Fish school biomass evaluation method for circulating water culture pond

Abstract

The invention discloses a circulating water culture pond fish school biomass evaluation method, which comprises the following steps of: firstly, acquiring a sonar image corresponding to each circle of rotary scanning of a circulating water culture pond by sonar through steps S1 and S2, and extracting a spatial three-dimensional coordinate of each identified cultured fish in each sonar image based on image identification; filtering noise points generated due to repeated identification in the identified cultured fish coordinate points through the step S3, and counting the number of fish schools corresponding to each sonar image according to the actual cultured fish coordinate points obtained through point cloud merging, and then obtaining a linear correlation curve Si (hi) of the effective area Si of the pool wall and the sampling depth hi through the step S4, and finally, performing integration on the two curves according to intervals through the step S5, performing layered calculation on the number of the fish schools in each interval, and performing accumulation to obtain an estimated value P of the fish school biomass in the circulating water culture pond, so that the measurement accuracy of the fish school biomass in the circulating water culture pond can be improved.

Description

Technical Field

[0001] The invention relates to an evaluation and measurement of aquaculture biomass, in particular to an evaluation method for fish biomass in a circulating aquaculture pond. Background Art

[0002] In modern aquaculture, closed systems allow for the artificial manipulation of water environmental parameters within the system, providing a suitable living environment and space for aquacultured organisms, making them a crucial aquaculture method. However, efficient and accurate monitoring of fish populations within recirculating aquaculture ponds remains a key issue in aquaculture management. Real-time monitoring of fish populations not only directly impacts feed intake, but also effectively reduces water pollution and promotes rapid and healthy fish growth. Traditional manual observation or monitoring methods, limited by water and lighting conditions and labor costs, often fail to provide accurate information on fish populations, hindering aquaculture yield assessment and feeding strategy optimization. Currently, methods for counting fish biomass in farmed fish primarily rely on optical cameras and acoustic sonar technology. As an advanced underwater detection technology, sonar technology is widely used in aquaculture monitoring due to its ability to penetrate water and provide real-time information on fish distribution. However, existing sonar monitoring technology mostly remains at the two-dimensional distribution recognition stage and is unable to accurately estimate the overall number of fish in recirculating aquaculture ponds. Especially in complex environments such as uneven fish density distribution, the farmed fish data obtained by sonar contains many noise points and the measurement error is large. Summary of the Invention

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

[0004] To solve the above technical problems, the technical solutions adopted by the present invention are as follows:

[0005] A method for assessing fish biomass in a circulating aquaculture pond, comprising:

[0006] Step S1, see Figure 2 The sonar is used to scan the circulating aquaculture pond in the following manner: the sonar is arranged at the center of the circulating aquaculture pond and dives from the water surface to the bottom of the circulating aquaculture pond or rises from the bottom of the circulating aquaculture pond to the water surface at a fixed speed v in the vertical direction. At the same time, the sonar performs rotational scanning in the horizontal direction at a constant angular velocity ω; thereby, the vertical movement of the sonar and the horizontal rotation scanning are performed synchronously, and the circulating aquaculture pond is scanned as a whole to obtain complete sonar image data of the aquaculture pond; wherein, the fixed speed v of the sonar for rising and falling needs to be moderate, that is, the overlapping part of each scan should not be too large, and large areas should not be left unscanned.

[0007] Step S2, see Figure 3, from the complete sonar image data of the aquaculture pond, based on each time the sonar completes a rotation scan, extracting the sonar image corresponding to each rotation scan; wherein the complete sonar image data of the aquaculture pond can be saved in a video format, and the sonar image can be extracted based on the time T required for the sonar to complete one rotation scan, or based on other parameters;

[0008] Also, see Figure 4 , each extracted sonar image is recognized by the image recognition model, and the spatial three-dimensional coordinates of each identified farmed fish in the sonar image are obtained, which are recorded as the identified farmed fish coordinate point P(x, y, h), where x and y represent the horizontal plane abscissa and horizontal plane ordinate of the identified farmed fish at the time of identification (i.e., the time when the sonar scanning obtains the corresponding sonar image), and h represents the sampling depth of the identified farmed fish at the time of identification, which can be calculated based on the uniform speed of the sonar and the angular velocity of the sonar completing a rotation scan;

[0009] The pixel coordinates in the sonar image are converted into x and y in the three-dimensional spatial coordinates by using the conversion method in the prior art.

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

[0011] Step S3-1, calculate the coordinate points P of any two identified farmed fish i (x q ,y q ,h q ) and P j (x j ,y j ,h j ) between the Euclidean distance D:

[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 point cloud satisfies the condition D<△h;

[0014] △h=T·v;

[0015] Wherein, Δh is the single circle depth of the sonar in the vertical direction corresponding to one circle of rotation scanning in step S1; T is the time corresponding to each circle of rotation scanning of the sonar;

[0016] Step S3-3, merging each point cloud into an actual coordinate point of the farmed fish, by calculating the average horizontal coordinate x, the average horizontal coordinate y, and the average sampling depth h of all the identified farmed fish coordinate points in the point cloud, and using these as the horizontal coordinate x, horizontal coordinate y, and sampling depth of the actual coordinate point of the farmed fish corresponding to the point cloud;

[0017] Step S3-4, see Figure 6 According to the sampling depth of the actual coordinate points of farmed fish, the number of actual coordinate points of farmed fish contained in each sonar image is counted and recorded as the number of fish in the sonar image; the number of fish and the sampling depth of the sonar image corresponding to the i-th rotation scan are recorded as N respectively. 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 single circle depth of the sonar in the vertical direction corresponding to one rotation scan completed in step S1;

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

[0019] Step S4-1, see Figure 7 According to the following formula, when the sonar performs each rotation scan, the effective area of ​​the wall of the circulating aquaculture pond corresponding to the rotation scan during the rotation scan period is approximately expressed as the irregular shape of the wall of the circulating aquaculture pond and / or its deformation changes in the water body:

[0020]

[0021] Where S i M represents the effective area of ​​the wall of the circulating water aquaculture pond corresponding to the i-th rotation scan, i It represents the total number of pixels in the inner area of ​​the wall of the circulating water aquaculture pond in the sonar image corresponding to the i-th rotation scan, see Figure 6 The square area in the image; T represents the total number of pixels of the sonar image corresponding to the i-th rotation scan; S represents the rotation scanning area of ​​the sonar, which is determined by the scanning radius r of the sonar, according to πr 2 calculate;

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

[0023] Step S4-3: Convert the number of fish obtained in step S3-4 into fish density according to 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 ):

[0024]

[0025] Where, ρ i represents the fish density of the sonar image corresponding to the i-th rotation scan;

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

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

[0028] Step S5: According to the following formula, the fish biomass assessment value P in the circulating aquaculture pond is calculated by integral accumulation:

[0029]

[0030] Therefore, the present invention first obtains the sonar image corresponding to each rotation scan of the circulating water aquaculture pond by the sonar 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, the noise points generated by repeated identification in the coordinate points of the identified farmed fish are filtered out, and the number of fish corresponding to each sonar image is counted based on the actual coordinate points of the farmed fish obtained by merging the point clouds. Then, in step S4, the effective area S of the pool wall is obtained. i With sampling depth h i The linear correlation curve Si (h i ), and fish density ρ i With sampling depth h i The nonlinear correlation curve ρ i (h i ), and finally, step S5 integrates the two curves by interval, calculates the number of fish in each interval in layers, and accumulates the fish biomass assessment value P in the circulating aquaculture pond. Therefore, the present invention filters out noise points generated by repeated identification in the identified farmed fish coordinate points to adapt to the complex environment of uneven fish density distribution in the circulating aquaculture pond, and fully considers the non-uniformity of fish density and circulating aquaculture pond wall area with depth, which can improve the measurement accuracy of fish biomass in the circulating aquaculture pond, and has the advantages of high accuracy, high efficiency, rapidness and convenience.

[0031] Among them, since step S3 of the present invention regards the identified farmed fish coordinate points with a close distance as echo reflections of the same fish based on the spatial position relationship, and merges the actual coordinate points of the farmed fish, it can adapt to the situation of rapid acceleration or abnormal movement of fish schools in the circulating water aquaculture pond, and avoids the problem of using clustering to achieve merging in the prior art, namely: when the movement of the fish school is too fast or abnormal, the local area may not meet the minimum number of points required for clustering, resulting in the formation of isolated clusters, which makes the data biased high.

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

[0033]

[0034] Where H is the total depth of the circulating water 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 circulating water aquaculture pond by the sonar. Figure 2 The scannable length L is the length of the sonar's scanning range projected onto the circulating water aquaculture pond, which is calculated based on the sonar's opening angle α.

[0035] This ensures that the water area scanned by the sonar when it moves in the vertical direction is as non-repetitive as possible, while avoiding the situation where the sonar does not scan the water area.

[0036] Preferably, in step S1, the sonar is moved in the vertical direction by a motor-driven mechanical lifting device or by human power; and the sonar is an omnidirectional sonar, which can perform 360° rotation scanning in the horizontal direction by driving the scanning beam.

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

[0038] Step S2-1: Randomly select 500 sonar images from the existing dataset, divide them into a training set and a test set in a ratio of 7:3, and use YoloLabel software to label the fish positions on the sonar images.

[0039] Step S2-2: Using the labeled sonar images, the model is trained based on Ultralytics. The training parameters are: 300 training rounds, an initial learning rate of 0.001, and an image size of 800×600. Finally, the image recognition model is obtained, which can be used to identify the location of fish schools in sonar images. Figure 4 , count the fish positions and get the corresponding number of fish.

[0040] Therefore, in step S2, the present invention adopts the YOLOv8s model to efficiently and accurately count the number of fish schools, 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 adopts a univariate 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 number of samples, is the model's predicted value, y i is the true value.

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

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

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

[0048] where a i ,b i ,c i ,d i is the coefficient to be determined.

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

[0050] First, the present invention obtains the sonar image corresponding to each rotation scan of the circulating water aquaculture pond by the sonar 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, step S3 is used to filter the noise points generated by repeated identification in the coordinate points of the identified farmed fish, and the actual coordinate points of the farmed fish obtained by merging the point clouds are used to count the number of fish corresponding to each sonar image. Then, step S4 is used to obtain the effective area S of the pool 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 ), and finally, step S5 integrates the two curves by interval, calculates the number of fish in each interval in layers, and accumulates the fish biomass assessment value P in the circulating aquaculture pond. Therefore, the present invention filters out noise points generated by repeated identification in the identified farmed fish coordinate points to adapt to the complex environment of uneven fish density distribution in the circulating aquaculture pond, and fully considers the non-uniformity of fish density and circulating aquaculture pond wall area with depth, which can improve the measurement accuracy of fish biomass in the circulating aquaculture pond, and has the advantages of high accuracy, high efficiency, rapidness and convenience.

[0051] Among them, since step S3 of the present invention regards the identified farmed fish coordinate points with a close distance as echo reflections of the same fish based on the spatial position relationship, and merges the actual coordinate points of the farmed fish, it can adapt to the situation of rapid acceleration or abnormal movement of fish schools in the circulating water aquaculture pond, and avoids the problem of using clustering to achieve merging in the prior art, namely: when the movement of the fish school is too fast or abnormal, the local area may not meet the minimum number of points required for clustering, resulting in the formation of isolated clusters, which makes the data biased high.

[0052] Second, the fixed speed v for driving the sonar to move up and down in the vertical direction in step S1 of the present invention is calculated according to a formula based on the total depth H, time T, and scannable length L. This can ensure that the water body scanned by the sonar when moving in the vertical direction is as non-repetitive as possible, while avoiding the situation where the sonar does not scan.

[0053] Third, in step S2, the present invention uses the YOLOv8s model to efficiently and accurately count the number of fish schools, which greatly improves the data acquisition efficiency and accuracy compared to manual observation or traditional two-dimensional sonar analysis.

[0054] Fourth, in step S4-3, the present invention uses cubic spline interpolation to avoid overfitting and underfitting, and obtains the nonlinear correlation curve ρ i (h i ) of the optimal fitting effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

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

[0057] Figure 2 Schematic diagram of sonar scanning in water in step S1 of the present invention;

[0058] Figure 3 The sonar image corresponding to the i-th rotation scan in step S2 of the present invention;

[0059] Figure 4 This is a schematic diagram of image data obtained by the image recognition model in step S2 of the present invention when recognizing the sonar image;

[0060] Figure 5 Schematic diagram of noise points filtered in step S3 of the present invention;

[0061] Figure 6 The sampling depth h of all sonar images obtained in step S3-4 of the present invention is i and the number of fish N i The relationship curve diagram;

[0062] Figure 7 The total number of pixels M in the inner area of ​​the wall of the circulating aquaculture pond calibrated in step S4-1 of the present invention is i Schematic diagram of;

[0063] Figure 8 is the sampling depth h in step S4-2 of the present invention i and the effective area of ​​the pool wall S i And the fitted linear correlation curve S i (hi ) schematic diagram. DETAILED DESCRIPTION

[0064] The present invention is described in detail below in conjunction with the embodiments and the accompanying drawings to help those skilled in the art better understand the inventive concept of the present invention. However, the scope of protection of the claims of the present invention is not limited to the following embodiments. For those skilled in the art, all other embodiments obtained without creative work without departing from the inventive concept of the present invention shall fall within the scope of protection of the present invention.

[0065] like Figure 1 As shown, the present invention discloses a method for evaluating fish biomass in a circulating aquaculture pond, comprising:

[0066] Step S1, see Figure 2 The sonar is used to scan the circulating aquaculture pond in the following manner: the sonar is arranged at the center of the circulating aquaculture pond and dives from the water surface to the bottom of the circulating aquaculture pond or rises from the bottom of the circulating aquaculture pond to the water surface at a fixed speed v in the vertical direction. At the same time, the sonar performs rotational scanning in the horizontal direction at a constant angular velocity ω; thereby, the vertical movement of the sonar and the horizontal rotation scanning are performed synchronously, and the circulating aquaculture pond is scanned as a whole to obtain complete sonar image data of the aquaculture pond; wherein, the fixed speed v of the sonar for rising and falling needs to be moderate, that is, the overlapping part of each scan should not be too large, and large areas should not be left unscanned.

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

[0068]

[0069] Where H is the total depth of the circulating water 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 circulating water aquaculture pond by the sonar. Figure 2 The scannable length L is the length of the sonar's scanning range projected onto the circulating water aquaculture pond, which is calculated based on the sonar's opening angle α.

[0070] This ensures that the water area scanned by the sonar when it moves in the vertical direction is as non-repetitive as possible, while avoiding the situation where the sonar does not scan the water area.

[0071] Preferably, in step S1, the sonar is moved in the vertical direction by a motor-driven mechanical lifting device or by human power; and the sonar is an omnidirectional sonar, which can perform 360° rotation scanning in the horizontal direction by driving the scanning beam.

[0072] Step S2, see Figure 3, from the complete sonar image data of the aquaculture pond, based on each time the sonar completes a rotation scan, extracting the sonar image corresponding to each rotation scan; wherein the complete sonar image data of the aquaculture pond can be saved in a video format, and the sonar image can be extracted based on the time T required for the sonar to complete one rotation scan, or based on other parameters;

[0073] Also, see Figure 4 , each extracted sonar image is recognized by the image recognition model, and the spatial three-dimensional coordinates of each identified farmed fish in the sonar image are obtained, which are recorded as the identified farmed fish coordinate point P(x, y, h), where x and y represent the horizontal plane abscissa and horizontal plane ordinate of the identified farmed fish at the time of identification (i.e., the time when the sonar scanning obtains the corresponding sonar image), and h represents the sampling depth of the identified farmed fish at the time of identification, which can be calculated based on the uniform speed of the sonar and the angular velocity of the sonar completing a rotation scan;

[0074] The pixel coordinates in the sonar image are converted into x and y in the three-dimensional spatial coordinates by using the conversion method in the prior art.

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

[0076] Step S2-1: Randomly select 500 sonar images from the existing dataset, divide them into a training set and a test set in a ratio of 7:3, and use YoloLabel software to label the fish positions on the sonar images.

[0077] Step S2-2: Using the labeled sonar images, the model is trained based on Ultralytics. The training parameters are: 300 training rounds, an initial learning rate of 0.001, and an image size of 800×600. Finally, the image recognition model is obtained, which can be used to identify the location of fish schools in sonar images. Figure 4 , count the fish positions and get the corresponding number of fish.

[0078] Therefore, in step S2, the present invention adopts the YOLOv8s model to efficiently and accurately count the number of fish schools, which greatly improves the data acquisition efficiency and accuracy compared with manual observation or traditional two-dimensional sonar analysis.

[0079] Step S3, see Figure 5 , filtering the noise points generated by repeated identification in the identified farmed fish coordinate points obtained in step S2, including:

[0080] Step S3-1, calculate the coordinate points P of any two identified farmed fish 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: 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 point cloud satisfies the condition D<△h;

[0083] △h=T·v;

[0084] Wherein, Δh is the single circle depth of the sonar in the vertical direction corresponding to one circle of rotation scanning in step S1; T is the time corresponding to each circle of rotation scanning of the sonar;

[0085] Step S3-3, merging each point cloud into an actual coordinate point of the farmed fish, by calculating the average horizontal coordinate x, the average horizontal coordinate y, and the average sampling depth h of all the identified farmed fish coordinate points in the point cloud, and using these as the horizontal coordinate x, horizontal coordinate y, and sampling depth of the actual coordinate point of the farmed fish corresponding to the point cloud;

[0086] Step S3-4, see Figure 6 According to the sampling depth of the actual coordinate points of farmed fish, the number of actual coordinate points of farmed fish contained in each sonar image is counted and recorded as the number of fish in the sonar image; the number of fish and the sampling depth of the sonar image corresponding to the i-th rotation scan are recorded as N respectively. 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 single circle depth of the sonar in the vertical direction corresponding to one rotation scan completed in step S1;

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

[0088] Step S4-1, see Figure 7According to the following formula, when the sonar performs each rotation scan, the effective area of ​​the wall of the circulating aquaculture pond corresponding to the rotation scan during the rotation scan period is approximately expressed as the irregular shape of the wall of the circulating aquaculture pond and / or its deformation changes in the water body:

[0089]

[0090] Where S i M represents the effective area of ​​the wall of the circulating water aquaculture pond corresponding to the i-th rotation scan, i It represents the total number of pixels in the inner area of ​​the wall of the circulating water aquaculture pond in the sonar image corresponding to the i-th rotation scan, see Figure 6 The square area in the image; T represents the total number of pixels of the sonar image corresponding to the i-th rotation scan; S represents the rotation 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 of ​​the pool wall S i , the effective area of ​​the pool wall S is obtained by 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 a 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, is the model's predicted value, y i is the true value.

[0096] Step S4-3: Convert the number of fish obtained in step S3-4 into fish density according to 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] Where, ρ i represents the fish density of the sonar image corresponding to the i-th rotation scan;

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

[0100] Therefore, based on all fish density data points and interpolated fish density data points, the fish density ρ is fitted i With sampling depth h i The nonlinear correlation curve ρ i (h i );

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

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

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

[0104] where a i ,b i ,c i ,d i is the coefficient to be determined.

[0105] Step S5: According to the following formula, the fish biomass assessment value P in the circulating aquaculture pond is calculated by integral accumulation:

[0106]

[0107] Therefore, the present invention first obtains the sonar image corresponding to each rotation scan of the circulating water aquaculture pond by the sonar 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, the noise points generated by repeated identification in the coordinate points of the identified farmed fish are filtered out, and the number of fish corresponding to each sonar image is counted based on the actual coordinate points of the farmed fish obtained by merging the point clouds. Then, in step S4, the effective area S of the pool wall is obtained. 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 ), and finally, step S5 integrates the two curves by interval, calculates the number of fish in each interval in layers, and accumulates the fish biomass assessment value P in the circulating aquaculture pond. Therefore, the present invention filters out noise points generated by repeated identification in the identified farmed fish coordinate points to adapt to the complex environment of uneven fish density distribution in the circulating aquaculture pond, and fully considers the non-uniformity of fish density and circulating aquaculture pond wall area with depth, which can improve the measurement accuracy of fish biomass in the circulating aquaculture pond, and has the advantages of high accuracy, high efficiency, rapidness and convenience.

[0108] Among them, since step S3 of the present invention regards the identified farmed fish coordinate points with a close distance as echo reflections of the same fish based on the spatial position relationship, and merges the actual coordinate points of the farmed fish, it can adapt to the situation of rapid acceleration or abnormal movement of fish schools in the circulating water aquaculture pond, and avoids the problem of using clustering to achieve merging in the prior art, namely: when the movement of the fish school is too fast or abnormal, the local area may not meet the minimum number of points required for clustering, resulting in the formation of isolated clusters, which makes the data biased high.

[0109] The present invention is not limited to the above-mentioned specific implementation methods. According to the above content, in accordance with the common technical knowledge and customary means in this field, without departing from the above-mentioned basic technical ideas of the present invention, the present invention can also make other various forms of equivalent modifications, replacements or changes, all of which fall within the scope of protection of the present invention.

Claims

1. A method for evaluating fish biomass in a circulating aquaculture pond, characterized in that: include: Step S1, scanning the circulating aquaculture pond using a sonar in the following manner: the sonar is placed at the center of the circulating aquaculture pond and vertically descends from the water surface to the bottom of the circulating aquaculture pond or ascends from the bottom of the circulating aquaculture pond to the water surface at a fixed speed v. Simultaneously, the sonar rotates and scans horizontally at a constant angular speed ω; thereby, obtaining complete sonar image data of the aquaculture pond; Step S2: extracting the sonar image corresponding to each rotation scan from the complete sonar image data of the aquaculture pond, based on each time the sonar completes a rotation scan; Furthermore, each extracted sonar image is identified using an image recognition model to obtain the three-dimensional spatial coordinates of each identified farmed fish in the sonar image, which are recorded as the identified farmed fish coordinate point P(x, y, h), where x and y represent the horizontal coordinate and horizontal 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, filtering noise points generated by repeated identification in the identified farmed fish coordinate points obtained in step S2, including: Step S3-1, calculate the coordinate points P of any two identified farmed fish i (x q ,y q ,h q ) and P j (x j ,y j ,h j ) between the Euclidean distance D: 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 point cloud satisfies the condition D<△h; △h=T·v; Wherein, Δh is the single circle depth of the sonar in the vertical direction corresponding to one circle of rotation scanning in step S1; T is the time corresponding to each circle of rotation scanning of the sonar; Step S3-3, merging each point cloud into an actual coordinate point of the farmed fish, by calculating the average horizontal coordinate x, the average horizontal coordinate y, and the average sampling depth h of all the identified farmed fish coordinate points in the point cloud, and using these as the horizontal coordinate x, horizontal coordinate y, and sampling depth of the actual coordinate point of the farmed fish corresponding to the point cloud; Step S3-4: Count the number of actual coordinate points of farmed fish contained in each sonar image according to the sampling depth of the actual coordinate points of farmed fish, and record it as the number of fish in the sonar image; where the number of fish and the sampling depth of the sonar image corresponding to the i-th rotation scan are recorded as N respectively. 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 single circle depth of the sonar in the vertical direction corresponding to one rotation scan completed in step S1; Step S4: Obtain linear correlation curve S i (h i ) and the nonlinear correlation curve ρ i (h i ), the specific steps include: Step S4-1: When the sonar performs each rotation scan, the effective area of ​​the wall of the circulating aquaculture pond corresponding to the rotation scan within the rotation scan time period is calculated according to the following formula: Where S i M represents the effective area of ​​the wall of the circulating water aquaculture pond corresponding to the i-th rotation scan, i represents the total number of pixels in the inner area of ​​the wall of the circulating aquaculture pond in the sonar image corresponding to the i-th rotation scan; T represents the total number of pixels in the sonar image corresponding to the i-th rotation scan; S T represents the rotational scanning area of ​​the sonar; Step S4-2: Based on the sampling depth h corresponding to each rotation scan i and the effective area of ​​the pool wall S i , the effective area of ​​the pool wall S is obtained by linear fitting i With sampling depth h i The linear correlation curve S i (h i ); Step S4-3: Convert the number of fish obtained in step S3-4 into fish density according to 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 ): Where, ρ i represents the fish density of the sonar image corresponding to the i-th rotation scan; And, for each sampling depth interval [h i ,h i+1 ] corresponding to the two adjacent fish density data points (h i ,ρ i ) and (h i+1 ,ρ i+1 ), interpolation is performed using the same interpolation method to obtain a batch of interpolated fish density data points corresponding to each sampling depth interval; Therefore, based on all fish density data points and interpolated fish density data points, the fish density ρ is fitted i With sampling depth h i The nonlinear correlation curve ρ i (h i ); Step S5: Calculate the fish biomass assessment value P in the circulating aquaculture pond according to the following formula:

2. The method for evaluating fish biomass in a circulating aquaculture pond according to claim 1, wherein: In step S1, the fixed speed v is calculated according to the following formula: Where H is the total depth of the circulating water 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 circulating water aquaculture pond by the sonar.

3. The method for evaluating fish biomass in a circulating aquaculture pond according to claim 1, wherein: In step S1, the sonar is moved in the vertical direction by a motor-driven mechanical lifting device or by human power; and the sonar is an omnidirectional sonar, so as to be able to perform 360° rotation scanning in the horizontal direction by driving the scanning beam.

4. The method for evaluating fish biomass in a circulating aquaculture pond according to any one of claims 1 to 3, wherein: In step S2, the image recognition model adopts the YOLOv8s model, and its training method is as follows: Step S2-1: Randomly select 500 sonar images from the existing dataset, divide them into a training set and a test set in a ratio of 7:3, and use YoloLabel software to label the fish positions on the sonar images. Step S2-2: Using the labeled sonar images, model training is performed based on Ultralytics. The training parameters are: 300 training rounds, an initial learning rate of 0.001, and an image size of 800×600. Finally, the image recognition model is obtained, so that the image recognition model can be used to identify the position of fish schools in the sonar images, and the fish positions are counted to obtain the corresponding number of fish schools.

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

6. The method for evaluating fish biomass in a circulating aquaculture pond according to any one of claims 1 to 3, characterized in that: In step S4-3, the interpolation method adopts cubic spline interpolation.

Images