Intelligent breeding control method and system based on AI auxiliary decision making and Bluetooth networking

Abstract

The invention discloses an intelligent breeding control method and system based on AI auxiliary decision making and Bluetooth networking, and relates to the field of intelligent breeding, and the method comprises the steps: obtaining breeding environment historical monitoring data and animal image historical data collected by Bluetooth node equipment, and historical health state data of animals; animal growth feature data and behavior feature data are extracted according to animal image historical data based on an image feature extraction algorithm; constructing a training set and a test set according to the behavior characteristic data, the breeding environment historical data, the animal growth characteristic data and the animal historical health state data; training the LSTM neural network model by using the training set and the test set to obtain a growth trend prediction model; and inputting real-time data into the growth trend prediction model, predicting the growth trend of animals, adjusting a breeding scheme, and adjusting breeding facilities through Bluetooth networking. Resource allocation is optimized while the decision-making efficiency and accuracy of breeding are improved, and the labor intensity of workers and the production cost are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent breeding, and particularly to an intelligent breeding control method and system based on AI-assisted decision-making and Bluetooth networking. BACKGROUND

[0002] With the improvement of people's living standards, the demand for breeding products is increasing, and the traditional manual breeding method has been unable to meet the high efficiency and high quality requirements of modern breeding industry. Traditional breeding methods have many problems, such as poor accuracy of manual monitoring of breeding environment, high labor intensity, difficulty in achieving precise feeding and disease prevention, etc. In recent years, with the rapid development of artificial intelligence technology and Internet of Things technology, the application of these advanced technologies in the field of breeding is expected to solve many problems of traditional breeding and realize the intelligentization, automation and precision of breeding.

[0003] Therefore, the present application provides an intelligent breeding control method and system based on AI-assisted decision-making and Bluetooth networking to solve the above problems. SUMMARY

[0004] In order to solve the above problems, the present application provides an intelligent breeding control method and system based on AI-assisted decision-making and Bluetooth networking, which not only improves the decision-making efficiency and accuracy of breeding, but also realizes the optimization of resource allocation, reduces the labor intensity and reduces the production cost.

[0005] To achieve the above purpose, the present application provides an intelligent breeding control method based on AI-assisted decision-making and Bluetooth networking, comprising the following steps:

[0006] S1: obtaining the breeding environment historical monitoring data and animal image historical data collected by the Bluetooth node device, and the historical health status data of the animal;

[0007] S2: extracting animal growth feature data and behavior feature data from the animal image historical data based on an image feature extraction algorithm; the animal growth feature data includes weight and growth rate; taking the behavior feature data and the breeding environment historical data as input data, and taking the animal growth feature data and the historical health status data of the animal as output data, a training set and a test set are constructed;

[0008] S3: training the LSTM neural network model using the training set and the test set to obtain a growth trend prediction model;

[0009] S4: inputting real-time data into the growth trend prediction model to predict the animal growth trend;

[0010] S5: adjusting the breeding scheme based on the predicted animal growth trend; and adjusting the breeding facilities through Bluetooth networking based on the adjusted breeding scheme.

[0011] Preferably, the Bluetooth node device in S1 comprises a Bluetooth environment sensor, an animal individual monitoring device, an image acquisition device and a special health monitoring device.

[0012] Preferably, S3 specifically comprises:

[0013] S31: constructing a multi-objective optimization function;

[0014] S32: using the training set and the test set and the multi-objective optimization function, training and testing the LSTM neural network model by using the NSGA-II algorithm to obtain the trained LSTM neural network model;

[0015] S33: using the trained LSTM neural network model as a growth trend prediction model to predict the growth trend of the animal.

[0016] Preferably, the multi-objective optimization function is expressed as:

[0017]

[0018] wherein λ1 is a weight coefficient of the weight prediction error, λ2 is a weight coefficient of the growth rate prediction error, and λ3 is a weight coefficient of the health state prediction classification error, is the weight prediction loss, is the growth rate prediction loss, is the health state prediction classification loss.

[0019] Preferably, S32 specifically comprises:

[0020] S321: initializing the LSTM neural network model parameters, setting the maximum iteration number t_max, and setting the current iteration number t=0;

[0021] S322: flattening the parameters of the LSTM neural network model into a vector to generate an initial population P_t, wherein the population size is N, and each individual represents a set of LSTM parameters;

[0022] S323: decoding each individual in the population P_t into LSTM neural network model parameters, training the LSTM neural network model using the training set to obtain the trained LSTM neural network model corresponding to each individual;

[0023] S324: testing the trained LSTM neural network model corresponding to each individual using the test set, calculating the multi-objective optimization function value on the test set as the fitness value of each individual;

[0024] S325: according to the fitness value of each individual, performing non-dominated sorting on the individuals in P_t to generate a Pareto frontier hierarchical structure, and calculating the crowding degree of each individual;

[0025] S326: selecting the top N / 2 individuals from P_t to form a parent population Q_t according to the hierarchical structure of the Pareto front and the crowding degree;

[0026] S327: selecting, crossing and mutating the individuals in Q_t to generate an offspring population O_t;

[0027] S328: merging Q_t and O_t to obtain a temporary population R_t, performing non-dominated sorting and crowding degree selection on R_t, and retaining the top N individuals as the next generation population P_{t+1};

[0028] S329: if t≥T_max or there exists an individual satisfying the preset condition, stopping iteration and outputting the LSTM neural network model corresponding to the optimal individual; otherwise, setting t=t+1 and returning to execute S233, the preset condition being that the weight prediction loss is less than a weight setting threshold, the growth rate prediction loss is less than a growth rate setting threshold, and the health state classification prediction loss is less than a health state classification setting threshold.

[0029] Preferably, S325 specifically comprises:

[0030] S3251: comparing the individuals in the population P_t in pairs to generate a non-dominated hierarchical structure and obtain a hierarchical structure of the Pareto front;

[0031] S3252: calculating the crowding degree of the individuals in the same hierarchical structure of the Pareto front according to the Euclidean distance in the objective space.

[0032] Preferably, the crowding degree in S3252 is represented as:

[0033]

[0034] wherein M is the number of objective functions, the objective functions including the weight prediction loss, the growth rate prediction loss, and the health state prediction classification loss; f m (i+1) is the function value of individual i+1 on the mth objective, is the maximum value of the mth objective in the current non-dominated layer, is the minimum value of the mth objective in the current non-dominated layer.

[0035] Preferably, the weight prediction loss adopts a minimum mean square error and is represented as:

[0036]

[0037] wherein, is the real animal weight measurement value of the ith sample, The weight value predicted by the i-th sample model, N is the number of samples, and i is the index of the i-th sample.

[0038] The growth rate prediction loss minimizes the mean square error, denoted as:

[0039]

[0040] wherein, The i-th sample real animal growth rate measurement value, The i-th sample model predicted growth rate value.

[0041] Preferably, the health status loss minimizes the cross-entropy, denoted as:

[0042]

[0043] wherein, The i-th sample real health status label, 0 represents healthy, and 1 represents abnormal, The i-th sample model predicted health status probability, ranging from [0, 1].

[0044] An intelligent breeding control system based on AI assisted decision-making and Bluetooth networking, comprising:

[0045] A Bluetooth node device for collecting breeding environment data, animal image data, and animal health status data;

[0046] A data processing unit for obtaining breeding environment historical monitoring data, animal image historical data, and animal historical health status data collected by the Bluetooth node device;

[0047] A feature extraction and data set construction unit for extracting animal growth feature data and behavior feature data from animal image historical data based on an image feature extraction algorithm, the animal growth feature data including weight and growth rate, using the behavior feature data and breeding environment historical data as input data, and using the animal growth feature data and animal historical health status data as output data to construct a training set and a test set;

[0048] A model construction unit for training an LSTM neural network model using the training set and the test set to obtain a growth trend prediction model;

[0049] A prediction unit for inputting real-time data into the growth trend prediction model to predict animal growth trend;

[0050] A control unit for adjusting the breeding scheme based on the predicted animal growth trend, and adjusting the breeding facilities through Bluetooth networking based on the adjusted breeding scheme.

[0051] Therefore, the intelligent breeding control method and system based on AI auxiliary decision and Bluetooth networking have the following beneficial effects:

[0052] (1) Precise prediction of animal growth trend: by fusing breeding environment data, animal behavior characteristics (such as activity amount, feeding frequency) and growth indicators (such as body weight, growth rate), the LSTM neural network model is used to capture the time sequence dynamic change, which significantly improves the accuracy of growth trend prediction and avoids the lag and error of traditional static experience model.

[0053] (2) Multi-modal data collaborative analysis: combining image feature extraction and historical health data, the environmental, behavioral and physiological state correlation modeling is realized, which can early detect animal growth abnormalities (such as disease or malnutrition) and provide scientific basis for dynamic adjustment of breeding scheme.

[0054] (3) Dynamic optimization of breeding strategy: based on the prediction results, the feeding plan and environmental parameters (such as ventilation, illumination) are adjusted in real time to avoid resource waste (such as overfeeding) or environmental stress (such as high temperature and humidity), and to improve the breeding efficiency.

[0055] (4) Efficient control of low-power Bluetooth networking: through Bluetooth network, the wireless linkage of sensors and execution devices (such as feeding machine, fan) is realized, which reduces the deployment cost, supports stable access of mobile nodes (such as animals wearing tags) and ensures real-time execution of control instructions.

[0056] (5) End-to-end intelligent management: from data collection, model prediction to device control, a closed loop is formed to reduce manual intervention, which is suitable for large-scale and intensive breeding scenes, significantly reduces the operating cost and improves the economic benefit.

[0057] The technical solutions of the present application will be further described in detail below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 It is a flowchart of the intelligent breeding control method based on AI auxiliary decision and Bluetooth networking in the present application;

[0059] Figure 2 It is a module schematic diagram of the intelligent breeding control system based on AI auxiliary decision and Bluetooth networking in the present application. DETAILED DESCRIPTION

[0060] The following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0061] Unless otherwise defined, technical or scientific terms used in the present invention shall have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs.

[0062] The words “include” or “comprising” and similar words used in the present invention mean that the elements before the word include the elements listed after the word, and do not exclude the possibility of also including other elements. The orientation or position relationship indicated by the terms “inside”, “outside”, “upper”, “lower”, etc. is based on the orientation or position relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation of the present invention. When the absolute position of the described object changes, the relative position relationship may also change accordingly. In the present invention, unless otherwise clearly stipulated and limited, the terms such as “attachment” should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral whole; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0063] Example

[0064] An intelligent farming control method based on AI-assisted decision-making and Bluetooth networking includes the following steps:

[0065] S1: Acquire historical monitoring data of the breeding environment and animal images, as well as historical health status data of the animals, collected by Bluetooth node devices;

[0066] Bluetooth node devices include Bluetooth environmental sensors, individual animal monitoring devices, image acquisition devices, and dedicated health monitoring devices. Bluetooth environmental sensors regularly upload time-series data on environmental parameters; individual animal monitoring devices record each animal's physiological indicators, such as activity level, body temperature, and feeding frequency; image acquisition devices periodically capture images of the animal's posture and behavior, compressing and storing them as historical image datasets; and dedicated health monitoring devices generate health status data, such as weight and blood biochemical indicators. All data is aggregated into a cloud database via a Bluetooth gateway and linked by timestamp, device ID, and animal ID, forming a multi-dimensional historical dataset covering environment, behavior, and health, providing a data foundation for subsequent analysis.

[0067] S2: extracting animal growth feature data and behavior feature data from animal image historical data based on an image feature extraction algorithm; the animal growth feature data includes body weight and growth speed; taking the behavior feature data and the historical data of the breeding environment as input data, and taking the animal growth feature data and the historical health state data of the animal as output data, a training set and a test set are constructed;

[0068] Specifically, based on computer vision and deep learning technology, through target detection, 3D point cloud reconstruction and time series behavior analysis, growth features such as body weight and growth speed and behavior features such as feeding and movement are non-contact extracted from animal image data, and a multi-modal data set is constructed combining with environmental data, to provide high-precision input for the growth prediction model.

[0069] S3: training an LSTM neural network model using the training set and the test set to obtain a growth trend prediction model;

[0070] S3 specifically includes:

[0071] S31: constructing a multi-objective optimization function;

[0072] The multi-objective optimization function is expressed as:

[0073]

[0074] wherein λ1 is a weight coefficient of the body weight prediction error, λ2 is a weight coefficient of the growth speed prediction error, and λ3 is a weight coefficient of the health state prediction classification error.

[0075] is the body weight prediction loss, is the growth speed prediction loss, is the health state prediction classification loss, and the values of λ1, λ2 and λ3 are adjusted according to actual conditions.

[0076] The body weight prediction loss adopts a minimum mean square error, and is expressed as:

[0077]

[0078] wherein is a real animal body weight measurement value of the i-th sample, is a model-predicted body weight value of the i-th sample, N is the number of samples, and i is the index of the i-th sample.

[0079] The growth speed prediction loss adopts a minimum mean square error, and is expressed as:

[0080]

[0081] wherein ​a true animal growth rate measurement value for the i-th sample, a model-predicted growth rate value for the i-th sample.

[0082] The health status loss is minimized cross-entropy, denoted as:

[0083]

[0084] wherein, a true health status label for the i-th sample, 0 represents healthy, 1 represents abnormal, a model-predicted health status probability for the i-th sample, ranging from [0, 1].

[0085] S32: using the training set and the test set and the multi-objective optimization function, using the NSGA-II algorithm to train and test the LSTM neural network model, and obtaining the trained LSTM neural network model;

[0086] S32 specifically includes:

[0087] S321: initialize the LSTM neural network model parameters, set the maximum number of iterations t_max, and the current iteration number t=0;

[0088] S322: flatten the parameters of the LSTM neural network model into a vector, and generate an initial population P_t, wherein the population size is N, and each individual represents a set of LSTM parameters;

[0089] S323: decode each individual in the population P_t into LSTM neural network model parameters, train the LSTM neural network model using the training set, and obtain the trained LSTM neural network model corresponding to each individual;

[0090] S324: test the trained LSTM neural network model corresponding to each individual using the test set, calculate the multi-objective optimization function value on the test set, and use it as the fitness value of each individual;

[0091] S325: according to the fitness value of each individual, non-dominant sorting is performed on the individuals in P_t, a Pareto front hierarchical structure is generated, and the crowding degree of each individual is calculated;

[0092] S325 specifically includes:

[0093] S3251: compare the pairwise dominance relationship of the individuals in the population P_t, generate a non-dominant hierarchy, and obtain the Pareto front hierarchical structure;

[0094] Specifically, all individuals in the population are traversed to determine the dominance relationship between each other by comparing the objective function values. If individual A is not worse than individual B in all objectives and strictly better than individual B in at least one objective, it is said that A dominates B. Based on this, the first layer of non-dominated front F1 is composed of all individuals that are not dominated by any other individual, i.e. the optimal solution set in the current population; then, the individuals in F1 are temporarily removed, and the non-dominated individuals in the remaining individuals constitute the second layer of front F2, and so on, until all individuals are layered. This process filters the non-dominated solution set through iteration, forming a hierarchical Pareto front structure, which provides a basis for the selection operation of the multi-objective optimization algorithm.

[0095] S3252: For the individuals in the same layer of the hierarchical Pareto front structure, the crowding degree is calculated according to the Euclidean distance in the objective space.

[0096] Specifically, the process of calculating the crowding degree of the individuals in the same layer of the hierarchical Pareto front structure is as follows: first, the individuals in the current non-dominated layer F i are sorted in ascending order in each objective function dimension; for each individual, the difference between the function values of its adjacent individuals in the objective is calculated as the local crowding distance. Then, the local crowding distances of all objective dimensions are summed to obtain the total crowding degree of the individual. Boundary individuals (i.e. extreme points in a certain objective) are usually assigned an infinite crowding degree to ensure priority retention, while the crowding degree of intermediate individuals reflects the degree of sparsity of their distribution in the objective space - the greater the crowding degree, the more dispersed the surrounding individuals, which helps to maintain the diversity of the population.

[0097] where the crowding degree is represented as:

[0098]

[0099] M is the number of objective functions, including the body weight prediction loss, the growth rate prediction loss, and the health status prediction classification loss; f m (i+1) is the function value of individual i+1 in the mth objective, is the maximum value of the mth objective in the current non-dominated layer, f m min is the minimum value of the mth objective in the current non-dominated layer.

[0100] S326: According to the layer and crowding degree of the hierarchical Pareto front structure, select the first N / 2 individuals from P_t to form the parent population Q_t;

[0101] S327: Perform selection, crossover and mutation operations on the individuals in Q_t to generate the offspring population O_t;

[0102] S328: merge Q_t and O_t to obtain a temporary population R_t, perform non-dominated sorting and crowding selection on R_t, and retain the first N individuals as the next generation population P_{t+1};

[0103] S329: if t >= T_max or there is an individual satisfying a preset condition, stop iteration and output the LSTM neural network model corresponding to the optimal individual; otherwise, let t = t + 1, return to execute S233, and the preset condition is that the weight prediction loss is less than the weight set threshold, and the growth speed prediction loss is less than the growth speed set threshold, and the health state classification prediction loss is less than the health state classification set threshold.

[0104] S33: using the trained LSTM neural network model as a growth trend prediction model, the growth trend of the animal is predicted.

[0105] S4: input real-time data into the growth trend prediction model to predict the growth trend of the animal;

[0106] S5: based on the predicted growth trend of the animal, adjust the breeding scheme; and based on the adjusted breeding scheme, adjust the breeding facility through Bluetooth networking.

[0107] Specifically, when the prediction result shows that the growth speed deviates from the expectation, the health indicators are abnormal, or the weight is not ideal, the breeding scheme is adjusted. The scheme is issued to each intelligent terminal device through a low-power Bluetooth network: an environmental control node, such as a variable frequency fan, a heater, and an LED lighting system, automatically adjusts the environment according to the new parameters; the intelligent feeder adjusts the feed formula and feeding amount according to the breeding scheme; the water line controller optimizes the water supply; at the same time, the Bluetooth camera and image analysis terminal deployed in the farm continuously monitor the animal group behavior, such as the degree of aggregation and the activity mode, cross-verify the adjustment effect with the sensor data, and feedback to the central control system in real time, forming a closed-loop regulation and control system of "monitoring-prediction-optimization-execution-verification". All device states and animal response data are recorded in the blockchain breeding log, providing a basis for subsequent optimization and traceability, and finally realizing precise, adaptive, and traceable modern intelligent breeding management.

[0108] Embodiment 1

[0109] An intelligent breeding control system based on AI assisted decision-making and Bluetooth networking, comprising:

[0110] a Bluetooth node device for collecting breeding environment data, animal image data, and animal health status data;

[0111] a data processing unit for obtaining breeding environment historical monitoring data, animal image historical data, and animal historical health status data collected by the Bluetooth node device;

[0112] The feature extraction and dataset construction unit is configured to extract animal growth feature data and behavior feature data from historical animal image data based on an image feature extraction algorithm, the animal growth feature data including weight and growth rate, the behavior feature data and historical breeding environment data being input data, the animal growth feature data and historical health state data of the animal being output data, and a training set and a test set being constructed;

[0113] The model construction unit is configured to train an LSTM neural network model using the training set and the test set to obtain a growth trend prediction model;

[0114] The prediction unit is configured to input real-time data into the growth trend prediction model to predict the growth trend of the animal.

[0115] The control unit is configured to adjust a breeding scheme based on the predicted growth trend of the animal and to adjust the breeding facility through Bluetooth networking based on the adjusted breeding scheme.

[0116] Therefore, the intelligent breeding control method and system based on AI-assisted decision-making and Bluetooth networking improve the decision-making efficiency and accuracy of breeding, and achieve optimized resource allocation, reduced labor intensity and reduced production cost.

[0117] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application and not to limit them, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can still be modified or replaced by equivalents, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. An intelligent farming control method based on AI-assisted decision-making and Bluetooth networking, characterized by: The following steps are involved: S1: Acquire historical monitoring data of the breeding environment and animal images, as well as historical health status data of the animals, collected by Bluetooth node devices; S2: Extract animal growth and behavior characteristic data from historical animal image data using an image feature extraction algorithm; animal growth characteristic data includes weight and growth rate; use the behavior characteristic data and historical breeding environment data as input data, and the animal growth characteristic data and historical animal health status data as output data to construct training and test sets; S3: Use the training set and test set to train the LSTM neural network model to obtain a growth trend prediction model; S4: Inputting real-time data into a growth trend prediction model to predict animal growth trends; S5: Adjust the breeding plan based on the predicted animal growth trend; and adjust the breeding facilities through Bluetooth networking based on the adjusted breeding plan.

2. The intelligent farming control method based on AI-assisted decision-making and Bluetooth networking according to claim 1 is characterized in that: The Bluetooth node devices in S1 include Bluetooth environmental sensors, animal individual monitoring devices, image acquisition devices and dedicated health monitoring devices.

3. The intelligent farming control method based on AI-assisted decision-making and Bluetooth networking according to claim 2 is characterized in that: S3 specifically includes: S31: Construct multi-objective optimization function; S32: Using the training set and test set and the multi-objective optimization function, the NSGA-II algorithm is used to train and test the LSTM neural network model to obtain a trained LSTM neural network model; S33: Use the trained LSTM neural network model as a growth trend prediction model to predict the growth trend of animals.

4. The intelligent farming control method based on AI-assisted decision-making and Bluetooth networking according to claim 3 is characterized in that: The multi-objective optimization function is expressed as: Among them, λ1 is the weight coefficient of weight prediction error, λ2 is the weight coefficient of growth rate prediction error, and λ3 is the weight coefficient of health status prediction classification error. For weight loss prediction, For growth rate prediction loss, Classification loss for health state prediction.

5. The intelligent farming control method based on AI-assisted decision-making and Bluetooth networking according to claim 4 is characterized in that: S32 specifically includes: S321: Initialize the LSTM neural network model parameters, set the maximum number of iterations t_max, and the current number of iterations t=0; S322: Flatten the parameters of the LSTM neural network model into a vector to generate an initial population P_t, where the population size is N and each individual represents a set of LSTM parameters; S323: Decode each individual in the population P_t into LSTM neural network model parameters, train the LSTM neural network model using the training set, and obtain the trained LSTM neural network model corresponding to each individual; S324: Using the test set, the trained LSTM neural network model corresponding to each individual is tested, and the multi-objective optimization function value is calculated on the test set as the fitness value of each individual; S325: According to the fitness value of each individual, perform non-dominated sorting on the individuals in P_t, generate a Pareto front hierarchical structure, and calculate the crowding degree of each individual; S326: According to the level and crowding degree of the Pareto front hierarchical structure, the first N / 2 individuals are selected from P_t to form the parent population Q_t; S327: Perform selection, crossover and mutation operations on individuals in Q_t to generate the offspring population O_t; S328: Merge Q_t and O_t to obtain a temporary population R_t, perform non-dominated sorting and crowding selection on R_t, and retain the first N individuals as the next generation population P_{t+1}; S329: If t≥T_max or there is an individual that meets the preset conditions, stop the iteration and output the LSTM neural network model corresponding to the optimal individual; otherwise, let t=t+1 and return to execute S233. The preset conditions are that the weight prediction loss is less than the weight setting threshold, the growth rate prediction loss is less than the growth rate setting threshold, and the health status classification prediction loss is less than the health status classification setting threshold.

6. The intelligent farming control method based on AI-assisted decision-making and Bluetooth networking according to claim 5 is characterized in that: S325 specifically includes: S3251: Compare the dominance relationships between individuals in the population P_t, generate non-dominated hierarchies, and obtain the Pareto frontier hierarchical structure; S3252: For individuals at the same level in the Pareto front hierarchy, the crowding degree is calculated based on the Euclidean distance in the target space.

7. The intelligent farming control method based on AI-assisted decision-making and Bluetooth networking according to claim 6 is characterized in that: The congestion degree in S3252 is expressed as: M is the number of objective functions, including weight prediction loss, growth rate prediction loss, and health status prediction classification loss; f m (i+1) is the function value of individual i+1 on the mth target, is the maximum value of the mth target in the current non-dominated layer, is the minimum value of the mth target in the current non-dominated layer.

8. The intelligent farming control method based on AI-assisted decision-making and Bluetooth networking according to claim 7 is characterized in that: The weight prediction loss is minimized by the mean square error, which is expressed as: in, is the actual animal weight measurement value of the i-th sample, is the weight value predicted by the model for the i-th sample, N is the number of samples, and i is the index of the i-th sample; The growth rate prediction loss is minimized by the mean square error, which is expressed as: in, is the actual animal growth rate measurement value of the i-th sample, is the growth rate value predicted by the i-th sample model.

9. The intelligent farming control method based on AI-assisted decision-making and Bluetooth networking according to claim 8 is characterized in that: The health state loss is minimized by cross entropy, which is expressed as: in, is the true health status label of the i-th sample, 0 means healthy, 1 means abnormal, The health status probability predicted by the i-th sample model is in the range of [0,1].

10. Intelligent farming control system based on AI-assisted decision-making and Bluetooth networking, characterized by: include: Bluetooth node devices are used to collect breeding environment data, animal image data, and animal health status data; A data processing unit is used to obtain historical monitoring data of the breeding environment, historical animal image data, and historical health status data of the animals collected by the Bluetooth node device; A feature extraction and data set construction unit is used to extract animal growth feature data and behavioral feature data from animal image historical data based on an image feature extraction algorithm. The animal growth feature data includes weight and growth rate. The behavioral feature data and breeding environment historical data are used as input data, and the animal growth feature data and animal historical health status data are used as output data to construct a training set and a test set. A model building unit is used to train the LSTM neural network model using the training set and the test set to obtain a growth trend prediction model; A prediction unit, configured to input real-time data into the growth trend prediction model to predict the growth trend of the animal; The control unit is used to adjust the breeding plan based on the predicted animal growth trend, and adjust the breeding facilities through Bluetooth networking based on the adjusted breeding plan.

Images