Large-water-surface fish school number estimation method, system, equipment and medium

Abstract

The invention belongs to the technical field of intelligent fishery, and discloses a large water surface fish school number estimation method, system and device and a medium, and the method comprises the steps: calculating the installation position of a sonar monitoring device on a large water surface through a particle swarm algorithm; obtaining a sonar image of each sonar monitoring point in the water area to be evaluated; inputting each sonar image into a fish school quantity estimation model for quantity prediction to obtain fish school quantity data corresponding to each sonar monitoring point; wherein the fish school number estimation model comprises an encoder, a counter and a regression prediction layer which are connected in sequence; and executing Kriging interpolation based on the fish school quantity data corresponding to each sonar monitoring point to obtain fish school total quantity data of the to-be-evaluated water area. According to the technical scheme, the number of fish schools can be automatically estimated through the large-water-surface multi-position sonar images in combination with the deep learning technology, and the method has the advantages of being low in cost, high in efficiency, capable of achieving continuous measurement and the like.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart fisheries, and in particular relates to a method, system, equipment and medium for estimating the number of fish schools on a large water surface. Background Art

[0002] With the continuous growth of the global population and the increasing demand for aquatic products, the importance of aquaculture and fishery resource management has become increasingly prominent. Traditional fishery resource assessment methods mainly rely on manual surveys and fishing data, which is not only time-consuming and labor-intensive, but also has poor accuracy and real-time performance. In order to improve the efficiency and accuracy of fishery resource assessment, modern science and technology has gradually introduced sonar technology and artificial intelligence technology.

[0003] As an efficient underwater detection method, sonar technology has been widely used in fishery resource monitoring. By emitting sound waves and receiving reflected signals, sonar equipment can draw images of the underwater environment and then identify and locate fish schools. However, there are certain limitations to estimating the number of fish schools based solely on sonar images, such as image noise and the complexity of different water environments, which affect the accuracy of the estimation results. Summary of the invention

[0004] The purpose of the present invention is to provide a method, system, device and medium for estimating the number of fish schools on a large water surface to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides a method for estimating the number of fish schools on a large water surface, comprising:

[0006] Obtain sonar images of each sonar monitoring point in the waters to be assessed;

[0007] Input each of the sonar images into a fish population estimation model to perform population prediction, and obtain fish population data corresponding to each of the sonar monitoring points; wherein the fish population estimation model includes an encoder, a counter, and a regression prediction layer connected in sequence;

[0008] Kriging interpolation is performed based on the fish population data corresponding to each of the sonar monitoring points to obtain the total fish population data in the water area to be assessed.

[0009] Optionally, obtain sonar images of each sonar monitoring point in the water area to be evaluated, including:

[0010] The monitoring position of each sonar monitoring point is determined, a sonar device is installed based on the monitoring position, and a sonar image of the corresponding sonar monitoring point is obtained based on the sonar device.

[0011] Optionally, determine the monitoring location of each sonar monitoring point, including:

[0012] Initialize the particle swarm and randomly generate particle positions and particle velocities, where each particle represents the monitoring position of a sonar monitoring point;

[0013] Taking the coverage area of ​​each particle as a fitness function, calculating the fitness value of each particle;

[0014] Based on the fitness value set of each particle, the particle swarm algorithm iteratively updates the particle position and particle velocity of each particle in the particle swarm, and after iteratively updating to a preset number of iterations, outputs the global optimal particle position, which is the optimal monitoring position of each sonar monitoring point.

[0015] Optionally, the training process of the fish population estimation model specifically includes:

[0016] Acquiring training data, wherein the training data includes sonar training images and corresponding fish population data;

[0017] An initial fish population estimation model is constructed, the training data is input into the initial fish population estimation model for population prediction, and training is performed with the goal of minimizing the loss between the initial training result after the population prediction and the fish population data corresponding to the sonar training image to obtain the fish population estimation model.

[0018] Optionally, the processing process of the fish population estimation model specifically includes:

[0019] Inputting the sonar image into an encoder for feature extraction to obtain image feature data;

[0020] The image feature data is connected to the regression prediction layer through the counter to perform quantity prediction, thereby obtaining fish population data.

[0021] Optionally, performing Kriging interpolation based on the fish population data corresponding to each of the sonar monitoring points specifically includes:

[0022] Calculating the semivariogram values ​​between the sonar monitoring points and constructing a semivariogram based on the calculated semivariogram values;

[0023] fitting a semivariogram model based on the semivariogram;

[0024] Based on the semi-variation model, interpolation prediction is performed on each unknown point in the water area to be evaluated to obtain fish population data corresponding to each unknown point;

[0025] The fish population data corresponding to each sonar monitoring point and the fish population data corresponding to each unknown point are summed to obtain the total fish population data of the water area to be assessed.

[0026] A fish population estimation system, comprising:

[0027] A data acquisition module, used to obtain sonar images of each sonar monitoring point in the water area to be evaluated;

[0028] A fish population estimation module at a monitoring point is used to input each of the sonar images into a fish population estimation model for population prediction, and obtain fish population data corresponding to each of the sonar monitoring points; wherein the fish population estimation model includes an encoder, a counter, and a regression prediction layer connected in sequence;

[0029] The fish population estimation module is used to perform Kriging interpolation based on the fish population data corresponding to each sonar monitoring point to obtain the fish population data of the water area to be evaluated.

[0030] An electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a method for estimating the number of fish schools on a large water surface.

[0031] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements a method for estimating the number of fish schools on a large water surface.

[0032] The technical effects of the present invention are:

[0033] The present invention obtains images by installing sonar equipment at multiple locations in a large water surface, and counts the number of fish at each sonar installation location in combination with deep learning technology, thereby obtaining a method for estimating the number of overall fish schools by integrating the number of fish schools at different locations on the large water surface. This method makes it unnecessary to rely on traditional resource surveys to count the number of fish schools on large water surfaces, but can be automatically estimated through sonar images at multiple locations on large water surfaces combined with deep learning technology, with the advantages of low cost, high efficiency, and continuous measurement. This method can provide technical support for freshwater fish farming on large water surfaces. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0035] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0036] Figure 1A schematic diagram of a large surface fish population estimation algorithm flow that integrates particle swarm algorithm and deep learning in an embodiment of the present invention;

[0037] Figure 2 Schematic diagram of the structure of a sonar image fish counting model based on a deep regression network in an embodiment of the present invention;

[0038] Figure 3 Schematic diagram of the installation position of a large surface underwater sonar in an embodiment of the present invention. DETAILED DESCRIPTION

[0039] Various exemplary embodiments of the present invention will now be described in detail. This detailed description should not be considered as limiting the present invention, but should be understood as a more detailed description of certain aspects, features, and embodiments of the present invention.

[0040] It should be understood that the terms described in the present invention are only for describing special embodiments and are not intended to limit the present invention. In addition, for the numerical range in the present invention, it should be understood that each intermediate value between the upper and lower limits of the scope is also specifically disclosed. Each smaller range between the intermediate value in any stated value or stated range and any other stated value or intermediate value in the described range is also included in the present invention. The upper and lower limits of these smaller ranges can be independently included or excluded in the scope.

[0041] Unless otherwise indicated, all technical and scientific terms used herein have the same meanings as those of ordinary skill in the art generally understood by the present invention. Although the present invention describes only preferred methods, any method similar or equivalent to that described herein may also be used in the implementation or testing of the present invention. All documents mentioned in this specification are incorporated by reference to disclose and describe the methods related to the documents. In the event of a conflict with any incorporated document, the content of this specification shall prevail.

[0042] It will be apparent to those skilled in the art that various modifications and variations may be made to the specific embodiments of the present invention description without departing from the scope or spirit of the present invention. Other embodiments derived from the present invention description will be apparent to those skilled in the art. The present application description and examples are exemplary only.

[0043] The words “include,” “including,” “have,” “contain,” etc. used in this article are open-ended terms, meaning including but not limited to.

[0044] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0045] Embodiment 1

[0046] like Figure 1 - Figure 3 As shown, in this embodiment, a method for estimating the number of fish schools in a large water surface is provided, comprising: obtaining sonar images of each sonar monitoring point in the water area to be evaluated; inputting each of the sonar images into a fish school quantity estimation model for quantity prediction, and obtaining fish school quantity data corresponding to each of the sonar monitoring points; wherein the fish school quantity estimation model comprises an encoder, a counter and a regression prediction layer connected in sequence; and performing Kriging interpolation based on the fish school quantity data corresponding to each of the sonar monitoring points to obtain the total fish school data in the water area to be evaluated.

[0047] This embodiment proposes a large water surface fish population estimation system that integrates particle swarm algorithm and deep learning. The system uses particle swarm algorithm to solve the installation position of large water surface sonar, and combines deep learning technology to process sonar images to achieve real-time statistics of the number of fish at each sonar installation position, and on this basis, estimates the number of fish on the entire large water surface. The specific workflow of the system is as follows:

[0048] (1) Calculation of large surface sonar installation location based on particle swarm algorithm: Traditional methods are blind and uncertain in determining the number and location of sonar equipment, which may lead to excessive equipment, uneven coverage or low efficiency. Particle swarm algorithm is an optimization algorithm. In order to improve the layout efficiency of sonar equipment, reduce the number of equipment and ensure coverage, a method based on particle swarm optimization algorithm is proposed. Through particle swarm algorithm, the large surface installation location of sonar equipment can be effectively determined, thereby improving the representativeness and coverage of sonar image data.

[0049] Particle swarm algorithm is an intelligent optimization algorithm that simulates the foraging behavior of bird flocks in nature and is often used to solve complex multi-dimensional optimization problems. Applying particle swarm algorithm to the calculation of large surface sonar installation location can help find the best installation location so that the sonar system can cover the key areas of the large surface. The following is a basic step to use particle swarm algorithm to calculate the installation location of large surface sonar:

[0050] 1) Problem definition: The objective function of the particle swarm-based large surface sonar installation location calculation method is set to maximize the coverage area, the variables are the two-dimensional spatial coordinates (x, y) of the sonar installation location, and the constraints are set to the restricted large surface area, and try to control the coverage of different sonars to not overlap.

[0051] 2) Initialize the particle swarm: Each particle is represented by the installation position of each sonar device, and the particle swarm is initialized by randomly generating a group of particles with certain positions and velocities.

[0052] 3) Fitness function evaluation: The coverage area of ​​each particle is used as the fitness function.

[0053] 4) Particle position and velocity update:

[0054] Velocity update: Update the velocity of each particle according to the formula, taking into account the current velocity, individual optimal position, and global optimal position.

[0055]

[0056] In the formula, v i (t) is the velocity of particle i at time t, x i (t) is the position of particle i at time t, is the historical optimal position of particle i, g best is the global optimal position, ω is the inertia weight, c1, c2 are learning factors, r1, r2 are random numbers in the range [0, 1].

[0057] Position update: Update the position of the particle according to the formula.

[0058] x i (t+1)=x i (t)+v i (t+1)

[0059] 5) Iteration and convergence: Repeat the fitness evaluation and position update process until the preset maximum number of iterations is reached or the convergence condition is met (such as the fitness change is less than a certain threshold).

[0060] 6) Optimal installation position output: Output the global optimal position in the particle swarm as the optimal installation position of the sonar system.

[0061] (2) Fish counting model of multi-point sonar images on large water surface based on deep regression network: After obtaining the installation position of the sonar in the large water surface through the particle swarm algorithm, the sonar equipment is installed on site and images are collected. Afterwards, the fish in the collected sonar images are annotated by point annotation, and a large water surface multi-point fish counting sonar image dataset is constructed on this basis. The large water surface multi-point fish counting sonar image dataset is divided into training set, validation set and test set in a ratio of 6:2:2, which are used for training, validation and testing of the large water surface multi-point sonar image fish counting dataset respectively. For the sonar image data of the training set and validation set, data enhancement methods such as rotation, scaling and translation are used to improve the robustness of the trained model.

[0062] After completing the construction of the large water surface multi-point sonar image fish counting dataset, according to the characteristics of large water surface fish counting, a sonar image fish counting model based on deep regression network is designed, such as Figure 2As shown in the figure. The sonar image fish counting model based on deep regression network mainly includes three parts: encoder, counter and loss function. Among them, the encoder is mainly used to extract the features of underwater fish sonar images. Commonly used networks include LeNet, VGG16, ResNet, DenseNet and EfficientNet. The counter is mainly used to connect the output of feature extraction to the regression layer for predicting the number of fish. It mainly includes global average pooling layer, convolution layer and redundant counting layer.

[0063] The mean square error loss function is used as the loss function of the large water surface multi-point sonar image fish counting model to measure the gap between the predicted value and the true value during the model training process. The definition of the mean square error loss function is as follows:

[0064]

[0065] Where n is the number of underwater sonar images, i is the i-th underwater sonar image, and y i is the actual number of fish in the i-th underwater sonar image, is the number of fish schools in the i-th underwater sonar image estimated by the model.

[0066] The Adam optimizer was used in the model training process, and the learning rate was set to 0.001, and the early stopping technique was adopted to avoid overfitting.

[0067] This completes the design of the fish counting model for multi-point sonar images on large water surfaces. After that, we input the image data of each sonar installation point into the model to obtain the number of fish at each point, laying the foundation for the subsequent estimation of the number of fish on large water surfaces.

[0068] (4) Estimation of the number of fish schools on large water surfaces based on Kriging interpolation: Kriging interpolation is a geostatistical method used to spatially predict and estimate data values ​​of unknown points. It is widely used in fields such as geology, meteorology, hydrology, and environmental science. Through Kriging interpolation, the fish population data at each sonar installation location on a known large water surface can be used to estimate the spatial distribution and total number of fish schools on a large water surface. Kriging interpolation is based on the assumption of spatial correlation, that is, data values ​​at similar locations have a high correlation. This means that two large water surface fish population observation points that are close to each other are more likely to have similar fish populations, while observation points that are farther apart may have larger differences in fish populations. This study studies the estimation of large water surface fish populations based on Kriging interpolation, which mainly includes the following steps:

[0069] Data preparation: Collect the number of fish and their coordinates of observation points on the large water surface. These observation points should be distributed throughout the large water surface to ensure the accuracy of the interpolation results.

[0070] Calculate Semivariogram: Calculates the semivariogram values ​​between observation points and plots a semivariogram. The semivariogram shows the semivariogram values ​​at different distances.

[0071] The semivariogram is the core of Kriging interpolation and is used to describe the spatial correlation between data points. It reflects the degree of difference between data points as the distance increases. The calculation formula of the semivariogram is:

[0072]

[0073] Where γ(h) is the value of the semivariogram at distance h, N(h) is the number of point pairs at distance h, and z(x i ) and z(x i +h) are the observation values ​​of two points with a distance h.

[0074] Fit Semivariogram Model: Select and fit an appropriate semivariogram model, such as a spherical model, exponential model, or Gaussian model, based on the semivariogram. This model is used to describe the spatial correlation of fish populations.

[0075] Perform Kriging Interpolation: Use the fitted semivariogram model to interpolate and predict each unknown point in the large water surface. Kriging interpolation calculates an estimate of each unknown point based on the weights of the known observations. The weights are determined based on the semivariogram model to minimize the variance of the interpolation results.

[0076] Estimation of the total number of fish on a large water surface: Sum the number of fish at all interpolation points in the entire large water surface area to estimate the total number of fish on the large water surface. This can be achieved through methods such as integration or grid summation.

[0077] This embodiment proposes a method of obtaining images by installing sonar equipment at multiple locations on a large water surface, and counting the number of fish at each sonar installation location in combination with deep learning technology, thereby obtaining a method for estimating the number of fish schools as a whole by integrating the number of fish schools at different locations on the large water surface. This method eliminates the need to rely on traditional resource surveys to count the number of fish schools on large water surfaces, but can be directly estimated automatically through sonar images at multiple locations on large water surfaces combined with deep learning technology, with the advantages of low cost, high efficiency, and continuous measurement. This method can provide technical support for freshwater fish farming on large water surfaces.

[0078] A fish population estimation system, comprising:

[0079] A data acquisition module, used to obtain sonar images of each sonar monitoring point in the water area to be evaluated;

[0080] A fish population estimation module at a monitoring point is used to input each of the sonar images into a fish population estimation model for population prediction, and obtain fish population data corresponding to each of the sonar monitoring points; wherein the fish population estimation model includes an encoder, a counter, and a regression prediction layer connected in sequence;

[0081] The fish population estimation module is used to perform Kriging interpolation based on the fish population data corresponding to each sonar monitoring point to obtain the fish population data of the water area to be evaluated.

[0082] An electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a method for estimating the number of fish schools on a large water surface.

[0083] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements a method for estimating the number of fish schools on a large water surface.

[0084] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for estimating the number of fish schools in large waters. It is characterized in that include: Obtain sonar images of each sonar monitoring point in the waters to be assessed; Input each of the sonar images into a fish population estimation model to perform population prediction, and obtain fish population data corresponding to each of the sonar monitoring points; wherein the fish population estimation model includes an encoder, a counter, and a regression prediction layer connected in sequence; Kriging interpolation is performed based on the fish population data corresponding to each of the sonar monitoring points to obtain the total fish population data in the water area to be assessed.

2. A method for estimating the number of fish schools in a large water surface according to claim 1, It is characterized in that Obtain sonar images of each sonar monitoring point in the waters to be assessed, including: The monitoring position of each sonar monitoring point is determined, a sonar device is installed based on the monitoring position, and a sonar image of the corresponding sonar monitoring point is obtained based on the sonar device.

3. A method for estimating the number of fish schools in a large water surface according to claim 2, It is characterized in that Determine the monitoring location of each sonar monitoring point, including: Initialize the particle swarm and randomly generate particle positions and particle velocities, where each particle represents the monitoring position of a sonar monitoring point; Taking the coverage area of ​​each particle as a fitness function, calculating the fitness value of each particle; Based on the fitness value set of each particle, the particle swarm algorithm iteratively updates the particle position and particle velocity of each particle in the particle swarm, and after iteratively updating to a preset number of iterations, outputs the global optimal particle position, which is the optimal monitoring position of each sonar monitoring point.

4. A method for estimating the number of fish schools in a large water surface according to claim 1, It is characterized in that The training process of the fish population estimation model specifically includes: Acquiring training data, wherein the training data includes sonar training images and corresponding fish population data; An initial fish population estimation model is constructed, the training data is input into the initial fish population estimation model for population prediction, and training is performed with the goal of minimizing the loss between the initial training result after the population prediction and the fish population data corresponding to the sonar training image to obtain the fish population estimation model.

5. A method for estimating the number of fish schools in a large water surface according to claim 1, It is characterized in that The processing process of the fish population estimation model specifically includes: Inputting the sonar image into an encoder for feature extraction to obtain image feature data; The image feature data is connected to the regression prediction layer through the counter to perform quantity prediction, thereby obtaining fish population data.

6. A method for estimating the number of fish schools in a large water surface according to claim 1, It is characterized in that Kriging interpolation is performed based on the fish population data corresponding to each of the sonar monitoring points, specifically including: Calculating the semivariogram values ​​between the sonar monitoring points and constructing a semivariogram based on the calculated semivariogram values; fitting a semivariogram model based on the semivariogram; Based on the semi-variation model, interpolation prediction is performed on each unknown point in the water area to be evaluated to obtain fish population data corresponding to each unknown point; The fish population data corresponding to each sonar monitoring point and the fish population data corresponding to each unknown point are summed to obtain the total fish population data of the water area to be evaluated.

7. A system for estimating the number of fish schools on large water surfaces. It is characterized in that include: A data acquisition module, used to obtain sonar images of each sonar monitoring point in the water area to be evaluated; A fish population estimation module at a monitoring point is used to input each of the sonar images into a fish population estimation model for population prediction, and obtain fish population data corresponding to each of the sonar monitoring points; wherein the fish population estimation model includes an encoder, a counter, and a regression prediction layer connected in sequence; The fish population estimation module is used to perform Kriging interpolation based on the fish population data corresponding to each sonar monitoring point to obtain the fish population data of the water area to be evaluated.

8. An electronic device, It is characterized in that It comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a method for estimating the number of fish schools on a large water surface according to any one of claims 1-6.

9. A computer-readable storage medium, It is characterized in that It stores a computer program, which, when executed by a processor, implements a method for estimating the number of fish schools on a large water surface as described in any one of claims 1-6.

Images