Whole sea area sound scattering layer resource abundance prediction method and electronic equipment

Abstract

The invention provides a whole sea area sound scattering layer resource abundance prediction method and electronic equipment. The method comprises the following steps: constructing a standard sound scattering layer NASC data set; constructing a sound scattering layer resource abundance prediction model based on machine learning; predicting the resource abundance of the sound scattering layer based on the trained resource abundance prediction model of the sound scattering layer; and optimizing the predicted NASC value by adopting a Kriging interpolation method to obtain a resource distribution diagram of the sound scattering layer of the whole sea area. Aiming at the problem that the survey area cannot cover the whole sea area in the fishery resource acoustic survey process, the resource abundance of the acoustic scattering layer of the sea area without hydroacoustic survey can be predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of marine fishery resource prediction, and in particular to a method and electronic equipment for predicting resource abundance of an entire sea area acoustic scattering layer. Background Art

[0002] Hydroacoustic survey is one of the commonly used methods for marine fishery resource survey. The biological communities represented by small mid-water fish in the ocean generally live in the water depth range of 200m-1000m. When conducting hydroacoustic survey, when underwater sound waves pass through these organisms, the organisms resonate and scatter with the sound waves, causing a large volume reverberation, which appears as a dense long strip echo pattern on the sonar display interface, so it is called the "acoustic scattering layer".

[0003] The acoustic scattering layer is widely distributed in the ocean and contains rich biological resources. It plays an important role in the development of marine fishery resources and ecosystem research. Due to the economic considerations of hydroacoustic surveys, large-scale surveys are difficult to carry out, so surveys are usually concentrated in a certain area. At the same time, the acoustic data collected have dense sampling on the same survey route and large intervals between adjacent survey routes, resulting in uneven data spatial distribution. These two problems make it difficult to fully understand the distribution of the abundance of acoustic scattering layer resources in the entire sea area.

[0004] Therefore, it is of great practical significance to develop a method to predict the abundance of acoustic scattering layer resources in sea areas where no survey has been conducted based on existing acoustic survey data. Summary of the invention

[0005] In view of the defects in the prior art, the purpose of the present invention is to provide a method and electronic equipment for predicting the abundance of acoustic scattering layer resources in the entire sea area. The present invention aims to predict the abundance of acoustic scattering layer resources in the sea area where no hydroacoustic survey has been carried out, in order to solve the problem that the survey area cannot cover the entire sea area during the acoustic survey of fishery resources.

[0006] To solve the above problems, the technical solution of the present invention is:

[0007] A method for predicting resource abundance of an entire sea area acoustic scattering layer comprises the following steps:

[0008] Construct a standard acoustic scattering layer NASC dataset;

[0009] Construct a prediction model for resource abundance of acoustic scattering layer based on machine learning;

[0010] Predicting the resource abundance of the acoustic scattering layer based on the trained acoustic scattering layer resource abundance prediction model;

[0011] The Kriging interpolation method is used to optimize the predicted NASC value and obtain the resource distribution map of the acoustic scattering layer in the entire sea area.

[0012] Preferably, the step of constructing a standard NASC data set for the acoustic scattering layer specifically comprises: firstly, resampling the NASC data set, processing the acquired sampling point data into surface data through two resampling processes in the vertical direction, the horizontal direction and the gridding process, solving the problem of uneven spatial distribution of the original data, and then standardizing the data set, performing K-means clustering analysis, and finally obtaining the standard NASC data set required for establishing the acoustic scattering layer resource abundance prediction model.

[0013] Preferably, in the step of constructing an acoustic scattering layer resource abundance prediction model based on machine learning, the acoustic scattering layer resource abundance prediction model based on machine learning is an SSA-XGBoost model.

[0014] Preferably, the steps of constructing the SSA-XGBoost model specifically include:

[0015] Build the mathematical model of XGBoost;

[0016] Construct the mathematical model of sparrow search algorithm SSA;

[0017] Taking advantage of the strong optimization ability and fast convergence speed of the sparrow search algorithm (SSA), the core parameters of XGBoost, such as the number of iterations, tree depth and learning rate, are optimized to build the optimal SSA-XGBoost prediction model.

[0018] Preferably, the training of the SSA-XGBoost model specifically includes: dividing 70% of the constructed standard acoustic scattering layer NASC data set into a training set, and the remaining 30% constitutes a validation set, using the longitude, latitude, and influencing factors in the training set as input data, and the category as output data, to train the constructed SSA-XGBoost model, using the longitude, latitude, and influencing factors in the validation set as input data, and the category as output data, to verify the accuracy of the trained SSA-XGBoost model.

[0019] Preferably, the verification of the SSA-XGBoost model specifically includes: using accuracy AC to evaluate the prediction accuracy of the SSA-XGBoost model, the accuracy refers to the proportion of correct prediction results in all samples, and the calculation formula is: AC=(TP+TN) / (TP+FN+FP+TN), wherein TP is the number of true positives, that is, the number of samples classified as positive and actually positive; FP is the number of false positives, that is, the number of samples classified as positive but actually negative; FN is the number of false negatives, that is, the number of samples classified as negative but actually positive; TN is the number of true negatives, that is, the number of samples classified as negative and actually negative.

[0020] Preferably, the step of predicting the resource abundance of the acoustic scattering layer based on the trained acoustic scattering layer resource abundance prediction model specifically includes: predicting the resource abundance of the acoustic scattering layer in the sea area where no investigation has been carried out based on the SSA-XGBoost model whose prediction accuracy meets the requirements.

[0021] Preferably, in the step of using the Kriging interpolation method to optimize the predicted NASC value to obtain a resource distribution map of the acoustic scattering layer in the entire sea area, the step of using the Kriging interpolation method to optimize the predicted NASC value is: using a Gaussian model, according to the distance between two known points, by weighted averaging the NASC value of the entire sea area, the weight of the weighted average is determined by the Gaussian model and the distance between the two points, the closer the distance is, the greater the weight of the known point, and the farther the distance is, the smaller the weight of the known point.

[0022] Furthermore, the present invention also provides an electronic device, comprising one or more processors, one or more memories, one or more programs and a data acquisition device; the data acquisition device is used to obtain data of the sea area acoustic scattering layer, and the one or more programs are stored in the memory. When the one or more programs are executed by the processor, the electronic device executes the above-mentioned method for predicting the abundance of the acoustic scattering layer resources in the entire sea area.

[0023] Compared with the prior art, the prediction method of the present invention first resamples the NASC sequence before constructing the prediction model of the resource abundance of the acoustic scattering layer, transforms the data distribution from point data to surface data, solves the problem of uneven spatial distribution of the original data, constructs a more reasonable NASC sequence, and overcomes the error caused by the original data to the prediction model of the resource abundance of the acoustic scattering layer; and the present invention selects a machine learning model that performs well in big data prediction, couples a parameter optimization algorithm with strong optimization ability and fast convergence speed, and after completing the prediction of the model, further considers the spatial autocorrelation characteristics of the prediction results, and uses the Kriging interpolation method to optimize the resource abundance of the acoustic scattering layer in the entire sea area. The method of the present invention not only ensures the prediction accuracy of the model, but also improves the rationality of the prediction results, can effectively serve the development of fishery resources and marine ecological research, and has great application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0025] Figure 1 It is a flowchart of the method for predicting the abundance of acoustic scattering layer resources in the entire sea area of ​​the present invention;

[0026] Figure 2It is a detailed flow chart of the method for predicting the abundance of acoustic scattering layer resources in the entire sea area of ​​the present invention;

[0027] Figure 3 A schematic diagram of the process of constructing the SSA-XGBoost model of the present invention;

[0028] Figure 4 It is a schematic diagram of the structure of the electronic device of the present invention. DETAILED DESCRIPTION

[0029] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0030] Specifically, the present invention provides a method for predicting the abundance of acoustic scattering layer resources in the entire sea area, such as Figure 1 and Figure 2 As shown, the method comprises the following steps:

[0031] S1: Construct the standard acoustic scattering layer NASC dataset;

[0032] Specifically, the construction of a standard acoustic scattering layer NASC data set includes the following steps:

[0033] Data resampling:

[0034] The acoustic data collected in the survey of sea area B during time period A were processed to obtain the sea area scattering coefficient (NASC) to characterize the resource abundance of the acoustic scattering layer in sea area B. The NASC values ​​of the sampling points obtained by the hydroacoustic survey are three-dimensional, including the sampling time, the sampling location (longitude and latitude) and the sampling depth. The NASC values ​​were resampled for the first time in the horizontal and vertical directions. The NASC values ​​in the water depth range of 20-1000m were selected in the vertical direction and accumulated as the NASC values ​​of each sampling point in the horizontal direction. Then, one sampling point was selected for every 10 sampling points in the horizontal direction to complete the first resampling. The resampled sampling point data includes the sampling time, sampling location (longitude and latitude) and NASC value (the accumulated value of NASC in the water depth range of 20-1000m for each sampling point).

[0035] Based on the longitude, latitude, and time after the first resampling, the influencing factors with the same longitude, latitude, and time are matched with the NASC value to obtain a data set M1 consisting of the NASC value and the uniquely corresponding longitude, latitude, time, and influencing factor.

[0036] The data set M1 is resampled for the second time. First, the sea area B is divided into grids with a spatial scale of 20'×20'. The NASC value, time and impact factor in the grids with sampling points are calculated as the average values ​​of the NASC value, time and impact factor of the sea area covered by the grid, and the central longitude and latitude of the grid are used as the longitude and latitude of the sea area covered by the grid. After the second resampling, the data set M2 consisting of the NASC value of the grid and the longitude, latitude, time, impact factor, etc. uniquely corresponding to it is obtained.

[0037] Data Standardization:

[0038] The values ​​contained in the data set M2 are standardized, wherein the NASC values ​​are not standardized, to obtain a data set M3.

[0039] In a specific embodiment, the data set M2 is standardized by subtracting the mean from the original data and dividing it by the standard deviation, and finally converting the original data into standard normal distribution data with a mean of 0 and a standard deviation of 1, thereby eliminating the influence of different dimensions between variables. Those skilled in the art may also reasonably select other data standardization processing methods according to the actual situation of the data.

[0040] K-means clustering:

[0041] Specifically, the K-means clustering method is used to perform cluster analysis on the data set M3, and the data set is divided into 10 categories, specifically: 10 objects are randomly selected in the data set as the initial cluster centers, and then the distance between each object and each cluster center is calculated, and each object is assigned to the cluster center closest to it. The cluster center and the objects assigned to them represent a cluster. Each time a sample is assigned, the cluster center of the cluster will be recalculated according to the existing objects in the cluster, and this process is continuously iterated until the cluster center no longer changes, or the number of iterations reaches 100 times. The protection scope of the present invention is not limited to this, and those skilled in the art can reasonably select the number of cluster categories and the stop condition of the iteration according to the actual situation of the data. The clustering results are added to the data set M3, and each group of data in the data set M3 includes the NASC value and the longitude, latitude, time, influencing factors and categories uniquely corresponding to it. The influencing factors of the resource abundance include chlorophyll concentration, dissolved oxygen, pH, mixed layer depth, sea surface temperature, light and effective radiation, eddy kinetic energy, and water depth data. The protection scope of the present invention is not limited thereto, and those skilled in the art can reasonably select the influencing factors according to actual conditions.

[0042] Through two resampling, the resampling results are standardized and K-means cluster analysis is performed, and finally the standard NASC data set required for establishing the resource abundance prediction model of the acoustic scattering layer is obtained, and the NASC mean of each category is calculated.

[0043] S2: Construct a prediction model for resource abundance of acoustic scattering layer based on machine learning;

[0044] Specifically, Figure 3 As shown in the figure, the SSA-XGBoost model for predicting the abundance of acoustic scattering layer resources based on machine learning is constructed, including the following steps:

[0045] 1. Construct the mathematical model of XGBoost. The extreme gradient boosting model (eXtreme Gradient Boosting Model, XGBoost) is an integrated learning algorithm in machine learning. It builds a powerful integrated classifier by gradually training multiple weak classifiers. First, set several values ​​for each parameter based on experience, and then select the combination with the highest accuracy by calculating the accuracy of each combination. Then, within a certain range of combinations, perform grid search to select the optimal combination.

[0046] The mathematical model of XGBoost is as follows:

[0047] Assume D1 = {(x i ,y i )} is a data set consisting of n samples and m eigenvalues. The additional function z is used by the ensemble tree model to approximate the system response as follows:

[0048]

[0049] Where F is the function space containing Z trees and is defined as:

[0050] F={f(x)=ω q(x)}(q:R m →T,ω∈R T ) (4)

[0051] In the formula, q represents the structure of the tree, T is the number of leaves, and ω is the weight of the leaf. q(x) is the score of leaf node q, f(x) is an independent tree, f z It is a function related to q, ω and associated with the independent tree.

[0052] In order to optimize the prediction performance of the ensemble tree, the objective function of XGBoost is defined as:

[0053]

[0054] Where: L is a convex loss function showing the prediction error, y i is the true value, and k is the number of iterations of the error minimization process.

[0055] 2. Construct a mathematical model of the Sparrow Search Algorithm SSA. The Sparrow Search Algorithm (SSA) is an algorithm that simulates the foraging process of a sparrow group. It has strong optimization ability and fast convergence speed. This algorithm can effectively reduce the search space, improve search efficiency, and avoid local optimal solutions in the search space.

[0056] The mathematical model of SSA is as follows:

[0057] Assign 70% of the sparrows as producers. In each iteration cycle, the position update rule of the discoverer is as follows:

[0058]

[0059] In the formula, t is the current iteration number, X ij represents the position of the i-th sparrow in the j-th dimension, α is a random number in (0,1], t max represents the maximum number of iterations, the safety value ST is a random number in the interval [0.5, 1]. The warning value R2 is a random number in [0, 1]. The random number Q follows a Gaussian distribution, and L is a 1×d matrix.

[0060] The position update formula is:

[0061]

[0062] Where, X b and X w are the global optimal position and the global worst position searched by the finder, respectively. A represents a matrix of size 1×d, in which each element is randomly set to -1 or 1, and A+=A T (AA T )-1. When i≤l / 2, the joiner competes for food with the producer who finds the best position. Otherwise, the joiner with lower fitness is in a hungry state and needs to find a new position for food.

[0063] When aware of danger from a predator, the sparrows change position according to the following rules:

[0064]

[0065] Where, X best is the current global optimal position, K is a random number in [-1,1]. β is the step length control parameter that follows the normal distribution. i , f g , f ware the fitness value of the current sparrow, the global best fitness value and the global worst fitness value respectively. ψ is set to 10 -8 , to avoid the denominator being zero.

[0066] 3. Taking advantage of SSA's strong optimization ability and fast convergence speed, the three core parameters of XGBoost: number of iterations, tree depth and learning rate are optimized to build the optimal SSA-XGBoost prediction model.

[0067] 4. Model training and verification:

[0068] The data set M3 is divided into categories, and 70% of each category is randomly selected to form the training set M4, and the remaining 30% of each category constitutes the verification set M5. The longitude, latitude, and influencing factors in the training set M4 are used as input data, and the categories are used as output data to train the constructed SSA-XGBoost model. The longitude, latitude, and influencing factors in the verification set M5 are used as input data, and the categories are used as output data to verify the accuracy of the trained SSA-XGBoost model. If the accuracy meets the requirements, the training is completed and the model is selected for the next step of predicting the abundance of acoustic scattering layer resources. Otherwise, the parameters in the model are optimized and the model training is repeated until the accuracy of the model meets the requirements.

[0069] In this embodiment, the accuracy (AC) is used to evaluate the prediction accuracy of the SSA-XGBoost model. The accuracy refers to the proportion of correct prediction results in all samples. The calculation formula is as follows:

[0070] AC=(TP+TN) / (TP+FN+FP+TN)

[0071] Among them, TP is the number of true positives, that is, the number of samples classified as positive and actually positive; FP is the number of false positives, that is, the number of samples classified as positive but actually negative; FN is the number of false negatives, that is, the number of samples classified as negative but actually positive; TN is the number of true negatives, that is, the number of samples classified as negative and actually negative.

[0072] S3: predicting the resource abundance of the acoustic scattering layer based on the trained acoustic scattering layer resource abundance prediction model;

[0073] Specifically, based on the SSA-XGBoost model whose prediction accuracy meets the requirements, the resource abundance of the acoustic scattering layer in the sea area where no investigation has been carried out is predicted. First, a standard data set M6 of the predicted sea area is constructed, which contains time, longitude, latitude, and influencing factors, and is processed according to the resampling and standardization requirements in step S1; then the data set M6 is input into the SSA-XGBoost model, which outputs the predicted category, that is, the category of the K-mean clustering analysis in step S1; the output predicted category is matched with the NASC mean of the corresponding category in step S1 to obtain the resource abundance value of the acoustic scattering layer in the predicted sea area.

[0074] S4: The predicted NASC value is optimized using the Kriging interpolation method to obtain the resource distribution map of the acoustic scattering layer in the entire sea area.

[0075] Specifically, the predicted NASC value and the corresponding longitude and latitude are extracted, and the Kriging interpolation method is used to optimize the NASC value, and finally the abundance distribution map of the acoustic scattering layer resources in the predicted sea area within a certain period of time is obtained.

[0076] The steps of optimizing the predicted NASC value using the Kriging interpolation method are as follows: using a Gaussian model, according to the distance between two known points, by weighted averaging the NASC value of the entire sea area, the weight of the weighted average is determined by the Gaussian model and the distance between the two points, the closer the known point is, the greater the weight is, and the farther the known point is, the smaller the weight is.

[0077] In summary, before constructing the prediction model of resource abundance of the acoustic scattering layer, the prediction method of the present invention first resamples the NASC sequence, transforms the data distribution from point data to surface data, solves the problem of uneven spatial distribution of the original data, constructs a more reasonable NASC sequence, and overcomes the error caused by the original data to the prediction model of resource abundance of the acoustic scattering layer; and the present invention selects a machine learning model that performs well in big data prediction, couples a parameter optimization algorithm with strong optimization ability and fast convergence speed, and after completing the prediction of the model, further considers the spatial autocorrelation characteristics of the prediction results, and uses the Kriging interpolation method to optimize the resource abundance of the acoustic scattering layer in the entire sea area. The method of the present invention not only ensures the prediction accuracy of the model, but also improves the rationality of the prediction results, can effectively serve the development of fishery resources and marine ecological research, and has great application prospects.

[0078] Another embodiment of the present invention provides an electronic device, such as Figure 4As shown, it includes one or more processors, one or more memories, one or more programs and a data acquisition device; the data acquisition device is used to obtain data of the sea area acoustic scattering layer, and the one or more programs are stored in the memory. When the one or more programs are executed by the processor, the electronic device executes the method for predicting the abundance of the acoustic scattering layer resources in the entire sea area as described above.

[0079] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for predicting the abundance of acoustic scattering layer resources in the entire sea area, characterized in that: The method comprises the following steps: Construct a standard acoustic scattering layer NASC dataset; Construct a prediction model for resource abundance of acoustic scattering layer based on machine learning; Predicting the resource abundance of the acoustic scattering layer based on the trained acoustic scattering layer resource abundance prediction model; The Kriging interpolation method is used to optimize the predicted NASC value and obtain the resource distribution map of the acoustic scattering layer in the entire sea area.

2. The method for predicting the abundance of acoustic scattering layer resources in the entire sea area according to claim 1 is characterized in that: The steps of constructing a standard NASC data set for an acoustic scattering layer specifically include: firstly, resampling the NASC data set, processing the acquired sampling point data into surface data through two resampling processes in the vertical direction, the horizontal direction and the gridding process, solving the problem of uneven spatial distribution of the original data, and then standardizing the data set and performing K-means clustering analysis, and finally obtaining the standard NASC data set required for establishing an acoustic scattering layer resource abundance prediction model.

3. The method for predicting resource abundance of the entire sea area acoustic scattering layer according to claim 1, characterized in that: In the step of constructing the acoustic scattering layer resource abundance prediction model based on machine learning, the acoustic scattering layer resource abundance prediction model based on machine learning is an SSA-XGBoost model.

4. The method for predicting the abundance of acoustic scattering layer resources in the entire sea area according to claim 3 is characterized in that: The construction steps of the SSA-XGBoost model specifically include: Build the mathematical model of XGBoost; Construct the mathematical model of sparrow search algorithm SSA; Taking advantage of the strong optimization ability and fast convergence speed of the sparrow search algorithm (SSA), the core parameters of XGBoost, such as the number of iterations, tree depth and learning rate, are optimized to build the optimal SSA-XGBoost prediction model.

5. The method for predicting the abundance of acoustic scattering layer resources in the entire sea area according to claim 4 is characterized in that: The training of the SSA-XGBoost model specifically includes: dividing 70% of the constructed standard acoustic scattering layer NASC data set into a training set, and the remaining 30% constitutes a validation set, using the longitude, latitude, and influencing factors in the training set as input data, and the category as output data, to train the constructed SSA-XGBoost model, using the longitude, latitude, and influencing factors in the validation set as input data, and the category as output data, to verify the accuracy of the trained SSA-XGBoost model.

6. The method for predicting the abundance of acoustic scattering layer resources in the entire sea area according to claim 4 is characterized in that: The verification of the SSA-XGBoost model specifically includes: using the accuracy AC to evaluate the prediction accuracy of the SSA-XGBoost model, the accuracy refers to the proportion of correct prediction results in all samples, and the calculation formula is: AC=(TP+TN) / (TP+FN+FP+TN), wherein TP is the number of true positives, that is, the number of samples classified as positive but actually positive; FP is the number of false positives, that is, the number of samples classified as positive but actually negative; FN is the number of false negatives, that is, the number of samples classified as negative but actually positive; TN is the number of true negatives, that is, the number of samples classified as negative but actually negative.

7. The method for predicting resource abundance of the entire sea area acoustic scattering layer according to claim 1, characterized in that: The step of predicting the resource abundance of the acoustic scattering layer based on the trained acoustic scattering layer resource abundance prediction model specifically includes: predicting the resource abundance of the acoustic scattering layer in the sea area where no investigation has been carried out based on the SSA-XGBoost model whose prediction accuracy meets the requirements.

8. The method for predicting resource abundance of the entire sea area acoustic scattering layer according to claim 1, characterized in that: In the step of using the Kriging interpolation method to optimize the predicted NASC value to obtain a resource distribution map of the acoustic scattering layer in the entire sea area, the step of using the Kriging interpolation method to optimize the predicted NASC value is: using a Gaussian model, according to the distance between two known points, by weighted averaging the NASC value of the entire sea area, the weight of the weighted average is determined by the Gaussian model and the distance between the two points, the closer the known point is, the greater the weight is, and the farther the known point is, the smaller the weight is.

9. An electronic device, characterized in that: The electronic device includes one or more processors, one or more memories, one or more programs and a data acquisition device; the data acquisition device is used to obtain data of the sea area acoustic scattering layer, and the one or more programs are stored in the memory. When the one or more programs are executed by the processor, the electronic device executes the method for predicting the abundance of the entire sea area acoustic scattering layer resources as described in any one of claims 1 to 8.

Images