonar in the scanning process of step S1, and △h is the depth of a single rotation in the vertical direction corresponding to the completion of one rotation scan by the sonar in step S1; Step S4: Obtain the linear correlation curve S i (h i ) and nonlinear correlation curve ρ i (h i The specific steps include: Step S4-1: According to the following formula, during each rotational scan of the sonar, the effective area of ​​the pool wall corresponding to the part of the rotational scan during the time period of that rotational scan is: In the formula, S i M represents the effective area of ​​the recirculating aquaculture pond wall corresponding to the i-th rotation scan. i S represents the total number of pixels in the internal region of the recirculating aquaculture pond wall in the sonar image corresponding to the i-th rotational scan; T represents the total number of pixels in the sonar image corresponding to the i-th rotational scan; S T This represents the rotating scanning area of ​​the sonar; Step S4-2: Based on the sampling depth h corresponding to each rotation scan i and the effective area S of the pool wall i The effective area S of the pool wall was obtained through linear fitting. i With sampling depth h i The linear correlation curve S i (h i ); Step S4-3: Convert the fish population obtained in step S3-4 into fish density using the following formula, so as to obtain a fish density data point based on each sonar image, and record the fish density data point of the sonar image corresponding to the i-th rotation scan as (h i ,ρ i ): In the formula, ρ i This represents the fish density in the sonar image corresponding to the i-th rotational scan. Furthermore, for each sampling depth interval [h] i ,h i+1 The corresponding two adjacent fish density data points (h) i ,ρ i ) and (h i+1 ,ρ i+1 The same interpolation method is used to interpolate, so as to obtain a batch of interpolated fish density data points for each sampling depth interval; Therefore, based on all fish density data points and interpolated fish density data points, the fish density ρ is obtained by fitting. i With sampling depth h i The nonlinear correlation curve ρ i (h i ); Step S5: Calculate the biomass assessment value P of the fish population in the recirculating aquaculture pond according to the following formula:

2. The method for assessing fish biomass in recirculating aquaculture ponds according to claim 1, characterized in that: In step S1, the fixed speed v is calculated according to the following formula: In the formula, H is the total depth of the recirculating aquaculture pond, T is the time required for the sonar to rotate and scan one revolution at the stated angular velocity, and L is the scannable length of the recirculating aquaculture pond by the sonar.

3. The method for assessing fish biomass in recirculating aquaculture ponds according to claim 1, characterized in that: In step S1, the sonar moves vertically via a motor-driven mechanical lifting device or by manual driving; and the sonar is an omnidirectional sonar, capable of performing a 360° rotational scan in the horizontal direction by driving the scanning beam.

4. The method for assessing fish biomass in recirculating aquaculture ponds according to any one of claims 1 to 3, characterized in that: In step S2, the image recognition model uses the YOLOv8s model, and its training method is as follows: Step S2-1: Randomly select 500 sonar images from the existing images, divide them into training and testing sets in a 7:3 ratio, and use the YoloLabel software to label the fish schools in the sonar images. Step S2-2: Using the labeled sonar images, train the model based on Ultralytics. The training parameters are: 300 training rounds, initial learning rate of 0.001, and image size of 800×600. Finally, the image recognition model is obtained, which can identify the location of fish in the sonar image and count the number of fish at the location.

5. The method for assessing fish biomass in recirculating aquaculture ponds according to any one of claims 1 to 3, characterized in that: In step S4-2, the linear fitting adopts univariate linear regression analysis, and the regression model is optimized by minimizing the mean square error.

6. The method for assessing fish biomass in recirculating aquaculture ponds according to any one of claims 1 to 3, characterized in that: In step S4-3, cubic spline interpolation is used as the interpolation method.

Images