Intelligent control system and method applied to breeding field

Abstract

The invention discloses an intelligent control system and method applied to the breeding field, and belongs to the technical field of artificial intelligence. Environmental data of a breeding water area are collected in real time through the sensor, the activity condition of crabs is monitored through the camera, and image data of the activity of the crabs are collected; the collected environment data and image data are preprocessed, and for the image data, a convolutional neural network is used to identify the motion of the crab; dividing a breeding water area into a plurality of areas according to the distribution of breeding equipment, and associating environment data of different areas with the actions of crabs; predicting water quality change trends of different areas in the aquaculture water area through an LSTM time sequence model; combining a support vector machine classification algorithm with water quality change trends of different areas to generate an influence function of environmental data on crab behaviors, and judging whether the crabs in the aquaculture water area have abnormal behaviors or not; and according to whether the crabs in different areas have abnormal behaviors, equipment in corresponding areas is automatically regulated and controlled.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an intelligent control system and method applied in the field of breeding. Background Art

[0002] Crab farming is an aquaculture activity that requires meticulous management. Water quality is a key factor affecting crab growth, health, and production. Fluctuations in environmental parameters such as water temperature, pH value, dissolved oxygen concentration, and ammonia nitrogen concentration will directly affect the physiological state of crabs and even cause disease or death. Therefore, farmers must monitor water quality in real time and take timely measures, such as oxygenation, water changes, and temperature adjustments, to maintain a stable farming environment. At the same time, the activity status of crabs is an important indicator of their health status. Abnormal behavior may indicate poor water quality or the risk of disease outbreaks.

[0003] Traditional water quality monitoring mainly relies on regular sampling and manual testing, which has a slow response speed and often fails to capture instantaneous changes in water quality in a timely manner, which can easily lead to sudden environmental deterioration not being discovered in time. Most aquaculture management relies on the experience of farmers for regulation. For example, operations such as oxygenation and water changes after abnormal water quality lack scientific basis, which can easily lead to overreaction or insufficient regulation, resulting in water quality fluctuations or waste of resources. Even if the water quality is monitored in real time through sensors, many systems do not have an intelligent feedback control mechanism and cannot automatically control the aquaculture equipment based on real-time data, resulting in delayed environmental regulation and affecting the growth environment of crabs. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent control system and method for use in the field of aquaculture to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An intelligent control method applied to the field of aquaculture, the method comprising the following steps:

[0007] S100, collecting environmental data of the aquaculture waters in real time through sensors, including: water temperature, pH value, dissolved oxygen concentration, ammonia nitrogen concentration, air temperature and air humidity; using cameras to monitor the activities of crabs and collect image data of crab activities;

[0008] S200, preprocessing the collected environmental data and image data, filtering out noise data and redundant data; using a convolutional neural network to build an action recognition model based on an image data set covering different crab actions, and identifying the crab's actions; using a Voronoi algorithm to divide the aquaculture waters into several areas according to the distribution of aquaculture equipment, and using an action-environment association model to associate the environmental data of different areas with the crab's actions;

[0009] S300, using the LSTM time series model, the environmental data of the aquaculture waters are analyzed based on the historical water quality data to predict the water quality change trends in different areas of the aquaculture waters; using the support vector machine classification algorithm combined with the water quality change trends in different areas, the influence function of the environmental data on the crab behavior is generated to determine whether the crabs in the aquaculture waters have abnormal behavior;

[0010] S400, according to whether the crabs in different areas have abnormal behavior, automatically adjust the aerator, water exchange system and heating equipment in the corresponding area to ensure that the water quality change trend in each area meets the healthy state of the crabs.

[0011] According to step S100, a water temperature sensor is used to monitor the temperature change of the water body in the aquaculture water area in real time, and the real-time value of the water temperature is recorded in degrees Celsius; a pH sensor is used to monitor the acidity and alkalinity of the water body and record the pH value of the water body; a dissolved oxygen sensor detects the oxygen content in the water body and records the dissolved oxygen concentration in the water in milligrams per liter; an ammonia nitrogen sensor detects the ammonia nitrogen content in the water and records the ammonia nitrogen concentration in milligrams per liter; an air temperature sensor monitors the temperature of the air around the aquaculture water area and records the air temperature value in degrees Celsius; an air humidity sensor monitors the relative humidity of the air in the aquaculture area and records the percentage value of the air humidity.

[0012] According to step S200, an image dataset covering different crab actions is established, wherein the crab actions include crawling, swimming, eating, resting, attacking and hiding;

[0013] A convolutional neural network is used to input image data of crab activities of a fixed size. Local features in the image are gradually extracted through several convolutional layers. The convolutional layers are used to detect the features of the crab's body parts and limb movements. The first convolutional layer detects the edge contours of the crab's limbs. The second and subsequent convolutional layers gradually extract the movement of the crab's claws, rotation of the body, and changes in posture. The output of the convolutional layer is downsampled through the pooling layer to reduce the image size and retain relevant information about the action. After the convolutional layer and the pooling layer, a fully connected layer is added. The fully connected layer converts the extracted action features into classification features related to the crab action category. The output of the model is converted into crab actions through the Softmax activation function.

[0014] On the image dataset covering different crab actions, the corresponding action of each image is annotated to form a training set and a test set; the annotated images are input into the convolutional neural network model and trained using the supervised learning method. The training process requires optimizing the model parameters so that the output action prediction accuracy is gradually improved; the cross entropy loss function is used to evaluate the error between the model output and the true label, and the Adam optimization algorithm is used to gradually adjust the model parameters; the learning rate is dynamically adjusted according to the training process;

[0015] After the training is completed, the accuracy of the model is verified and evaluated; the test data that has not participated in the training is input into the model, and the recognition accuracy and recall rate of the model on different actions are calculated to further optimize the model;

[0016] The collected image data of crab activities is transmitted in real time and input into the trained convolutional neural network model. Based on the feature extraction and classification algorithm, the current movements of different crabs are obtained.

[0017] According to step S200, the aquaculture waters are divided into several areas, each area is divided according to the distribution of aquaculture equipment, and the principle of area division is as follows: according to the coverage of the equipment, the aquaculture waters are divided into N sub-areas, each area Z i (i=1, 2, ..., N) all have independent environmental monitoring and control equipment;

[0018] Each area Z i Environmental data i Including water temperature T i pH i , dissolved oxygen concentration O i , Ammonia nitrogen concentration A i , air temperature TAir i and air humidity i ; Monitor the crabs in each area through the camera to obtain the movements of the crabs in the area. The movement data of the crabs in each area is expressed as: A i =(a 1 , a 2 , ..., a m ), where a 1 to a m Indicates area Z i The actions of the 1st to mth crabs in the area, where m is the number of monitored crabs in the area;

[0019] The environmental data of different regions i Action data with crabs i Perform association analysis and build an action-environment association model, which is expressed using the following formula:

[0020] Actioni =Effect+∈, where Effect is the influence function of environmental data on crab behavior, and ∈ is the random error term.

[0021] The coverage of each area depends on the effective radius of the equipment, which includes aerators, water exchange systems and heating equipment. The effective radius of the aerator is R oxy , the radius of action of the water exchange system is R water , the effective radius of the heating device is R heat , calculate the coverage area of ​​the aerator Area oxy , Water exchange system coverage area water Area of ​​coverage of heating equipment heat ; When dividing the area, the coverage of the equipment is set as the priority to ensure that all crabs in each sub-area are within the range of the corresponding equipment; if the coverage of the equipment is greater than or equal to Q% of the total area of ​​the water area, the coverage of the equipment is considered sufficient and there is no need to further subdivide the sub-areas; if the coverage of the equipment is less than Q% of the total area, the coverage of the equipment is considered insufficient and it is necessary to use an algorithm based on the Voronoi diagram to divide the sub-areas, where Q is a positive integer set by the user;

[0022] In the Voronoi algorithm, the locations of all equipment in the aquaculture waters are determined. The locations of the equipment will be used as the center points of the Voronoi division. The center point set is S = {s 1 ,s 2 , ..., s K}, where s k =(k=1, 2, ..., K) represents the location coordinates of the kth device, K is the number of devices, and the Euclidean distance d(P, s k )Measure the distance from any point P to s in the water area k The distance is as follows: Among them, x′ P is the horizontal coordinate of the point P in the water area, y′ P is the ordinate of the midpoint P in the water area, is the horizontal coordinate of the kth device, is the ordinate of the kth device; the Voronoi diagram divides the water area into several sub-areas, each of which is a Voronoi cell V(s k ), represents the area controlled by the kth device, and each Voronoi cell is defined as follows:

[0023] Among them, R 2 Refers to the physical space where the crab farming waters are located, using two-dimensional coordinates to represent the specific location of each point.k is the effective radius of the device;

[0024] The control area of ​​each device is determined by calculating the distance between any point P in the water area and all devices; the Voronoi diagram divides the control area into polygons, and the boundary of each polygon is on the perpendicular bisector between the two devices, ensuring that each point is in the same area as its nearest device.

[0025] According to step S300, the historical water quality data of each region is used as an input sequence and input into the LSTM time series model, where the historical water quality data includes a number of consecutive time steps, reflecting the change of water quality over time;

[0026] Use the collected historical water quality data to train the LSTM time series model, learn the law of water quality parameters changing over time, and capture the dependency of water quality changes; establish independent LSTM models for different regions and adjust the model's hyperparameters;

[0027] The historical data is divided into several time steps and represented as a multidimensional time series, where the data of each time step is a multidimensional vector containing different water quality parameters. LSTM is a recurrent neural network that uses a gating mechanism to memorize and forget long-term dependency information in the time series. The input of LSTM is the historical time series, and the output is the predicted value of the water quality parameter. To establish independent LSTM models for different regions, it is necessary to adjust the hyperparameters of the model, including: the number of hidden layer neurons, time step, learning rate, and batch size. The mean square error (MSE) is used as the loss function to measure the difference between the predicted value and the true value. The gradient descent method Adam optimizer is used to minimize the loss function and adjust the model parameters.

[0028] The trained LSTM model can predict water quality parameters in the future, and the model outputs the predicted water quality values ​​in several future time steps.

[0029] The parameters of the water quality change trend include: water temperature change trend ΔT w , pH value change trend ΔpH, dissolved oxygen change trend ΔDO, ammonia nitrogen concentration change trend ΔNH 3 ; According to the crab's actions, the crab's behavioral characteristics are extracted, including: activity frequency AF and feeding behavior EB; the input feature vector of the support vector machine consists of the water quality change trend and the crab's behavioral data. The goal is to predict the impact function f(E) of environmental data on crab behavior based on the input feature vector X i ), describing how water quality changes affect crab behavior, expressed as: Effect = f(ΔT w );

[0030] By finding an optimal hyperplane, the input feature vector X is classified into two categories: "normal behavior" and "abnormal behavior". The classification decision function of the support vector machine is expressed as: F = w T X+b, where w is the weight vector of the support vector machine model, representing the weight of the impact of different input features on the crab behavior, and b is the bias term, representing the intercept of the classification;

[0031] The training goal of the support vector machine model is to maximize the classification interval while minimizing the classification error. The optimization objective function is:

[0032]

[0033] Among them, C is the regularization parameter, which controls the trade-off between the classification interval and misclassification, and y i′ is the label of the i′th sample, indicating whether the crab behavior corresponding to the sample is normal or abnormal, X i′ is the feature vector of the i′th sample;

[0034] After the model training is completed, the new water quality change trend and crab behavior data are used as input to generate the impact function of environmental data on crab behavior and determine whether the crab behavior is abnormal.

[0035] According to step S400, based on the influence function of environmental data on crab behavior, the dissolved oxygen concentration in the water is monitored. i and crab activity frequency AF, if O i If the concentration of dissolved oxygen in the water is lower than the set threshold and the crab activity frequency is abnormal, a control signal is generated; the oxygenator is started or the working intensity is increased to increase the dissolved oxygen concentration in the water, and the dissolved oxygen changes are monitored in real time until O i Reach normal level; when the ammonia nitrogen concentration A i When the crab's feeding behavior EB is abnormal, if A i If the set threshold is exceeded and the crabs are eating abnormally, a control signal is generated to start the water exchange system, replace the polluted water, and monitor the changes in ammonia nitrogen concentration in real time until it returns to the threshold level; when the water temperature T i When the water temperature is lower than the set threshold and the crab activity frequency AF is abnormal, start or increase the temperature setting of the heating device to raise the water temperature to a range suitable for crab growth.

[0036] An intelligent control system applied to the field of breeding, comprising:

[0037] Data acquisition module: including: sensor data acquisition unit and image data acquisition unit; wherein the sensor data acquisition unit collects environmental data of the aquaculture waters in real time through sensors, including: water temperature, pH value, dissolved oxygen concentration, ammonia nitrogen concentration, air temperature and air humidity; the image data acquisition unit uses a camera to monitor the activities of crabs and collect image data of crab activities;

[0038] Data association analysis module: including: data preprocessing unit, convolutional neural network action recognition unit, Voronoi region division unit and action-environment association analysis unit; wherein, the data preprocessing unit preprocesses the collected environmental data and image data, filters out noise data and redundant data, the convolutional neural network action recognition unit uses a convolutional neural network to build an action recognition model based on image data sets covering different crab actions for image data, and recognizes the crab's actions, the Voronoi region division unit uses the Voronoi algorithm to divide the aquaculture waters into several areas according to the distribution of aquaculture equipment, and the action-environment association analysis unit uses the action-environment association model to associate environmental data of different areas with the crab's actions;

[0039] Water quality prediction and behavior analysis module: including: LSTM water quality prediction unit and SVM behavior analysis unit; the LSTM water quality prediction unit uses the LSTM time series model to analyze the environmental data of the aquaculture waters based on historical water quality data, and predicts the water quality change trends in different areas of the aquaculture waters. The SVM behavior analysis unit uses the support vector machine classification algorithm to combine the water quality change trends in different areas, generate the impact function of environmental data on crab behavior, and determine whether the crabs in the aquaculture waters have abnormal behavior;

[0040] Equipment control module: including: equipment control unit; wherein, the equipment control unit automatically controls the aerator, water exchange system and heating equipment in the corresponding area according to whether the crabs in different areas have abnormal behavior, so as to achieve the water quality change trend in each area to meet the health status of the crabs.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The present invention uses convolutional neural networks to accurately identify the behavior of crabs, and combines water quality data to establish an action-environment association model. By analyzing the abnormal behavior of crabs through support vector machines, dual monitoring of water quality and crab health status is achieved, further improving the accuracy and reliability of monitoring.

[0043] 2. According to the behavior of crabs and water quality in different areas, the present invention can automatically adjust the aerator, water exchange system and heating equipment to achieve dynamic environmental adjustment, ensure that the water quality in each sub-area meets the health needs of crabs, and reduce the workload and errors of manual intervention.

[0044] 3. The present invention introduces the Voronoi algorithm to divide the aquaculture waters into multiple sub-areas according to the equipment coverage and water quality change trend, ensuring that each area is within the effective control range of the equipment. This regional management improves the utilization efficiency of the equipment and avoids waste of resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the steps of an intelligent control method applied to the breeding field of the present invention;

[0046] Figure 2 It is a system structure diagram of an intelligent control system applied to the breeding field of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution.

[0049] According to one embodiment of the present invention, Figure 1 As shown in the schematic diagram of the steps of an intelligent control method applied to the field of aquaculture, an intelligent control method applied to the field of aquaculture comprises the following steps:

[0050] S100, collecting environmental data of the aquaculture waters in real time through sensors, including: water temperature, pH value, dissolved oxygen concentration, ammonia nitrogen concentration, air temperature and air humidity; using cameras to monitor the activities of crabs and collect image data of crab activities;

[0051] S200, preprocessing the collected environmental data and image data, filtering out noise data and redundant data; using a convolutional neural network to build an action recognition model based on an image data set covering different crab actions, and identifying the crab's actions; using a Voronoi algorithm to divide the aquaculture waters into several areas according to the distribution of aquaculture equipment, and using an action-environment association model to associate the environmental data of different areas with the crab's actions;

[0052] S300, using the LSTM time series model, the environmental data of the aquaculture waters are analyzed based on the historical water quality data to predict the water quality change trends in different areas of the aquaculture waters; using the support vector machine classification algorithm combined with the water quality change trends in different areas, the influence function of the environmental data on the crab behavior is generated to determine whether the crabs in the aquaculture waters have abnormal behavior;

[0053] S400, according to whether the crabs in different areas have abnormal behavior, automatically adjust the aerator, water exchange system and heating equipment in the corresponding area to ensure that the water quality change trend in each area meets the healthy state of the crabs.

[0054] According to step S100, a water temperature sensor is used to monitor the temperature change of the water body in the aquaculture water area in real time, and the real-time value of the water temperature is recorded in degrees Celsius; a pH sensor is used to monitor the acidity and alkalinity of the water body and record the pH value of the water body; a dissolved oxygen sensor detects the oxygen content in the water body and records the dissolved oxygen concentration in the water in milligrams per liter; an ammonia nitrogen sensor detects the ammonia nitrogen content in the water and records the ammonia nitrogen concentration in milligrams per liter; an air temperature sensor monitors the temperature of the air around the aquaculture water area and records the air temperature value in degrees Celsius; an air humidity sensor monitors the relative humidity of the air in the aquaculture area and records the percentage value of the air humidity.

[0055] According to step S200, an image dataset covering different crab actions is established, wherein the crab actions include crawling, swimming, eating, resting, attacking and hiding;

[0056] A convolutional neural network is used to input image data of crab activities of a fixed size. Local features in the image are gradually extracted through several convolutional layers. The convolutional layers are used to detect the features of the crab's body parts and limb movements. The first convolutional layer detects the edge contours of the crab's limbs. The second and subsequent convolutional layers gradually extract the movement of the crab's claws, rotation of the body, and changes in posture. The output of the convolutional layer is downsampled through the pooling layer to reduce the image size and retain relevant information about the action. After the convolutional layer and the pooling layer, a fully connected layer is added. The fully connected layer converts the extracted action features into classification features related to the crab action category. The output of the model is converted into crab actions through the Softmax activation function.

[0057] On the image dataset covering different crab actions, the corresponding action of each image is annotated to form a training set and a test set; the annotated images are input into the convolutional neural network model and trained using the supervised learning method. The training process requires optimizing the model parameters so that the output action prediction accuracy is gradually improved; the cross entropy loss function is used to evaluate the error between the model output and the true label, and the Adam optimization algorithm is used to gradually adjust the model parameters; the learning rate is dynamically adjusted according to the training process;

[0058] After the training is completed, the accuracy of the model is verified and evaluated; the test data that has not participated in the training is input into the model, and the recognition accuracy and recall rate of the model on different actions are calculated to further optimize the model;

[0059] The collected image data of crab activities is transmitted in real time and input into the trained convolutional neural network model. Based on the feature extraction and classification algorithm, the current movements of different crabs are obtained.

[0060] According to step S200, the aquaculture waters are divided into several areas, each area is divided according to the distribution of aquaculture equipment, and the principle of area division is as follows: according to the coverage of the equipment, the aquaculture waters are divided into N sub-areas, each area Z i (i=1, 2, ..., N) all have independent environmental monitoring and control equipment;

[0061] Each area Z i Environmental data i Including water temperature T i pH i , dissolved oxygen concentration O i , Ammonia nitrogen concentration A i , air temperature TAir i and air humidity i ; Monitor the crabs in each area through the camera to obtain the movements of the crabs in the area. The movement data of the crabs in each area is expressed as: A i =(a 1 , a 2 , ..., a m ), where a 1 to a m Indicates area Z i The actions of the 1st to mth crabs in the area, where m is the number of monitored crabs in the area;

[0062] The environmental data of different regions i Action data with crabs i Perform association analysis and build an action-environment association model, which is expressed using the following formula:

[0063] Action i =Effect+∈, where Effect is the influence function of environmental data on crab behavior, and ∈ is the random error term.

[0064] In this embodiment, the area of ​​the aquaculture water area is 5,000 square meters, and 5 aerators, 3 water exchange systems and 2 heating devices are arranged in the aquaculture water area.

[0065] The coverage of each area depends on the effective radius of the equipment, which includes aerators, water exchange systems and heating equipment. The effective radius of the aerator is R oxy = 30 meters, the effective radius of the water exchange system is R water = 40 meters, the effective radius of the heating equipment is R heat =50 meters, calculate the coverage area of ​​the aerator oxy =π*30 2 ≈2827, Water exchange system coverage area water =π*40 2 ≈5026 and the coverage area of ​​heating equipment heat =π*50 2 ≈7854; the total coverage area of ​​5 aerators is 14135 square meters, the total coverage area of ​​3 water exchange systems is 15078 square meters, and the total coverage area of ​​2 heating equipment is 15708 square meters. When dividing the area, the coverage range of the equipment is set first to ensure that all crabs in each sub-area are within the range of the corresponding equipment; if the coverage range of the equipment is greater than or equal to 80% of the total area compared with the total area of ​​the water area, it is considered that the coverage range of the equipment is sufficient and there is no need to further subdivide the sub-areas. If the coverage range of the equipment is less than 80% of the total area, it is considered that the coverage range of the equipment is insufficient and it is necessary to use an algorithm based on the Voronoi diagram to divide the sub-areas;

[0066] In the Voronoi algorithm, the locations of all equipment in the aquaculture waters are determined. The locations of the equipment will be used as the center points of the Voronoi division. The center point set is S = {s 1 ,s 2 , ..., s K}, where s k =(k=1, 2, ..., K) represents the location coordinates of the kth device, K is the number of devices, and the Euclidean distance d(P, s k )Measure the distance from any point P to s in the water area k The distance is as follows: Among them, x′ P is the horizontal coordinate of the point P in the water area, y′ P is the ordinate of the midpoint P in the water area, is the horizontal coordinate of the kth device, is the ordinate of the kth device; the Voronoi diagram divides the water area into several sub-areas, each of which is a Voronoi cell V(s k ), represents the area controlled by the kth device, and each Voronoi cell is defined as follows:

[0067] in,

[0068] R 2 Refers to the physical space where the crab farming waters are located, using two-dimensional coordinates to represent the specific location of each point. k is the effective radius of the device;

[0069] The control area of ​​each device is determined by calculating the distance between any point P in the water area and all devices; the Voronoi diagram divides the control area into polygons, and the boundary of each polygon is on the perpendicular bisector between the two devices, ensuring that each point is in the same area as its nearest device.

[0070] According to step S300, the historical water quality data of each region is used as an input sequence and input into the LSTM time series model, where the historical water quality data includes a number of consecutive time steps, reflecting the change of water quality over time;

[0071] Use the collected historical water quality data to train the LSTM time series model, learn the law of water quality parameters changing over time, and capture the dependency of water quality changes; establish independent LSTM models for different regions and adjust the model's hyperparameters;

[0072] In this embodiment, the input data of the LSTM time series model is the historical water quality data of each region, including water temperature, pH value, dissolved oxygen concentration and ammonia nitrogen concentration, with a collection frequency of once a day and a time span of 180 days. The model settings are as follows: time step: 30 days, hidden layer neurons: 64, learning rate: 0.001, batch size: 32, optimizer: Adam, loss function: mean square error MSE.

[0073] The historical data is divided into several time steps and represented as a multidimensional time series, where the data of each time step is a multidimensional vector containing different water quality parameters. LSTM is a recurrent neural network that uses a gating mechanism to memorize and forget long-term dependency information in the time series. The input of LSTM is the historical time series, and the output is the predicted value of the water quality parameter. To establish independent LSTM models for different regions, it is necessary to adjust the hyperparameters of the model, including: the number of hidden layer neurons, time step, learning rate, and batch size. The mean square error (MSE) is used as the loss function to measure the difference between the predicted value and the true value. The gradient descent method Adam optimizer is used to minimize the loss function and adjust the model parameters.

[0074] After 500 iterations of training, the mean square error (MSE) of the LSTM model on the test data set reached 0.015, indicating that the model has a high accuracy in predicting future water quality parameters. The errors between the predicted results and the actual values ​​of water temperature, pH value, dissolved oxygen and ammonia nitrogen concentration in the next 10 days are: water temperature ±0.5℃, pH value ±0.02, dissolved oxygen concentration ±0.1mg / L, and ammonia nitrogen concentration ±0.05mg / L.

[0075] The trained LSTM model can predict water quality parameters in the future, and the model outputs the predicted water quality values ​​in several future time steps.

[0076] The parameters of the water quality change trend include: water temperature change trend ΔT w , pH value change trend ΔpH, dissolved oxygen change trend ΔDO, ammonia nitrogen concentration change trend ΔNH 3 ; According to the crab's movements, the crab's behavioral characteristics are extracted, including: activity frequency AF and feeding behavior EB; the input feature vector of the support vector machine consists of the water quality change trend and the crab's behavioral data. The goal is to predict the impact function of environmental data on crab behavior based on the input feature vector X

[0077] f(E i ), describing how changes in water quality affect crab behavior, is expressed as:

[0078] Effect=f(ΔT w );

[0079] By finding an optimal hyperplane, the input feature vector X is classified into two categories: "normal behavior" and "abnormal behavior". The classification decision function of the support vector machine is expressed as: F = w T X+b, where w is the weight vector of the support vector machine model, representing the weight of the impact of different input features on the crab behavior, and b is the bias term, representing the intercept of the classification;

[0080] The settings of the support vector machine are as follows: Sample data: A total of 540 sets of data were collected, each set of data consisted of the above 6 features, and the label was the classification result of the crab behavior, including normal behavior and abnormal behavior. Regularization parameter C: 1.0, kernel function: linear kernel.

[0081] The training goal of the support vector machine model is to maximize the classification interval while minimizing the classification error. The optimization objective function is:

[0082]

[0083] Among them, C is the regularization parameter, which controls the trade-off between the classification interval and misclassification, and y i′is the label of the i′th sample, indicating whether the crab behavior corresponding to the sample is normal or abnormal, X i′ is the feature vector of the i′th sample;

[0084] The accuracy of the training set was 98.2%, and the accuracy of the test set was 96.5%. In the weight vector w of the classification decision function, the weight of the water temperature change trend was 0.35, the weight of the pH value change trend was 0.25, the weight of the dissolved oxygen change trend was 0.20, the weight of the ammonia nitrogen concentration change trend was 0.15, the weight of the activity frequency AF was 0.10, and the weight of the feeding behavior EB was 0.05. The support vector machine model can accurately distinguish between normal and abnormal behaviors of crabs, and is particularly effective in areas with large changes in water quality.

[0085] After the model training is completed, the new water quality change trend and crab behavior data are used as input to generate the impact function of environmental data on crab behavior and determine whether the crab behavior is abnormal.

[0086] Through the LSTM and SVM models, we can derive the impact function of environmental data on crab behavior. We found that when the water temperature suddenly rises by more than 2°C, the pH value changes by more than 0.2 units, or the dissolved oxygen concentration drops by more than 0.5 mg / L, the abnormal behavior of crabs increases significantly. The model can predict possible water quality deterioration 3 days in advance and remind managers to intervene, reducing the probability of abnormal crab behavior caused by water quality changes and improving the success rate of breeding.

[0087] According to step S400, based on the influence function of environmental data on crab behavior, the dissolved oxygen concentration in the water is monitored. i and crab activity frequency AF, if O i If the concentration of dissolved oxygen in the water is lower than the set threshold and the crab activity frequency is abnormal, a control signal is generated; the oxygenator is started or the working intensity is increased to increase the dissolved oxygen concentration in the water, and the dissolved oxygen changes are monitored in real time until O i Reach normal level; when the ammonia nitrogen concentration A i When the crab's feeding behavior EB is abnormal, if A i If the set threshold is exceeded and the crabs are eating abnormally, a control signal is generated to start the water exchange system, replace the polluted water, and monitor the changes in ammonia nitrogen concentration in real time until it returns to the threshold level; when the water temperature T i When the water temperature is lower than the set threshold and the crab activity frequency AF is abnormal, start or increase the temperature setting of the heating device to raise the water temperature to a range suitable for crab growth.

[0088] According to another embodiment of the present invention, Figure 2 As shown in the system structure diagram of an intelligent control system applied to the field of aquaculture, an intelligent control system applied to the field of aquaculture includes:

[0089] Data acquisition module: including: sensor data acquisition unit and image data acquisition unit; wherein the sensor data acquisition unit collects environmental data of the aquaculture waters in real time through sensors, including: water temperature, pH value, dissolved oxygen concentration, ammonia nitrogen concentration, air temperature and air humidity; the image data acquisition unit uses a camera to monitor the activities of crabs and collect image data of crab activities;

[0090] Data association analysis module: including: data preprocessing unit, convolutional neural network action recognition unit, Voronoi region division unit and action-environment association analysis unit; wherein, the data preprocessing unit preprocesses the collected environmental data and image data, filters out noise data and redundant data, the convolutional neural network action recognition unit uses a convolutional neural network to build an action recognition model based on image data sets covering different crab actions for image data, and recognizes the crab's actions, the Voronoi region division unit uses the Voronoi algorithm to divide the aquaculture waters into several areas according to the distribution of aquaculture equipment, and the action-environment association analysis unit uses the action-environment association model to associate environmental data of different areas with the crab's actions;

[0091] Water quality prediction and behavior analysis module: including: LSTM water quality prediction unit and SVM behavior analysis unit; the LSTM water quality prediction unit uses the LSTM time series model to analyze the environmental data of the aquaculture waters based on historical water quality data, and predicts the water quality change trends in different areas of the aquaculture waters. The SVM behavior analysis unit uses the support vector machine classification algorithm to combine the water quality change trends in different areas, generate the impact function of environmental data on crab behavior, and determine whether the crabs in the aquaculture waters have abnormal behavior;

[0092] Equipment control module: including: equipment control unit; wherein, the equipment control unit automatically controls the aerator, water exchange system and heating equipment in the corresponding area according to whether the crabs in different areas have abnormal behavior, so as to achieve the water quality change trend in each area to meet the health status of the crabs.

[0093] In this embodiment, three areas of crab breeding waters are selected, and each area is equipped with an aerator, a water exchange system and a heating device. When the dissolved oxygen concentration is lower than 6.0 mg / L, the aerator is automatically started until the dissolved oxygen concentration returns to above 6.5 mg / L. When the ammonia nitrogen concentration is higher than 0.5 mg / L, the water exchange system is started to control the ammonia nitrogen concentration below 0.3 mg / L. When the water temperature is lower than 18°C ​​or higher than 28°C, the heating or cooling system is started to maintain the water temperature between 20°C and 26°C.

[0094] The water temperature in area 1 is 27.5℃, pH is 7.5, dissolved oxygen concentration is 5.5mg / L, and ammonia nitrogen concentration is 0.4mg / L; the water temperature in area 2 is 20℃, pH is 7.2, dissolved oxygen concentration is 6.0mg / L, and ammonia nitrogen concentration is 0.2mg / L; the water temperature in area 3 is 17℃, pH is 7.1, dissolved oxygen concentration is 6.5mg / L, and ammonia nitrogen concentration is 0.6mg / L; through the SVM classification model detection, the crab behaviors in each area are as follows: the crab activity frequency AF in area 1 is abnormal, the crab feeding behavior EB in area 3 is abnormal, and the crab behavior in area 2 is normal.

[0095] The equipment control unit controls the area 1: the dissolved oxygen concentration is too low and the crab activity frequency is abnormal, the aerator starts automatically, and the dissolved oxygen concentration rises from 5.5mg / L to 6.5mg / L after 10 minutes, and the crab activity returns to normal. For area 2: the water quality and crab behavior are normal, and there is no need to control the equipment. For area 3: the ammonia nitrogen concentration is high and the feeding behavior is abnormal, the water exchange system is started, and the ammonia nitrogen concentration drops from 0.6mg / L to 0.3mg / L after 30 minutes, and the feeding behavior returns to normal.

[0096] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An intelligent control method applied to the field of breeding, characterized in that: The method comprises the following steps: S100, collecting environmental data of the aquaculture waters in real time through sensors, including: water temperature, pH value, dissolved oxygen concentration, ammonia nitrogen concentration, air temperature and air humidity; using cameras to monitor the activities of crabs and collect image data of crab activities; S200, preprocessing the collected environmental data and image data, filtering out noise data and redundant data; using a convolutional neural network to build an action recognition model based on an image data set covering different crab actions, and identifying the crab's actions; using a Voronoi algorithm to divide the aquaculture waters into several areas according to the distribution of aquaculture equipment, and using an action-environment association model to associate the environmental data of different areas with the crab's actions; S300, using the LSTM time series model, the environmental data of the aquaculture waters are analyzed based on the historical water quality data to predict the water quality change trends in different areas of the aquaculture waters; using the support vector machine classification algorithm combined with the water quality change trends in different areas, the influence function of the environmental data on the crab behavior is generated to determine whether the crabs in the aquaculture waters have abnormal behavior; S400, according to whether the crabs in different areas have abnormal behavior, automatically adjust the aerator, water exchange system and heating equipment in the corresponding area to ensure that the water quality change trend in each area meets the healthy state of the crabs.

2. The intelligent control method applied to the aquaculture field according to claim 1, characterized in that: According to step S100, a water temperature sensor is used to monitor the temperature change of the water body in the aquaculture water area in real time, and the real-time value of the water temperature is recorded, and the temperature unit is Celsius; The pH sensor is used to monitor the acidity and alkalinity of the water body and record the pH value of the water body; the dissolved oxygen sensor detects the oxygen content in the water body and records the dissolved oxygen concentration in the water in milligrams per liter; the ammonia nitrogen sensor detects the ammonia nitrogen content in the water and records the ammonia nitrogen concentration in milligrams per liter; the air temperature sensor monitors the temperature of the air around the aquaculture waters and records the air temperature value in degrees Celsius; the air humidity sensor monitors the relative humidity of the air in the aquaculture area and records the percentage value of the air humidity.

3. The intelligent control method applied to the aquaculture field according to claim 1, characterized in that: According to step S200, an image dataset covering different crab actions is established, wherein the crab actions include crawling, swimming, eating, resting, attacking and hiding; A convolutional neural network is used to input image data of crab activities of a fixed size. Local features in the image are gradually extracted through several convolutional layers. The convolutional layers are used to detect the features of the crab's body parts and limb movements. The first convolutional layer detects the edge contours of the crab's limbs. The second and subsequent convolutional layers gradually extract the movement of the crab's claws, rotation of the body, and changes in posture. The output of the convolutional layer is downsampled through the pooling layer to reduce the image size and retain relevant information about the action. After the convolutional layer and the pooling layer, a fully connected layer is added. The fully connected layer converts the extracted action features into classification features related to the crab action category. The output of the model is converted into crab actions through the Softmax activation function.

4. The intelligent control method applied to the aquaculture field according to claim 3 is characterized in that: On the image dataset covering different crab actions, the corresponding action of each image is annotated to form a training set and a test set; the annotated images are input into the convolutional neural network model and trained using the supervised learning method. The training process requires optimizing the model parameters so that the output action prediction accuracy is gradually improved; the cross entropy loss function is used to evaluate the error between the model output and the true label, and the Adam optimization algorithm is used to gradually adjust the model parameters; the learning rate is dynamically adjusted according to the training process; After the training is completed, the accuracy of the model is verified and evaluated; the test data that has not participated in the training is input into the model, and the recognition accuracy and recall rate of the model on different actions are calculated to further optimize the model; The collected image data of crab activities is transmitted in real time and input into the trained convolutional neural network model. Based on the feature extraction and classification algorithm, the current movements of different crabs are obtained.

5. The intelligent control method applied to the aquaculture field according to claim 4 is characterized in that: According to step S200, the aquaculture waters are divided into several areas, each area is divided according to the distribution of aquaculture equipment, and the principle of area division is as follows: according to the coverage of the equipment, the aquaculture waters are divided into N sub-areas, each area Z i (i=1, 2, ..., N) all have independent environmental monitoring and control equipment; Each area Z i Environmental data i Including water temperature T i pH i , dissolved oxygen concentration O i , ammonia nitrogen concentration A i , air temperature TAir i and air humidity i ; Monitor the crabs in each area through the camera to obtain the movements of the crabs in the area. The movement data of the crabs in each area is expressed as: A i =(a1, a2, ..., a m ), where a1 to a m Indicates area Z i The actions of the 1st to mth crabs in the area, where m is the number of monitored crabs in the area; The environmental data of different regions i Action data with crabs i Perform association analysis and build an action-environment association model, which is expressed using the following formula: Action i =Effect+∈, where Effect is the influence function of environmental data on crab behavior, and ∈ is the random error term.

6. The intelligent control method applied to the aquaculture field according to claim 5, characterized in that: The coverage of each area depends on the effective radius of the equipment, which includes aerators, water exchange systems and heating equipment. The effective radius of the aerator is R oxy , the radius of action of the water exchange system is R water , the effective radius of the heating device is R heat , calculate the coverage area of ​​the aerator Area oxy , Water exchange system coverage area water Area covered by heating equipment heat ; When dividing the area, the coverage of the equipment is set as the priority to ensure that all crabs in each sub-area are within the range of the corresponding equipment; if the coverage of the equipment is greater than or equal to Q% of the total area of ​​the water area, the coverage of the equipment is considered sufficient and there is no need to further subdivide the sub-areas; if the coverage of the equipment is less than Q% of the total area, the coverage of the equipment is considered insufficient and it is necessary to use an algorithm based on the Voronoi diagram to divide the sub-areas, where Q is a positive integer set by the user; In the Voronoi algorithm, the locations of all equipment in the aquaculture waters are determined. The locations of the equipment will be used as the center points of the Voronoi division. The center point set is S = {s1, s2, ..., s K },in, s k =(k=1, 2, ..., K) represents the location coordinates of the kth device, K is the number of devices, and the Euclidean distance d(P, s k )Measure the distance from any point P to s in the water area k The distance is as follows: Among them, x′ P is the horizontal coordinate of the point P in the water area, y′ P is the ordinate of the midpoint P in the water area, is the horizontal coordinate of the kth device, is the ordinate of the kth device; the Voronoi diagram divides the water area into several sub-areas, each of which is a Voronoi cell V(s k ), represents the area controlled by the kth device, and each Voronoi cell is defined as follows: in, R 2 Refers to the physical space where the crab farming waters are located, using two-dimensional coordinates to represent the specific location of each point. k is the effective radius of the device; The control area of ​​each device is determined by calculating the distance between any point P in the water area and all devices; the Voronoi diagram divides the control area into polygons, and the boundary of each polygon is on the perpendicular bisector between the two devices, ensuring that each point is in the same area as its nearest device.

7. The intelligent control method applied to the aquaculture field according to claim 1, characterized in that: According to step S300, the historical water quality data of each region is used as an input sequence and input into the LSTM time series model, where the historical water quality data includes a number of consecutive time steps, reflecting the change of water quality over time; Use the collected historical water quality data to train the LSTM time series model, learn the law of water quality parameters changing over time, and capture the dependency of water quality changes; establish independent LSTM models for different regions and adjust the model's hyperparameters; The trained LSTM model can predict water quality parameters in the future, and the model outputs the predicted water quality values ​​in several future time steps.

8. The intelligent control method applied to the aquaculture field according to claim 7, characterized in that: The parameters of the water quality change trend include: water temperature change trend ΔT w , pH value change trend ΔpH, dissolved oxygen change trend ΔDO, ammonia nitrogen concentration change trend ΔNH3; according to the crab's actions, the crab's behavioral characteristics are extracted, including: activity frequency AF and feeding behavior EB; the input feature vector of the support vector machine is composed of water quality change trend and crab behavior data. The goal is to predict the impact function f(E) of environmental data on crab behavior based on the input feature vector X i ), describing how changes in water quality affect crab behavior, is expressed as: Effect=f(ΔT w ); By finding an optimal hyperplane, the input feature vector X is classified into two categories: "normal behavior" and "abnormal behavior". The classification decision function of the support vector machine is expressed as: F = w T X+b, where w is the weight vector of the support vector machine model, representing the weight of the impact of different input features on the crab behavior, and b is the bias term, representing the intercept of the classification; The training goal of the support vector machine model is to maximize the classification interval while minimizing the classification error. The optimization objective function is: Among them, C is the regularization parameter, which controls the trade-off between the classification interval and misclassification, and y i is the label of the i′th sample, indicating whether the crab behavior corresponding to the sample is normal or abnormal. X i′ is the feature vector of the i′th sample; After the model training is completed, the new water quality change trend and crab behavior data are used as input to generate the impact function of environmental data on crab behavior and determine whether the crab behavior is abnormal.

9. The intelligent control method applied to the aquaculture field according to claim 1, characterized in that: According to step S400, based on the influence function of environmental data on crab behavior, the dissolved oxygen concentration in the water is monitored. i and crab activity frequency AF, if O i If the concentration of dissolved oxygen in the water is lower than the set threshold and the crab activity frequency is abnormal, a control signal is generated; the oxygenator is started or the working intensity is increased to increase the dissolved oxygen concentration in the water, and the dissolved oxygen changes are monitored in real time until O i Reach normal level; when the ammonia nitrogen concentration A i When the crab's feeding behavior EB is abnormal, if A i If the set threshold is exceeded and the crabs are eating abnormally, a control signal is generated to start the water exchange system, replace the polluted water, and monitor the changes in ammonia nitrogen concentration in real time until it returns to the threshold level; when the water temperature T i When the water temperature is lower than the set threshold and the crab activity frequency AF is abnormal, start or increase the temperature setting of the heating device to raise the water temperature to a range suitable for crab growth.

10. An intelligent control system applied to the aquaculture field, using an intelligent control method applied to the aquaculture field according to any one of claims 1 to 9, characterized in that: include: Data acquisition module: including: sensor data acquisition unit and image data acquisition unit; wherein the sensor data acquisition unit collects environmental data of the aquaculture waters in real time through sensors, including: water temperature, pH value, dissolved oxygen concentration, ammonia nitrogen concentration, air temperature and air humidity; the image data acquisition unit uses a camera to monitor the activities of crabs and collect image data of crab activities; Data association analysis module: including: data preprocessing unit, convolutional neural network action recognition unit, Voronoi region division unit and action-environment association analysis unit; wherein, the data preprocessing unit preprocesses the collected environmental data and image data, filters out noise data and redundant data, the convolutional neural network action recognition unit uses a convolutional neural network to build an action recognition model based on image data sets covering different crab actions for image data, and recognizes the crab's actions, the Voronoi region division unit uses the Voronoi algorithm to divide the aquaculture waters into several areas according to the distribution of aquaculture equipment, and the action-environment association analysis unit uses the action-environment association model to associate environmental data of different areas with the crab's actions; Water quality prediction and behavior analysis module: including: LSTM water quality prediction unit and SVM behavior analysis unit; the LSTM water quality prediction unit uses the LSTM time series model to analyze the environmental data of the aquaculture waters based on historical water quality data, and predicts the water quality change trends in different areas of the aquaculture waters. The SVM behavior analysis unit uses the support vector machine classification algorithm to combine the water quality change trends in different areas, generate the impact function of environmental data on crab behavior, and determine whether the crabs in the aquaculture waters have abnormal behavior; Equipment control module: including: equipment control unit; wherein, the equipment control unit automatically controls the aerator, water exchange system and heating equipment in the corresponding area according to whether the crabs in different areas have abnormal behavior, so as to achieve the water quality change trend in each area to meet the health status of the crabs.

Images