A hippocampus health state intelligent identification method and grading seed control system

Abstract

The present application relates to the technical field of intelligent aquaculture or precision aquaculture, and specifically relates to a sea horse health state intelligent identification method and grading seed control system, which comprises the following steps: collecting time sequence behavior data and aquaculture environment state data of a sea horse individual; quantifying the time sequence behavior data to generate behavior characteristics; quantifying the aquaculture environment state data to generate environment characteristics; combining the behavior characteristics and the environment characteristics to generate a dynamic health index; generating a pathological risk index based on the dynamic health index; combining the dynamic health index and the pathological risk index to generate a comprehensive seed value index; determining the grade of the sea horse individual according to the comprehensive seed value index and a preset grading threshold; and generating a sorting control instruction according to the grade. The present application realizes intelligent decision-making from simple optimization to optimization and risk avoidance, and significantly improves the screening efficiency of excellent germplasm.

Description

Technical Field

[0001] This invention relates to the field of intelligent aquaculture or precision aquaculture technology, specifically a method for intelligent identification of seahorse health status and a graded breeding control system. Background Technology

[0002] In modern intensive seahorse farming, the selection and breeding of superior germplasm is the core link to ensure the sustainable development of the industry. The health status of seahorses directly determines their value as germplasm. However, this status is dynamically affected by multiple factors such as behavior, physiology and farming environment, and its complexity brings significant challenges to accurate assessment.

[0003] Currently, the assessment of seahorse health status and the decision on retaining them for breeding mainly rely on the experience of aquaculture technicians and manual visual inspection. This traditional method is highly subjective, difficult to standardize, and inefficient, and cannot meet the continuous monitoring needs of large-scale aquaculture. In addition, judgments based on intermittent observations often lag behind the actual changes in individual health status, making it difficult to detect potential disease risks early. Furthermore, the manual screening operation itself can easily cause stress to seahorses.

[0004] Therefore, how to achieve objective, continuous, and accurate automated assessment and prediction of the health status of hippocampi in order to improve the efficiency and accuracy of the selection of superior germplasm and effectively prevent the spread of diseases in the population has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides an intelligent identification method for hippocampal health status and a graded breeding control system. Specifically, the technical solution of this invention is as follows:

[0006] A method for intelligent identification and graded control of hippocampal health status for breeding includes the following steps:

[0007] S1. Collect temporal behavioral data and aquaculture environment status data of individual seahorses;

[0008] S2. Quantify time-series behavioral data to generate behavioral characteristics; quantify aquaculture environment status data to generate environmental characteristics;

[0009] S3. Combine behavioral and environmental characteristics to generate a dynamic health index;

[0010] S4. Generate a pathological risk index based on the dynamic health index;

[0011] S5. Combine dynamic health index and pathological risk index to generate comprehensive seed retention value index;

[0012] S6. Determine the grade of the individual seahorse based on the comprehensive seed retention value index and the preset grading threshold.

[0013] S7. Generate sorting control instructions based on the grade level.

[0014] Preferably, S2 and S3 specifically include:

[0015] S21. Process the time-series behavioral data into feeding efficiency factor, activity entropy factor, and stress posture frequency factor;

[0016] S22, combining feeding efficiency factor, activity entropy factor and stress posture frequency factor to generate behavioral characteristics;

[0017] S23. Integrate aquaculture environment status data into environmental stress factors as environmental characteristics;

[0018] S31. Calculate the basic behavioral health index based on behavioral characteristics;

[0019] S32. Combine basic behavioral health indices with environmental characteristics to generate dynamic health indices.

[0020] Preferably, S4 specifically includes:

[0021] S41. Compare the dynamic health index with the preset health benchmark value to determine the static risk;

[0022] S42. Calculate the time derivative of the dynamic health index to determine dynamic risk;

[0023] S43. Combine static risk and dynamic risk to generate a pathological risk index.

[0024] Preferably, S5 specifically includes:

[0025] S51. Adjust the pathological risk index based on the preset risk aversion coefficient and determine the risk adjustment item;

[0026] S52. Subtract the risk adjustment item from the dynamic health index to generate the comprehensive seed retention value index.

[0027] Preferably, S6 specifically includes:

[0028] S61. Preset high-bit threshold and low-bit threshold;

[0029] S62. If the comprehensive seed retention value index is higher than the high threshold, the corresponding level will be determined as the seed retention level.

[0030] S63. If the comprehensive seed retention value index is not higher than the high threshold and not lower than the low threshold, then the level is determined as the observation level.

[0031] S64. If the comprehensive seed retention value index is lower than the low threshold, the level will be determined as the elimination level.

[0032] Preferably, S7 specifically includes:

[0033] S71. If the grade is the breeding grade, then generate a sorting control instruction to sort the seahorse individuals to the preferred breeding area.

[0034] S72. If the level is observation level, generate a sorting control instruction to sort to the isolation observation area;

[0035] S73. If the grade is a rejection grade, a sorting control instruction to remove the main breeding area will be generated.

[0036] Preferred, including:

[0037] The data acquisition module is used to collect time-series behavioral data of individual seahorses and data on the state of their breeding environment;

[0038] The feature quantization module is used to quantify time-series behavioral data to generate behavioral features and to quantify aquaculture environment status data to generate environmental features.

[0039] The health assessment module is used to generate a dynamic health index by combining behavioral and environmental characteristics.

[0040] The risk prediction module is used to generate a pathological risk index based on a dynamic health index.

[0041] The value assessment module is used to combine dynamic health index and pathological risk index to generate a comprehensive seed retention value index;

[0042] The grading decision module is used to determine the grade of an individual seahorse based on the comprehensive seed retention value index and the preset grading threshold.

[0043] The instruction generation module is used to generate sorting control instructions based on the classification level.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. This invention achieves objectivity and precision in health status assessment. Existing technologies rely on the experience of breeders for subjective judgment, resulting in inconsistent standards and susceptibility to misjudgment and omission. This invention uses a data acquisition module to continuously and frequently collect temporal behaviors and data on the breeding environment of individual seahorses. The feature quantification module then transforms the raw data into standardized behavioral and environmental characteristics. The health assessment module further dynamically integrates behavioral characteristics from multiple dimensions, such as feeding efficiency, activity complexity, and stress response, with environmental stress characteristics that incorporate multiple parameters, including water temperature and dissolved oxygen, to generate a dynamic health index that comprehensively reflects the individual's physiological state and environmental adaptability. This multi-dimensional data-driven assessment method completely eliminates subjective factors, making the health assessment results not only objective and reproducible, but also profoundly revealing the biological logic behind behavioral manifestations through nonlinear modeling of key indicators such as activity entropy. Its assessment accuracy far surpasses that of manual observation.

[0046] 2. This invention introduces a forward-looking pathological risk prediction capability. Traditional methods can only take action after obvious symptoms appear in seahorses, often missing the optimal intervention time and easily leading to the spread of disease. This invention uniquely constructs a risk prediction module, which not only assesses the static risk constituted by the current low health index, but more importantly, determines the dynamic risk by calculating the time change trend of the health index. This comprehensive risk assessment mechanism, which combines health stock and health change rate, can keenly capture the deterioration trend of individual health status, thereby identifying those individuals who are in a sub-healthy state or tending to become diseased before pathological characteristics appear. This shifts the focus of aquaculture management from post-event response to pre-event warning, winning a valuable time window for preventive isolation and intervention, and greatly improving the proactive prevention and control capability of the population against diseases.

[0047] 3. A scientific and adaptive breeding decision-making model has been established. Traditional breeding selection is mostly based on single indicators such as body shape and vitality for extensive selection, lacking consideration of individual future risks. This invention integrates a dynamic health index representing immediate benefits and a pathological risk index representing future costs through a value assessment module to generate a comprehensive breeding value index. This model introduces a risk aversion coefficient, allowing managers to flexibly adjust the risk weight of decisions based on production goals at different stages, such as population safety and propagation speed, realizing intelligent decision-making that balances selection and risk avoidance from simple selection. The grading decision module, based on this value index, compares it with a grading threshold dynamically calibrated based on historical data to accurately classify individuals into breeding, observation, or rejection levels, making the selection criteria more scientific and flexible.

[0048] 4. A fully automated closed-loop control system from identification to execution has been constructed. From data acquisition, feature quantification, health assessment, risk prediction, value assessment, hierarchical decision-making to final instruction generation, this invention integrates previously fragmented manual processes into a highly efficient and interconnected technical chain. The instruction generation module seamlessly transforms high-level logical decisions into precise control instructions for physical sorting equipment, achieving automated physical isolation of seahorses of different grades. This closed-loop system not only liberates managers from heavy, inefficient, and repetitive labor, significantly improving the efficiency of selecting superior breeds, but also avoids stress damage to seahorses caused by manual harvesting, providing a complete systematic technical solution for achieving large-scale, refined, and low-interference healthy seahorse farming. Attached Figure Description

[0049] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0050] Figure 1 This is a flowchart of the method of the present invention;

[0051] Figure 2 This is a structural block diagram of the system of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0053] Example 1:

[0054] Please see Figure 1 A method for intelligent identification and graded breeding control of hippocampal health status includes the following steps:

[0055] S1. Collect temporal behavioral data and aquaculture environment status data of individual seahorses;

[0056] S2. Quantify time-series behavioral data to generate behavioral characteristics; quantify aquaculture environment status data to generate environmental characteristics;

[0057] S3. Combine behavioral and environmental characteristics to generate a dynamic health index;

[0058] S4. Generate a pathological risk index based on the dynamic health index;

[0059] S5. Combine dynamic health index and pathological risk index to generate comprehensive seed retention value index;

[0060] S6. Determine the grade of the individual seahorse based on the comprehensive seed retention value index and the preset grading threshold.

[0061] S7. Generate sorting control instructions based on the grade level.

[0062] This embodiment discloses a method for intelligent identification and graded breeding control of seahorse health status. As a complete technical closed loop, this method aims to achieve full automation from status monitoring to breeding decision-making in seahorse farming. Its operation process includes the following steps:

[0063] S1. Collect time-series behavioral data and aquaculture environment status data of individual seahorses. In this embodiment, this step is completed through a multimodal dynamic data acquisition system. Time-series behavioral data refers to dynamic information that is continuously recorded over time and reflects the physiological and activity status of seahorses. Its purpose is to provide raw material for subsequent behavioral analysis. It is obtained by continuously capturing the two-dimensional movement trajectory, feeding action details, and body posture information of the seahorse group at a rate of no less than 50 frames per second using high-frame-rate industrial cameras deployed at the top and sides of the aquaculture water body. Aquaculture environment status data refers to the quantitative recording of key physicochemical indicators of the seahorse's living environment. Its purpose is to assess the potential impact of the environment on the seahorse's physiology. It is obtained by acquiring and recording water temperature in real time through a multi-parameter water quality monitoring system deployed in the aquaculture water body. Dissolved oxygen ,salinity Core environmental parameters;

[0064] S2. Quantify time-series behavioral data to generate behavioral features; quantify aquaculture environment status data to generate environmental features; this step aims to transform the heterogeneous, raw data streams collected in step S1 into standardized, dimensionless feature factors that can be used for unified measurement and calculation.

[0065] S3. Combine behavioral and environmental characteristics to generate a dynamic health index. This step aims to construct a single, intuitive health measure that can comprehensively reflect the internal physiological state of the hippocampus and the influence of the external environment. It does not simply add the characteristics together, but uses a modified model to treat environmental characteristics as a moderating factor on the health status reflected by behavioral characteristics, thereby obtaining a health assessment result that is more in line with biological reality.

[0066] S4. Generate a pathological risk index based on the dynamic health index. This step aims to deepen the understanding from status assessment to trend prediction. The simple health index only reflects the current status, while the pathological risk index aims to quantify the likelihood of an individual developing a disease or experiencing a rapid deterioration in health status in the future, especially to identify potentially high-risk individuals whose current health index is acceptable but has shown a rapid downward trend.

[0067] S5. Combine the dynamic health index and the pathological risk index to generate a comprehensive seed retention value index; this step is the core link in decision-making; its purpose is to establish a final decision basis, which is to weigh the dynamic health index representing current benefits and the pathological risk index representing future costs to obtain a risk-adjusted comprehensive value score for an individual.

[0068] S6. Based on the comprehensive seed retention value index and the preset grading threshold, determine the grade of the individual seahorse; this step maps the continuous value index calculated in the previous step to the discrete management grade, providing a clear classification label for subsequent physical sorting.

[0069] S7. Generate sorting control instructions according to the grade; this step is the final execution link to realize the technical closed loop; the system converts the logical grade determined in step S6 into a standardized control signal that can be recognized and executed by the physical sorting equipment, thereby realizing the automated physical isolation of different grades of seahorses.

[0070] This invention constructs a complete technology chain from multimodal data acquisition, multidimensional feature quantification, dynamic health assessment, forward-looking risk prediction to final automated hierarchical decision-making. Compared with traditional methods that rely on manual observation, this invention achieves objective, continuous, and accurate assessment and prediction of seahorse health status, significantly improves the efficiency and accuracy of screening superior germplasm, and can effectively prevent the spread of diseases in the population, providing a systematic technical solution for realizing intelligent and refined management of seahorse farming.

[0071] Example 2:

[0072] S2 and S3 specifically include:

[0073] S21. Process the time-series behavioral data into feeding efficiency factor, activity entropy factor, and stress posture frequency factor;

[0074] S22, combining feeding efficiency factor, activity entropy factor and stress posture frequency factor to generate behavioral characteristics;

[0075] S23. Integrate aquaculture environment status data into environmental stress factors as environmental characteristics;

[0076] S31. Calculate the basic behavioral health index based on behavioral characteristics;

[0077] S32. Combine basic behavioral health indices with environmental characteristics to generate dynamic health indices.

[0078] This embodiment is a further explanation of the specific implementation of steps S2 and S3 in embodiment 1. Its core lies in defining the specific technical path for feature quantification and health index construction.

[0079] S21. Process the time-series behavioral data into feeding efficiency factor, activity entropy factor and stress posture frequency factor. The purpose of this step is to extract the three core behavioral indicators most directly related to health status from complex video data.

[0080] Feeding efficiency factor The purpose is to quantify the initiative and success rate of seahorse feeding; in this embodiment, it is calculated using a double-standardized formula:

[0081]

[0082] in, : No. A seahorse The standardized feeding efficiency factor at any given time is a dimensionless relative value, and the input variables for its calculation are provided in real time by the data acquisition module and the preceding calculations in this step.

[0083] : within the preset time window Within, individuals identified through image recognition algorithms The number of times food was successfully sucked into the mouth;

[0084] : in the same time window Inside, individual The total number of times a nod or sucking motion is made toward the food;

[0085] Regularization of small quantities, such as an extremely small positive number (e.g. This is used to prevent calculation interruption due to a zero denominator, and is preset according to the system's calculation accuracy.

[0086] : Population reference feeding efficiency, a healthy baseline, is determined by statistical analysis of feeding data of population batches historically marked as optimal by experts over their entire life cycle, resulting in an average feeding efficiency benchmark.

[0087] : Individual historical average feeding efficiency, individual Historical average feeding efficiency since monitoring began, used to eliminate individual differences;

[0088] : Stability regularization term, a very small positive number (e.g.) ), used to prevent when an individual's historical average feeding efficiency When the value approaches zero, the calculation results diverge, thus enhancing the robustness of the model;

[0089] The formula first calculates an individual's current raw feeding success rate, then standardizes it using its own historical average and the population-optimal benchmark. This method not only fairly compares feeding abilities among different individuals, but also introduces a stability regularization term. This ensures the computational stability of the model under various extreme input conditions.

[0090] Activity entropy factor The purpose is to quantify the complexity and spatial occupancy of seahorse movement trajectories. High activity entropy is generally associated with healthy exploratory behavior, while excessively high or low entropy may indicate stress or insufficient vitality. In this embodiment, it is calculated through the following steps: First, the aquaculture water area is divided into two-dimensional planes. Two grids of the same size; secondly, statistics are performed within a time window. Inside, individual The trajectory point falls on the first Number of times within each grid Finally, the activity entropy is calculated using the information entropy formula:

[0091]

[0092] in, , representing an individual Appears in the grid within the time window The probability. This formula quantifies complex motion trajectories into a dimensionless index that reflects the complexity of their behavioral patterns.

[0093] Stress Posture Frequency Factor The purpose is to quantify the frequency of abnormal physiological postures exhibited by the hippocampus as a direct indicator of deteriorating health; in this embodiment, it is calculated through normalization:

[0094]

[0095] in, : No. A seahorse The standardized stress posture frequency factor at any given time is dimensionless, and its calculation input variable comes from the recognition results of the individual's real-time posture.

[0096] : within the time window Inside, an individual is detected using a pre-trained pose recognition deep learning model. The total number of times a predefined stress posture, such as excessive tail curling or abnormal body tremors, was observed.

[0097] Risk reference frequency, the benchmark for danger signals, is determined based on historical data analysis as the average stress posture frequency of the population observed in the week before the outbreak of a group disease.

[0098] This formula calculates the observed individual stress frequency ( ), and correlate it with the risk reference frequency Dividing them yields a standardized, dimensionless risk indicator factor;

[0099] S22. Combine the feeding efficiency factor, activity entropy factor, and stress posture frequency factor to generate behavioral features; here, the behavioral features are a vector that contains the above three quantified factors, providing direct input for subsequent calculation of the basic behavioral health index;

[0100] S23. Integrate the aquaculture environment status data into environmental stress factors, which serve as environmental characteristics; these environmental characteristics are the environmental stress factors. The purpose is to integrate multiple environmental parameters with different dimensions into a single dimensionless index that can intuitively reflect the overall stress level of the current environment on the hippocampus; in this embodiment, it is calculated using a weighted summation formula:

[0101]

[0102] in, : The comprehensive environmental stress factor at time t, and the input variables for its calculation. Provided in real time by the water quality monitoring system;

[0103] The total number of environmental parameters monitored, which is 3 in this example (water temperature, dissolved oxygen, salinity).

[0104] : No. Real-time monitoring values ​​of various environmental parameters;

[0105] The preset optimal value of this parameter is derived from the publicly published "Technical Specifications for Seahorse Farming" or determined through preliminary experiments;

[0106] The physiological tolerance range of this parameter, that is, the difference between the maximum and minimum values ​​at which the hippocampus can survive, is derived from the same source as above.

[0107] : No. The stress weights of the parameters satisfy This reflects the importance of different parameters at specific growth stages, and is set based on the knowledge of domain experts or optimized and calibrated through correlation analysis of historical data;

[0108] It should be noted that the linear weighted model used in this embodiment is a simplification of complex biological responses, aiming to balance computational efficiency and model accuracy. In specific applications, to more accurately simulate the nonlinear and asymmetric responses of organisms to environmental changes, such as the different degrees of harm caused by excessively high and low water temperatures, the linear deviation term can be adjusted. Replace it with a nonlinear penalty function fitted based on physiological experimental data.

[0109] S31. Calculate the basic behavioral health index based on behavioral characteristics; the purpose of this step is to assess an individual's health status solely based on their own behavioral performance, without considering environmental influences; in this embodiment, the basic behavioral health index... Calculated using the following nonlinear model:

[0110]

[0111] in, :individual exist Basic behavioral health index at any time;

[0112] : These are the weighting coefficients for food intake, activity, and stress, respectively. Their function is to adjust the contribution of each behavioral factor to the health index. They are obtained by optimizing and training a multivariate logistic regression model on a historical dataset.

[0113] These are the feeding efficiency factor, activity entropy factor, and stress posture frequency factor calculated in the previous step, respectively.

[0114] : Optimal activity entropy factor, a quantitative representation of the healthiest activity pattern, obtained by Gaussian fitting of the distribution of historical activity entropy data of a large number of healthy hippocampal samples;

[0115] The activity entropy tolerance parameter determines the sensitivity of the health state to deviations from the optimal activity entropy, and is the standard deviation obtained from the Gaussian fitting above.

[0116] The formula has a clear biological logic: the numerator represents the positive contribution to health resulting from feeding efficiency and ideal activity patterns (modeled via a Gaussian function); the denominator serves as a penalty term, applied when the frequency of stress postures... When the denominator increases, the overall health index decreases, reflecting the negative impact of stress on health.

[0117] S32. Combining the basic behavioral health index with environmental characteristics, a dynamic health index is generated; this step modifies the basic behavioral health index using environmental stress factors to obtain the final assessment result; in this embodiment, the dynamic health index... Calculated using the following formula:

[0118]

[0119] This formula introduces environmental impact as a multiplicative correction term, which applies when the environment is ideal. Approaching 0, the final health index is approximately equal to the behavioral health index; when the environment is harsh, An increase in environmental stress will proportionally reduce the final health index, thus reflecting the depletion of health reserves caused by environmental stress.

[0120] This multiplicative correction model is an effective way to assess the interaction between the environment and individual health. In other embodiments, other coupling models, such as subtraction models, can be selected depending on the actual application scenario. ,in The environmental impact coefficient is used in this model to treat environmental stress as an independent deduction from the health index.

[0121] The specific scheme described in this embodiment accurately transforms the original multimodal data stream into feature factors with clear biological significance. In particular, the Gaussian function processing of activity entropy and the multiplicative correction of environmental stress factors make the generated dynamic health index not only structured and standardized, but also more profound and accurate in reflecting the true and comprehensive health status of the hippocampus in a specific environment. This provides a high-quality data foundation for subsequent risk prediction and decision-making, thereby improving the scientificity and accuracy of the assessment.

[0122] Example 3:

[0123] S4 specifically includes:

[0124] S41. Compare the dynamic health index with the preset health benchmark value to determine the static risk;

[0125] S42. Calculate the time derivative of the dynamic health index to determine dynamic risk;

[0126] S43. Combine static risk and dynamic risk to generate a pathological risk index.

[0127] This embodiment is a further explanation of the specific implementation of step S4 in embodiment 1. Its core lies in defining a calculation method for vectorizing instantaneous health status into a prospective pathological risk index.

[0128] S41. Compare the dynamic health index with a preset health benchmark value to determine the static risk. Static risk aims to quantify the degree of poorness or poorness in an individual's current health status; it focuses on the absolute value level of the health index. In this embodiment, this risk item is determined by the formula... To calculate; where, It is a health benchmark, its function is to define an acceptable lower limit for health level, and its source can be the 80th percentile of the population's historical health index, or a threshold set by experts based on production goals; only when the individual's health index is within a certain range can the health level be considered acceptable. The static risk item is positive only when the value is below this benchmark, and the greater the difference, the greater the risk value.

[0129] S42. Calculate the time derivative of the dynamic health index to determine dynamic risk; dynamic risk aims to quantify the deterioration trend of an individual's health status; it focuses on the rate of change of the health index; in this embodiment, this risk term is determined by calculating the time derivative of the dynamic health index. To determine; in discrete time series, this can be approximated as Dynamic risk is considered to exist only when the derivative is negative.

[0130] S43. Combining static and dynamic risks, a pathological risk index is generated; this step involves weighted summation of the two types of risks to obtain the final pathological risk index; in this embodiment, the pathological risk index... Calculated using the following formula:

[0131]

[0132] in, : The pathological risk index of individual i at time t, which is dimensionless and non-negative;

[0133] βd is the weighting coefficient for static risk, and βstatic is the weighting coefficient for dynamic risk; both are dimensionless normal numbers. Their role is to balance the importance of the current state and future trends in risk assessment. They are selected through receiver operating characteristic (ROC) curve analysis on historical disease outbreak data to obtain the coefficients that most sensitively predict future disease status. combination;

[0134] The time scale factor, whose physical meaning is time, has the dimension of time. Its function is to eliminate the influence of the time dimension of the derivative term, so that the dynamic risk term and the static risk term are consistent in dimension. Its source is a constant set according to the data sampling period and response sensitivity requirements.

[0135] The formula's design ensures that the risk index only increases when the health level is below a benchmark or shows a downward trend. This will increase the risk by precisely quantifying the two sources of risk: existing disease and disease progression.

[0136] Through the risk calculation method described in this embodiment, the present invention is no longer limited to the static assessment of the current health status, but introduces the dynamic prediction of the trend of health status change. This comprehensive assessment model, which combines existing risks and incremental risks, can identify those sub-healthy individuals whose health status is deteriorating earlier, thereby effectively advancing the early warning window, buying time for intervention measures and preventing disease outbreaks, and thus improving the proactive prevention and control capabilities of aquaculture risks.

[0137] Example 4:

[0138] S5 specifically includes:

[0139] S51. Adjust the pathological risk index based on the preset risk aversion coefficient and determine the risk adjustment item;

[0140] S52. Subtract the risk adjustment item from the dynamic health index to generate the comprehensive seed retention value index.

[0141] This embodiment is a further explanation of the specific implementation of step S5 in embodiment 1. Its core lies in defining how to integrate the health index, which represents current value, and the risk index, which represents future costs, to generate the final decision-making basis.

[0142] S51. Adjust the pathological risk index based on a preset risk aversion coefficient to determine the risk adjustment term; this step aims to transform the abstract pathological risk into a quantifiable cost that can be directly calculated with the health index; in this embodiment, the risk adjustment term is defined as... ;in, It is the risk aversion coefficient, which reflects the decision-maker's tolerance for risk. It is a positive value that can be set by the user on the system interface according to production goals.

[0143] S52. Subtract the risk adjustment term from the dynamic health index to generate the comprehensive seed retention value index; this step completes the final value assessment calculation; in this embodiment, the comprehensive seed retention value index... Calculated using the following formula:

[0144]

[0145] in, : The comprehensive seed retention value index of individual i at time t, dimensionless;

[0146] The dynamic health index of an individual represents its immediate benefits as a germplasm resource.

[0147] An individual's pathological risk index represents the potential losses they may suffer.

[0148] Risk aversion coefficient; can be lowered when rapid proliferation is required. Tolerating certain risks and preserving more individuals; when facing the threat of disease and prioritizing population safety, then raising the level. Implement stricter screening criteria;

[0149] The value assessment method described in this embodiment establishes a decision-making model that combines biological status assessment with risk management strategies. This model no longer simply selects the best based on health status, but rather prioritizes risk mitigation, making seed retention decisions more scientific and flexible. Users can dynamically adjust the risk aversion coefficient according to different production stages and risk environments. This achieves an effective balance between population expansion rate and population health stability, realizing intelligent and adaptive decision-making.

[0150] Example 5:

[0151] S6 specifically includes:

[0152] S61. Preset high-bit threshold and low-bit threshold;

[0153] S62. If the comprehensive seed retention value index is higher than the high threshold, the corresponding level will be determined as the seed retention level.

[0154] S63. If the comprehensive seed retention value index is not higher than the high threshold and not lower than the low threshold, then the level is determined as the observation level.

[0155] S64. If the comprehensive seed retention value index is lower than the low threshold, the level will be determined as the elimination level.

[0156] This embodiment is a further explanation of the specific implementation of step S6 in embodiment 1. Its core lies in defining specific rules for mapping continuous value indices to three discrete management levels.

[0157] S61. Preset high-level and low-level thresholds; In order to perform hierarchical processing, the system requires two key decision thresholds; high-level threshold and low-bit threshold ;

[0158] Its purpose is to select the top-tier core breeding population; its numerical values ​​are based on the best batches of individuals in history. Distribution statistics, for example, taking the 25th percentile of the distribution, ensure that only the most stable and outstanding individuals are selected;

[0159] Its purpose is to distinguish between acceptable health fluctuations and potential risks that require vigilance; its numerical value is based on the historical data of individuals who eventually developed the disease in the week preceding the onset of illness. The distribution is analyzed, for example, by taking the 90th percentile of the distribution, to ensure that most individuals who perform poorly later on can be effectively blocked;

[0160] S62. If the comprehensive seed retention value index is higher than the high threshold, then the corresponding level is determined as the seed retention level; the specific judgment logic is as follows: if If the individual is selected, it is systematically marked as a first-level breeding stock; these individuals exhibit consistently high health levels and virtually zero pathological risk.

[0161] S63. If the comprehensive seed retention value index is not higher than the high threshold and not lower than the low threshold, then the corresponding level is determined as the observation level; the specific judgment logic is as follows: if If the individual is classified as a Level 2 observation, then the individual is marked by the system as such; these individuals are in relatively good health, but may have slight fluctuations or a very low tendency to deteriorate.

[0162] S64. If the comprehensive seed retention value index is lower than the low threshold, the corresponding level will be determined as the elimination level; the specific judgment logic is as follows: if If the individual is marked as Level 3 excluded by the system, then the individual is considered to be in significantly poor health or is in a rapidly deteriorating state and has a high risk of disease transmission.

[0163] The threshold grading strategy described in this embodiment clearly transforms complex continuous evaluation results into three distinct levels to guide production operations. This method of setting thresholds based on historical data statistics and risk analysis ensures the scientific nature and dynamic adaptability of the grading standards, avoiding misjudgments caused by using fixed and rigid thresholds. It makes the management of large-scale populations more hierarchical, allowing resources to be more accurately allocated to high-value individuals, while decisively removing high-risk sources, thus achieving refined and differentiated management of the hippocampus population.

[0164] Example 6:

[0165] S7 specifically includes:

[0166] S71. If the grade is the breeding grade, then generate a sorting control instruction to sort the seahorse individuals to the preferred breeding area.

[0167] S72. If the level is observation level, generate a sorting control instruction to sort to the isolation observation area;

[0168] S73. If the grade is a rejection grade, a sorting control instruction to remove the main breeding area will be generated.

[0169] This embodiment is a further explanation of the specific implementation of step S7 in embodiment 1. Its core lies in defining how to transform logical levels into actions in the physical world.

[0170] S71. If the grade is for breeding, a sorting control instruction is generated to sort the seahorse individual to the preferred breeding area. When the grading decision module determines that an individual is grade 1 for breeding, the instruction generation module will immediately generate a specific control instruction. This instruction is sent to the automated sorting system in the breeding pond. For example, it controls an underwater robotic arm to gently capture the seahorse individual and transfer it to the physically isolated preferred breeding area, or it controls a set of water flow guide valves to change the direction of the water flow and guide the individual to the designated area.

[0171] S72. If the level is observation level, a sorting control instruction is generated to sort the individual to the isolation observation area. Similarly, for individuals identified as level 2 observation, the instruction generation module generates another different instruction to drive the sorting system to transfer them to the isolation observation area. In this area, these individuals will undergo more frequent health monitoring in order to track changes in their status in a timely manner.

[0172] S73. If the level is culling level, a sorting control instruction to remove it from the main breeding area is generated. For individuals identified as being culling level three, the instruction generation module will generate the highest priority instruction to drive the sorting system to quickly and safely remove them from the main breeding area to prevent potential disease transmission and to process them separately.

[0173] This embodiment constructs a complete automated closed-loop control system that links hierarchical decision-making with physical execution mechanisms. It seamlessly transforms the complex intelligent identification and value assessment results at the front end into clear and efficient physical sorting actions at the back end, completely replacing the inefficient and stressful manual retrieval and screening operations. This not only significantly improves the efficiency and processing capacity of hierarchical seed retention, but also reduces human interference with the seahorses and lowers their stress response, thereby further ensuring the overall health of the population.

[0174] Example 7:

[0175] Please see Figure 2 A smart identification and graded breeding control system for hippocampal health status includes:

[0176] The data acquisition module is used to collect time-series behavioral data of individual seahorses and data on the state of their breeding environment;

[0177] The feature quantization module is used to quantify time-series behavioral data to generate behavioral features and to quantify aquaculture environment status data to generate environmental features.

[0178] The health assessment module is used to generate a dynamic health index by combining behavioral and environmental characteristics.

[0179] The risk prediction module is used to generate a pathological risk index based on a dynamic health index.

[0180] The value assessment module is used to combine dynamic health index and pathological risk index to generate a comprehensive seed retention value index;

[0181] The grading decision module is used to determine the grade of an individual seahorse based on the comprehensive seed retention value index and the preset grading threshold.

[0182] The instruction generation module is used to generate sorting control instructions based on the classification level.

[0183] This embodiment provides an intelligent identification and graded breeding control system for hippocampal health status. This system is the physical carrier for implementing the above method, including:

[0184] In this embodiment, the data acquisition module consists of at least two high frame rate industrial cameras and a multi-parameter online water quality monitor, which are used to collect real-time, non-contact temporal behavioral data of individual seahorses and data on the state of the breeding environment.

[0185] In this embodiment, the feature quantization module is a software module deployed on an edge computing server or in the cloud, with built-in image processing algorithms and data standardization procedures. It receives raw data streams from the data acquisition module, calculates in parallel the feeding efficiency factor, activity entropy factor, stress posture frequency factor, and environmental stress factor for each hippocampus, and finally outputs structured behavioral and environmental features.

[0186] The health assessment module, as one of the core algorithm modules of the software system, receives the output of the feature quantification module, calculates the basic behavioral health index of each hippocampus, and corrects it by combining environmental stress factors to generate the final dynamic health index.

[0187] The risk prediction module maintains a time-series database of dynamic health indices for each hippocampus; it combines the current health index value with its historical rate of change to calculate a pathological risk index that characterizes the risk of future disease development.

[0188] The value assessment module receives dynamic health index and pathological risk index, and allows users to set risk aversion coefficient through the interface; it then calculates the comprehensive seed retention value index for final decision-making.

[0189] The hierarchical decision-making module has built-in high and low grading thresholds. It compares the comprehensive breeding value index output by the value assessment module with these two thresholds to determine the level of each seahorse to be kept, observed, or removed.

[0190] The instruction generation module is the interface between the system and the physical world. Based on the level determined by the hierarchical decision module, it generates sorting control instructions that conform to a specific hardware protocol and sends them to the physical sorting equipment via industrial bus or wireless network.

[0191] This system, through its modular design, clearly maps complex methods and processes to specific software and hardware functional units. Each module performs its own function and works closely together to form an efficient and reliable automated system. This system not only realizes all the functions of the methods described in the embodiments, but also has the advantages of high integration, high automation and good scalability. It can be applied as a standardized product to modern seahorse farming scenarios, thereby improving production efficiency and economic benefits.

[0192] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention; any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0193] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent identification and graded breeding control of hippocampal health status, characterized in that, Includes the following steps: S1. Collect temporal behavioral data and aquaculture environment status data of individual seahorses; S2. Quantify time-series behavioral data to generate behavioral features; Quantify aquaculture environment status data to generate environmental characteristics; S3. Combine behavioral and environmental characteristics to generate a dynamic health index; S4. Generate a pathological risk index based on the dynamic health index; S5. Combine dynamic health index and pathological risk index to generate comprehensive seed retention value index; S6. Determine the grade of the individual seahorse based on the comprehensive seed retention value index and the preset grading threshold. S7. Generate sorting control instructions based on the classification level; S2 and S3 specifically include: S21. Process the time-series behavioral data into feeding efficiency factor, activity entropy factor, and stress posture frequency factor; Feeding efficiency factor The purpose is to quantify the initiative and success rate of seahorse feeding; this is calculated using a double-standardized formula: ； in, : No. A seahorse The standardized feeding efficiency factor at any given time is a dimensionless relative value, and the input variables for its calculation are provided in real time by the data acquisition module and the preceding calculations in this step. : within the preset time window Within, individuals identified through image recognition algorithms The number of times food was successfully sucked into the mouth; : in the same time window Inside, individual The total number of times a nod or sucking motion is made toward the food; Regularization small quantity: a very small positive number used to prevent calculation interruption caused by a denominator of zero; preset according to the system's calculation accuracy. Population reference feeding efficiency, a healthy baseline, is determined by statistical analysis of feeding data from population batches historically marked as optimal by experts over their entire life cycle, resulting in an average feeding efficiency benchmark. : Individual historical average feeding efficiency, individual Historical average feeding efficiency since monitoring began, used to eliminate individual differences; Stability regularization term, a very small positive number, used to prevent the occurrence of inconsistencies in the individual's historical average feeding efficiency. When the value approaches zero, the calculation results diverge, thus enhancing the robustness of the model; Activity entropy factor The purpose is to quantify the complexity and spatial occupancy of seahorse movement trajectories. High activity entropy is usually associated with healthy exploratory behavior, while excessively high or low entropy may indicate stress or insufficient vitality. The calculation is performed through the following steps: First, the aquaculture water area is divided into two-dimensional planes... Two grids of the same size; secondly, statistics are performed within a time window. Inside, individual The trajectory point falls on the first Number of times within each grid Finally, the activity entropy is calculated using the information entropy formula: ； in, , representing an individual Appears in the grid within the time window The probability; this formula quantifies complex motion trajectories into a dimensionless index that reflects the complexity of their behavioral patterns. Stress posture frequency factor The purpose is to quantify the frequency of abnormal physiological postures exhibited by the hippocampus as a direct indicator of deteriorating health; calculations are performed through normalization. ； in, : No. A seahorse The standardized stress posture frequency factor at any given time is dimensionless, and its calculation input variable comes from the recognition results of the individual's real-time posture. : within the time window Inside, an individual is detected using a pre-trained pose recognition deep learning model. The total number of times a predefined stress posture, such as excessive tail curling or abnormal body tremors, was observed. Risk reference frequency, the benchmark for danger signals, is determined based on historical data analysis as the average stress posture frequency of the population observed in the week before the outbreak of a group disease. This formula calculates the observed individual stress frequency ( ), and correlate it with the risk reference frequency Dividing them yields a standardized, dimensionless risk indicator factor; S22, combining feeding efficiency factor, activity entropy factor and stress posture frequency factor to generate behavioral characteristics; S23. Integrate aquaculture environment status data into environmental stress factors as environmental characteristics; S31. Calculate the basic behavioral health index based on behavioral characteristics; S32. Combine basic behavioral health indices with environmental characteristics to generate dynamic health indices; S4 specifically includes: S41. Compare the dynamic health index with the preset health benchmark value to determine the static risk; S42. Calculate the time derivative of the dynamic health index to determine dynamic risk; S43. Combine static risk and dynamic risk to generate a pathological risk index.

2. The method for intelligent identification and graded breeding control of hippocampal health status according to claim 1, characterized in that, S5 specifically includes: S51. Adjust the pathological risk index based on the preset risk aversion coefficient and determine the risk adjustment item; S52. Subtract the risk adjustment item from the dynamic health index to generate the comprehensive seed retention value index.

3. The method for intelligent identification and graded breeding control of hippocampal health status according to claim 1, characterized in that, S6 specifically includes: S61. Preset high and low thresholds; S62. If the comprehensive seed retention value index is higher than the high threshold, then the corresponding level is determined as the seed retention level; S63. If the comprehensive seed retention value index is not higher than the high threshold and not lower than the low threshold, then the level is determined as the observation level. S64. If the comprehensive seed retention value index is lower than the low threshold, the level will be determined as the elimination level.

4. The method for intelligent identification and graded breeding control of hippocampal health status according to claim 1, characterized in that, S7 specifically includes: S71, if the grade is the breeding grade, then generate a sorting control instruction to sort the seahorse individuals to the preferred breeding area; S72. If the level is observation level, generate a sorting control instruction to sort to the isolation observation area; S73. If the grade is a rejection grade, a sorting control instruction to remove the main breeding area will be generated.

5. A smart identification and graded breeding control system for hippocampal health status, based on the smart identification and graded breeding control method for hippocampal health status as described in any one of claims 1-4, characterized in that, include: The data acquisition module is used to collect time-series behavioral data of individual seahorses and data on the state of their breeding environment; The feature quantization module is used to quantify time-series behavioral data to generate behavioral features and to quantify aquaculture environment status data to generate environmental features. The health assessment module is used to generate a dynamic health index by combining behavioral and environmental characteristics. The risk prediction module is used to generate a pathological risk index based on a dynamic health index. The value assessment module is used to combine dynamic health index and pathological risk index to generate a comprehensive seed retention value index; The grading decision module is used to determine the grade of an individual seahorse based on the comprehensive seed retention value index and the preset grading threshold. The instruction generation module is used to generate sorting control instructions based on the classification level.

Images