Automated control method for intelligent farming of terrestrial sea cucumbers based on AI-assisted decision-making

Abstract

The present invention relates to an automated control method for intelligent breeding of terrestrial sea cucumbers based on AI-assisted decision-making, comprising the following steps: Step 1: Acquire monitoring water quality and environmental data and sea cucumber images, and upload them to the cloud; Step 2: Establish a relationship model between water quality and environmental data and the growth status of sea cucumbers, and predict the growth trend of sea cucumbers based on the relationship model and real-time data; Step 3: Detect the health status of sea cucumbers based on a health status detection model; Step 4: Formulate the best breeding plan based on the predicted growth trend and health status of sea cucumbers; and automatically adjust the breeding facilities based on the best breeding plan; Step 5: Build a digital twin model of the breeding system, monitor the actual breeding status in real time, establish a feedback mechanism, and adjust the relationship model and the best breeding plan according to the system execution effect. The present invention not only improves the decision-making efficiency and accuracy of sea cucumber breeding, but also optimizes resource allocation and reduces production costs.

Description

Technical Field

[0001] The present invention relates to the field of intelligent farming, and in particular to an automated control method for intelligent farming of terrestrial sea cucumbers based on AI-assisted decision-making. Background Art

[0002] Land-based sea cucumber farming is a challenging industry that involves a variety of complex environmental factors and biological characteristics. Traditional farming methods rely on manual monitoring and management, which not only consumes a lot of human resources, but also is difficult to achieve refined management due to limitations in data collection and processing, often leading to waste of resources and environmental pollution. In addition, sea cucumbers are a species that is sensitive to environmental changes, and their growth and health status are affected by multiple parameters such as water temperature, salinity, pH, and dissolved oxygen levels. Therefore, achieving efficient and precise environmental control is crucial to improving sea cucumber production and quality. Summary of the invention

[0003] In order to solve the above problems, the purpose of the present invention is to provide an automated control method for intelligent farming of terrestrial sea cucumbers based on AI-assisted decision-making, which not only improves the decision-making efficiency and accuracy of sea cucumber farming, but also optimizes resource allocation and reduces production costs.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] An automated control method for intelligent farming of terrestrial sea cucumbers based on AI-assisted decision-making, comprising the following steps:

[0006] Step 1: Deploy various sensors in the aquaculture environment to monitor water quality and environmental data in real time, obtain sea cucumber images through image acquisition devices, and use IoT technology to upload water quality data, environmental data and sea cucumber image data to the cloud;

[0007] Step 2: Based on the machine learning algorithm, according to the collected water quality data, environmental data and sea cucumber image data, the sea cucumber growth characteristic data is extracted from the sea cucumber image based on the image feature extraction algorithm, and a relationship model between the water quality and environmental data and the growth status of the sea cucumber is established. Based on the relationship model and real-time data, the growth trend of the sea cucumber is predicted;

[0008] Step 3: Detect the health status of the sea cucumber based on the health status detection model;

[0009] Step 4: Develop the best breeding plan based on the predicted growth trend and health status of the sea cucumbers; and automatically adjust the breeding facilities based on the best breeding plan, specifically:

[0010] The goal of the optimal breeding program is to maximize the growth rate and health status of sea cucumbers while minimizing the cost of bait and environmental regulation. The objective function is:

[0011] ;

[0012] in, is the growth rate; for health status; and is the parameter that adjusts the weight of the optimization objective, Represents the cost of control and resource investment; It is a combination of parameters, including bait release amount F, temperature T, and water quality W;

[0013] Constraints include that the amount of bait cannot exceed the ecological carrying capacity and the temperature range is limited;

[0014] Use simulated annealing to solve the above objective function:

[0015] ① Define the initial parameter combination and initial temperature, set the cooling factor α (usually between 0.8 and 0.99) and the stop condition;

[0016] ② For the current status , generate a new state of the neighborhood , and calculate the objective function difference ;

[0017] ③If , accept the new status ,like , then with probability Accept the new status ;

[0018] ④Use the cooling formula T=α×T;

[0019] ⑤Terminate when the temperature drops to Tmin or the maximum number of iterations is reached;

[0020] ⑥ Apply the final parameters to the breeding system and adjust them through automated equipment;

[0021] Step 5: Build a digital twin model of the breeding system, monitor the actual breeding conditions in real time, establish a feedback mechanism, and adjust the relationship model and optimal breeding plan based on the system execution effect.

[0022] Further, step 1 is as follows:

[0023] The various sensors include temperature sensors, humidity sensors, pH sensors, dissolved oxygen sensors, turbidity sensors, light sensors and salinity sensors; the temperature sensors are distributed at different depths and positions in the aquaculture pond, and the dissolved oxygen sensors use optical dissolved oxygen sensors to measure the oxygen content of the water body and are placed in different areas of the aquaculture pond;

[0024] The image acquisition device includes an underwater camera and a surface camera, supporting high-definition video and low-light environment image acquisition;

[0025] The sensors and image acquisition devices are connected to an integrated IoT gateway, which is responsible for collecting data from all devices and uploading water quality data, environmental data and sea cucumber image data to the cloud using Internet of Things (IoT) technology.

[0026] Furthermore, step 2 is specifically as follows:

[0027] Step 2.1: Acquire water quality data, environmental data and sea cucumber image data, wherein the water quality data includes pH value, dissolved oxygen, turbidity and salinity; the environmental data includes temperature and light intensity; and acquire sea cucumber growth characteristic data based on the sea cucumber image data;

[0028] Step 2.2: Process the data to remove missing values ​​and outliers, and then standardize or normalize the data to ensure that the dimensions of different features are consistent;

[0029] Step 2.3: Use correlation analysis to select features that are highly correlated with the growth status of sea cucumbers, and apply recursive feature elimination algorithm to further screen important features to obtain the final training feature set;

[0030] Step 2.4: Build a relationship model based on the support vector regression model, and train based on the final training feature set, adjust the error limit, regularization parameter and kernel parameter to obtain a fit that meets the preset requirements and obtain the final relationship model.

[0031] Further, step 2.3 is as follows:

[0032] Step 2.3.1: Calculate all water quality and environmental characteristics Growth characteristics of sea cucumbers Pearson correlation coefficient between ;

[0033] For variables and , and its Pearson correlation coefficient ρ is defined as:

[0034] ;

[0035] in, and The variables are and The mean of ; U is the number of samples; For variables The u-th sample of For variables The u-th sample of

[0036] Step 2.3.2: Select a threshold based on the absolute value of the correlation coefficient, filter out features with absolute values ​​higher than the threshold, and obtain a filtered feature set;

[0037] Step 2.3.3: Use RFE combined with the base learner for evaluation on the filtered feature set;

[0038] Step 2.3.4: Eliminate the least important features in each iteration and record the model performance of each iteration;

[0039] Step 2.3.5: Use cross-validation to evaluate the impact of feature sets on model performance after each iteration;

[0040] Step 2.3.6: Finally, select the feature set that performs best on the evaluation metric for training the model.

[0041] Further, step 2.4 is as follows:

[0042] The SVR model is as follows:

[0043] ;

[0044] Where n is the number of feature vectors; i represents the index;

[0045] The constraints are:

[0046]

[0047] Among them, w, b are the weight and bias of the model; C is the regularization parameter, is the precision parameter; , is the slack variable; x i is the ith input feature vector, is the corresponding true value;

[0048] Based on the final training feature set training, k-fold cross validation is used to evaluate the SVR model performance, and grid search is combined to adjust the error limit, regularization parameter and kernel parameter to obtain a fit that meets the preset requirements and obtain the final relationship model.

[0049] Furthermore, the kernel function of the SVR model adopts the RBF kernel function:

[0050] ;

[0051] in, is the kernel parameter; x j is the jth input feature vector;

[0052] The dual problem of SVR is solved based on the kernel function. The original problem is converted into a dual form based on the Cramérod condition by the Lagrange multiplier method:

[0053] First, we introduce the Lagrange multiplier To construct the Lagrangian function:

[0054] ;

[0055] in, is the error variable;

[0056] According to the KKT condition, the optimal solution must meet the following conditions:

[0057] ① Gradient conditions:

[0058] ;

[0059] ②Original constraints:

[0060] ;

[0061] ③Complementary relaxation conditions:

[0062] ;

[0063] By analyzing the KKT conditions and eliminating the original variables, we get the dual problem:

[0064] ;in, , is the Lagrangian function;

[0065] Constraints:

[0066] .

[0067] Furthermore, the sea cucumber growth characteristic data is extracted from the sea cucumber image based on the image feature extraction algorithm, as follows:

[0068] Acquire sea cucumber images based on an image acquisition device and perform preprocessing, including removing water ripples and other random noise in the image, adjusting white balance and color, and restoring the color of the real growth environment;

[0069] Extracting the biological characteristics of sea cucumbers, including shape characteristics, color characteristics and texture characteristics, based on the preprocessed sea cucumber images;

[0070] The shape features include length, width, and surface concavity. The Canny edge detection algorithm is used to locate the edge of the sea cucumber, and the findContours function of OpenCV is used to identify the external contour and important feature points of the sea cucumber, and to obtain the length, width, and surface concavity; color features are identified through color histogram analysis; and texture features are extracted using the grayscale co-occurrence matrix method.

[0071] Furthermore, the health status detection model is constructed using random forest, as follows;

[0072] Acquire a data set related to the health status of sea cucumbers to construct an original training data set, where the original training data set includes biological characteristics, environmental characteristics, and annotated health status;

[0073] Randomly extract samples with replacement from the original training data set to form multiple sub-sample sets. Each sub-sample set is used to train a decision tree, and the unselected data is regarded as out-of-bag data.

[0074] For each tree, when splitting a node, a subset is randomly selected from all features, and the best split point is selected in this subset for splitting;

[0075] The tree building process continues until the stopping criterion is reached, and a random forest model with N trees is obtained;

[0076] For an input sample x, each tree Given a prediction, the final category is determined by voting, and the category with the most votes is selected, i.e. the health status :

[0077] ;

[0078] in, is an indicator function, indicating Whether the tree predicts category k;

[0079] The trained random forest model is validated based on out-of-bag data. The prediction results of out-of-bag data samples are compared with their true labels, and the out-of-bag error rate is calculated to evaluate the generalization ability of random forest.

[0080] Furthermore, 5 is specifically:

[0081] Use 3D modeling software (such as Unity 3D or Blender) to create a simulation of the sea cucumber farming environment and connect sensors and image acquisition devices to update data in real time;

[0082] Use real-time data streams to drive the simulator, providing mapping and simulation of virtual and actual environments;

[0083] Integrate simulation predictions and real-time data and feed them back to the operator interface as the basis for adjustment plans and equipment control;

[0084] Based on real-time data, the relationship model and the optimal breeding plan are adjusted in real time through reinforcement learning.

[0085] The present invention has the following beneficial effects:

[0086] 1. The present invention not only improves the decision-making efficiency and accuracy of sea cucumber farming, but also optimizes resource allocation and reduces production costs;

[0087] 2. The present invention uses a machine learning algorithm to accurately establish a complex relationship model between water quality, environmental conditions and sea cucumber health and growth, which can efficiently predict the growth trend of sea cucumbers, and detect the health status of sea cucumbers based on the health status detection model, and formulate the best breeding plan based on the predicted growth trend and health status of sea cucumbers, reduce unnecessary resource input, improve bait utilization and water resource utilization efficiency, and effectively improve the survival rate and quality of sea cucumber breeding;

[0088] 3. The present invention combines the simulated annealing optimization method to significantly improve the efficiency and sustainable management capabilities of the sea cucumber farming system without increasing the complexity, which can not only promote scientific decision-making, but also improve the success rate and economic benefits of actual production. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 is a flow chart of the method of the present invention;

[0090] Figure 2 The figure is a schematic diagram of a data processing flow in one embodiment of the present invention. DETAILED DESCRIPTION

[0091] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0092] refer to Figure 1 In this embodiment, an automated control method for intelligent farming of sea cucumbers on land based on AI-assisted decision-making is provided, comprising the following steps:

[0093] Step 1: Deploy various sensors in the aquaculture environment to monitor water quality and environmental data in real time, obtain sea cucumber images through image acquisition devices, and use IoT technology to upload water quality data, environmental data and sea cucumber image data to the cloud;

[0094] Step 2: Based on the machine learning algorithm, according to the collected water quality data, environmental data and sea cucumber image data, the sea cucumber growth characteristic data is extracted from the sea cucumber image based on the image feature extraction algorithm, and a relationship model between the water quality and environmental data and the growth status of the sea cucumber is established. Based on the relationship model and real-time data, the growth trend of the sea cucumber is predicted;

[0095] Step 3: Detect the health status of the sea cucumber based on the health status detection model;

[0096] Step 4: Develop the best breeding plan based on the predicted growth trend and health status of the sea cucumbers; and automatically adjust the breeding facilities based on the best breeding plan;

[0097] Step 5: Build a digital twin model of the breeding system, monitor the actual breeding conditions in real time, establish a feedback mechanism, and adjust the relationship model and optimal breeding plan based on the system execution effect.

[0098] In this embodiment, step 1 is as follows:

[0099] The various sensors include temperature sensors, humidity sensors, pH sensors, dissolved oxygen sensors, turbidity sensors, light sensors and salinity sensors; the temperature sensors are distributed at different depths and positions in the culture pond, and the dissolved oxygen sensors use optical dissolved oxygen sensors to measure the oxygen content of the water body and are placed in different areas of the culture pond;

[0100] The image acquisition device includes an underwater camera and a surface camera, supporting high-definition video and low-light environment image acquisition;

[0101] The sensors and image acquisition devices are connected to an integrated IoT gateway, which is responsible for collecting data from all devices and uploading water quality data, environmental data and sea cucumber image data to the cloud using Internet of Things (IoT) technology.

[0102] In this embodiment, step 2 is specifically as follows:

[0103] Step 2.1: Obtain historical water quality data, environmental data and sea cucumber image data, wherein the water quality data includes pH value, dissolved oxygen, turbidity and salinity; the environmental data includes temperature and light intensity; and obtain sea cucumber growth characteristic data based on the sea cucumber image data;

[0104] Step 2.2: Process the data to remove missing values ​​and outliers, and then standardize or normalize the data to ensure that the dimensions of different features are consistent;

[0105] Step 2.3: Use correlation analysis to select features that are highly correlated with the growth status of sea cucumbers, and apply recursive feature elimination algorithm to further screen important features to obtain the final training feature set;

[0106] Step 2.4: Build a relationship model based on the support vector regression model, and train based on the final training feature set, adjust the error limit, regularization parameter and kernel parameter to obtain a fit that meets the preset requirements and obtain the final relationship model.

[0107] In this embodiment, step 2.3 is as follows:

[0108] Step 2.3.1: Calculate all water quality and environmental characteristics Growth characteristics of sea cucumbers Pearson correlation coefficient between ;

[0109] For variables and , and its Pearson correlation coefficient ρ is defined as:

[0110] ;

[0111] in, and The variables are and The mean of ; U is the number of samples; For variables The u-th sample of For variables The u-th sample of

[0112] Step 2.3.2: Select a threshold based on the absolute value of the correlation coefficient, filter out features with absolute values ​​higher than the threshold, and obtain a filtered feature set;

[0113] Step 2.3.3: Use RFE combined with the base learner for evaluation on the filtered feature set;

[0114] Step 2.3.4: Eliminate the least important features in each iteration and record the model performance of each iteration;

[0115] Step 2.3.5: Use cross-validation to evaluate the impact of feature sets on model performance after each iteration;

[0116] Step 2.3.6: Finally, select the feature set that performs best on the evaluation metric for training the model.

[0117] In this embodiment, step 2.4 is as follows:

[0118] The SVR model is as follows:

[0119] ;

[0120] Where n is the number of feature vectors; i represents the index;

[0121] The constraints are:

[0122]

[0123] Among them, w, b are the weight and bias of the model; C is the regularization parameter, is the precision parameter; , is the slack variable; x i is the ith input feature vector, is the corresponding true value;

[0124] Based on the final training feature set training, k-fold cross validation is used to evaluate the SVR model performance, and grid search is combined to adjust the error limit, regularization parameter and kernel parameter to obtain a fit that meets the preset requirements and obtain the final relationship model.

[0125] In this embodiment, the SVR model kernel function adopts the RBF kernel function:

[0126] ;

[0127] in, is the kernel parameter; x j is the jth input feature vector;

[0128] The dual problem of SVR is solved based on the kernel function. The original problem is converted into a dual form based on the Cramérod condition by the Lagrange multiplier method:

[0129] First, we introduce the Lagrange multiplier To construct the Lagrangian function:

[0130] ;

[0131] in, is the error variable;

[0132] According to the KKT condition, the optimal solution must meet the following conditions:

[0133] ① Gradient conditions:

[0134] ;

[0135] ②Original constraints:

[0136] ;

[0137] ③Complementary relaxation conditions:

[0138] ;

[0139] By analyzing the KKT conditions and eliminating the original variables, we get the dual problem:

[0140]

[0141] in, , is the Lagrangian function;

[0142] Constraints:

[0143] .

[0144] In the SVR model, the kernel function The introduction of can be optimized in high-dimensional space without explicitly calculating these high-dimensional features, which not only simplifies the calculation but also enhances the model's ability to handle nonlinear data.

[0145] In this embodiment, the sea cucumber growth characteristic data is extracted from the sea cucumber image based on the image feature extraction algorithm, as follows:

[0146] Acquire sea cucumber images based on an image acquisition device and perform preprocessing, including removing water ripples and other random noise in the image, adjusting white balance and color, and restoring the color of the real growth environment;

[0147] Extracting the biological characteristics of sea cucumbers, including shape characteristics, color characteristics and texture characteristics, based on the preprocessed sea cucumber images;

[0148] The shape features include length, width, and surface concavity. The Canny edge detection algorithm is used to locate the edge of the sea cucumber, and the findContours function of OpenCV is used to identify the external contour and important feature points of the sea cucumber, and to obtain the length, width, and surface concavity; color features are identified through color histogram analysis; and texture features are extracted using the grayscale co-occurrence matrix method.

[0149] In this embodiment, the health status detection model is constructed using random forest, as follows;

[0150] Obtaining data sets related to the health status of sea cucumbers to construct an original training data set, which includes biological characteristics (shape characteristics, color characteristics, and texture characteristics), environmental characteristics (such as water temperature and salinity), and annotated health status (healthy or unhealthy);

[0151] Samples are randomly selected from the original training data set with replacement to form multiple sub-sample sets. Each sub-sample set is used to train a decision tree, and the unselected data is used as out-of-bag data. For each tree, when a node is split, a subset is randomly selected from all features, and the best split point is selected in this subset for splitting.

[0152] The tree building process continues until the stopping criterion (e.g., maximum depth, or minimum number of leaf node samples) is reached, and a random forest model with N trees is obtained;

[0153] For an input sample x, each tree Given a prediction, the final category is determined by voting, and the category with the most votes is selected, i.e. the health status :

[0154] ;

[0155] in, is an indicator function, indicating Whether the tree predicts category k;

[0156] The trained random forest model is validated based on out-of-bag data. The prediction results of out-of-bag data samples are compared with their true labels, and the out-of-bag error rate is calculated to evaluate the generalization ability of random forest.

[0157] In this embodiment, step 4 is specifically as follows:

[0158] The control goal is to maximize the growth rate and health status of sea cucumbers while minimizing the cost of bait and environmental regulation. The objective function is:

[0159] ;

[0160] in, is the growth rate; for health status; and is the parameter that adjusts the weight of the optimization objective, Represents the cost of control and resource investment; It is a combination of parameters, including bait release amount F, temperature T, and water quality W parameters;

[0161] Constraints include that the amount of bait cannot exceed the ecological carrying capacity and the temperature range is limited;

[0162] Use simulated annealing to solve the above objective function:

[0163] ① Define the initial parameter combination and initial temperature, set the cooling factor α (usually between 0.8 and 0.99) and the stop condition;

[0164] ② For the current status , generate a new state of the neighborhood , and calculate the objective function value difference ;

[0165] ③If , accept the new status ,like , then with probability Accept the new status ;

[0166] ④Use the cooling formula T=α×T;

[0167] ⑤Terminate when the temperature drops to Tmin or the maximum number of iterations is reached;

[0168] ⑥ Apply the final parameters to the breeding system and adjust them through automated equipment.

[0169] In this embodiment, 5 specifically refers to:

[0170] Use 3D modeling software (such as Unity 3D or Blender) to create a simulation of the sea cucumber farming environment and connect sensors and image acquisition devices to update data in real time;

[0171] Use real-time data streams to drive the simulator, providing mapping and simulation of virtual and actual environments;

[0172] Integrate simulation predictions and real-time data and feed them back to the operator interface as the basis for adjustment plans and equipment control;

[0173] Based on real-time data, the relationship model and the optimal breeding plan are adjusted in real time through reinforcement learning.

[0174] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0175] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0176] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0177] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0178] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. An automated control method for intelligent breeding of sea cucumbers on land based on AI-assisted decision-making, characterized in that: The following steps are involved: Step 1: Deploy various sensors in the aquaculture environment to monitor water quality and environmental data in real time, obtain sea cucumber images through image acquisition devices, and use IoT technology to upload water quality data, environmental data and sea cucumber image data to the cloud; Step 2: Based on the machine learning algorithm, according to the collected water quality data, environmental data and sea cucumber image data, the sea cucumber growth characteristic data is extracted from the sea cucumber image based on the image feature extraction algorithm, and a relationship model between the water quality and environmental data and the growth status of the sea cucumber is established. Based on the relationship model and real-time data, the growth trend of the sea cucumber is predicted; Step 3: Detect the health status of the sea cucumber based on the health status detection model; Step 4: Develop the best breeding plan based on the predicted growth trend and health status of the sea cucumbers; and automatically adjust the breeding facilities based on the best breeding plan, specifically: The goal of the optimal breeding program is to maximize the growth rate and health status of sea cucumbers while minimizing the cost of bait and environmental regulation. The objective function is: ; in, is the growth rate; for health status; and is the parameter that adjusts the weight of the optimization objective, Represents the cost of control and resource investment; It is a combination of parameters, including bait release amount F, temperature T, and water quality W; Constraints include that the amount of bait cannot exceed the ecological carrying capacity and the temperature range is limited; Use simulated annealing to solve the above objective function: ① Define the initial parameter combination and initial temperature, set the cooling factor α and stop condition; ② For the current status , generate a new state of the neighborhood , and calculate the objective function difference ; ③If , accept the new state ,like , then with probability Accept the new status ; ④Use the cooling formula T=α×T; ⑤Terminate when the temperature drops to Tmin or the maximum number of iterations is reached; ⑥ Apply the final parameters to the breeding system and adjust them through automated equipment; Step 5: Build a digital twin model of the farming system, monitor the actual farming conditions in real time, establish a feedback mechanism, and adjust the relationship model and optimal farming plan based on the system execution effect; The step 2 is specifically as follows: Step 2.1: Acquire water quality data, environmental data and sea cucumber image data, wherein the water quality data includes pH value, dissolved oxygen, turbidity and salinity; the environmental data includes temperature and light intensity; and acquire sea cucumber growth characteristic data based on the sea cucumber image data; Step 2.2: Process the data to remove missing values ​​and outliers, and then standardize or normalize the data to ensure that the dimensions of different features are consistent; Step 2.3: Use correlation analysis to select features that are highly correlated with the growth status of sea cucumbers, and apply recursive feature elimination algorithm to further screen important features to obtain the final training feature set; Step 2.4: Build a relationship model based on the support vector regression model, and train based on the final training feature set, adjust the error limit, regularization parameter and kernel parameter, obtain a fit that meets the preset requirements, and obtain the final relationship model; The health status detection model is constructed using random forest Acquire a data set related to the health status of sea cucumbers to construct an original training data set, where the original training data set includes biological characteristics, environmental characteristics, and annotated health status; Randomly extract samples with replacement from the original training data set to form multiple sub-sample sets. Each sub-sample set is used to train a decision tree, and the unselected data is regarded as out-of-bag data. For each tree, when splitting a node, a subset is randomly selected from all features, and the best split point is selected in this subset for splitting; The tree building process continues until the stopping criterion is reached, resulting in a random forest model with N trees.

2. The automated control method for intelligent breeding of sea cucumbers on land based on AI-assisted decision-making according to claim 1, characterized in that: The step 1 is specifically as follows: The various sensors include temperature sensors, humidity sensors, pH sensors, dissolved oxygen sensors, turbidity sensors, light sensors and salinity sensors; the temperature sensors are distributed at different depths and positions in the aquaculture pond, and the dissolved oxygen sensors use optical dissolved oxygen sensors to measure the oxygen content of the water body and are placed in different areas of the aquaculture pond; The image acquisition device includes an underwater camera and a surface camera, supporting high-definition video and low-light environment image acquisition; The sensors and image acquisition devices are connected to an integrated IoT gateway, which is responsible for collecting data from all devices and uploading water quality data, environmental data and sea cucumber image data to the cloud using IoT technology.

3. The automated control method for intelligent breeding of sea cucumbers on land based on AI-assisted decision-making according to claim 1 is characterized in that: The step 2.3 is as follows: Step 2.3.1: Calculate all water quality and environmental characteristics Growth characteristics of sea cucumbers Pearson correlation coefficient between ; For variables and , and its Pearson correlation coefficient ρ is defined as: ； in, and The variables are and The mean of ; U is the number of samples; For variables The u-th sample of ; For variables The u-th sample of Step 2.3.2: Select a threshold based on the absolute value of the correlation coefficient, filter out features with absolute values ​​higher than the threshold, and obtain a filtered feature set; Step 2.3.3: Use RFE combined with the base learner for evaluation on the filtered feature set; Step 2.3.4: Eliminate the least important features in each round of iteration and record the model performance of each iteration; Step 2.3.5: Use cross-validation to evaluate the impact of feature sets on model performance after each iteration; Step 2.3.6: Finally, select the feature set that performs best on the evaluation metric for training the model.

4. The automated control method for intelligent breeding of sea cucumbers on land based on AI-assisted decision-making according to claim 3 is characterized in that: The step 2.4 is as follows: The SVR model is as follows: ； Where n is the number of feature vectors; i represents the index; The constraints are: Among them, w, b are the weight and bias of the model; C is the regularization parameter, is the precision parameter; , is the slack variable; x i is the ith input feature vector, is the corresponding true value; Based on the final training feature set training, k-fold cross validation is used to evaluate the SVR model performance, and grid search is combined to adjust the error limit, regularization parameter and kernel parameter to obtain a fit that meets the preset requirements and obtain the final relationship model.

5. The automated control method for intelligent breeding of terrestrial sea cucumbers based on AI-assisted decision-making according to claim 4 is characterized in that: The kernel function of the SVR model adopts the RBF kernel function: ； in, is the kernel parameter; x j is the jth input feature vector; The dual problem of SVR is solved based on the kernel function. The original problem is converted into a dual form based on the Cramérod condition by the Lagrange multiplier method: First, we introduce the Lagrange multiplier To construct the Lagrangian function: ； in, is the error variable; According to the KKT condition, the optimal solution must meet the following conditions: ① Gradient conditions: ； ②Original constraints: ； ③Complementary relaxation conditions: ； By analyzing the KKT conditions and eliminating the original variables, we get the dual problem: ;in, , is the Lagrangian function; Constraints: 。 6. The automated control method for intelligent breeding of sea cucumbers on land based on AI-assisted decision-making according to claim 1, characterized in that: The image feature extraction algorithm is used to extract sea cucumber growth feature data according to the sea cucumber image, specifically as follows: Acquire sea cucumber images based on an image acquisition device and perform preprocessing, including removing water ripples and other random noise in the image, adjusting white balance and color, and restoring the color of the real growth environment; Extracting the biological characteristics of sea cucumbers, including shape characteristics, color characteristics and texture characteristics, based on the preprocessed sea cucumber images; The shape features include length, width, and surface concavity. The Canny edge detection algorithm is used to locate the edge of the sea cucumber, and the findContours function of OpenCV is used to identify the external contour and important feature points of the sea cucumber, and to obtain the length, width, and surface concavity; color features are identified through color histogram analysis; and texture features are extracted using the grayscale co-occurrence matrix method.

7. The automated control method for intelligent breeding of sea cucumbers on land based on AI-assisted decision-making according to claim 6 is characterized in that: The random forest is specifically as follows: For an input sample x, each tree Given a prediction, the final category is determined by voting, and the category with the most votes is selected, i.e. the health status : ； in, is an indicator function, indicating Whether the tree predicts category k; The trained random forest model is validated based on out-of-bag data. The prediction results of out-of-bag data samples are compared with their true labels, and the out-of-bag error rate is calculated to evaluate the generalization ability of random forest.

8. The automated control method for intelligent breeding of sea cucumbers on land based on AI-assisted decision-making according to claim 1, characterized in that: The step 5 is specifically as follows: Use 3D modeling software to create a simulation of the sea cucumber farming environment, and connect sensors and image acquisition devices to update data in real time; Use real-time data streams to drive the simulator, providing mapping and simulation of virtual and actual environments; Integrate simulation predictions and real-time data and feed them back to the operator interface as the basis for adjustment plans and equipment control; Based on real-time data, the relationship model and the optimal breeding plan are adjusted in real time through reinforcement learning.

Images