Real-time monitoring integrated system for health and reproduction conditions of dairy cows

Abstract

The invention discloses a real-time monitoring integrated system for health and reproduction conditions of dairy cows, and relates to the field of intelligent animal husbandry, and the system comprises a model construction module, a health abnormity early warning module, a fusion synchronization module, a file establishment module and an interactive feedback module. According to the method, the digital twin of each dairy cow is constructed, so that the whole-process and full-dimension dynamic mapping of the physiological state, the behavior track and the reproduction cycle of the dairy cow is realized, and the management precision and the response efficiency are improved. According to the method, multi-modal behavior characteristics such as tail vibration amplitude, step number change rate and sound frequency are combined, and a deep learning model is fused for intelligent reasoning, so that the misjudgment rate and the omission ratio are effectively reduced, and the disease risk and the estrus window period are accurately identified. Based on physiological, behavioral and historical health data, quantifiable health scores are generated for each dairy cow, and a trend chart is formed to assist in finding potential diseases, monitoring a recovery process or identifying superior and inferior breeding individuals.

Description

Technical Field

[0001] This invention relates to the field of smart animal husbandry technology, and in particular to an integrated system for real-time monitoring of the health and reproductive status of dairy cows. Background Technology

[0002] With the accelerating trend of large-scale and intensive development in animal husbandry, the precision requirements for individual health monitoring, disease early warning, and reproductive management in dairy farming are increasing. Traditional manual inspection methods are not only labor-intensive and inefficient, but also lack continuity and accuracy, often failing to detect sub-health conditions or optimal estrus windows in dairy cows in a timely manner, easily leading to problems such as decreased reproductive success rates, delayed disease detection, and reduced milk production performance. In recent years, information technologies such as the Internet of Things, artificial intelligence, and big data have been gradually applied to the construction of smart farms. Using wearable sensors and environmental monitoring equipment to collect data from dairy cows has become an important means of achieving intelligent management of dairy cows. However, existing technologies still have the following shortcomings: Currently, most systems collect data in a single dimension, lacking the ability to integrate physiological, behavioral, and reproductive information for modeling, and thus failing to form a complete digital model of the individual's life cycle. Although some platforms have data display functions, they have not yet achieved a dynamic digital twin linked to the actual individual's status, making it difficult to intuitively reflect the process of individual health changes, and also unable to support real-time intervention and suggestion pushes. Current estrus detection methods rely on manual observation or single behavioral indicators, resulting in widespread issues of missed or false estrus detections. There is also a lack of intelligent recommendation mechanisms based on multi-source data fusion and reasoning.

[0003] Therefore, there is an urgent need for a dairy cow health and reproduction management method that integrates multi-source sensing, intelligent analysis, digital mapping, and intelligent push. By constructing a digital twin, it is possible to achieve accurate modeling and real-time dynamic reflection of the individual status of dairy cows, thereby improving the health level and reproductive efficiency of dairy cows and promoting the management upgrade of smart ranches. Summary of the Invention

[0004] In view of the existing problems such as data silos, delayed identification, and lack of intervention, this invention is proposed.

[0005] Therefore, the problem this invention aims to solve is how to achieve comprehensive monitoring, multimodal fusion analysis, real-time status updates, and closed-loop management of dairy cow health and reproductive behavior, thereby improving the digital, intelligent, and refined management level of dairy farms. To solve the above technical problems, this invention provides the following technical solution:

[0006] In a first aspect, embodiments of the present invention provide a real-time monitoring and integration system for the health and reproductive status of dairy cows, which includes a model building module for constructing a dairy cow estrus behavior recognition model based on a historical manually labeled dataset, using a fusion of temporal convolutional neural network and long short-term memory network, and combining tail vibration amplitude, daily step change rate and vocal frequency pattern for multi-dimensional reasoning and judgment.

[0007] The health abnormality early warning module is used to classify and model the health status of dairy cows using an integrated random forest and Bayesian discriminant algorithm, identify disease early warning signals, and if the dynamic threshold is exceeded, the corresponding dairy cow number will be pushed to the ranch management terminal and an alarm will be triggered.

[0008] The fusion and synchronization module is used to perform full-cycle archiving, analysis and modeling, and health trend visualization of the filtered data;

[0009] The profile creation module is used to generate a dynamic digital twin for each dairy cow on the cloud platform, mapping its physiological state, behavioral trajectory and reproductive records, so as to realize visualized management and individualized intervention strategy formulation;

[0010] The interactive feedback module is used to push real-time monitoring results, early warning notices and breeding recommendation windows to ranch managers, and provide individual health scores and production performance trend analysis.

[0011] As a preferred embodiment of the real-time monitoring and integration system for dairy cow health and reproductive status described in this invention, the system comprises: constructing a dairy cow estrus behavior recognition model based on a historical manually labeled dataset using a fusion of temporal convolutional neural network and long short-term memory network, including:

[0012] The TCN model is used to model time series data and extract short-term behavioral patterns. In TCN, the convolution operation formula is as follows:

[0013]

[0014] Among them, f t Let w be the output feature of the TCN module at time t. i Let x be the weight of the i-th convolutional kernel. t-d·i The input is the data at time t, d·i time steps ahead, where k is the size of the convolution kernel;

[0015] The feature sequence extracted by TCN is used as input and fed into LSTM to further learn the behavioral evolution trend and time dependence. The output features of LSTM are fed into a fully connected layer, and finally the probability of belonging to the estrus state is output through the Softmax activation function, as shown in the following formula:

[0016]

[0017] Among them, h t is the hidden state output of the LSTM module at time t, and w0 and w1 are the classification weight vectors used for non-estrus and estrus states.

[0018] As a preferred embodiment of the real-time monitoring and integration system for dairy cow health and reproductive status described in this invention, the system incorporates multi-dimensional reasoning and judgment based on tail vibration amplitude, daily step rate of change, and vocalization frequency patterns, including:

[0019] Vibration signals of the tail in the horizontal and vertical directions were collected using an accelerometer attached to the tail. The variance and rate of change of amplitude of the tail's motion acceleration per unit time were extracted, and the following indices were calculated:

[0020]

[0021] Among them, a i Let be the tail acceleration at time i. This represents the tail-end average acceleration within that time window.

[0022] A tail To indicate the level of activity of the tail vibration;

[0023] Total steps S taken in 24 hours are obtained via leg or collar sensors. t Calculate the rate of change in daily steps:

[0024]

[0025] Among them, S t S represents the number of steps taken that day. t-1 R represents the number of steps taken the previous day, ∈ is a small constant to prevent the denominator from being zero. s To represent the growth rate of daily steps, microphones worn around the neck were used to record segments of cow mooing over 24 hours; audio signal processing algorithms were used to identify cow mooing and filter out non-estrus-related sounds; the number of mooings per hour was counted to construct a mooing time distribution map.

[0026] As a preferred embodiment of the real-time monitoring and integration system for the health and reproductive status of dairy cows described in this invention, the system further includes: multi-dimensional reasoning and judgment based on tail vibration amplitude, daily step rate of change, and vocalization frequency pattern;

[0027] Tail vibration amplitude, daily step rate of change, and vocal frequency pattern were standardized and mapped to an estrus risk score. Empirical weights were assigned to each feature, and an overall estrus index E was synthesized. score :

[0028]

[0029] Where α, β, and γ are the weight coefficients of each feature. This is the normalized value of the tail vibration amplitude. This is the normalized value of the daily step rate of change. This is the normalized value of the call frequency;

[0030] If E score A score ≥0.75 indicates the cow is in estrus. If the score is ≥0.75 for two consecutive days, and at least one day the score is ≥0.85, a strong estrus warning is triggered. If tail tremors are greater than the threshold, the step count change rate is greater than 15%, and the daily average call frequency is greater than 50 times, the cow is placed in an estrus observation state. If the tail is still, the step count is unchanged, and the call frequency is low, the cow is considered to be in a non-estrus state. Continuity analysis is introduced, and if the score is higher than 0.8 three times within 12 hours, the behavioral trend is judged to be continuous, enhancing the reliability of the prediction.

[0031] As a preferred embodiment of the real-time monitoring and integration system for dairy cow health and reproductive status described in this invention, the system employs an integrated random forest and Bayesian discriminant algorithm to classify and model the health status of dairy cows and identify disease warning signals, including:

[0032] Disease features are modeled as class-conditional probability distributions and combined with prior probabilities; Bayesian formulas are used to infer the posterior probability of the class; multivariate Gaussian distributions are constructed for all health states, and combined with the actual feature distributions, a Bayesian classifier is used to determine the probability that a sample belongs to an abnormal state; if the Bayesian model and the random forest prediction results are highly consistent, the confidence level is increased; if the discrepancy is large, it is marked as "requiring manual review";

[0033] The random forest and Bayesian results are combined. If the two predictions are consistent, the final judgment is directly output. If the predictions are inconsistent, the posterior consistency weight is calculated and the confidence level is adjusted. If the system judges the cow as "disease" for two consecutive days, the cow is automatically marked as an "observation object" and an individual health behavior abnormality map is generated.

[0034] As a preferred embodiment of the real-time monitoring and integration system for dairy cow health and reproductive status described in this invention, the filtered data undergoes full-cycle archiving, analysis, modeling, and health trend visualization, including:

[0035] The indicators on the continuous time axis are smoothed to construct body temperature curves, daily step count change curves, rumination behavior frequency curves, and disease confidence trend lines; a healthy stability band is established for each indicator, and any deviations from the trend are detected and marked as "potential health fluctuation period" or "recovery period monitoring" status.

[0036] By using regression or time series models to model the indicator curves, we can predict the trend of health parameters for several days to come.

[0037] As a preferred embodiment of the real-time monitoring and integration system for the health and reproductive status of dairy cows described in this invention, a dynamic digital twin is generated for each dairy cow on a cloud platform, mapping its physiological state, behavioral trajectory, and reproductive records to achieve visualized management and individualized intervention strategy formulation, including:

[0038] A unique identifier ID is generated for each cow in the cloud platform database, and a digital twin object is created using this ID as an index to continuously receive and map multi-source data for that cow;

[0039] The cloud platform updates the various attributes of the digital twin in real time based on newly uploaded data, archives historical data daily to form a continuous time series, and judges the estrus status, pregnancy progress and sub-health risk status of dairy cows based on the built-in model and updates the status labels in the twin.

[0040] As a preferred embodiment of the real-time monitoring and integration system for dairy cow health and reproductive status described in this invention, the system includes: pushing real-time monitoring results, early warning notifications, and recommended reproductive windows to farm managers, and providing individual health scores and production performance trend analysis, including:

[0041] A health scoring model is established for each dairy cow, with a daily output score range of 0-100, graded as follows: 85-100 points: good health; 70-85 points: sub-health warning; below 70 points: high risk of disease; the platform automatically calculates the health score of each dairy cow for the previous 24 hours every night and records it in its digital twin, displaying a health trend chart;

[0042] The system identifies suspected estrus signals in dairy cows based on a fusion structure of long short-term memory and temporal convolutional network models. If the model outputs an estrus confidence score ≥ 0.85 and more than 21 days have passed since the last mating, the system determines it as a "recommended mating window"; if the confidence score ≥ 0.95, it strongly suggests "immediate mating".

[0043] In a second aspect, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the real-time monitoring integrated system for the health and reproductive status of dairy cows as described in the first aspect of the present invention are implemented.

[0044] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the real-time monitoring integrated system for the health and reproductive status of dairy cows as described in the first aspect of the present invention are implemented.

[0045] The beneficial effects of this invention are as follows: By constructing a digital twin of each dairy cow, this invention achieves dynamic mapping of its physiological state, behavioral trajectory, and reproductive cycle across the entire process and dimensions, improving management accuracy and response efficiency. Combining multimodal behavioral characteristics such as tail vibration amplitude, step rate change rate, and vocalization frequency, and integrating deep learning models for intelligent reasoning, it effectively reduces false positive and false negative rates, accurately identifying disease risks and estrus windows. Based on physiological, behavioral, and historical health data, a quantifiable health score is generated for each dairy cow, and a trend chart is formed to assist in discovering potential diseases, monitoring the recovery process, or identifying superior and inferior breeding individuals. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 A schematic diagram of the structure of an integrated system for real-time monitoring of dairy cow health and reproductive status.

[0048] Figure 2 A diagram of computer equipment used in a real-time monitoring system for the health and reproductive status of dairy cows. Detailed Implementation

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0050] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0052] Example 1

[0053] Reference Figures 1-2 This is the first embodiment of the present invention, which provides an integrated system for real-time monitoring of the health and reproductive status of dairy cows, comprising:

[0054] S1: Based on historical manually labeled datasets, a dairy cow estrus behavior recognition model is constructed by fusing temporal convolutional neural networks and long short-term memory networks. The model is then used to make multi-dimensional inferences and judgments by combining tail vibration amplitude, daily step rate change rate, and vocal frequency patterns.

[0055] Preferably, the TCN model is used to model time series data and extract short-term behavioral patterns. In TCN, the convolution operation formula is as follows:

[0056]

[0057] Among them, f t Let w be the output feature of the TCN module at time t. i Let x be the weight of the i-th convolutional kernel. t-d.i The input is the data at time t, d·i time steps ahead, where k is the size of the convolution kernel;

[0058] The feature sequence extracted by TCN is used as input and fed into LSTM to further learn the behavioral evolution trend and time dependence. The output features of LSTM are fed into a fully connected layer, and finally the probability of belonging to the estrus state is output through the Softmax activation function, as shown in the following formula:

[0059]

[0060] Among them, h t is the hidden state output of the LSTM module at time t, and w0 and w1 are the classification weight vectors used for non-estrus and estrus states.

[0061] Preferably, an accelerometer worn on the tail is used to collect vibration signals of the tail in the horizontal and vertical directions, and the variance and amplitude change rate of the tail motion acceleration per unit time are extracted. The calculated indicators are as follows:

[0062]

[0063] Among them, a i Let be the tail acceleration at time i. This represents the tail-end average acceleration within that time window.

[0064] A tail To indicate the level of activity of the tail vibration;

[0065] Total steps S taken in 24 hours are obtained via leg or collar sensors. t Calculate the rate of change in daily steps:

[0066]

[0067] Among them, S t S represents the number of steps taken that day. t-1R represents the number of steps taken the previous day, ∈ is a small constant to prevent the denominator from being zero. s To represent the growth rate of daily steps, microphones worn around the neck were used to record segments of cow mooing over 24 hours; audio signal processing algorithms were used to identify cow mooing and filter out non-estrus-related sounds; the number of mooings per hour was counted to construct a mooing time distribution map.

[0068] Preferably, the tail vibration amplitude, daily step rate of change, and vocal frequency pattern are uniformly scaled and mapped to an estrus risk score. An empirical weight is assigned to each feature, and an overall estrus index E is synthesized. score :

[0069]

[0070] Where α, β, and γ are the weight coefficients of each feature. This is the normalized value of the tail vibration amplitude. This is the normalized value of the daily step rate of change. This is the normalized value of the call frequency;

[0071] If the Escore is ≥ 0.75, the cow is considered to be in estrus. If the score is ≥ 0.75 for two consecutive days and at least ≥ 0.85 for one day, a strong estrus warning is triggered. If the tail vibration is greater than the threshold, the step change rate is greater than 15%, and the daily average call frequency is greater than 50 times, the cow is placed in an estrus observation state. If the tail is still, the step count is unchanged, and the call frequency is low, the cow is considered to be in a non-estrus state. Continuity analysis is introduced, and if the score is higher than 0.8 three times within 12 hours, the behavioral trend is judged to be continuous, which enhances the reliability of the prediction.

[0072] Furthermore, the TCN output sequence {f t The sequence is passed as input to the LSTM module to capture long-term behavioral evolution trends.

[0073] h t =LSTM(f1,f2,...,f t )

[0074] Among them, h t The output of the hidden state of the LSTM at time step t;

[0075] The LSTM hidden state h at the final moment T The input is fed into a fully connected layer, and the Softmax activation function is used to output the estrus probability:

[0076] P estrus =Softmax(W0·h T +b0)

[0077] Where W0 is the weight matrix of the fully connected layer, and b0 is the bias term.

[0078] Furthermore, the variance and amplitude of the tail acceleration within a unit time window are extracted from the tail acceleration signal to form a comprehensive vibration index:

[0079]

[0080] Where N is the number of samples, and ε is a small constant to prevent the denominator from being zero;

[0081] The audio segments throughout the day were analyzed, and the spectrum was extracted using short-time Fourier transform. A custom-designed model for recognizing estrus calls in cows was used to filter out background noise and non-estrus calls. The number of valid calls was counted hourly to construct a call time distribution map and extract peak frequency bands and loudness patterns.

[0082] Furthermore, if the daily rating is E score A score ≥0.75 indicates a potential estrus state; this is used to alert observers or activate the automatic camera analysis module. If the score is ≥0.75 for two consecutive days, and at least one day is ≥0.85, a strong estrus warning is triggered; the system automatically pushes a breeding suggestion window to the management terminal. If the following three conditions are met simultaneously: tail vibration amplitude > threshold (e.g., normalized value > 0.7); step change rate > 15%; daily average vocalization frequency > 50 times; the system will list the cow in the "to be observed" state and continuously track behavioral changes. If the tail vibration is detected to be static for a long period, the step count shows no significant change (e.g., growth rate < 5%), and vocalization is infrequent (e.g., < 10 times / day), it is determined to be a non-estrus state.

[0083] S2: The integrated random forest and Bayesian discriminant algorithm are used to classify and model the health status of dairy cows, identify disease warning signals, and if the dynamic threshold is exceeded, the corresponding dairy cow number is pushed to the ranch management terminal and an alarm is triggered.

[0084] Preferably, disease characteristics are modeled as class-conditional probability distributions and combined with prior probabilities; Bayesian formulas are used to infer the posterior probability of the class; multivariate Gaussian distributions are constructed for all health states, and combined with the actual feature distributions, a Bayesian classifier is used to determine the probability that a sample belongs to an abnormal state; if the Bayesian model and the random forest prediction results are highly consistent, the confidence level is increased; if the discrepancy is large, it is marked as "requiring manual review";

[0085] The random forest and Bayesian results are combined. If the two predictions are consistent, the final judgment is directly output. If the predictions are inconsistent, the posterior consistency weight is calculated and the confidence level is adjusted. If the system judges the cow as "disease" for two consecutive days, the cow is automatically marked as an "observation object" and an individual health behavior abnormality map is generated.

[0086] Furthermore, a set of health status categories is established, and for each category, a feature conditional probability density model is built:

[0087]

[0088] Where, μ c Let Σc be the mean vector under state c, and let Σc be the covariance matrix.

[0089] If the Bayesian model and the random forest predict the same result, then that category is directly used as the final health status output, and the system records the mean of the probabilities predicted by the two models as the final confidence level; if the two models predict differently: the system calculates the posterior consistency weight between the predicted distributions of the two models.

[0090] ω=1-KL(P Bayes ||P RF )

[0091] Where KL is the Kullback-Leibler divergence. If ω < 0.7, it indicates a large divergence. In this case, the output is "uncertain state" and it is marked as requiring manual review. If the system continues to output an "abnormal" state for two days, the system marks the individual as an "observation object".

[0092] S3: Perform full-cycle archiving, analysis, modeling, and health trend visualization on the filtered data.

[0093] Preferably, the indicators on the continuous time axis are smoothed to construct body temperature curves, daily step count change curves, rumination behavior frequency curves, and disease confidence trend lines; a healthy stability band is established for each indicator, and any deviations from the trend are detected and marked as "potential health fluctuation period" or "recovery period monitoring" status;

[0094] By using regression or time series models to model the indicator curves, we can predict the trend of health parameters for several days to come.

[0095] Furthermore, an individualized stable health band is established for each behavioral indicator, calculated from its mean ± confidence interval based on historical health period data:

[0096] StabilityBand(x) = [μ x -λ·σ x ,μ x +λ·σ x ]

[0097] Where, μ x σ is the historical mean. x Let λ be the standard deviation and λ be the tolerance interval coefficient.

[0098] If the indicator value continues to exceed the upper or lower limit of the stability zone for more than 2 days, it is marked as "potential health fluctuation period"; if it was previously in an abnormal state and the current value gradually returns to the stability zone, it is marked as "recovery period monitoring"; the system automatically overlays the markings onto the curves of each indicator.

[0099] S4: Generate a dynamic digital twin for each dairy cow on the cloud platform, mapping its physiological state, behavioral trajectory and reproductive records, to achieve visualized management and individualized intervention strategy formulation.

[0100] Preferably, a unique identifier ID is generated for each cow in the cloud platform database, and a digital twin object is created using this ID as an index to continuously receive and map multi-source data for that cow;

[0101] The cloud platform updates the various attributes of the digital twin in real time based on newly uploaded data, archives historical data daily to form a continuous time series, and judges the estrus status, pregnancy progress and sub-health risk status of dairy cows based on the built-in model and updates the status labels in the twin.

[0102] Furthermore, multi-source data is integrated, including tail vibration, steps, vocalizations, rumination frequency, body temperature, heart rate, milk volume, etc., from multiple sensors (accelerometers, audio acquisition, infrared thermometry, collar terminals, etc.). Data is standardized to a unified protocol and uploaded to the cloud. A real-time attribute refresh mechanism is implemented: the system receives sensor data packets every 5 minutes; based on field mapping rules, it refreshes the following in the digital twin in real time: HealthState after health model analysis; and ReproState for estrus and pregnancy identification. The system packages and archives the data from the previous 24 hours into the HistoricalData field at midnight each day.

[0103] The format is a multidimensional time series, which is convenient for trend analysis. The model runs daily at a set time (or is triggered after new data is introduced); the output results are status labels + probability confidence scores.

[0104]

[0105] If the current state probability confidence level is >0.75, the current label of the digital twin is automatically updated; if the state labels are the same for two consecutive days, it enters the "trend stable" state, which is used to assist in intervention recommendations; the results are synchronized to the management interface for state visualization and batch sorting management.

[0106] S5: Pushes real-time monitoring results, early warning notices, and recommended breeding windows to ranch managers, and provides individual health scores and production performance trend analysis.

[0107] Preferably, a health scoring model is established for each dairy cow, with a daily output score range of 0-100, graded as follows: 85-100 points: good health; 70-85 points: sub-health warning; below 70 points: high risk of disease; the platform automatically calculates the health score of each dairy cow for the previous 24 hours every night and records it in its digital twin, displaying a health trend chart;

[0108] The system identifies suspected estrus signals in dairy cows based on a fusion structure of long short-term memory and temporal convolutional network models. If the model outputs an estrus confidence score ≥ 0.85 and more than 21 days have passed since the last mating, the system determines it as a "recommended mating window"; if the confidence score ≥ 0.95, it strongly suggests "immediate mating".

[0109] Furthermore, if the model outputs the estrus confidence score P... Estrus If the estrus confidence score is ≥0.85 and the current time is ≥21 days from the last mating date, the system determines it to be within the "recommended mating window" and displays a reminder on the digital twin page. If the estrus confidence score is ≥0.95, the system issues a strong "immediate mating recommendation" prompt, which is pushed to the management terminal via platform message. The system generates a 7-day health score line chart for each dairy cow; each node is automatically colored according to the health level (green, yellow, red); when the health score suddenly drops, the system issues a "health turning point prompt".

[0110] {

[0111] "Cow_ID":"Cow_2024A123",

[0112] "CurrentScore": 89,

[0113] "Estrus_Confidence":0.93,

[0114] "Days_Since_Last_Mating":23,

[0115] Recommendation: Immediate mating suggested.

[0116] }

[0117] Administrators can click "View Details" to enter the individual twin page and view more behavioral analysis and suggestions;

[0118] If the cow is in good health and shows clear signs of estrus: If the cow's current health score is 85 or above, and the estrus recognition model outputs an estrus confidence score greater than or equal to 0.85, and more than 21 days have passed since the last mating, the system will output the suggestion: "Recommended to enter the mating window period", prompting farm managers to arrange mating operations as soon as possible.

[0119] If the confidence level of estrus is greater than or equal to 0.95: the system will automatically issue a high-priority prompt of "immediate mating suggestion", which will be highlighted in red and pinned to the top of the digital twin interface.

[0120] If the cow's current health score is between 70 and 85, and the estrus confidence level exceeds 0.85, the system will prompt: "Observational breeding suggestion," recommending that the health trend be monitored first to assess whether breeding is suitable. If the health score continues to decline, the system will automatically delay the recommended breeding time.

[0121] If a cow's health score is below 70 and it is in any estrus state: If a cow's current score is below 70, it is in a high-risk disease zone, and the system will not recommend mating. A "Health Intervention Priority" prompt will be issued, requiring priority examination and treatment, and mating recommendations will be suspended. Simultaneously, the cow will be automatically added to the "Health Observation List." If the health score continuously declines (e.g., a drop of more than 10 points within 3 days), the system identifies this as a "downward health trend." Even if the score is not below 70, a "Health Trend Warning" will still be issued, prompting managers to intervene in advance and check for chronic diseases or stress responses.

[0122] If a dairy cow has just recovered from an illness and its current score is 75-85, the system outputs: "Recovery Period Monitoring" recommendation, suggesting that it is not recommended to breed immediately, but to continue to observe it for more than 3 days to ensure that its health status is stable.

[0123] If the health score is within the good range, but no obvious estrus behavior is detected for 30 consecutive days, the system will prompt: "There may be hidden estrus or reproductive disorder." It is recommended to arrange a physical examination or hormone level monitoring to rule out reproductive problems.

[0124] If the score is high but the behavioral indicators fluctuate abnormally (such as loud vocalizations but insufficient steps and tail activity), the system will output a message: "Behavior and score are inconsistent," suggesting manual verification to prevent recognition errors or sensor malfunctions.

[0125] This embodiment also provides a computer device suitable for a real-time monitoring integrated system for dairy cow health and reproductive status, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the real-time monitoring integrated system for dairy cow health and reproductive status as proposed in the above embodiment.

[0126] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0127] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the integrated system for real-time monitoring of the health and reproductive status of dairy cows as proposed in the above embodiments.

[0128] In summary, this invention constructs a digital twin of each dairy cow, enabling dynamic mapping of its physiological state, behavioral patterns, and reproductive cycle across all dimensions, thereby improving management precision and response efficiency. By combining multimodal behavioral characteristics such as tail vibration amplitude, step rate change rate, and vocalization frequency, and integrating deep learning models for intelligent reasoning, it effectively reduces false positive and false negative rates, accurately identifying disease risks and estrus windows. Based on physiological, behavioral, and historical health data, a quantifiable health score is generated for each cow, and a trend chart is created to assist in identifying potential diseases, monitoring the recovery process, or identifying superior and inferior breeding individuals.

[0129] Example 2

[0130] This embodiment is the second embodiment of the present invention. This embodiment provides an integrated system for real-time monitoring of the health and reproductive status of dairy cows. In order to verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0131] Specifically, 1. A certain ranch selected 50 dairy cows in the lactating period, numbered Cow001 to Cow050; equipment was deployed: a three-axis accelerometer was attached to the tail, a sound sensor and thermometer were attached to the neck, and a pedometer was attached to the legs; sampling period: data was collected every 10 minutes; experimental period: 30 days.

[0132] II. After each cow is equipped with a sensor, a unique digital twin object (CowID) is generated for it in the background database. The following behavioral and physiological data are continuously collected daily: tail acceleration data, total steps per 24 hours, frequency of vocal segments per hour, and auxiliary indicators such as body temperature and rumination time. The raw data is standardized, and the following tail activity amplitude index A is calculated: tail Daily step change rate R s Frequency of calls per unit time F call All features are normalized to [0,1].

[0133] 3. Based on the above-collected data, a multi-feature fusion health scoring model (such as random forest or deep regression network) is trained using a historical artificial health scoring training set. Every night at 23:00, the system calls the model to score the day's behavioral and physiological data, with a score range of 0-100. The system determines the health level based on the score: ≥85 points: good health; 70-85 points: sub-health; <70 points: disease risk. The health score is written into the digital twin of the corresponding dairy cow, and a continuous score trend chart is plotted.

[0134] IV. Input 24-hour behavioral time-series data of each cow into the fusion model (TCN+LSTM); TCN extracts local behavioral surge patterns; LSTM captures behavioral evolution trends; the model outputs the estrus confidence score P. estrus ∈[0,1];

[0135] If P estrus If the value is ≥0.85 and the last mating was ≥21 days ago, the system will display "Recommended mating window"; P estrus If the value is ≥0.95, the system will pop up a "breeding suggestion" window; the recommendation result will be written to the twin object and simultaneously displayed on the ranch management terminal.

[0136] 5. The system performs regression modeling on continuous data such as health score, daily steps, and rumination time for each dairy cow; predicts whether the health indicators will show a downward trend in the next three days; if abnormal fluctuations are found (such as increased body temperature + decreased steps + continuous decline in health score), the system marks it as a "potential risk period" and generates intervention suggestions.

[0137] VI. The system will visualize the analysis results daily on the twin panel: health score chart; estrus confidence chart; abnormal behavior annotation; the system will automatically classify and manage suggestions based on the score and model: if the health score is high and the estrus confidence is high, then mating is recommended; if the health score is low, then intervention is prioritized; if the behavior is inconsistent with the score, then manual review is recommended; all suggestions are recorded in a log, which can be exported for managers or veterinarians to review.

[0138] VII. Within 30 days, the accuracy rate of the system reached 92% when comparing the overlap between artificial insemination records and the system's recommended window period; the health score prediction results were highly correlated with veterinary inspection records, with an average of 2.3 days earlier in identifying abnormalities; feedback from the management side indicated that the system effectively improved: estrus recognition efficiency by 28%; disease warning advance by 35%; and reduced manual inspection time by 40%.

[0139] 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, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A real-time monitoring and integrated system for the health and reproductive status of dairy cows, characterized in that, include: The model building module is used to construct a dairy cow estrus behavior recognition model based on historical manually labeled datasets, using a fusion of temporal convolutional neural networks and long short-term memory networks, and combining tail vibration amplitude, daily step change rate and vocal frequency pattern for multi-dimensional reasoning and judgment. The health abnormality early warning module is used to classify and model the health status of dairy cows using an integrated random forest and Bayesian discriminant algorithm, identify disease early warning signals, and if the dynamic threshold is exceeded, the corresponding dairy cow number will be pushed to the ranch management terminal and an alarm will be triggered. The fusion and synchronization module is used to perform full-cycle archiving, analysis and modeling, and health trend visualization of the filtered data; The profile creation module is used to generate a dynamic digital twin for each dairy cow on the cloud platform, mapping its physiological state, behavioral trajectory and reproductive records, so as to realize visualized management and individualized intervention strategy formulation; The interactive feedback module is used to push real-time monitoring results, early warning notices and breeding recommendation windows to ranch managers, and provide individual health scores and production performance trend analysis.

2. The real-time monitoring and integration system for dairy cow health and reproductive status as described in claim 1, characterized in that, The model for recognizing estrus behavior in dairy cows, constructed using a fusion of temporal convolutional neural networks and long short-term memory networks based on historical manually labeled datasets, includes: The TCN model is used to model time series data and extract short-term behavioral patterns. In TCN, the convolution operation formula is as follows: Among them, f t Let w be the output feature of the TCN module at time t. i Let x be the weight of the i-th convolutional kernel. t-d·i The input is the data at time t, d·i time steps ahead, where k is the size of the convolution kernel; The feature sequence extracted by TCN is used as input and fed into LSTM to further learn the behavioral evolution trend and time dependence. The output features of LSTM are fed into a fully connected layer, and finally the probability of belonging to the estrus state is output through the Softmax activation function, as shown in the following formula: Among them, h t is the hidden state output of the LSTM module at time t, and w0 and w1 are the classification weight vectors used for non-estrus and estrus states.

3. The real-time monitoring and integration system for dairy cow health and reproductive status as described in claim 1, characterized in that: The multi-dimensional reasoning and judgment, which combines tail vibration amplitude, daily step rate of change, and vocal frequency pattern, includes: Vibration signals of the tail in the horizontal and vertical directions were collected using an accelerometer attached to the tail. The variance and rate of change of amplitude of the tail's motion acceleration per unit time were extracted, and the following indices were calculated: Among them, a i Let be the tail acceleration at time i. This represents the tail-end average acceleration within that time window. A tail To indicate the level of activity of the tail vibration; Total steps S taken in 24 hours are obtained via leg or collar sensors. t Calculate the rate of change in daily steps: Among them, S t S represents the number of steps taken that day. t-1 R represents the number of steps taken the previous day, ∈ is a small constant to prevent the denominator from being zero. s To represent the growth rate of daily steps, microphones worn around the neck were used to record segments of cow mooing over 24 hours; audio signal processing algorithms were used to identify cow mooing and filter out non-estrus-related sounds; the number of mooings per hour was counted to construct a mooing time distribution map.

4. The real-time monitoring and integration system for dairy cow health and reproductive status as described in claim 1 or 3, characterized in that: The multi-dimensional reasoning and judgment based on tail vibration amplitude, daily step rate of change, and vocal frequency pattern also includes: Tail vibration amplitude, daily step rate of change, and vocal frequency pattern were standardized and mapped to an estrus risk score. Empirical weights were assigned to each feature, and an overall estrus index E was synthesized. score : Where α, β, and γ are the weight coefficients of each feature. This is the normalized value of the tail vibration amplitude. This is the normalized value of the daily step rate of change. This is the normalized value of the call frequency; If E score A score ≥0.75 indicates the cow is in estrus. If the score is ≥0.75 for two consecutive days, and at least one day the score is ≥0.85, a strong estrus warning is triggered. If tail tremors are greater than the threshold, the step count change rate is greater than 15%, and the daily average call frequency is greater than 50 times, the cow is placed in an estrus observation state. If the tail is still, the step count is unchanged, and the call frequency is low, the cow is considered to be in a non-estrus state. Continuity analysis is introduced, and if the score is higher than 0.8 three times within 12 hours, the behavioral trend is judged to be continuous, enhancing the reliability of the prediction.

5. The real-time monitoring and integration system for dairy cow health and reproductive status as described in claim 1, characterized in that, The method employs an integrated random forest and Bayesian discriminant algorithm to classify and model the health status of dairy cows and identify disease warning signals, including: Disease features are modeled as class-conditional probability distributions and combined with prior probabilities; Bayesian formulas are used to infer the posterior probability of the class; multivariate Gaussian distributions are constructed for all health states, and combined with the actual feature distributions, a Bayesian classifier is used to determine the probability that a sample belongs to an abnormal state; if the Bayesian model and the random forest prediction results are highly consistent, the confidence level is increased; if the discrepancy is large, it is marked as "requiring manual review"; The random forest and Bayesian results are combined. If the two predictions are consistent, the final judgment is directly output. If the predictions are inconsistent, the posterior consistency weight is calculated and the confidence is adjusted. If the system judges the cow as "disease" for two consecutive days, the cow is automatically marked as an "observation object" and an individual health behavior abnormality map is generated.

6. The real-time monitoring and integration system for dairy cow health and reproductive status as described in claim 1, characterized in that: The process of archiving, analyzing, modeling, and visualizing health trends of the filtered data throughout its entire lifecycle includes: The indicators on the continuous time axis are smoothed to construct body temperature curves, daily step count change curves, rumination behavior frequency curves, and disease confidence trend lines; a healthy stability band is established for each indicator, and any deviations from the trend are detected and marked as "potential health fluctuation period" or "recovery period monitoring" status; By using regression or time series models to model the indicator curves, we can predict the trend of health parameters for several days to come.

7. The real-time monitoring and integration system for dairy cow health and reproductive status as described in claim 1, characterized in that: The process of generating a dynamic digital twin for each dairy cow on the cloud platform, mapping its physiological state, behavioral trajectory, and reproductive records, enables visualized management and individualized intervention strategy development, including: A unique identifier ID is generated for each cow in the cloud platform database, and a digital twin object is created using this ID as an index to continuously receive and map multi-source data for that cow; The cloud platform updates the various attributes of the digital twin in real time based on newly uploaded data, archives historical data daily to form a continuous time series, and judges the estrus status, pregnancy progress and sub-health risk status of dairy cows based on the built-in model and updates the status labels in the twin.

8. The real-time monitoring and integration system for dairy cow health and reproductive status as described in claim 1, characterized in that: The system pushes real-time monitoring results, early warning notifications, and recommended breeding windows to ranch managers, and provides individual health scores and production performance trend analysis, including: A health scoring model is established for each dairy cow, with a daily output score range of 0-100, graded as follows: 85-100 points: good health; 70-85 points: sub-health warning; below 70 points: high risk of disease; the platform automatically calculates the health score of each dairy cow for the previous 24 hours every night and records it in its digital twin, displaying a health trend chart; The system identifies suspected estrus signals in dairy cows based on a fusion structure of long short-term memory and temporal convolutional network models. If the model outputs an estrus confidence score ≥ 0.85 and more than 21 days have passed since the last mating, the system determines it as a "recommended mating window". If the confidence score ≥ 0.95, it strongly suggests "immediate mating".

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the real-time monitoring and integration system for the health and reproductive status of dairy cows as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the real-time monitoring and integration system for the health and reproductive status of dairy cows as described in any one of claims 1 to 8.

Images