Marine ranching decision determination method, device, medium and product

Abstract

The application discloses a mariculture decision determination method and device, medium and product, and relates to the field of mariculture. The method comprises the following steps: acquiring multi-source heterogeneous data for a mariculture, wherein the multi-source heterogeneous data comprises sensor data, remote sensing data and log data; performing feature extraction on the sensor data, the remote sensing data and the log data respectively to obtain corresponding feature vectors, which are respectively denoted as a sensor feature vector, a remote sensing feature vector and a log feature vector; and sending the sensor feature vector, the remote sensing feature vector and the log feature vector to a fuzzy logic model, so that the fuzzy logic model outputs a mariculture decision suggestion. The application improves the speed and preparedness of determining the mariculture decision suggestion.

Description

Technical Field

[0001] This application relates to the field of marine ranching technology, and in particular to a method, equipment, medium and product for determining marine ranching decisions. Background Technology

[0002] Marine ranching, as an important direction for the development of modern fisheries, is a typical model for the comprehensive utilization of marine space resources and the implementation of intensive, ecological, and intelligent management of aquaculture. With the development of the Internet of Things, remote sensing monitoring, and automation technology, marine ranching increasingly relies on multi-source data sensing systems for ecological environment monitoring and aquaculture operation management.

[0003] Currently, such systems typically include multiple information sources, such as underwater sensor networks, satellite remote sensing platforms, UAV inspection systems, ship AIS track data, and manually entered aquaculture logs. These data vary significantly in type, spatiotemporal resolution, acquisition frequency, and reliability, constituting a typical "multi-source heterogeneous data" environment.

[0004] Due to the lack of data integration capabilities mentioned above, existing systems often struggle to provide accurate, reliable, and timely decision support. Most marine ranches still rely on subjective comparisons and judgments of various data based on human experience, which is not only inefficient but also prone to misjudgment. This can lead to significant economic losses and ecological impacts, such as oxygen depletion, red tides, and water quality deterioration caused by overfeeding. If decision-making delays or misjudgments prevent the timely activation of aeration equipment, large-scale fish suffocation and death may occur; improper control of feeding amounts can exacerbate eutrophication and trigger a chain reaction of ecological problems. Summary of the Invention

[0005] The purpose of this application is to provide a method, equipment, medium, and product for determining aquaculture decisions in marine ranching, in order to solve the problems of low efficiency and accuracy in determining aquaculture decisions as described in the prior art.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for determining marine ranching aquaculture decisions, including: Acquire multi-source heterogeneous data for marine ranches, including sensor data, remote sensing data, and log data; Feature vectors are obtained by performing feature extraction on the sensor data, the remote sensing data, and the log data, respectively, and are denoted as sensor feature vector, remote sensing feature vector, and log feature vector. The sensor feature vector, the remote sensing feature vector, and the log feature vector are sent to the fuzzy logic model so that the fuzzy logic model can output aquaculture decision suggestions.

[0007] Optionally, the sensor data includes a data sequence of dissolved oxygen concentration, and the sensor feature vector is extracted using the following method: For any given time period, obtain the data sequence of the dissolved oxygen concentration within that time period; Within the time period, based on the data sequence of dissolved oxygen concentration, the dominant frequency feature of dissolved oxygen concentration fluctuation is determined, and the dominant frequency feature is determined as the sensor feature vector within the time period.

[0008] Optionally, the remote sensing data includes a sea surface temperature data sequence, and the remote sensing feature vector is extracted using the following method: For any given time period, obtain the sea surface temperature data sequence; Within the time period, the gradient vector of the sea surface temperature field is determined based on the sea surface temperature data sequence. The sea surface temperature gradient vector is converted into a temperature gradient magnitude feature, and the temperature gradient magnitude feature is determined as the remote sensing feature vector.

[0009] Optionally, the log data includes a sequence of feed amount data, and the log feature vector is extracted using the following method: For any given time period, obtain the sequence of feed amount data; Within the time period, based on the feeding amount data sequence, the peak points of each feeding amount within the time period are determined to obtain the peak period feature sequence, and the peak period feature sequence is determined as the log feature vector.

[0010] Optionally, after sending the sensor feature vector, the remote sensing feature vector, and the log feature vector to the fuzzy logic model so that the fuzzy logic model outputs aquaculture decision suggestions, the method further includes: Based on the aforementioned aquaculture decision recommendations, the dynamic characteristics of marine organisms are determined; Obtain environmental data of the marine ranch; The dynamic features and environmental data are fused into the static three-dimensional marine ranch model to obtain a dynamic three-dimensional marine model.

[0011] Optionally, after sending the sensor feature vector, the remote sensing feature vector, and the log feature vector to the fuzzy logic model so that the fuzzy logic model outputs aquaculture decision suggestions, the method further includes: The aquaculture decision suggestions are sent to the equipment control system so that the equipment control system can generate control commands based on the aquaculture decision suggestions.

[0012] Optionally, after acquiring multi-source heterogeneous data for marine ranches, the method includes: preprocessing the sensor data and the remote sensing data, wherein the preprocessing includes removing outliers from the sensor data and supplementing missing values ​​from the sensor data; the preprocessing also includes spatiotemporally aligning the remote sensing data and the sensor data to obtain spatiotemporally aligned data of the remote sensing data and the sensor data.

[0013] In a second aspect, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described in any one of the first aspects above.

[0014] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the first aspects above.

[0015] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any one of the first aspects above.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed: The marine ranching decision-making method provided in this application acquires multi-source heterogeneous data for marine ranching, including sensor data, remote sensing data, and log data. Feature vectors are extracted from the sensor data, remote sensing data, and log data respectively, denoted as sensor feature vector, remote sensing feature vector, and log feature vector. These feature vectors are then sent to a fuzzy logic model, which outputs aquaculture decision recommendations. In contrast, existing technologies rely on manual comparison, which is slow and prone to errors. Therefore, this application improves the speed and accuracy of determining aquaculture decision recommendations. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a diagram illustrating the application environment of the marine ranching decision-making method in one embodiment of this application. Figure 2 A flowchart illustrating a method for determining marine ranching aquaculture decisions according to an embodiment of this application; Figure 3 A flowchart illustrating a method for determining marine ranching aquaculture decisions according to an embodiment of this application; Figure 4 A flowchart illustrating a data cleaning method provided in another embodiment of this application; Figure 5 A flowchart illustrating a method for spatiotemporal alignment of data provided in an embodiment of this application; Figure 6 A flowchart illustrating a sensor feature vector extraction method provided in an embodiment of this application; Figure 7 A flowchart illustrating a method for extracting remote sensing feature vectors according to an embodiment of this application; Figure 8 A flowchart illustrating a method for extracting log feature vectors according to an embodiment of this application; Figure 9 A flowchart illustrating a method for determining a dynamic three-dimensional ocean model according to an embodiment of this application; Figure 10 A flowchart illustrating a method for visualizing a three-dimensional model and implementing decision recommendations, provided as an embodiment of this application; Figure 11 A flowchart illustrating a method for determining decision recommendations, provided in an embodiment of this application; Figure 12 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the contents of this application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Definitions: AIS (Automatic Identification System for Ships): An automatic identification system used to obtain information such as the location and course of ships related to marine ranching, and is one of the heterogeneous data sources in this solution.

[0022] The 3σ principle: In statistics, the outlier judgment principle based on the normal distribution characteristics refers to the judgment of data when the deviation from the mean exceeds three times the standard deviation. It is used for the anomaly cleaning of underwater sensor data in this scheme.

[0023] Multi-source heterogeneous data: data from different sources such as underwater sensors in marine ranches, satellite remote sensing platforms, aquaculture log databases, AIS, and drone inspection systems, and data with different data types or formats.

[0024] Fuzzy logic: A logical method for handling uncertainty problems. It maps feature values ​​to fuzzy sets such as "low frequency / high frequency" and "normal / floating head" through membership functions, and then outputs risk levels based on preset rules for decision-making fusion in this scheme.

[0025] Membership function: A function that describes the degree to which an input feature value belongs to a certain fuzzy set. In this scheme, a Gaussian type is used, and the membership distribution range is defined by the center value and the width parameter.

[0026] Power spectral density: A quantitative index that characterizes the distribution of signal energy at different frequencies by performing discrete Fourier transform on time series signals such as dissolved oxygen concentration, and is used to extract the frequency characteristics of dissolved oxygen concentration fluctuations.

[0027] Sea surface temperature gradient characteristics: extracted from satellite remote sensing data, and standardized characteristics reflecting local sea surface temperature thermodynamic differences are obtained by calculating the rate of change and gradient modulus of sea surface temperature in the x and y directions.

[0028] Dissolved oxygen concentration fluctuation frequency characteristics: Extracted from underwater sensor data, the main oscillation frequencies of dissolved oxygen concentration were determined by discrete Fourier transform and power spectral density analysis, reflecting the standardized characteristics of the water body's ecological stability.

[0029] Peak feeding period characteristics: Extracted from aquaculture log data, peak detection is performed using a sliding window (feeding amount exceeding 2 standard deviations of the local mean is marked as peak), and standardized characteristics of concentrated feeding periods are identified in binary (1 for peak, 0 for non-peak).

[0030] 3D geographic coordinate system: A three-dimensional coordinate system used to uniformly map spatial information such as seabed topography, spatial location of underwater sensors, and fish distribution in marine ranches, enabling spatial alignment of multi-source data.

[0031] A 3D digital twin model: Integrating spatiotemporally aligned data and decision-making recommendations, this model uses Web Graphics Library (WebGL) technology to render and display a real-time 3D virtual model of the marine ranch's terrain, sensor locations, fish distribution, and environmental data, corresponding to the visualization aspects of the solution. WebGL is a JavaScript API for rendering 2D and 3D graphics in web browsers, achieving high-performance graphics rendering without the need for plugins.

[0032] Control command frame: A data structure used to send control commands to the marine ranch execution equipment. It includes a frame header (0xAA55), device address, command code, parameter list, and CRC-8 checksum, realizing the conversion of decision suggestions into equipment actions.

[0033] CRC-8 algorithm: A cyclic redundancy check algorithm that calculates an 8-bit checksum by concatenating the device address, instruction code, and parameter list in a control command frame to ensure data transmission integrity.

[0034] Linear interpolation: A method used to fill data gaps after outlier removal or to estimate the value of remote sensing data at high-frequency time points. It calculates the target time / location value through the linear relationship between adjacent valid data points.

[0035] Discrete Fourier Transform: A mathematical transformation method that converts continuous time-domain signals such as dissolved oxygen concentration into frequency-domain signals. By calculating the signal amplitude at different angular frequencies, it provides a basis for extracting fluctuation frequency characteristics.

[0036] This application provides an environment in which the marine ranching decision-making method can be applied. See also: Figure 1 The application environment includes terminals and servers.

[0037] The data storage system stores the data that the server needs to process, such as the multi-source heterogeneous data mentioned below. The data storage system can be set up independently, integrated into the server, or located in the cloud or on other servers. Furthermore, the data storage system stores relevant data required for executing the marine ranching decision-making method, such as source data obtained from different data sources. Of course, the data storage system also stores intermediate data generated during the execution of the marine ranching decision-making method, so that it can be retrieved promptly when needed.

[0038] In this system, the terminal communicates with the server via a network. The terminal can send data to be processed to the server. After receiving the data, the server can store it and retrieve it from the storage location when processing is needed, or it can perform the processing task while storing the data. The server can also provide feedback on the obtained aquaculture decisions to the terminal.

[0039] In addition, in some embodiments, the marine ranching decision-making method can also be implemented by a server or a terminal alone. For example, the terminal can directly perform marine ranching decision-making processing on the data to be processed, or the server can obtain the data to be processed from the data storage system and perform marine ranching decision-making processing on the data to be processed.

[0040] The terminals can be, but are not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. Servers can be implemented using independent servers, server clusters composed of multiple servers, or cloud servers.

[0041] In one exemplary embodiment, see Figure 2 and Figure 3 As shown, a method for determining marine ranching aquaculture decisions is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking the server in the example, the following steps 101 to 103 are used as an example: Step 101: Obtain multi-source heterogeneous data for marine ranches, including sensor data, remote sensing data, and log data; Among them, sensor data is data collected by underwater sensors, remote sensing data is data collected by satellite remote sensing, and log data is log data recorded by the local server's aquaculture logs.

[0042] This step may involve: obtaining raw data from multiple heterogeneous data sources; the heterogeneous data sources include underwater sensors deployed in marine ranches, satellite remote sensing platforms that scan marine ranches, aquaculture log databases that record aquaculture operations, Automatic Identification System (AIS) for ships, and unmanned aerial vehicle (UAV) inspection systems.

[0043] Step 102: Extract features from the sensor data, the remote sensing data, and the log data to obtain corresponding feature vectors, which are denoted as sensor feature vector, remote sensing feature vector, and log feature vector, respectively. Among them, the sensor feature vector, remote sensing feature vector, and log feature vector are feature vectors used to determine aquaculture decision-making recommendations, or standardized feature vectors for ecological-aquaculture collaborative decision-making.

[0044] Step 103: Send the sensor feature vector, the remote sensing feature vector, and the log feature vector to the fuzzy logic model so that the fuzzy logic model can output aquaculture decision suggestions.

[0045] The fuzzy logic model is equipped with a fuzzy logic algorithm and a decision rule base. The decision rule base predefines the correspondence between various preset risks and aquaculture decision suggestions. The fuzzy logic algorithm can calculate the probability that a multi-vector belongs to each preset risk based on the input multi-vector.

[0046] Step 103 can be performed as follows: calculate the probability that the current multi-feature vector combination belongs to each preset risk using a fuzzy logic algorithm, and determine the risk with the highest probability; determine the aquaculture decision suggestion corresponding to the risk with the highest probability based on the decision rule base.

[0047] For example, when the dissolved oxygen concentration fluctuation frequency characteristic is high frequency, the UAV observation characteristic identifies fish surfacing, and the peak feeding time characteristic shows a recent feeding peak, the decision rule base outputs the decision suggestion of "risk of oxygen deficiency" and "turn on the oxygenation equipment and reduce feeding".

[0048] Furthermore, the decision rule base can predefine the mapping relationship between various feature vector combinations and corresponding aquaculture decision suggestions. This allows for direct querying of corresponding aquaculture decision suggestions based on various feature vector combinations, eliminating the need to calculate the probability of each preset risk.

[0049] Additionally, step 103 corresponds to Figure 3 The decision-making level in the process is integrated and processed.

[0050] The marine ranching decision-making method provided in this application converts multi-source heterogeneous data into different feature vectors, then fuses these different feature vectors using a fuzzy logic model, and finally outputs aquaculture decision recommendations through the fuzzy logic model. This reduces the influence of human factors and improves the speed and accuracy of determining aquaculture decision recommendations.

[0051] Optionally, see Figure 4 and Figure 5After step 101, the method includes: preprocessing the sensor data and the remote sensing data, wherein the preprocessing includes removing outliers from the sensor data and supplementing missing values ​​from the sensor data; the preprocessing also includes spatiotemporally aligning the remote sensing data and the sensor data to obtain spatiotemporally aligned data of the remote sensing data and the sensor data.

[0052] Preprocessing can remove unreasonable data and fill in missing data, thereby improving data quality.

[0053] Furthermore, an anomaly cleaning algorithm based on the 3σ principle is adopted to identify and remove faulty data and outliers in the underwater sensor data, resulting in a cleaned data sequence.

[0054] In the process of anomaly cleaning, the outlier cleaning algorithm based on the 3σ principle first receives the raw monitoring data sequence from the underwater sensor: in, Indicates the first Sensor readings at each moment, Let be the time index, and 'a' be the total number of sampling points within the time window. The first step of the algorithm is to calculate the statistical characteristics of the data sequence, obtaining the arithmetic mean μ and the standard deviation σ, expressed as: Here, μ represents the central trend of the sequence, reflecting the overall level of the sensor within the time window, while σ represents the dispersion of each sampling point from the arithmetic mean μ, used to measure the amplitude of data fluctuation. Then, the entire data sequence is traversed, and the readings at each time point are... Calculate the deviation from the arithmetic mean μ, if the following condition is met: Then this data point Points identified as outliers and removed are subject to the 3σ principle, which states that under the approximation of a normal distribution, points exceeding three standard deviations are considered outliers outside the confidence interval. For the gaps created by the removal, linear interpolation is used to fill them. Let the time of the gap be t, and the effective observation points before and after it be... and Then the interpolation at time t for: in, The valid reading before the empty defect, The effective reading after the empty defect, and For the corresponding time index, The values ​​are estimated by linear interpolation, also known as interpolated or imputed values. This method ensures the consistency of the time series in terms of numerical continuity and trend preservation, thus avoiding damage to the overall structure of the time series due to missing values ​​while cleaning outliers. After the above steps, the final cleaned and imputed complete data series is obtained: in, The output data sequence represents either the original readings retained after outlier removal (where there are multiple values ​​at a given moment, outliers were removed, and normal values ​​were retained) or the imputed values ​​after interpolation, where 'a' represents the total number of sampling points within the time window. This output data sequence serves as a reliable input basis for subsequent spatiotemporal alignment and feature extraction.

[0055] Further, see Figure 5 The spatiotemporal alignment includes: receiving the cleaned data sequence, satellite remote sensing data, and UAV observation data; performing spatiotemporal standardization processing on data from different spatiotemporal references; for the time dimension, aligning the low-frequency hourly data from satellite remote sensing and the high-frequency continuous monitoring data from underwater sensors to a unified high-frequency time series using a time interpolation algorithm (linear interpolation algorithm); for the spatial dimension, uniformly mapping the data from all data sources (sensors, remote sensing, and UAVs) to the three-dimensional geographic coordinate system of the marine ranch; and finally outputting the spatiotemporally aligned data sequence.

[0056] In the spatiotemporal alignment process, the goal of the time interpolation algorithm is to align low-frequency satellite remote sensing hourly observation data with high-frequency sensor acquisition data, enabling the fusion of data from different time scales under a unified time reference. The input includes a sequence of satellite remote sensing hourly data: in, Indicates the hour. Remote sensing observations Indicates total The input includes a sequence of satellite remote sensing data at specific times, and also includes a sequence of high-frequency sampling timestamps from the sensors: Each of them This indicates the specific data acquisition time of the sensor within the time window, where 'a' is the total number of sampling points within the time window. For any given sensor acquisition time... The remote sensing data value corresponding to that moment needs to be calculated using linear interpolation. The interpolation formula is: in, and These are two consecutive integer times. and Satellite remote sensing observations, For the moment when the sensor acquires data, the molecular part Indicates from to Linear estimation of changes in remote sensing values ​​over a time interval, denominator For the hourly intervals, the overall calculation results are given at the following times. Estimated value of remote sensing data at time By sampling all high-frequency times After performing the above interpolation operations, the finally obtained is an aligned remote sensing data sequence that corresponds one-to-one with the sensor sampling timestamps: in, Indicates the time of sensor acquisition The corresponding interpolated remote sensing values ​​ensure that low-frequency remote sensing observations and high-frequency sensor monitoring are perfectly matched in the time dimension, thereby guaranteeing that the data input for subsequent feature extraction and decision-level fusion has a unified time resolution and reliability.

[0057] Optionally, see Figure 6 In another exemplary embodiment of this application, the sensor data includes a data sequence of dissolved oxygen concentration. The sensor feature vector in step 102 is extracted by the following method: based on the dissolved oxygen concentration data sequence, the dominant frequency feature of the dissolved oxygen concentration fluctuation is determined, and the dominant frequency feature is determined as the sensor feature vector. Further, it can be extracted through the following steps 201-202: Step 201: For any given time period, obtain the data sequence of the dissolved oxygen concentration within that time period; Step 202: Within the time period, based on the data sequence of dissolved oxygen concentration, determine the main frequency feature of dissolved oxygen concentration fluctuation, and determine the main frequency feature as the sensor feature vector within the time period.

[0058] In the feature layer fusion processing step, the goal of dissolved oxygen concentration fluctuation frequency feature extraction is to reveal the main periodic patterns of dissolved oxygen changes over time in water bodies, thus serving as an important indicator reflecting the stability of the ecological environment. The input is a dissolved oxygen concentration time series after anomaly data cleaning. ,in, Indicates the first Dissolved oxygen concentration value at each moment, time index Corresponding to consecutive sampling points, the total number is To analyze the frequency characteristics of the sequence, we first perform a discrete Fourier transform on it, calculated using the following formula: Where ω represents the angular frequency, Δt represents the sensor sampling time interval, and the exponent term... The time-domain signal is transformed to the frequency domain. The Fourier transform result F(ω) is a complex number whose amplitude reflects the energy distribution of the signal at frequency ω. To quantitatively describe the energy distribution at different frequencies, the power spectral density is further calculated: in, The denominator represents the energy of the Fourier transform result. The signal is normalized to ensure the comparability of power spectra across different sampling lengths. Analysis of the power spectral density function reveals the principal components of the signal in the frequency domain, denoted as: in, This indicates the solution for the argument of a complex number. This represents the angular frequency at which the power spectral density reaches its maximum value, corresponding to the dominant frequency component of the dissolved oxygen concentration fluctuation. To convert this into a practically meaningful frequency characteristic, the output dominant frequency value is defined as: Where f represents the dominant frequency of dissolved oxygen concentration fluctuations, expressed in Hertz, which intuitively reflects the number of major oscillations in dissolved oxygen concentration per unit time. The resulting dominant frequency feature f is incorporated into the standardized feature vector as an important input component for decision-making layer fusion, providing a quantitative basis for identifying the risk of water hypoxia and abnormal fish behavior.

[0059] Optionally, see Figure 7 In another exemplary embodiment of this application, the remote sensing data includes a sea surface temperature data sequence, and the remote sensing feature vector in step 102 is extracted by the following method: determining the temperature gradient magnitude feature based on the sea surface temperature data sequence, and determining the temperature gradient magnitude feature as the remote sensing feature vector; further, it can be extracted by the following steps 301 to 303: Step 301: For any given time period, obtain the sea surface temperature data sequence; Step 302: Within the time period, determine the gradient vector of the sea surface temperature field based on the sea surface temperature data sequence; Step 303: The sea surface temperature gradient vector is converted into a temperature gradient magnitude feature, and the temperature gradient magnitude feature is determined as the remote sensing feature vector.

[0060] In the feature layer fusion processing step, the goal of sea surface temperature gradient feature extraction is to quantitatively calculate the spatial rate of change of satellite-sensed sea surface temperature, thereby reflecting the thermodynamic differences in local sea areas and providing effective environmental features for subsequent collaborative decision-making in ecological aquaculture. The input is a spatiotemporally aligned sea surface temperature data sequence: in Indicates at time The sea surface temperature value, This represents the total number of co-humidity data sequences. Since temperature not only varies over time but also exhibits spatial non-uniformity, it is necessary to calculate the gradient vector of the sea surface temperature field. in, This represents the rate of change of temperature in the horizontal x-axis. This represents the rate of change of temperature in the vertical direction y. The spatial partial derivative is approximated using the central difference method, and its formula is: in, and , respectively, represent the step size of the spatial grid in the x and y directions. and This represents the sea surface temperature values ​​in the horizontal direction for adjacent grid points. and This represents the sea surface temperature value of adjacent grid points in the vertical direction. The calculated gradient vector components characterize the intensity of temperature variation in different directions. To transform the spatial gradient features into a single quantifiable index that can be directly used for feature fusion, the temperature gradient magnitude feature is further calculated: The result G represents the overall variation in sea surface temperature within a local spatial unit, reflecting potential temperature fronts or local anomalies in the sea area. The final output gradient magnitude feature G is incorporated into the standardized feature vector as an important input component for decision-making layer fusion, thereby ensuring that the spatial variability of the marine environment can be comprehensively considered during risk assessment and decision calculation.

[0061] Optionally, see Figure 8 In another exemplary embodiment of this application, the log data includes a feeding amount data sequence, and the log feature vector in step 102 is extracted by the following method: determining the peak feeding point based on the feeding amount data sequence to obtain a peak period feature sequence; and determining the peak period feature sequence as the log feature vector. Further, it can be extracted through steps 401 and 402: Step 401: For any given time period, obtain the feeding amount data sequence; Step 402: Within the time period, based on the feeding amount data sequence, determine the peak points of each feeding amount within the time period to obtain the peak period feature sequence, and determine the peak period feature sequence as the log feature vector.

[0062] In the feature layer fusion processing step, the goal of extracting peak feeding periods is to identify abnormally concentrated periods of feeding behavior during aquaculture operations, thereby providing a quantitative basis for the correlation analysis between environmental monitoring data and aquaculture activities. The input is the time series of feeding amounts recorded in the aquaculture log: in Indicates the first Feeding amount within a time unit, time index The range is the entire observation period. This represents the total number of time units. The feature extraction process uses a sliding window peak detection algorithm, with a window width of w, meaning that at each time point, the window expands forward and backward from its center, covering a subsequence of w data points. Statistics are calculated for the data within this window, including the window mean: And window standard deviation: in, This indicates the center level of the amount of feed placed within the window. This indicates the fluctuation range of the feeding amount within this local area. During peak detection, the feeding amount at each moment is... Compared with local statistical characteristics, the following conditions are met: This moment is marked as the peak feeding point. This threshold rule means that points where the feeding amount exceeds two standard deviations from the local mean are considered significantly abnormally high. To facilitate subsequent feature vector standardization, features from the peak period are extracted: in, It is a binary identifier, at time When detected as a peak point ,otherwise . This represents the total number of features, resulting in the final peak period feature sequence. The study accurately characterized the concentrated range of feeding behavior in aquaculture operations. This result, as an important component of the standardized feature vector, provides input support for identifying the coupling relationship between feeding and the risk of water hypoxia in decision-making fusion.

[0063] Optionally, see Figure 9 and Figure 10 In another exemplary embodiment of this application, after step 103, the method further includes steps 501 to 503: Step 501: Obtain the dynamic characteristics of the marine organisms corresponding to the aquaculture decision suggestions from the pre-stored correspondence between aquaculture decision suggestions and the dynamic characteristics of marine organisms; Step 502: Obtain environmental data of the marine ranch; Step 503: The dynamic features and environmental data are fused into the static three-dimensional marine ranch model to obtain a dynamic three-dimensional marine model.

[0064] The correspondence between aquaculture decision-making recommendations and the dynamic characteristics of marine organisms can be stored in advance so that they can be retrieved in a timely manner when needed.

[0065] Among them, the three-dimensional model is also called the three-dimensional digital twin model.

[0066] In this step, the rendering process of the 3D digital twin model first uses the spatiotemporally aligned data sequence as the basic input. Further, the spatiotemporally aligned data sequence is defined as: in, Indicates a point in time At that time, in spatial location The environmental data collected includes multi-dimensional information such as water temperature, dissolved oxygen concentration, and salinity. g represents the total number of data sequences after spatiotemporal alignment, and the decision suggestion vector is defined as: in, This indicates the generation of the first [unit / item] based on intelligent analysis and prediction models. Management or control recommendations, such as feed distribution, aeration device start-up and shutdown strategies, and instructions for adjusting cage layout, etc. Indicates shared ownership Recommendations for management or regulation.

[0067] Furthermore, during the rendering process, WebGL technology converts the geographic information of the marine ranch into a set of three-dimensional mesh vertices: in For the first The three-dimensional coordinates of each grid point are consistent with the geographical boundaries of the pasture, seabed topography, and water profile data. This represents the total number of grid points.

[0068] Furthermore, the visualization of environmental data relies on color mapping functions. ,in D represents the environmental data value at a certain spatiotemporal sampling point. This corresponds to the color vector. (Function) The structure follows the principle of continuous monotonic mapping. When the data value is in a low range, it is mapped to a cool color tone; as the data value gradually increases, the color gradually transitions to a warm color tone, thus providing a visual representation of the water temperature gradient, dissolved oxygen concentration levels, and salinity distribution. The dynamic representation of fish school distribution is based on a particle system method. The position of each particle is obtained by combining multi-view observation data from UAVs with a three-dimensional interpolation algorithm, ensuring that the spatial distribution of particles is consistent with the actual state of the fish school. The particle motion state is calculated by the fluid dynamics equations, the basic form of which is: in, The velocity of a fluid element as a function of time is called acceleration. For velocity vector, For local water pressure, The viscosity coefficient of the fluid. The resultant force is composed of environmental disturbances and the self-driving force of the fish school, with t representing time. This equation enables the updating of fish particle velocity and trajectory evolution, thus creating a dynamically changing fish school movement effect over time in a 3D twin model. Finally, the decision suggestion vector R and the environmental data visualization results are overlaid and rendered in the same 3D scene. Managers can intuitively observe the correspondence between environmental status, fish distribution, and decision execution effects in the interactive interface, thereby achieving visualized management and optimization decision support for the marine ranch's operational status.

[0069] The decision recommendations are simultaneously sent to the data visualization interface and the equipment control system; the data visualization interface integrates the spatiotemporally aligned data sequence with the decision recommendations, and renders and displays the ocean seabed topography, sensor spatial location, fish distribution and environmental data in real time in the three-dimensional digital twin model.

[0070] Optionally, in another exemplary embodiment of this application, after step 103, the method further includes step 601: Step 601: Send the aquaculture decision suggestion to the equipment control system so that the equipment control system generates control commands based on the aquaculture decision suggestion.

[0071] The equipment control system generates control commands based on the received decision suggestions and automatically sends them to the corresponding execution devices in the marine ranch. The execution devices then perform the corresponding aquaculture operations according to the control commands.

[0072] In this step, the specific implementation process of the data-device linkage control protocol is as follows: First, the system receives the decision suggestion vector output from the previous stage: in, This indicates the generation of the first [unit / item] based on intelligent analysis and prediction models. This is a management or regulation suggestion, or is considered to be the first The protocol analyzes the decision recommendation value based on the risk level or operational dimension. It parses the operation type item by item from this vector and maps it to the corresponding control instruction code Cmd, directly binding the decision result to the device's execution action. The protocol then constructs a control instruction frame, which consists of a frame header, device address, instruction code, parameter list, and checksum. The frame header is fixed at hexadecimal value 0xAA55, used by the device to quickly identify the start of the data frame; the device address (Addr) is a 4-byte unsigned integer used to uniquely identify the target execution device; the instruction code (Cmd) is a 2-byte integer, maintaining a one-to-one mapping with the operation type corresponding to each element in the decision recommendation vector, ensuring that the decision result can be unambiguously converted into a device action instruction; the parameter list (Params) is dynamically generated according to operational requirements. For example, when the operation type is water aeration, the protocol will assign the risk level value... The formula for converting this to an aerator power value is as follows: in, The final power parameters will be issued. This is a preset proportional coefficient used to achieve dimensional mapping between water quality risk level values ​​and power regulation amounts. This corresponds to the water quality risk level value in the decision-making recommendations. After generating the parameter list, to ensure data transmission integrity and anti-interference capabilities, the protocol uses the CRC-8 algorithm to verify and calculate the concatenated sequence of the device address Addr, instruction code Cmd, and parameter list Params. The formula is: in This is a binary concatenation operation; the checksum is a single-byte result appended to the end of the frame as a verification field. Thus, the complete control command frame format is expressed as follows: Head=0xAA55 ensures that the receiving end can accurately distinguish the data frame boundaries. Finally, the constructed complete instruction frame is sent to the corresponding execution device via the communication interface. The device drives the action based on the parsing result, completes the physical execution of the decision suggestion, and thus realizes a data-device linkage closed loop.

[0073] Optionally, see Figure 11 In another exemplary embodiment of this application, step 103 includes fuzzification, fuzzy rule reasoning, and defuzzification in sequence.

[0074] The fuzzification process relies on a fuzzy logic decision rule base. This base transforms the standardized feature vectors (humidity gradient, dissolved oxygen frequency, and feeding peak) output from the feature layer into risk levels and corresponding decision recommendations. In other words, the standardized feature vectors output from the feature layer serve as input for the fuzzification process. Furthermore, the standardized feature vectors can be expressed as: in The j-th feature value is represented by the feature value of the sea surface temperature gradient, the feature value of the frequency of dissolved oxygen concentration fluctuation, and the feature value of the peak feeding period, etc., and m represents the total number of feature values.

[0075] The blurring process is as follows: For each feature Define fuzzy sets and use membership function To represent this feature Belongs to fuzzy set The degree of each fuzzy element in the model ranges from [0,1]. The closer the value is to 1, the more the feature matches the semantic description of that fuzzy element.

[0076] The fuzzy rule reasoning is as follows: The pre-defined fuzzy decision rule base includes multiple fuzzy rules, referred to as IF-THEN rules. Each fuzzy rule takes the following form: in, i This represents the i-th fuzzy rule. In the i-th fuzzy rule, the feature is... The corresponding fuzzy set, This represents the risk level mapped by the i-th fuzzy rule.

[0077] For the input vector V, the activation degree of each fuzzy rule is calculated using a product inference engine, with the following formula: in Activation, used to represent the input vector The degree of matching with the i-th fuzzy rule. Let represent the k-th input vector. If all features highly conform to the corresponding fuzzy set, then The value is relatively large.

[0078] Deblurring process: After calculating the activation values ​​of all fuzzy rules, a weighted average method is used for defuzzification, and the comprehensive risk level value Y is output: in The total number of fuzzy rules, Let Y be the risk level value corresponding to the i-th fuzzy rule. The numerator represents the comprehensive result of the risk level of each fuzzy rule under the weighting of its activation degree, and the denominator is the total activation degree. The result is then normalized. This output Y directly corresponds to the risk level under the current multi-feature combination, thereby triggering corresponding decision suggestions. This realizes the integration of multi-source environmental features and aquaculture behavior features into a unified risk assessment result.

[0079] Furthermore, during the fuzzification process, a Gaussian membership function can be used to perform fuzzy mapping on the input features, and its form is as follows: Where v represents the input feature value, Indicates the input value to the fuzzy set membership degree For fuzzy sets The central value of determines the peak position of the membership function. The width parameter determines the smoothness and coverage of the Gaussian function. For example, considering the frequency characteristics of dissolved oxygen fluctuations, three fuzzy sets—low-frequency, mid-frequency, and high-frequency—are defined, with their center values ​​being, respectively, [values ​​to be filled in]. , , By setting appropriate By maintaining partial overlap among the three fuzzy sets, a smooth division of the continuous frequency range is achieved, ensuring a natural transition between low, medium, and high frequencies in the fuzzified input. For the fish school activity state characteristics, two fuzzy sets, "normal" and "surfacing," are defined, with their center values ​​being... and The input feature values ​​are obtained through standardized processing of fish behavior monitoring data. Normal behavior corresponds to low-level values, while fish surfacing behavior corresponds to high-level values. A Gaussian membership function is used to achieve a smooth fuzzy mapping from normal to surfacing behavior, avoiding abrupt classification. For the feeding amount feature, three fuzzy sets are defined: low-peak period, stable period, and peak period, with their center values ​​being... , , The input feature values ​​are derived from the normalized results of the actual baiting amount. Three fuzzy sets are overlapped by setting appropriate width parameters, ensuring a continuous membership distribution as the baiting amount changes from low to high, thus avoiding misjudgments caused by boundary effects. Through the definition of fuzzy sets and the mapping mechanism of Gaussian functions, the input features are transformed into continuous and comparable membership values ​​during the fuzzification stage, providing accurate and smooth input support for the subsequent execution of fuzzy inference rules.

[0080] Based on the above, this application has the following key points: Multi-source heterogeneous data fusion architecture: This invention constructs a complete multi-source heterogeneous data fusion architecture for marine ranches, covering the fusion processing of data layer, feature layer and decision layer, realizing the efficient transformation from raw data to decision information. This unique architecture is the foundation for realizing deep data fusion and intelligent decision-making in marine ranches. It is different from the traditional simple data aggregation method, and can fully explore the complex correlation and complementary information between multi-source data, improving the comprehensiveness and accuracy of data processing.

[0081] Data layer fusion processing (preprocessing): In data layer fusion processing, an outlier cleaning algorithm based on the 3σ principle and a spatiotemporal alignment algorithm are adopted. The outlier cleaning algorithm based on the 3σ principle can accurately identify and remove faulty data and outliers in underwater sensor data, while using linear interpolation to fill in missing positions, ensuring the integrity and reliability of the data sequence. The spatiotemporal alignment algorithm aligns data from different time scales to a unified high-frequency time series through time interpolation, and maps all data to a unified three-dimensional geographic coordinate system of the marine ranch, solving the inconsistency problem of multi-source data in spatiotemporal reference, and providing a unified spatiotemporal framework for subsequent data processing and analysis.

[0082] Feature layer fusion processing: In feature layer fusion processing, representative features are extracted from satellite remote sensing data, underwater sensor data, and aquaculture log data. For example, sea surface temperature gradient features are extracted to reflect local thermodynamic differences in the sea area; discrete Fourier transform and power spectral density analysis are used to extract dissolved oxygen concentration fluctuation frequency features; and a sliding window peak detection algorithm is employed to extract peak feeding periods. These feature extraction algorithms effectively mine key information from the data, generating standardized feature vectors for eco-aquaculture collaborative decision-making, providing crucial data support for the decision-making level. Compared with traditional single feature extraction methods, this approach more comprehensively reflects the ecological and aquaculture status of marine ranches.

[0083] Decision-making layer fusion processing: The decision-making layer employs a fuzzy logic-based decision rule base for fusion calculation. By defining fuzzy sets and membership functions, standardized feature vectors are transformed into risk levels and corresponding decision recommendations. A product inference engine is used to calculate rule activations, followed by defuzzification using a weighted average method to output a comprehensive risk level value. This decision-making model fully considers the uncertainty and fuzziness of multi-source data features, enabling accurate judgment and scientific decision-making regarding complex situations in marine ranches. Compared to traditional deterministic decision-making models, it exhibits greater adaptability and flexibility.

[0084] Decision Execution and Visualization (Data-Equipment Linkage and Visualization): This feature enables the linkage between decision results and equipment control. By constructing a data-equipment linkage control protocol, decision suggestions are transformed into executable control commands for the equipment, achieving automated control of marine ranching equipment. Simultaneously, data visualization is achieved using 3D digital twin models and WebGL technology. This integrates spatiotemporally aligned data sequences with decision suggestions, providing managers with intuitive and comprehensive information on the operational status of the marine ranch. This facilitates real-time monitoring and decision adjustments, enhancing the intelligence and visualization level of marine ranch management.

[0085] Furthermore, this application has the following effects: Compared with traditional marine ranching data processing and decision-making technologies, this invention has significant advantages, effectively solving many problems in existing technologies and providing strong support for the scientific management and sustainable development of marine ranches.

[0086] In terms of data processing precision, traditional methods often fall short in handling multi-source heterogeneous data. Simple data aggregation and analysis methods merely superficially integrate various data types, failing to delve into the underlying information or handle noise and outliers, thus significantly compromising data accuracy and reliability. This invention, however, employs an outlier cleaning algorithm based on the 3σ principle, accurately identifying and removing faulty data and outliers from underwater sensor data, effectively improving data quality. Simultaneously, the spatiotemporal alignment algorithm successfully resolves the inconsistency in time and spatial scales between different data sources, enabling data fusion and analysis within a unified spatiotemporal framework. Through the application of these advanced algorithms, this invention greatly improves data processing precision, providing a solid and reliable data foundation for subsequent decision analysis, and enabling a more accurate understanding of the actual conditions of marine ranches.

[0087] From the perspective of decision-making accuracy and timeliness, traditional experience-based models have significant limitations. Because experience-based models are often built upon limited past experience and data, they struggle to comprehensively cover the complex and ever-changing conditions of marine ranches. When faced with emergencies such as sudden marine disasters or aquaculture diseases, traditional methods cannot quickly and accurately assess the severity of the problem and provide effective decision-making suggestions, leading to delayed decisions, missed optimal response opportunities, and severe economic losses to marine ranches. This invention, by performing hierarchical fusion processing of multi-source data, fully explores the potential connections and patterns between the data. The decision rule base based on fuzzy logic can comprehensively consider the uncertainty and fuzziness of multiple factors, accurately identify and assess various risk factors existing in marine ranches, and provide timely and scientifically sound decision-making suggestions. This intelligent decision-making approach greatly improves the accuracy and timeliness of decisions, enhances the ability of marine ranches to cope with emergencies, and effectively reduces losses caused by risks.

[0088] In terms of resource utilization and management efficiency, traditional marine ranching management methods mainly rely on manual experience and simple data recording, resulting in extensive management and difficulty in optimizing resource allocation. For example, in feeding management, the lack of accurate data may lead to overfeeding or underfeeding, wasting resources and affecting the growth of farmed organisms. This invention, however, enables precise management of aquaculture activities based on the real-time status and data analysis results of the marine ranch. By rationally adjusting stocking density and optimizing feeding programs, it improves the utilization efficiency of aquaculture resources, reduces aquaculture costs, and minimizes negative environmental impacts. This invention also achieves automated control and remote management of marine ranch equipment, reducing manual intervention, improving management efficiency, and making marine ranch operations more efficient and intelligent.

[0089] From the perspective of promoting the sustainable development of marine ranches, traditional technologies are relatively weak in terms of ecological environmental protection. As the scale of marine ranches continues to expand, traditional methods struggle to monitor changes in the marine environment in real time, leading to insufficient timely detection and handling of ecological and environmental problems, which can easily cause damage to the marine ecosystem. This invention places great emphasis on the protection and monitoring of the marine ranch's ecological environment. By collecting and analyzing marine environmental data in real time, it can promptly identify ecological and environmental problems and take corresponding measures for protection and restoration. Through scientific decision-making and management, it achieves the rational utilization of marine ranch resources, protects marine biodiversity, maintains marine ecological balance, and provides a strong guarantee for the long-term stable development of marine ranches.

[0090] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram can be found in [reference needed]. Figure 12 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores relevant data for marine ranching decision-making methods. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it can implement a marine ranching decision-making method.

[0091] Those skilled in the art will understand, see Figure 12 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0092] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0093] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0094] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0095] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data that have been agreed to by the user or have been fully agreed to by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0096] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. In the embodiments provided in this application, any reference to memory, database, or other media can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0097] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units, etc., and are not limited to these.

[0098] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0099] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for determining marine ranching aquaculture decisions, characterized in that, include: Acquire multi-source heterogeneous data for marine ranches, including sensor data, remote sensing data, and log data; Feature vectors are obtained by performing feature extraction on the sensor data, the remote sensing data, and the log data, respectively, and are denoted as sensor feature vector, remote sensing feature vector, and log feature vector. The sensor feature vector, the remote sensing feature vector, and the log feature vector are sent to the fuzzy logic model so that the fuzzy logic model can output aquaculture decision suggestions. The sensor data includes a data sequence of dissolved oxygen concentration, and the process of extracting the sensor feature vector includes: determining the main frequency feature of dissolved oxygen concentration fluctuation based on the dissolved oxygen concentration data sequence, and determining the main frequency feature as the sensor feature vector; The remote sensing data includes a sea surface temperature data sequence, and the process of extracting the remote sensing feature vector includes: determining the temperature gradient magnitude feature based on the sea surface temperature data sequence, and determining the temperature gradient magnitude feature as the remote sensing feature vector. The log data includes a sequence of feeding amount data. The process of extracting the log feature vector includes: determining the peak point of feeding amount based on the feeding amount data sequence to obtain the peak period feature sequence; and determining the peak period feature sequence as the log feature vector.

2. The marine ranching decision-making method according to claim 1, characterized in that, The sensor data includes a data sequence of dissolved oxygen concentration, and the sensor feature vector is extracted using the following method: For any given time period, obtain the data sequence of the dissolved oxygen concentration within that time period; Within the time period, based on the data sequence of dissolved oxygen concentration, the dominant frequency feature of dissolved oxygen concentration fluctuation is determined, and the dominant frequency feature is determined as the sensor feature vector within the time period.

3. The marine ranching decision-making method according to claim 1, characterized in that, The remote sensing data includes a sea surface temperature data sequence, and the remote sensing feature vector is extracted using the following method: For any given time period, obtain the sea surface temperature data sequence; Within the time period, the gradient vector of the sea surface temperature field is determined based on the sea surface temperature data sequence. The sea surface temperature gradient vector is converted into a temperature gradient magnitude feature, and the temperature gradient magnitude feature is determined as the remote sensing feature vector.

4. The marine ranching aquaculture decision-making method according to claim 1, characterized in that, The log data includes a sequence of feeding amounts, and the log feature vector is extracted using the following method: For any given time period, obtain the sequence of feed amount data; Within the time period, based on the feeding amount data sequence, the peak points of each feeding amount within the time period are determined to obtain the peak period feature sequence, and the peak period feature sequence is determined as the log feature vector.

5. The marine ranching decision-making method according to claim 1, characterized in that, After sending the sensor feature vector, the remote sensing feature vector, and the log feature vector to the fuzzy logic model so that the fuzzy logic model outputs aquaculture decision suggestions, the method further includes: The dynamic characteristics of marine organisms corresponding to the aquaculture decision suggestions are obtained from the pre-stored correspondence between aquaculture decision suggestions and the dynamic characteristics of marine organisms. Obtain environmental data of the marine ranch; The dynamic features and environmental data are fused into the static three-dimensional marine ranch model to obtain a dynamic three-dimensional marine model.

6. The marine ranching aquaculture decision-making method according to claim 5, characterized in that, After sending the sensor feature vector, the remote sensing feature vector, and the log feature vector to the fuzzy logic model so that the fuzzy logic model outputs aquaculture decision suggestions, the method further includes: The aquaculture decision suggestions are sent to the equipment control system so that the equipment control system can generate control commands based on the aquaculture decision suggestions.

7. The marine ranching aquaculture decision-making method according to any one of claims 1-6, characterized in that, After acquiring multi-source heterogeneous data for marine ranches, the method includes: preprocessing the sensor data and the remote sensing data, wherein the preprocessing includes removing outliers from the sensor data and supplementing missing values ​​from the sensor data; the preprocessing also includes spatiotemporally aligning the remote sensing data and the sensor data to obtain spatiotemporally aligned data of the remote sensing data and the sensor data.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the marine ranching decision-making method according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the marine ranching aquaculture decision-making method according to any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the marine ranching aquaculture decision-making method according to any one of claims 1-7.

Images