Macrobrachium rosenbergii breeding management system based on artificial intelligence

Abstract

The invention relates to the technical field of aquaculture, and discloses a macrobrachium rosenbergii breeding management system based on artificial intelligence. The system integrates five modules of data acquisition, intelligent diagnosis, prediction and early warning, automatic control and user interaction; the method comprises the following steps: continuously collecting multi-dimensional data streams such as physical and chemical parameters of a seedling culture water body and behavior images of shrimp seedling groups in real time, analyzing the vitality, distribution and form of the shrimp seedlings by utilizing a deep learning image recognition model, and performing prospective judgment on the future trend and disease outbreak risk of key water quality indexes based on a time sequence model. And the automatic control module automatically generates and executes an accurate closed-loop control instruction according to the diagnosis result and the early warning information. According to the invention, the traditional passive response type seedling culture management is converted into active prevention type intelligent closed-loop control, so that the objectivity of diagnosis, the accuracy of intervention and the automation level of management are improved, and the survival rate of seedlings is effectively guaranteed.

Description

Technical Field

[0001] This invention relates to the field of aquaculture technology, specifically to an artificial intelligence-based seedling management system for giant freshwater prawns. Background Technology

[0002] As a key freshwater aquaculture species globally, the giant freshwater prawn industry is rapidly transforming towards intensive, high-density, factory-style seedling production. While this model increases yield per unit area, it also makes the aquatic environment for seedling production exceptionally fragile. Even a minor imbalance can trigger a devastating disease outbreak, leading to large-scale losses of seedlings. Therefore, developing an intelligent management system capable of early warning, precise diagnosis, and automated intervention has become a core technological requirement for overcoming the industry's development bottlenecks and ensuring successful aquaculture. The purpose of this invention is to provide such a solution, through data-driven closed-loop control, to elevate the management model of aquaculture seedling production from a passive response to a proactive prediction approach.

[0003] In existing aquaculture practices, the industry has developed a set of routine management methods. Technicians use handheld or online sensors to measure key indicators in the seedling water at fixed points or times, providing quantitative readings for maintaining basic aquatic environmental stability. Simultaneously, through microscopic examination or laboratory pathological analysis, suspected samples can be tested after disease outbreaks, providing precise identification criteria for the final confirmation of pathogens and subsequent drug selection. The combined application of these technologies constitutes the basic framework for disease control in current giant freshwater prawn seedling production.

[0004] However, the inherent limitations of existing technological frameworks in addressing the complex challenges of high-density seedling cultivation are becoming increasingly apparent. First, the discontinuous nature of existing monitoring methods in the time domain leads to significant data blind spots. Sudden deterioration of water quality parameters often occurs between two manual tests, causing managers to miss the opportunity to respond proactively. Second, early disease identification relies heavily on subjective experience and lacks objective, quantitative diagnostic evidence. It is difficult for the human eye to effectively capture and continuously track key information such as subtle abnormalities in group behavior or early symptoms in individuals. More importantly, the existing technological framework is essentially a delayed response mechanism. It can only react to indicators exceeding limits or symptoms that have already appeared, and cannot proactively predict potential risks. In terms of intervention measures, the extensive model of whole-pond application is commonly used, which not only wastes drugs and pollutes the environment but also subjects healthy individuals to unnecessary drug stress. Finally, from an overall architectural perspective, data monitoring, manual diagnosis, and manual control are isolated, lacking effective closed-loop linkage between data, diagnosis, and control, resulting in low management efficiency and a high susceptibility to human error. To address these issues, those skilled in the art propose an artificial intelligence-based management system for giant freshwater prawns. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an artificial intelligence-based management system for giant freshwater prawns, which solves the technical defects of existing technologies, such as discontinuous monitoring, subjective diagnosis, delayed early warning, extensive intervention, and fragmentation of various management links.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a giant freshwater prawn (Macrobrachium rosenbergii) seedling management system based on artificial intelligence, the system comprising:

[0007] The data acquisition module is used to acquire environmental data, biological behavior data and non-contact physiological data of the giant freshwater prawns in the breeding water in real time.

[0008] The intelligent diagnostic module is used to assess the health status of the giant freshwater prawns based on environmental data, biological behavior data, and non-contact physiological data acquired by the data acquisition module, and to identify and track abnormal individuals in the nursery water.

[0009] The prediction and early warning module is used to predict the water quality change trend and disease risk in the future period based on historical and real-time environmental data, biological behavior data and non-contact physiological data, and generate early warning information.

[0010] An automated control module is used to automatically perform environmental regulation or precise intervention operations in response to the early warning information generated by the prediction and early warning module, or based on the evaluation results of the intelligent diagnosis module.

[0011] The user interaction module is used to display the data acquired by the data acquisition module, the evaluation results generated by the intelligent diagnosis module, the early warning information generated by the prediction and early warning module, and the control status of the automation control module.

[0012] Preferably, the data acquisition module is specifically used for:

[0013] The environmental data is acquired in real time by a water quality sensor, including water temperature, dissolved oxygen, pH value, ammonia nitrogen, nitrite and key ion concentrations.

[0014] The biological behavior data is obtained by acquiring images or video streams through a high-definition camera, including population distribution, activity level, and feeding behavior.

[0015] The non-contact physiological data is acquired using infrared thermal imaging or spectral analysis equipment, and the non-contact physiological data includes body surface temperature or fluorescence intensity of specific biomarkers.

[0016] Preferably, the intelligent diagnostic module is specifically used for:

[0017] Using a deep learning image recognition model, the developmental stage and health status of the giant freshwater prawn were assessed based on the biological behavior data;

[0018] The abnormal individuals are identified based on their body surface spots, body color, or behavioral characteristics, and the real-time location of the abnormal individuals in the seedling water is continuously monitored using a tracking algorithm.

[0019] A multimodal fusion model is constructed, which combines the environmental data, biological behavior data, and non-contact physiological data to generate disease risk assessment results.

[0020] Preferably, the prediction and early warning module is specifically used for:

[0021] Using time series prediction models, based on historical and real-time multidimensional data sequences, the changing trends of the environmental data, biological behavior data, and non-contact physiological data over a future period are predicted.

[0022] When the predicted trend exceeds a preset safety threshold, or when the disease risk assessment result generated by the intelligent diagnostic module exceeds a preset risk threshold, the warning information is generated and sent.

[0023] Preferably, the automation control module is specifically used for:

[0024] Based on water quality analysis and prediction results, calculate the dosage of substances required for environmental control and control the intelligent dosing equipment to perform the dosing operation.

[0025] Perform non-pharmacological interventions, including controlling a programmable lighting module or a variable water flow device;

[0026] The precise drug delivery intervention operation involves controlling a fixed-point drug delivery device to deliver medication to the local area where the abnormal individual is located, based on the real-time location of the abnormal individual.

[0027] Preferably, the user interaction module is specifically used for:

[0028] The environmental data, biological behavior data, non-contact physiological data, health status assessment results, disease risk assessment results, and early warning information are displayed in a visual manner.

[0029] Provides a remote control interface for manual intervention in the equipment controlled by the automated control module;

[0030] Record and store all collected data and intervention operations to form a historical database for model optimization.

[0031] In one specific embodiment, the intelligent diagnostic module assesses the group behavior of the giant freshwater prawns by calculating a population distribution index. The population distribution index is calculated as follows:

[0032] Obtain the center coordinates of all identified giant freshwater prawn individuals in the image, calculate the Euclidean distance between the center coordinates of each individual and the average of the center coordinates of all individuals, and sum all the Euclidean distances and take the average.

[0033] The population distribution index D group The calculation formula is:

[0034]

[0035] Where N is the total number of giant freshwater prawns identified in the image, (x i ,y i Let be the center coordinates of the i-th individual giant freshwater prawn. This represents the average center coordinates of all individual giant freshwater prawns.

[0036] In one specific embodiment, the time series prediction model is a long short-term memory network model or a gated recurrent unit model.

[0037] The multidimensional data sequence is an input sequence composed of environmental data, biological behavior data, and non-contact physiological data from a predetermined number of time steps in the past.

[0038] Based on the input sequence, the model outputs predicted values ​​of the environmental data, biological behavior data, and non-contact physiological data for a predetermined number of time steps in the future.

[0039] In one specific embodiment, the steps of the automated control module performing the precision drug delivery intervention include:

[0040] The intelligent diagnostic module receives the real-time location coordinates of the identified abnormal individual.

[0041] Based on the real-time location coordinates, the target area and drug dosage for precise intervention are determined;

[0042] Control the targeted drug delivery device to apply the drug dose to the target area.

[0043] In one specific embodiment, the non-pharmacological intervention performed by the automated control module includes:

[0044] In response to a diagnosed or predicted pathogen risk, the programmable illumination module is controlled to illuminate with specific wavelengths, intensities, and durations to inhibit pathogen growth.

[0045] In response to a diagnosed decrease in vitality or group aggregation, the variable flow device is controlled to adjust the flow speed and direction to induce activity in the giant freshwater prawns and improve water circulation.

[0046] This invention provides an artificial intelligence-based management system for giant freshwater prawns. It offers the following advantages:

[0047] 1. This invention constructs a multimodal comprehensive diagnostic model by integrating environmental data, biological behavior data, and non-contact physiological data, enabling early and accurate assessment of the health status of giant freshwater prawns. Compared with existing technologies that rely on monitoring single water quality parameters or manual visual inspection, this invention solves the shortcomings of delayed detection of diseases or stress states and lack of objective basis for judgment.

[0048] 2. This invention utilizes an intelligent diagnostic module to identify and track abnormal individuals in the seedling water body, and links an automated control module to execute precise, localized intervention operations. This limits the scope of intervention measures to a specific area, unlike existing technologies that uniformly administer drugs or treat the entire seedling water body. This solves the technical defects of existing technologies, such as drug waste due to excessive intervention scope, unnecessary stress on healthy individuals, and increased burden on the aquatic environment.

[0049] 3. This invention realizes the transformation of seedling management mode from passive response to active prevention. It uses a time series prediction model to predict the trend of water quality environment and disease risk, and can perform control operations in advance based on this, such as non-pharmaceutical light or water flow intervention. Existing technologies usually use threshold alarm mechanisms, which only trigger the response when the parameters have deteriorated. This invention solves the problem that existing technologies are difficult to maintain the long-term stability of the seedling environment due to the lag in intervention measures. Attached Figure Description

[0050] Figure 1 This is a system functional block diagram of the present invention;

[0051] Figure 2 This is a schematic diagram illustrating the functional structure of the data acquisition module of the present invention;

[0052] Figure 3 This is a schematic diagram illustrating the functional structure of the intelligent diagnostic module of the present invention;

[0053] Figure 4 This is a schematic diagram illustrating the functional structure of the prediction and early warning module of the present invention;

[0054] Figure 5 This is a schematic diagram illustrating the functional structure of the automation control module of the present invention;

[0055] Figure 6 This is a schematic diagram illustrating the functional structure of the user interaction module of the present invention;

[0056] Figure 7 This is a schematic diagram illustrating the change of the disease risk index over time and the system intervention of the present invention.

[0057] Among them, 10 is the data acquisition module; 20 is the intelligent diagnosis module; 30 is the prediction and early warning module; 40 is the automatic control module; and 50 is the user interaction module. Detailed Implementation

[0058] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] See attached document Figure 1 , Figure 1 This is a system functional block diagram according to an embodiment of the present invention. The present invention provides an artificial intelligence-based macrophage seedling management system, which may include: a data acquisition module 10, an intelligent diagnosis module 20, a prediction and early warning module 30, an automated control module 40, and a user interaction module 50.

[0060] In this embodiment, the data acquisition module 10 is configured to connect to multiple sensors and data acquisition devices in the seedling aquatic environment to acquire environmental data, biological behavior data of the giant freshwater prawn, and non-contact physiological data in real time. This acquired data, as a raw data stream, is transmitted to the intelligent diagnosis module 20 and the prediction and early warning module 30.

[0061] The intelligent diagnostic module 20 is connected to the data acquisition module 10 and is used to receive and process raw data streams. Based on the received data, the intelligent diagnostic module 20 assesses the health status of the giant freshwater prawns and identifies and tracks abnormal individuals in the nursery water. The generated assessment results and abnormal individual data are transmitted to the automation control module 40 and the user interaction module 50.

[0062] The prediction and early warning module 30 is also connected to the data acquisition module 10 to receive raw data streams. Based on historical and real-time multidimensional data, this module predicts water quality trends and disease risks over a future period. The generated prediction results and early warning information are transmitted to the automation control module 40 and the user interaction module 50.

[0063] The automated control module 40 is connected to the intelligent diagnostic module 20 and the prediction and early warning module 30, respectively, and is used to receive assessment results, abnormal individual data, prediction results, and early warning information. In response to the received information, the automated control module 40 generates specific control commands and outputs them to external environmental control equipment or precision intervention equipment to perform environmental control or precision intervention operations.

[0064] The user interaction module 50 establishes data connections with the data acquisition module 10, the intelligent diagnostic module 20, the prediction and early warning module 30, and the automation control module 40. The user interaction module 50 is used to centrally display all input data, processing results, early warning information, and equipment control status. It also receives user input commands and transmits these commands to the automation control module 40.

[0065] The system's workflow is as follows: Data acquisition module 10 serves as the data source, continuously collecting multi-dimensional data. Intelligent diagnostic module 20 and prediction / early warning module 30 process the collected data in parallel, analyzing it from two perspectives: current state assessment and future trend prediction. Automation control module 40 synthesizes the analysis results, makes control decisions, and drives external devices to perform specific operations. User interaction module 50 provides an interface for monitoring and manual intervention, thus forming a data-driven, automatically operating closed-loop control system.

[0066] To further clarify the technical solutions of each component in the above system, the specific composition, working principle and implementation method of the data acquisition module 10, intelligent diagnosis module 20, prediction and early warning module 30, automatic control module 40 and user interaction module 50 will be explained in detail below.

[0067] See attached document Figure 2 , Figure 2 This is a functional configuration diagram of the data acquisition module 10 according to an embodiment of the present invention. The data acquisition module 10 is the entry point for the system to acquire external information, and it is responsible for comprehensively and in real time acquiring multi-dimensional data from the seedling environment.

[0068] In one specific embodiment, the data acquisition module 10 acquires environmental data through multiple water quality sensors deployed in the seedling water body. These water quality sensors include a temperature sensor for measuring water temperature, a dissolved oxygen sensor for measuring dissolved oxygen concentration, a pH electrode for measuring pH value, an ammonia nitrogen sensor for measuring ammonia nitrogen concentration, and a nitrite sensor for measuring nitrite concentration. Furthermore, the module also includes ion-selective electrodes, such as calcium and magnesium ion electrodes, for acquiring the concentrations of key ions. All sensors are connected to a data logger or IoT gateway to acquire data at a fixed time frequency.

[0069] The data acquisition module 10 also includes a high-definition camera deployed above or on the side wall of the nursery pond to acquire biological behavior data. This high-definition camera continuously collects images or videos of the nursery water to record the population distribution, individual activity, and feeding behavior of the giant freshwater prawns. To ensure consistent image quality under different lighting conditions, the high-definition camera can be used in conjunction with a supplemental lighting module, which automatically turns on or off based on ambient light intensity or a preset schedule.

[0070] To acquire non-contact physiological data, the data acquisition module 10 includes an infrared thermal imaging device and a spectral analysis device. The infrared thermal imaging device is configured to capture infrared radiation from the surface of the giant freshwater prawn and calculate its surface temperature accordingly. The spectral analysis device, such as a fluorescence spectrometer, emits excitation light of a specific wavelength into the rearing water or onto the surface of the prawn and receives the emitted fluorescence signal to measure the fluorescence intensity of specific biomarkers.

[0071] The data acquisition module 10 integrates a data aggregation unit. All raw data collected by the aforementioned sensors and devices, including environmental data, image data, and physiological data, are transmitted to this data aggregation unit. This unit performs preliminary processing on the received data, including data denoising, format standardization, and adding a unified timestamp. The processed structured data is then transmitted in real-time to the intelligent diagnostic module 20 and the predictive early warning module 30 via wired or wireless communication.

[0072] See attached document Figure 3 , Figure 3 This is a functional configuration diagram of the intelligent diagnostic module 20 according to an embodiment of the present invention. The intelligent diagnostic module 20 is connected to the data aggregation unit of the data acquisition module 10 and is used to receive and process the transmitted structured data to perform health status assessment and abnormal individual identification and tracking of giant freshwater prawns.

[0073] In one specific embodiment, the intelligent diagnostic module 20 includes a deep learning image recognition model. This model is based on a convolutional neural network (CNN) architecture, and its input is the image or video stream acquired by the data acquisition module 10. After training, the model can analyze the input images and output multi-dimensional evaluation results: first, based on the morphological characteristics of the giant freshwater prawn, it outputs the developmental stage category; second, based on image features such as body color, appendage integrity, and activity posture, it outputs a quantitative health score; third, through target detection methods, it identifies and locates individuals in the image with abnormal surface spots, abnormal body color, or lethargic behavior, and outputs the location coordinates of these individuals.

[0074] Based on the recognition results of the deep learning image recognition model, the intelligent diagnosis module 20 further calculates the population distribution index to quantitatively assess the group behavior of the giant freshwater prawn. The calculation steps of the population distribution index are as follows: First, obtain the center coordinates of all identified giant freshwater prawn individuals in a single frame image; then, calculate the Euclidean distance between the center coordinates of each individual and the arithmetic mean of the center coordinates of all individuals; finally, sum all the calculated Euclidean distances and take the average.

[0075] Population distribution index D group The calculation formula is:

[0076]

[0077] Where N is the total number of giant freshwater prawns identified in the image, (x i ,y i Let be the center coordinates of the i-th individual giant freshwater prawn. This represents the average center coordinates of all individual giant freshwater prawns.

[0078] Once the deep learning image recognition model identifies an abnormal individual, the intelligent diagnosis module 20 initiates a tracking algorithm to continuously track the location of these abnormal individuals. This tracking algorithm can employ a Kalman filter or a Siamese Region Proposal Network (SiamRPN) model. The algorithm takes the location of the abnormal individual identified in the current frame as input and combines it with motion trajectory information from previous frames to predict its location in the next frame, thereby achieving continuous monitoring of the motion trajectory of specific abnormal individuals in the seedling cultivation water body. The tracking results, i.e., the real-time location data of the abnormal individuals, are transmitted to the automation control module 40.

[0079] The intelligent diagnostic module 20 also constructs a multimodal fusion model to generate comprehensive disease risk assessment results. This model receives an input vector V composed of data from multiple dimensions. input The input vector V input This includes: water temperature, dissolved oxygen, pH, ammonia nitrogen, nitrite, key ion concentration, body surface temperature, and biomarker fluorescence intensity acquired by data acquisition module 10, as well as health score and population distribution index D calculated by this module. group The multimodal fusion model outputs a disease risk index R. risk Its calculation can be expressed as:

[0080] R risk =f(V input );

[0081] Among them, V input Let f be the input vector containing all the above parameters, and let f represent the trained multimodal fusion model function. The disease risk index R... riskIt is a quantitative value that represents the overall risk level of disease outbreaks in the current seedling system. This index is transmitted to the prediction and early warning module 30 and the automation control module 40.

[0082] See attached document Figure 4 , Figure 4 This is a functional configuration diagram of the prediction and early warning module 30 according to an embodiment of the present invention. The prediction and early warning module 30 is connected to the data acquisition module 10 and the intelligent diagnosis module 20, and is used to receive real-time and historical data, and perform future trend prediction and risk warning.

[0083] In one specific embodiment, the core of the prediction and early warning module 30 is a time series prediction model. This model employs a recurrent neural network architecture using either a Long Short-Term Memory (LSTM) network or a Gated Recurrent Unit (GRU). This architecture is chosen to process the multidimensional time series data in this invention and to capture long-term dependencies between data points.

[0084] The input to this time series forecasting model is a multidimensional data series S. input This sequence is constructed using a sliding window approach, combining multidimensional data from a predetermined number (k) of time steps. At each time step t, its data vector X... t This includes: environmental data, biological behavioral data, non-contact physiological data, and health scores and disease risk indices generated by the intelligent diagnostic module 20. Therefore, the input sequence can be represented as:

[0085] S input =[X t-k+1 ,X t-k+2 ,…,X t ];

[0086] Among them, X i Let be the multidimensional data vector at the i-th time step.

[0087] The model is based on the input sequence S input Output a predicted sequence S output This sequence contains multidimensional data predictions for a predetermined number of time steps (denoted as m). The output sequence can be represented as:

[0088]

[0089] in, Let be the predicted value of the multidimensional data vector at the j-th future time step. The model is trained on a large amount of historical seedling data and optimized using mean squared error as the loss function to make its predicted values ​​close to the true values.

[0090] The prediction and early warning module 30 also includes an early warning generation unit. This unit generates early warning information based on two independent conditions.

[0091] The first condition is that the predicted value exceeds the limit: the unit will predict sequence S. output The predicted value of each parameter in the sequence (e.g., the predicted ammonia nitrogen value at the j-th step) is compared with a pre-set safety threshold for that parameter. An alert is triggered when any predicted value in the sequence exceeds its corresponding safety threshold.

[0092] The second condition is that the current risk exceeds the limit: This unit receives the disease risk index R calculated in real time by the intelligent diagnosis module 20. risk It is then compared with a preset risk threshold. When the disease risk index R... risk When the value exceeds the preset risk threshold, an alert will also be triggered.

[0093] When an early warning is triggered, the early warning generation unit generates structured early warning information. This information includes the early warning level, risk type (e.g., water quality deterioration warning or disease risk warning), the specific parameter name that triggered the warning, the predicted exceedance value, and the expected occurrence time. The generated early warning information is simultaneously transmitted to the automation control module 40 and the user interaction module 50.

[0094] See attached document Figure 5 , Figure 5 This is a functional configuration diagram of an automation control module 40 according to an embodiment of the present invention. The automation control module 40, as the system's execution terminal, is connected to the intelligent diagnostic module 20 and the prediction and early warning module 30, respectively, for receiving diagnostic results, prediction data, and early warning information, and generating control commands accordingly to drive external execution devices.

[0095] When the automation control module 40 receives a water quality deterioration warning from the prediction and early warning module 30, it initiates environmental control operations. In one specific embodiment, the module calculates the dosage of substances required for environmental control based on the predicted deviation of water quality parameters. For example, when it is predicted that the pH value will fall below the safety threshold at a future time, the module calculates the required dosage M of an alkaline regulator (such as sodium bicarbonate). The calculation formula can be:

[0096] M = K·(V) target -V predicted )·V water ;

[0097] Where K is the dosage coefficient associated with a specific regulator; V target Target pH value; V predicted For the predicted pH value; V waterThis represents the total volume of the seedling cultivation water. After calculation, the automated control module 40 sends control commands to the intelligent dosing device connected to the system (e.g., a peristaltic pump driven by a stepper motor). The commands include the total dosing amount and the dosing speed to execute the dosing operation.

[0098] The automated control module 40 also performs non-pharmacological interventions. When it receives a specific pathogen risk warning from the predictive warning module 30, the module queries a preset control strategy database and controls the programmable illumination module. For example, for a Vibrio risk warning, the module controls the illumination module to provide short-term irradiation with ultraviolet light of a specific wavelength (e.g., 254 nm) to inhibit the proliferation of pathogens in the water. When it receives a warning from the intelligent diagnostic module 20 indicating a decrease in the vitality of Macrobrachium rosenbergii or a population distribution index D... group When an abnormal (indicating group aggregation) diagnosis is received, the module controls variable flow equipment (e.g., variable frequency pumps or multi-angle nozzles) to change the speed and direction of the water flow by adjusting the operating frequency of the pump or changing the angle of the nozzles, in order to induce the activity of giant freshwater prawns and improve water circulation.

[0099] In addition, the automated control module 40 performs precise drug delivery intervention. The execution process of this operation is as follows:

[0100] Step 1: The module receives the real-time location coordinates (x, y) of the continuously tracked abnormal individual from the intelligent diagnosis module 20.

[0101] Step two: Based on the location coordinates, the module determines a precise intervention target area, for example, a circular area centered at (x,y) with a radius of R. Simultaneously, based on the diagnostic conclusion made by the intelligent diagnostic module 20 for the abnormal individual (e.g., a specific type of surface attachment), the module retrieves and determines the required drug type and dosage D from the drug database.

[0102] Step three: The module converts the target area coordinates and drug dosage D into specific motion and execution commands for the targeted drug delivery device. This device can be an XY-axis robotic arm deployed above the seedling pond. The module controls the robotic arm to move directly above the target area and drives the micro-injection pump at its end to apply the drug dosage D to the target area.

[0103] See attached document Figure 6 , Figure 6 This is a functional configuration diagram of a user interaction module 50 according to an embodiment of the present invention. The user interaction module 50 provides a centralized monitoring and management interface for the system, and it establishes data connections with the data acquisition module 10, the intelligent diagnosis module 20, the prediction and early warning module 30, and the automatic control module 40.

[0104] In one specific embodiment, the user interaction module 50 displays the system's operating status visually through a graphical user interface (GUI). For data acquired by the data acquisition module 10, the interface presents it in a multi-window or regional format: environmental data is displayed as real-time values ​​and historical trend curves (e.g., 24-hour dissolved oxygen change curve); biological behavior data is displayed by playing a video stream captured by a high-definition camera in real time, with annotation information generated by the intelligent diagnosis module 20 superimposed on the video screen, such as drawing bounding boxes for each identified giant freshwater prawn individual and marking individuals identified as abnormal with a highlighted color or a specific label; non-contact physiological data is displayed in the form of pseudo-color thermal imaging or fluorescence intensity histograms.

[0105] The user interaction module 50 also visualizes the analysis results generated by the intelligent diagnosis module 20 and the prediction and early warning module 30. For example, the health status assessment result is quantified into a percentage health index and displayed in a dashboard-style control; the disease risk assessment result is visually represented by a color bar (e.g., from green to red) to indicate the current risk level. When an early warning message is received, a separate early warning window pops up on the interface, clearly listing the warning level, type, content, and suggested actions.

[0106] The user interaction module 50 also provides a remote control interface for manual intervention in the equipment controlled by the automation control module 40. This interface includes virtual control components corresponding to each physical device, such as a slider for adjusting the dosing speed of the intelligent dosing device, a switch button for turning the programmable lighting module on or off, and a text box for inputting target coordinates to control the movement of the fixed-point dosing device. User commands issued through these components are converted into standardized control commands by this module and sent to the automation control module 40 for execution.

[0107] In addition, the user interaction module 50 includes a data recording unit connected to a background database. All raw data collected by the data acquisition module 10, intermediate and final results generated by the intelligent diagnosis module 20 and the prediction and early warning module 30, and every control operation (including automatic execution and manual intervention) performed by the automation control module 40 are precisely timestamped and recorded in the background database. This forms a complete historical database, which is not only used for event tracing and data querying, but also serves as a dataset for subsequent retraining and optimization of the algorithm models in the intelligent diagnosis module 20 and the prediction and early warning module 30.

[0108] To further illustrate the collaborative working process of the technical solution of this invention, a specific working scenario example will be used below.

[0109] See attached document Figure 7 , Figure 7 This is a graph showing the changes in ammonia nitrogen concentration and disease risk index over time during the seedling stage according to an embodiment of the present invention.

[0110] The system of this invention operates continuously during the middle and later stages of a giant freshwater prawn breeding cycle. In the initial stage ( Figure 7 During the time period from t0 to t1, the data acquisition module 10 acquires various data in real time. At this time, the ammonia nitrogen concentration measured by the water quality sensor remains below the safe threshold, and the intelligent diagnosis module 20 analyzes and derives the population distribution index D. group Both the health score and other indicators are within the normal range, and the calculated disease risk index R... risk It remains at a low level below 0.2.

[0111] At time t1, the data acquisition module 10 detected a continuous, slight upward trend in the ammonia nitrogen concentration in the water. Simultaneously, the acquired high-definition video stream showed a decrease in the activity frequency of some giant freshwater prawns. Upon receiving this continuous data, the time-series prediction model within the prediction and early warning module 30, based on this upward trend and historical data, outputs a prediction: the ammonia nitrogen concentration will exceed the preset first-level safety threshold within the next 24 hours.

[0112] At the same time, the intelligent diagnostic module 20 analyzes the received image data and calculates the population distribution index D. group The 10% increase compared to the baseline indicates a clustering trend within the population. Its multimodal fusion model incorporates inputs such as a slight increase in ammonia nitrogen concentration, changes in the population distribution index, and non-contact physiological data (e.g., a slight increase in average body surface temperature) to calculate the disease risk index R. risk It rose to 0.4, exceeding the Level 1 risk threshold of 0.3. Figure 7 (between t1 and t2).

[0113] Based on the above prediction and diagnosis results, the prediction and early warning module 30 generates a level-one early warning message and transmits it to the automation control module 40 and the user interaction module 50. The early warning message is then displayed on the interface of the user interaction module 50, showing the predicted trend of ammonia nitrogen concentration and the real-time changes of the disease risk index in the form of a curve graph.

[0114] Upon receiving a Level 1 warning, the automated control module 40 executes a preset non-pharmaceutical intervention strategy. It sends control commands to the variable flow equipment to increase the water circulation speed by 15% to enhance water reoxygenation efficiency and promote the decomposition of organic matter.

[0115] At time t2, the system detected that the upward trend of ammonia nitrogen concentration had slowed but not stopped, and the intelligent diagnostic module 20, through its deep learning image recognition model, successfully identified and located an individual with tiny abnormal spots on its body surface. The module then activated its tracking algorithm, continuously outputting the real-time location coordinates of the abnormal individual. At this time, the disease risk index R... risk It rose further to 0.6, triggering a level-two warning.

[0116] Upon receiving the secondary early warning information and the real-time coordinates of the abnormal individual, the automated control module 40 immediately initiates a precise drug delivery intervention. It controls the fixed-point drug delivery device to move directly above the location of the abnormal individual and delivers a preset dose of antibacterial drug to the target area.

[0117] After performing the above series of interventions ( Figure 7 After time t2, the system continued monitoring. Data showed that the ammonia nitrogen concentration began to decline after briefly exceeding the first-level safety threshold. Monitoring results from the intelligent diagnostic module 20 showed that no new abnormal individuals appeared in the water body, and the population distribution index D remained unchanged. group Gradually recovering to normal levels, the disease risk index R risk This also caused it to drop below the safety threshold. All data, diagnostic results, warning records, and control operations for the entire event were recorded by the user interaction module 50 and stored in the historical database.

Claims

1. A macrophage seedling management system based on artificial intelligence, characterized in that, include: The data acquisition module is used to acquire environmental data, biological behavior data and non-contact physiological data of the giant freshwater prawns in the breeding water in real time. The intelligent diagnostic module is used to assess the health status of the giant freshwater prawns based on environmental data, biological behavior data, and non-contact physiological data acquired by the data acquisition module, and to identify and track abnormal individuals in the nursery water. The prediction and early warning module is used to predict the water quality change trend and disease risk in the future period based on historical and real-time environmental data, biological behavior data and non-contact physiological data, and generate early warning information. An automated control module is used to automatically perform environmental regulation or precise intervention operations in response to the early warning information generated by the prediction and early warning module, or based on the evaluation results of the intelligent diagnosis module. The user interaction module is used to display the data acquired by the data acquisition module, the evaluation results generated by the intelligent diagnosis module, the early warning information generated by the prediction and early warning module, and the control status of the automation control module.

2. The artificial intelligence-based macrophage seedling management system according to claim 1, characterized in that, The data acquisition module is specifically used for: The environmental data is acquired in real time by a water quality sensor, including water temperature, dissolved oxygen, pH value, ammonia nitrogen, nitrite and key ion concentrations. The biological behavior data is obtained by acquiring images or video streams through a high-definition camera. The biological behavior data includes population distribution, activity level, and feeding behavior. The non-contact physiological data is acquired using infrared thermal imaging or spectral analysis equipment, and the non-contact physiological data includes body surface temperature or fluorescence intensity of specific biomarkers.

3. The artificial intelligence-based macrophage seedling management system according to claim 1, characterized in that, The intelligent diagnostic module is specifically used for: Using a deep learning image recognition model, the developmental stage and health status of the giant freshwater prawn were assessed based on the biological behavior data; The abnormal individuals are identified based on their body surface spots, body color, or behavioral characteristics, and the real-time location of the abnormal individuals in the seedling water is continuously monitored using a tracking algorithm. A multimodal fusion model is constructed, which combines the environmental data, biological behavior data, and non-contact physiological data to generate disease risk assessment results.

4. The artificial intelligence-based macrophage seedling management system according to claim 1, characterized in that, The prediction and early warning module is specifically used for: Using time series prediction models, based on historical and real-time multidimensional data sequences, the changing trends of the environmental data, biological behavior data, and non-contact physiological data over a future period are predicted. When the predicted trend exceeds a preset safety threshold, or when the disease risk assessment result generated by the intelligent diagnostic module exceeds a preset risk threshold, the warning information is generated and sent.

5. The artificial intelligence-based macrophage seedling management system according to claim 1, characterized in that, The automated control module is specifically used for: Based on water quality analysis and prediction results, calculate the dosage of substances required for environmental control and control the intelligent dosing equipment to perform the dosing operation. Perform non-pharmacological interventions, including controlling a programmable lighting module or a variable water flow device; The precise drug delivery intervention operation involves controlling a fixed-point drug delivery device to deliver medication to the local area where the abnormal individual is located, based on the real-time location of the abnormal individual.

6. The artificial intelligence-based macrophage seedling management system according to claim 1, characterized in that, The user interaction module is specifically used for: The environmental data, biological behavior data, non-contact physiological data, health status assessment results, disease risk assessment results, and early warning information are displayed in a visual manner. Provides a remote control interface for manual intervention in the equipment controlled by the automated control module; Record and store all collected data and intervention operations to form a historical database for model optimization.

7. The artificial intelligence-based macrophage seedling management system according to claim 3, characterized in that, The intelligent diagnostic module assesses the group behavior of the giant freshwater prawns by calculating the population distribution index. The population distribution index is calculated as follows: obtain the center coordinates of all identified giant freshwater prawn individuals in the image, calculate the Euclidean distance between the center coordinates of each individual and the average of the center coordinates of all individuals, and sum all the Euclidean distances and take the average. The population distribution index D group The calculation formula is: Where N is the total number of giant freshwater prawns identified in the image, (x i ,y i Let be the center coordinates of the i-th individual giant freshwater prawn. This represents the average center coordinates of all individual giant freshwater prawns.

8. The artificial intelligence-based macrophage seedling management system according to claim 4, characterized in that, The time series prediction model is a long short-term memory network model or a gated recurrent unit model. The multidimensional data sequence is an input sequence composed of environmental data, biological behavior data, and non-contact physiological data of a predetermined number of time steps in the past; Based on the input sequence, the model outputs predicted values ​​of the environmental data, biological behavior data, and non-contact physiological data for a predetermined number of time steps in the future.

9. A giant freshwater prawn seedling management system based on artificial intelligence according to claim 5, characterized in that, The steps by which the automated control module performs precise drug delivery intervention include: The intelligent diagnostic module receives the real-time location coordinates of the identified abnormal individual. Based on the real-time location coordinates, the target area and drug dosage for precise intervention are determined; Control the targeted drug delivery device to apply the drug dose to the target area.

10. A giant freshwater prawn seedling management system based on artificial intelligence according to claim 5, characterized in that, The non-pharmacological interventions performed by the automated control module include: In response to a diagnosed or predicted pathogen risk, the programmable illumination module is controlled to illuminate with specific wavelengths, intensities, and durations to inhibit pathogen growth. In response to a diagnosed decrease in vitality or group aggregation, the variable flow device is controlled to adjust the flow speed and direction to induce activity in the giant freshwater prawns and improve water circulation.

Images