Health monitoring and alarming system and method for animal individuals in zoo

By using a collaborative monitoring system with multiple non-invasive sensors, edge computing, and cloud analysis layers in the zoo, the problems of strong subjectivity in manual observation and low accuracy of equipment monitoring in existing technologies have been solved. Accurate assessment of animal health status and timely alarms are achieved, the false alarm rate is reduced, and efficient health management support is provided.

CN120781245APending Publication Date: 2025-10-14BEIJING HUYUAN TECH CO LTD

Patent Information

Application Number
CN202510840830.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-23
Publication Date
2025-10-14

AI Technical Summary

Technical Problem

Existing technologies for zoo animal health monitoring have problems such as strong subjectivity in manual observation, inability to monitor around the clock, low equipment monitoring accuracy, and high false alarm rates, making it difficult to accurately assess the health status of animals and provide timely alarms.

Method used

A variety of non-invasive sensors (magnetic array sensors, infrared thermal imaging cameras, sound collection devices, behavior monitoring cameras and weight monitoring platforms) are used for data collection. Combined with a variety of animal health analysis algorithms at the edge computing layer and cloud analysis layer, an individualized health baseline parameter set is established, and health monitoring is carried out through anomaly detection and graded alarm decision-making.

Benefits of technology

It achieves accurate assessment of animal health status and timely alarm, reduces the false alarm rate to 5-8%, can identify potential health risks 12-24 hours in advance, and provides efficient and accurate health management support.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120781245A_ABST
    Figure CN120781245A_ABST
Patent Text Reader

Abstract

The invention discloses a health monitoring and alarming system and method for animal individuals in a zoo, belongs to the field of zoo animal health monitoring, and solves the technical bottleneck through cooperation of a sensing layer, an edge computing layer and a cloud analysis layer. The system particularly sets differentiated threshold values for animals of different age groups, genders and health conditions, the dynamic baseline adjustment and multi-source data fusion algorithm is combined, the anomaly detection precision is remarkably improved, meanwhile, a grading alarm mechanism and feedback driven model optimization form a closed loop, the conversion from passive response to active early warning is achieved, and the system is suitable for the animal health monitoring and early warning of animals of different age groups, genders and health conditions. Potential health risks can be identified 12-24 hours in advance, and efficient and accurate technical support is provided for animal health management.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of zoo animal health monitoring, and particularly relates to a health monitoring and alarm system and method for individual zoo animals. BACKGROUND

[0002] In the management of zoo animals, the existing technology mainly relies on traditional manual observation and simple equipment monitoring. Breeders observe the daily behavior, mental state and diet of animals based on experience, and veterinarians conduct physical examinations on animals regularly to check the physical condition of animals through auscultation, palpation and other means. In addition, some zoos use some basic monitoring equipment, such as a thermometer to monitor the body temperature of animals, a camera to record the behavior of animals, etc.

[0003] However, the traditional manual observation method has many shortcomings. On the one hand, manual observation has strong subjectivity, and there are differences in experience and judgment standards between different breeders or veterinarians, resulting in inaccurate and inconsistent evaluation of the health status of animals. On the other hand, manual observation can only be carried out within a limited time, and cannot realize all-weather monitoring of animals, which is easy to miss the critical period of changes in the health status of animals.

[0004] Although the application of basic monitoring equipment to some extent makes up for the shortcomings of manual observation, there are still limitations. For example, the traditional body temperature monitoring equipment can only be used during the veterinarian examination, and cannot obtain the body temperature data of animals in real time. Although the ordinary camera can record the behavior of animals, it cannot analyze the behavior in depth, and it is difficult to accurately judge whether the animal has abnormal behavior. The traditional system usually relies on a single parameter fixed threshold to make abnormality judgment, and it is difficult to cope with complex scenes such as environmental fluctuations, individual differences and multi-parameter coordinated changes, with high false alarm rate and missed alarm rate. SUMMARY

[0005] In view of the shortcomings of the existing technology, the application provides a health monitoring and alarm system and method for individual zoo animals.

[0006] The first aspect of the application provides a health monitoring and alarm system for individual zoo animals, which comprises a perception layer, an edge computing layer, a cloud analysis layer, a communication network and a user terminal:

[0007] The perception layer is composed of a variety of non-invasive sensors, including a magnetic array sensor, an infrared thermal imaging camera, a sound collecting device, a behavior monitoring camera and a weight monitoring platform, which are used to collect animal vital sign data.

[0008] The edge computing layer is configured to perform preliminary anomaly detection analysis and anomaly detection on the animal vital sign data collected by the plurality of non-invasive sensors in combination with a basic threshold database, and trigger a local alarm if necessary.

[0009] The cloud analysis layer is configured to perform deep analysis, comprehensive health assessment, and hierarchical alarm decision on the data transmitted by the edge computing layer, including establishing an individualized health baseline parameter set for the physiological parameter time series data and the basic threshold data in the edge computing layer, performing anomaly detection on the individualized health baseline parameter set by an anomaly detection algorithm, and triggering an alarm based on the anomaly detection result and the hierarchical alarm decision;

[0010] The communication network is configured to connect the perception layer, the edge computing layer, and the cloud analysis layer, support wired and wireless backup, and have a network security protection mechanism.

[0011] The user terminal includes a Web application, a mobile application, and a large-screen display, and is configured to receive alarm information, view health reports, and perform system management.

[0012] The perception layer, the edge computing layer, and the cloud analysis layer together form a three-layer architecture for data collection, local processing, and cloud collaborative analysis, and perform long-term health monitoring based on a dynamic optimization mechanism combining preset thresholds and machine learning models.

[0013] In some embodiments, the edge computing layer includes a heart rate data analysis module and a temperature data analysis module.

[0014] The heart rate data analysis module is configured to input raw data of a magnetic array sensor, perform calculation by a multi-channel adaptive magnetic field heart rate detection algorithm, and output a heart rate value, heart rate variability, and a signal quality index.

[0015] The temperature data analysis module is configured to input an infrared thermal imaging video stream, perform calculation by a multi-region thermal imaging body temperature estimation algorithm, and output an animal core body temperature, a body surface temperature distribution feature, and a temperature abnormal region.

[0016] In some embodiments, the heart rate data analysis module includes a signal preprocessing unit, a heart rate detection calculation unit, a quality evaluation unit, and a basic threshold comparison unit.

[0017] The signal preprocessing unit is configured to apply a notch filter to remove power frequency interference, apply a band-pass filter to extract a heart magnetic field frequency band, apply a wavelet denoising algorithm to each sensor channel to calculate the signal-to-noise ratio of each channel, select the top 25% channels with the highest signal-to-noise ratio, apply independent component analysis to separate the heart signal from the selected channels, and select the independent component that best matches the heart template as the heart signal.

[0018] The heart rate detection calculation unit is configured to detect R-wave peaks by applying a Pan-Tompkins algorithm, calculate adjacent R-R intervals, filter abnormal R-R intervals by applying an adaptive threshold, calculate average R-R intervals within a preset time window, and calculate heart rate=60 / average R-R interval, and calculate R-R interval standard deviation as a heart rate variability indicator.

[0019] The quality assessment unit is configured to calculate a signal quality index (SQI) representing signal quality based on signal-to-noise ratio (SNR) and R-wave detection consistency, and the SQI calculation formula is:

[0020]

[0021] wherein, SNR represents signal-to-noise ratio, measuring the strength of the cardiac signal relative to background noise, and W1 and W2 represent respective weight coefficients, RR represents the current R-R interval, which is the time between adjacent heartbeats, mean_RR represents the average R-R interval,

[0022] When SQI<0.6, it is marked as low quality, and the heart rate value is unreliable, when 0.6≤SQI<0.8, it is marked as medium quality, and the heart rate value is usable but needs to be cautious, and when SQI≥0.8, it is marked as high quality, and the heart rate value is reliable.

[0023] The baseline threshold comparison unit is configured to obtain the heart rate baseline threshold range [HR_min, HR_max] of the current animal from a baseline threshold database, calculate the deviation percentage of the current heart rate from the threshold, and mark it as potential abnormality when the deviation exceeds ±15%, and mark it as definite abnormality when the deviation exceeds ±30%.

[0024] In some embodiments, the temperature data analysis module includes an image preprocessing unit, an image detection and segmentation unit, a key area identification unit, and a temperature analysis unit.

[0025] The image preprocessing unit is configured to apply Gaussian filtering to reduce thermal noise, perform geometric correction to compensate for lens distortion, and apply radiation correction to compensate for environmental reflection.

[0026] The image detection and segmentation unit is configured to identify the animal in the image using a target detection algorithm, apply a semantic segmentation network to separate the animal contour from the background, and generate a heat map mask of the animal body.

[0027] The key area identification unit is configured to identify the medial canthus area, the ear base area, and the nose area.

[0028] The temperature analysis unit is used for extracting corresponding temperatures in the key area, wherein the body temperature of the inner corner of the eye area is the core body temperature, the correlation is the highest, the temperature of the ear base area is the secondary correlation, and the temperature of the nose area is greatly affected by the environment temperature, which is used for temperature calibration, and the temperature extraction formula is:

[0029]

[0030] wherein, represents the temperature of the inner corner of the eye area, represents the temperature of the ear base area, represents the weight coefficient of the temperature of the inner corner of the eye area, represents the weight coefficient of the temperature of the ear base area, represents a bias term, which is determined based on experimental data;

[0031] The environment temperature compensation formula is:

[0032]

[0033] wherein, represents the current environment temperature, represents the standard environment temperature, represents an environment compensation coefficient.

[0034] In some embodiments, the cloud analysis layer includes an individualized health baseline establishment module, an anomaly detection module, and a hierarchical alarm decision module;

[0035] The individualized health baseline establishment module is used for inputting animal ID, physiological parameter time series data, expert-labeled health status and basic threshold data, calculating through an adaptive health baseline generation algorithm, and outputting an individual health baseline parameter set;

[0036] The anomaly detection module is used for inputting the individual health baseline parameter set and the basic threshold data, calculating through an anomaly detection algorithm, and outputting an anomaly detection result;

[0037] The hierarchical alarm decision module is used for inputting the anomaly detection result, historical alarm records, user feedback data, time environment information and basic threshold deviation, calculating through a context-aware hierarchical alarm decision algorithm, and outputting an alarm level, a notification object and alarm content.

[0038] In some embodiments, the individualized health baseline establishment module includes a basic threshold initialization unit, a data segmentation filtering unit, a basic statistical feature extraction unit, a time series feature extraction unit, an environment factor correlation analysis unit and a baseline parameter set generation unit;

[0039] The basic threshold initialization unit is configured to obtain an initial threshold from a basic threshold database according to an animal species, an age, and a gender, and generate an initial baseline parameter set according to an animal-specific adjustment factor;

[0040] The data segment filtering unit is configured to determine, based on an animal moving speed and a posture, an activity state segment according to an activity state, the activity state including rest, light activity, moderate activity, and high activity, and filter out time period data marked as unhealthy by an expert;

[0041] The basic statistical feature extraction unit is configured to calculate a mean value and a standard deviation for each physiological parameter in each activity state, divide 24 hours into multiple time periods, and calculate a mean value, a maximum value, a minimum value, and a variation coefficient of each time period, the mean value and the standard deviation being calculated according to the following formula:

[0042]

[0043] The variation coefficient is calculated according to the following formula:

[0044]

[0045] wherein, represents an i-th physiological parameter measurement value, represents a sample number, represents a mean value, represents a standard deviation; is a variation coefficient, representing a percentage of the standard deviation relative to the mean value;

[0046] The time sequence feature extraction unit is configured to extract time sequence features by applying an autoregressive integrated moving average model, including calculating short-term, medium-term, and long-term trend features, and obtaining periodic features.

[0047] The environmental factor correlation analysis unit is configured to calculate a correlation coefficient of a physiological parameter and an environmental temperature, and establish a linear regression model:

[0048]

[0049]

[0050] wherein, represents an environmental-adjusted physiological parameter value, represents an originally measured physiological parameter value, represents an environmental adjustment coefficient, represents a current environmental temperature, represents a reference environmental temperature, represents an i-th environmental temperature sample, represents an average environmental temperature, represents an i-th physiological parameter sample, represents the average physiological parameter value;

[0051] The baseline parameter set generation unit is configured to generate a normal range for each activity state and each physiological parameter, ensure that the generated normal range does not exceed the safety boundary of the basic threshold, and when individual data is insufficient, the basic threshold is used preferentially, and as data accumulates, the weight of individualized data is adaptively increased to obtain a final baseline parameter, wherein the calculation formula of the normal range is:

[0052]

[0053] The weight adaptive adjustment calculation formula is:

[0054]

[0055]

[0056] The final baseline parameter calculation formula is:

[0057]

[0058] wherein, represents the weight of individualized data, represents the weight of basic threshold data, represents the amount of accumulated individual data, represents a threshold value for completely using individualized data, represents the final baseline parameter value, represents a parameter value calculated from individual data, represents a parameter value in the basic threshold.

[0059] In some embodiments, the anomaly detection module includes a single-parameter anomaly detection unit, a multi-parameter correlation analysis unit, a time sequence pattern analysis unit, an anomaly rating unit, and an anomaly rating and report generation unit.

[0060] The single-parameter anomaly detection unit is configured to obtain individualized baseline parameters in a current activity state, and calculate a Z-score of each physiological parameter, with the formula being:

[0061]

[0062] The basic threshold value is applied as a safety boundary, and parameters that exceed the individualized baseline or the basic threshold value are marked;

[0063] The multi-parameter correlation analysis is configured to apply a Bayesian network model to analyze conditional dependency relationships between physiological parameters, detect combined anomalies of physiological parameters, and calculate a multi-physiological parameter anomaly score.

[0064] The time series pattern analysis unit is configured to analyze parameter variation trends by using a sliding window, detect abnormal trends, and compare the abnormal trends with historical time series patterns to identify situations deviating from normal patterns, wherein the abnormal trends include continuous rising / falling, mutation, and fluctuation increase.

[0065] The abnormal rating unit is configured to calculate a basic abnormal score based on single-parameter abnormality, multi-parameter correlation abnormality, and time series abnormality, according to the following formula:

[0066]

[0067]

[0068] wherein, is a Z-score of the i-th parameter, indicating a standard deviation number of an observation value deviating from a mean value, represents an observation value of the i-th parameter, represents a basic abnormal score, represents a weight of the i-th parameter, represents a time factor reflecting an influence of abnormal duration, represents a variation rate factor reflecting an influence of parameter variation speed, represents abnormal duration, represents a reference time, represents a variation rate of the Z-score;

[0069] An abnormal level is determined according to the abnormal score and the duration, including a slight abnormality, a moderate abnormality, a severe abnormality, and an urgent abnormality.

[0070] The abnormal rating and report generation unit is configured to determine an abnormal level according to the number of abnormal parameters, the deviation degree, and the duration, generate an abnormal description including a starting time, an affected parameter, and a severity, and generate a visual report including an abnormal parameter trend chart and comparison data.

[0071] In some embodiments, the hierarchical alarm decision module includes a final abnormal score calculation unit, an alarm level determination unit, a notification object determination unit, and an alarm content generation unit.

[0072] The final abnormal score calculation unit is configured to calculate a context adjustment coefficient by combining a time factor, a historical factor, and an animal-specific factor, and calculate a final abnormal score based on the calculated context adjustment coefficient, according to the following formula:

[0073]

[0074]

[0075]

[0076] in, represents the final anomaly score, represents the context adjustment coefficient, represents the time-dependent adjustment factor, Historically relevant adjustment factors, Indicates the special state adjustment factor, Indicates the maximum absolute value of the Z score among all parameters, represents the time factor weight of the contextual factor, represents the historical factor weights of contextual factors, Special state weights representing contextual factors, represents the urgency factor, which is associated with the largest Z score;

[0077] The alarm level determination unit is used to determine the alarm level using fuzzy logic rules, including when <3.0 or the deviation of a single parameter exceeds 10% of the basic threshold, it is determined to be a warning level alarm. <3.0 or the deviation of a single parameter exceeds 10% to 20% of the basic threshold, it is determined to be a concern level alarm. <9.0 or the deviation of a single parameter exceeds 20% to 30% of the basic threshold, it is determined to be a warning level alarm. ≥9.0 or the deviation of a single parameter exceeds 30% of the basic threshold, it is determined to be an emergency alarm;

[0078] The notification object determination unit is used to determine the notification object according to the alarm level. When it is a prompt-level alarm, the record is displayed in the system; when it is a concern-level alarm, the record is displayed in the system and the breeder is notified; when it is a warning-level alarm, the record is displayed in the system and the breeder and the on-duty veterinarian are notified; when it is an emergency-level alarm, the record is displayed in the system and the breeder, on-duty veterinarian and management personnel are notified;

[0079] The alarm content generation unit is used to generate a concise alarm title including the animal ID and the abnormality type, generate detailed alarm content including abnormal parameters, values ​​and duration, add historical comparison information, and display parameter change trends.

[0080] In a second aspect, the present application proposes a health monitoring and alarm method for individual zoo animals, comprising the following steps:

[0081] Continuously collect animal vital sign data through perception layer sensors;

[0082] The edge computing layer pre-processes the data and extracts parameters such as heart rate, respiratory rate, body temperature, and activity level;

[0083] Preliminary anomaly detection based on a preset threshold triggers a local alarm;

[0084] The processed data is transmitted to a cloud analysis layer for deep analysis combined with a machine learning model;

[0085] A health report is generated according to the anomaly detection result, and an alarm is triggered based on a grading rule.

[0086] In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0087] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, wherein the computer program is executed by a processor to implement the steps of the above method.

[0088] Advantages of the present application:

[0089] Through the synergy of the perception layer (high-sensitivity magnetic array sensor, infrared thermal imaging camera, and other non-invasive devices), the edge computing layer (real-time data preprocessing and preliminary analysis), and the cloud analysis layer (deep machine learning and comprehensive evaluation), the above technical bottlenecks are solved. The system sets differentiated thresholds for animals of different age groups, genders, and health conditions, combined with dynamic baseline adjustment and multi-source data fusion algorithms, significantly improving the accuracy of anomaly detection (false positive rate reduced to 5-8%). At the same time, the grading alarm mechanism and feedback-driven model optimization form a closed loop, realizing the transition from "passive response" to "active early warning", which can identify potential health risks 12-24 hours in advance, providing efficient and accurate technical support for animal health management. BRIEF DESCRIPTION OF DRAWINGS

[0090] Figure 1 The system principle block diagram of the present application.

[0091] Figure 2 The overall flowchart of the present application. DETAILED DESCRIPTION

[0092] Exemplary embodiments of the present application will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein; on the contrary, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0093] In view of the above deficiencies of the prior art, the present application provides a health monitoring and alarm system and method for individual animals in a zoo.

[0094] The first aspect of the present application proposes a health monitoring and alarm system for individual zoo animals, as shown, including a perception layer, an edge computing layer, a cloud analysis layer, a communication network and a user terminal: Figure 1

[0095] The perception layer is composed of various non-invasive sensors, including magnetic array sensors, infrared thermal imaging cameras, sound collection devices, behavior monitoring cameras and body weight monitoring platforms, for collecting animal vital sign data.

[0096] The magnetic array sensor has the function of monitoring heart rate and respiratory rate, and the technical feature of high-sensitivity magnetic resistance sensor array can detect the weak magnetic field changes generated by animal heart activity. It is installed under the ground in the animal resting area, and when the animal is a young one, the layout is considered according to the characteristics of small size of young animals.

[0097] The infrared thermal imaging camera has the function of monitoring body surface temperature distribution, and the technical feature of high-precision thermal imaging technology can non-contact measure animal body surface temperature. It is installed on the top or side wall of the animal cage.

[0098] The sound collection device has the function of monitoring breathing sound and abnormal sound, and the technical feature of directional microphone array has noise suppression and sound source positioning function. It is installed on the walls around the cage.

[0099] The behavior monitoring camera has the function of monitoring activity state and behavior mode, and the technical feature of high-definition camera supports infrared night vision function. It is installed at two opposite positions on the top of the cage.

[0100] The edge computing layer is used to preliminarily detect and analyze the animal vital sign data collected by various non-invasive sensors through various animal health analysis algorithms combined with basic threshold database, and trigger local alarm when necessary.

[0101] The edge computing layer has an edge computing unit built-in, which is installed in the technical equipment room near the animal cage and is responsible for the preliminary processing and analysis of sensor data.

[0102] Hardware features: multi-core processor, sufficient memory and storage space, support for multiple interfaces including Ethernet, Wi-Fi, RS485, etc., industrial design, suitable for harsh environment, support for uninterrupted power supply.

[0103] Software configuration: custom Linux operating system, containerized microservices, time series database, lightweight machine learning framework.

[0104] ​Further, the system adopts a scientific basic threshold setting method to provide precise health monitoring benchmarks for animals of different age groups. The system establishes a comprehensive basic threshold database, including:

[0105] Species-specific thresholds:

[0106] Differential thresholds are set according to animal taxonomy groups (families, genera, species). Specialized threshold sets for large cats are considered for subspecies differences (such as Siberian tigers, Bengal tigers, etc.) and threshold adjustments.

[0107] Age group thresholds (with special attention to juveniles and sub-adults):

[0108] Juvenile stage (0-1 years old): specific thresholds for growth and development period;

[0109] Sub-adult stage (1-3 years old): threshold for development transition period;

[0110] Adult stage (3-10 years old): stable period threshold;

[0111] Old age stage (>10 years old): specific thresholds for aging period;

[0112] It also includes gender difference thresholds, male-specific thresholds, female-specific thresholds, special thresholds for breeding period, health status thresholds, health status benchmark thresholds, recovery period adjustment thresholds, chronic disease management thresholds, and key physiological parameter basic thresholds. The system sets differential thresholds for animals of different age groups, with special attention to the special physiological characteristics of juveniles and sub-adults, such as:

[0113] 1. Heart rate threshold (times / minute):

[0114] According to age classification: juvenile, sub-adult, adult, old;

[0115] According to activity state classification: resting state, light activity, moderate activity, high activity;

[0116] The basic heart rate of juvenile animals is generally 20-40% higher than that of adult animals;

[0117] The basic heart rate of sub-adult animals is 10-25% higher than that of adult animals;

[0118] 2. Respiratory rate threshold (times / minute):

[0119] Classification according to age and activity state;

[0120] The resting state respiratory rate benchmark range of juveniles is 50-70% higher than that of adult animals;

[0121] The resting state respiratory rate of sub-adults is 30-40% higher than that of adult animals;

[0122] Environmental temperature adjustment: increase correction coefficient in high temperature environment;

[0123] 3. Body temperature threshold (°C):

[0124] Different age groups set different core body temperature normal ranges;

[0125] Juvenile and subadult set specific fever warning and emergency thresholds;

[0126] Consider the normal range of intra-day temperature fluctuations;

[0127] 4. Activity threshold (activity time percentage / hour):

[0128] Set different standards according to circadian rhythm and age group;

[0129] Juvenile animals generally have higher daytime activity time percentage than adult animals;

[0130] Set warning thresholds for low activity and overactivity;

[0131] 5. Growth and development threshold (designed for juveniles in particular):

[0132] Monitor the compliance of body weight gain rate with the standard curve;

[0133] Set a warning threshold for deviation from the body weight gain curve;

[0134] Record and evaluate key activity ability development milestones;

[0135] It also includes threshold management and updating mechanism: when a new animal enters the zoo, it is automatically assigned an initial threshold value according to species, age, and gender. The veterinarian confirms and adjusts the initial threshold value according to individual characteristics, and sets an adaptation period (usually 30 days), during which the threshold sensitivity is reduced by 20%;

[0136] Threshold dynamic adjustment: adjust the base threshold value according to historical data trends, adjust according to changes in health status, seasonal changes, age growth, etc. The veterinarian can manually adjust the threshold value of a specific animal.

[0137] In some embodiments, the edge computing layer includes a heart rate data analysis module and a temperature data analysis module;

[0138] The heart rate data analysis module inputs the raw data of the magnetic array sensor, calculates through a multi-channel adaptive magnetic field heart rate detection algorithm, and outputs the heart rate value, heart rate variability, and signal quality index;

[0139] In some embodiments, the heart rate data analysis module includes a signal preprocessing unit, a heart rate detection calculation unit, a quality assessment unit, and a basic threshold comparison unit;

[0140] The signal preprocessing unit is configured to apply a notch filter to remove power frequency interference, apply a band-pass filter to extract a cardiac magnetic field frequency band, apply a wavelet denoising algorithm to each sensor channel to calculate a signal-to-noise ratio of each channel, select the top 25% channels with the highest signal-to-noise ratio, apply independent component analysis to separate cardiac signals from the selected channels, and select an independent component that best matches a cardiac template as a cardiac signal.

[0141] The heart rate detection and calculation unit is configured to apply a Pan-Tompkins algorithm to detect R-wave peaks, calculate adjacent R-R intervals, apply an adaptive threshold to filter abnormal R-R intervals, calculate an average R-R interval within a preset time window, calculate a heart rate as 60 / average R-R interval, and calculate an R-R interval standard deviation as a heart rate variability indicator.

[0142] The quality assessment unit is configured to calculate a signal quality index (SQI) based on a signal-to-noise ratio (SNR) and R-wave detection consistency, and the SQI calculation formula is:

[0143]

[0144] wherein, SNR represents the signal-to-noise ratio, which measures the strength of the cardiac signal relative to the background noise, and W1 and W2 represent respective weight coefficients, RR represents a current R-R interval, which is the time between adjacent heartbeats, mean_RR represents an average R-R interval,

[0145] When SQI < 0.6, it is marked as low quality, and the heart rate value is unreliable; when 0.6 ≤ SQI < 0.8, it is marked as medium quality, and the heart rate value is usable but needs to be cautious; and when SQI ≥ 0.8, it is marked as high quality, and the heart rate value is reliable.

[0146] The baseline threshold comparison unit is configured to obtain a heart rate baseline threshold range [HR_min, HR_max] of the current animal from a baseline threshold database, calculate a deviation percentage of the current heart rate from the threshold, and mark it as a potential abnormality when the deviation exceeds ±15%, and mark it as a definite abnormality when the deviation exceeds ±30%.

[0147] wherein, an example calculation of the heart rate data analysis module is:

[0148] Given data: signal amplitude: 1.2 mV, noise amplitude: 0.2 mV, current R-R interval: 0.8 s, average R-R interval: 0.75 s, the calculation process is as follows:

[0149]

[0150] The temperature data analysis module is configured to input an infrared thermal imaging video stream, calculate the core body temperature of the animal, the surface temperature distribution characteristics of the animal, and the temperature abnormal area of the animal by using a multi-region thermal imaging body temperature estimation algorithm.

[0151] In some embodiments, the temperature data analysis module comprises an image preprocessing unit, an image detection and segmentation unit, a key region identification unit, and a temperature analysis unit.

[0152] The image preprocessing unit is configured to apply Gaussian filtering to reduce thermal noise, perform geometric correction to compensate for lens distortion, apply radiation correction to compensate for environmental reflection, and the like.

[0153] The image detection and segmentation unit is configured to identify the animal in the image using a target detection algorithm, apply a semantic segmentation network to separate the animal contour from the background, and generate a thermal map mask of the animal body.

[0154] The key region identification unit is configured to identify the medial canthus region, the ear base region, and the nose region.

[0155] The temperature analysis unit is configured to extract the corresponding temperature in the key region, wherein the body temperature of the medial canthus region is the core body temperature with the highest correlation, the temperature of the ear base region is the secondary correlation, and the temperature of the nose region is greatly affected by the environmental temperature, and is used for temperature calibration.

[0156]

[0157] wherein, T1 represents the temperature of the medial canthus region, T2 represents the temperature of the ear base region, W1 represents the weight coefficient of the temperature of the medial canthus region, W2 represents the weight coefficient of the temperature of the ear base region, B represents a bias term, which is determined based on experimental data.

[0158] The environmental temperature compensation formula is:

[0159]

[0160] wherein, T represents the current environmental temperature, T0 represents the standard environmental temperature, W represents the environmental compensation coefficient.

[0161] In some embodiments, the temperature data analysis module comprises an image preprocessing unit, an image detection and segmentation unit, a key region identification unit, and a temperature analysis unit.

[0162] Given data: eye corner temperature (Teye): 38.5°C, ear temperature (Tea): 38.0°C, current environment temperature: 25°C, standard environment temperature: 20°C, weights: w1=0.7, w2=0.3, bias: b=0.5, environment compensation coefficient: k=0.08, the calculation process is as follows:

[0163]

[0164] The cloud analysis layer is configured to perform deep analysis, comprehensive health assessment, and hierarchical alarm decision on the data transmitted by the edge computing layer, including establishing an individualized health baseline parameter set for the physiological parameter time series data and the basic threshold data in the edge computing layer, performing abnormality detection on the individualized health baseline parameter set by an abnormality detection algorithm, and triggering an alarm based on the abnormality detection result and the hierarchical alarm decision;

[0165] In some embodiments, the cloud analysis layer includes an individualized health baseline establishment module, an abnormality detection module, and a hierarchical alarm decision module.

[0166] The individualized health baseline establishment module is configured to input an animal ID, physiological parameter time series data, expert-labeled health status, and basic threshold data, perform calculation by an adaptive health baseline generation algorithm, and output an individual health baseline parameter set.

[0167] In some embodiments, the individualized health baseline establishment module includes a basic threshold initialization unit, a data segmentation filtering unit, a basic statistical feature extraction unit, a time series feature extraction unit, an environmental factor correlation analysis unit, and a baseline parameter set generation unit.

[0168] The basic threshold initialization unit is configured to obtain initial threshold values from a basic threshold database according to animal species, age, and gender, and generate an initial baseline parameter set according to animal-specific adjustment factors.

[0169] The data segmentation filtering unit is configured to segment and determine based on animal movement speed and posture according to activity states, including rest, light activity, moderate activity, and high activity, and filter out time period data labeled as unhealthy by experts.

[0170] The basic statistical feature extraction unit is configured to calculate the mean and standard deviation of each physiological parameter in each activity state, divide 24 hours into multiple time periods, and calculate the average, maximum, minimum, and coefficient of variation of each time period. The mean and standard deviation calculation formula is:

[0171]

[0172] The coefficient of variation calculation formula is:

[0173]

[0174] wherein, represents the i-th physiological parameter measurement value, represents the sample number, represents the mean value, represents the standard deviation; is the coefficient of variation, representing the percentage of the standard deviation relative to the mean value;

[0175] The time sequence feature extraction unit is configured to apply an autoregressive integrated moving average model to extract time sequence features, including calculating short-term, medium-term and long-term trend features to obtain periodic features.

[0176] The environmental factor correlation analysis unit is configured to calculate a correlation coefficient of the physiological parameter and the environmental temperature, and establish a linear regression model:

[0177]

[0178]

[0179] wherein, represents the physiological parameter value adjusted according to the environment, represents the original measured physiological parameter value, represents the environmental adjustment coefficient, represents the current environmental temperature, represents the reference environmental temperature, represents the i-th environmental temperature sample, represents the average environmental temperature, represents the i-th physiological parameter sample, represents the average physiological parameter value;

[0180] The baseline parameter set generation unit is configured to generate a normal range for each activity state and each physiological parameter, and ensure that the generated normal range does not exceed the safety boundary of the basic threshold. When the individual data is insufficient, the basic threshold is used as a bias. With the accumulation of data, the weight of individualized data is adaptively increased to obtain the final baseline parameter. The calculation formula of the normal range is:

[0181]

[0182] The weight adaptive adjustment calculation formula is:

[0183]

[0184]

[0185] The final baseline parameter calculation formula is:

[0186]

[0187] wherein, represents the weight of individualized data, represents the weight of the base threshold data, represents the cumulative amount of individual data, represents the threshold for complete adoption of individualized data, represents the final baseline parameter value, represents the parameter value calculated from individual data, represents the parameter value in the base threshold.

[0188] wherein, the example calculation of the individualized health baseline establishment module is:

[0189] Given data: heart rate data (times / minute): 75, 78, 72, 76, 74, 77, 73, 75, 76, 74, the calculation process is as follows:

[0190]

[0191] Given data: ambient temperature (°C): 20, 22, 24, 26, 28; heart rate (times / minute): 72, 74, 76, 78, 80, the calculation process is as follows:

[0192]

[0193] At this time, when the ambient temperature changes by 1°C, the heart rate changes by about 1 time / minute.

[0194] Assuming that an animal has accumulated 15 days of data, and the threshold is set to 30 days:

[0195]

[0196] If: P individual (individual calculation of mean heart rate) = 75, P base (base threshold of mean heart rate) = 80, then:

[0197]

[0198] The anomaly detection module is configured to input the individual health baseline parameter set and the base threshold data, calculate through an anomaly detection algorithm, and output an anomaly detection result.

[0199] In some embodiments, the anomaly detection module includes a single-parameter anomaly detection unit, a multi-parameter correlation analysis unit, a time series pattern analysis unit, an anomaly rating unit, and an anomaly rating and report generation unit.

[0200] The single-parameter anomaly detection unit is configured to obtain individualized baseline parameters in a current active state, and calculate a Z-score of each physiological parameter according to the following formula:

[0201]

[0202] The application of the basic threshold as a safety boundary marks the parameters exceeding the individualized baseline or the basic threshold;

[0203] The multi-parameter correlation analysis is configured to analyze conditional dependency relationships among the physiological parameters by using a Bayesian network model, detect a combination anomaly of the physiological parameters, and calculate a multi-physiological parameter anomaly score;

[0204] The time-series pattern analysis unit is configured to analyze a parameter change trend by using a sliding window, detect an abnormal trend, and compare the abnormal trend with historical time-series patterns to identify a situation deviating from a normal pattern, wherein the abnormal trend includes a continuous rise / fall, a mutation, and an increased fluctuation.

[0205] The anomaly rating unit is configured to calculate a basic anomaly score by comprehensively considering the single-parameter anomaly, the multi-parameter correlation anomaly, and the time-series anomaly according to the following formula:

[0206]

[0207]

[0208] wherein, Z i represents a Z-score of the i th parameter, and indicates a standard deviation number of an observation value deviating from a mean value, X i represents an observation value of the i th parameter, A represents a basic anomaly score, W i represents a weight of the i th parameter, T represents a time factor reflecting an influence of an abnormal duration, V represents a change rate factor reflecting an influence of a parameter change speed, D represents an abnormal duration, T 0 represents a reference time, V Z represents a change rate of the Z-score;

[0209] An anomaly level is determined according to the anomaly score and the abnormal duration, and includes a slight anomaly, a moderate anomaly, a severe anomaly, and an urgent anomaly.

[0210] The anomaly rating and report generation unit is configured to determine an anomaly level according to a number of abnormal parameters, a deviation degree, and a duration, generate an anomaly description including a starting time, an affected parameter, and a severity, and generate a visual report including an abnormal parameter trend chart and comparison data.

[0211] wherein, an example calculation of the anomaly detection module is as follows:

[0212] Monitoring data: heart rate: current value 95, mean value 80, standard deviation 10; body temperature: current value 39.2, mean value 38.5, standard deviation 0.3; weight: wheart rate=0.6, wbody temperature=0.4; abnormal duration: 2 hours (reference time 1 hour); heart rate change rate: 5 times / minute / min;

[0213] Calculation process: Zheart rate=(95-80) / 10=1.5; Zbody temperature=(39.2-38.5) / 0.3=2.33;

[0214] Ftime=1+log2(2 / 1)=2; Fspeed=1+0.5*|5 / 80|=1.03;

[0215] Basic abnormal score=(0.6*|1.5|+0.4*|2.33|)*2*1.03=3.77.

[0216] The hierarchical alarm decision module is configured to input abnormal detection results, historical alarm records, user feedback data, time environment information, and basic threshold deviation, calculate through a context-aware hierarchical alarm decision algorithm, and output an alarm level, a notification object, and alarm content.

[0217] In some embodiments, the hierarchical alarm decision module includes a final abnormal score calculation unit, an alarm level determination unit, a notification object determination unit, and an alarm content generation unit.

[0218] The final abnormal score calculation unit is configured to calculate a context adjustment coefficient in combination with time factors, historical factors, and animal-specific factors, and calculate a final abnormal score based on the calculated context adjustment coefficient, according to the formula:

[0219]

[0220]

[0221]

[0222] wherein, represents the final abnormal score, represents the context adjustment coefficient, represents a time-related adjustment factor, a historical-related adjustment factor, represents a special state adjustment factor, represents the maximum value of the absolute value of the Z-score in all parameters, represents the time factor weight of the context factor, represents the historical factor weight of the context factor, a special state weight representing a contextual factor, an urgency factor, related to the largest Z-score;

[0223] The alarm level determination unit is configured to determine the alarm level using fuzzy logic rules, including determining a prompt-level alarm when

[0224] The notification object determination unit is configured to determine the notification object according to the alarm level, recording in the system for display when it is a prompt-level alarm, recording in the system for display and notifying the breeder when it is a concern-level alarm, recording in the system for display and notifying the breeder and on-duty veterinarian when it is a warning-level alarm, and recording in the system for display and notifying the breeder, on-duty veterinarian and management personnel when it is an emergency-level alarm.

[0225] The alarm content generation unit is configured to generate a concise alarm title containing the animal ID and the abnormal type, generate detailed alarm content containing the abnormal parameter, value and duration, add historical comparison information, and display the parameter change trend.

[0226] wherein the example calculation of the hierarchical alarm decision module is as follows:

[0227] Given data: base abnormal score: 3.77, night period (time factor: 0.2), recent similar alarms (history factor: 0.3), animal in recovery period (special state factor: -0.2), weights: w1=0.3, w2=0.3, w3=0.4, maximum Z-score: 2.33;

[0228] Calculation process:

[0229] =1+0.3*0.2+0.3*0.3+0.4*(-0.2)=1.07;

[0230] =1+log2(2.33)=2.22;

[0231] ​​​​=3.77*1.07*2.22=8.93, at this time, it is determined as a warning level alarm, which needs to be displayed on the system interface, and the keeper and the on-duty veterinarian are notified through the mobile application and the short message respectively.

[0232] The communication network is used to connect the perception layer, edge computing layer and cloud analysis layer, supports wired and wireless backup, and has network security protection mechanism;

[0233] The user terminal includes a Web application, a mobile application and a large screen display, and is used for receiving alarm information, viewing health reports and performing system management;

[0234] The perception layer, edge computing layer and cloud analysis layer jointly form a three-layer architecture for data collection, local processing and cloud collaborative analysis, and perform long-term health monitoring based on a dynamic optimization mechanism combining preset thresholds and machine learning models.

[0235] The mechanism in which the threshold value cooperates with the machine learning includes:

[0236] The system relies on preset thresholds for monitoring at the initial stage of deployment, and gradually increases the weight of the machine learning model as data accumulates;

[0237] The basic threshold value serves as a safety boundary to ensure that the alarm is triggered when the parameter exceeds the threshold value, and is not affected by the model judgment;

[0238] The machine learning model is used to identify parameter change trends and multi-parameter correlation abnormalities, reduce false alarm rates and improve early warning capabilities.

[0239] The initial monitoring stage based on the threshold value includes:

[0240] At the initial stage of system deployment, the preset basic threshold value is mainly relied on for monitoring and alarm, and the system is pre-installed with standard physiological parameter thresholds in scientific literature and veterinarian manuals, which are stored according to species, age group and gender

[0241] For example: the normal range of heart rate of young male animals (3 months) in resting state is 90-120 times / minute

[0242] Animal file establishment: when a new animal enters the zoo, basic information such as species, subspecies, gender, age and weight is recorded, and the system automatically matches the most suitable basic threshold set from the threshold value library, and the veterinarian can manually adjust the initial threshold value according to the specific circumstances of the animal;

[0243] Threshold value grading: three threshold values are set for each physiological parameter:

[0244] Warning level: slightly deviates from the normal range (e.g. 0-10% higher heart rate);

[0245] Attention level: deviates from the normal range (e.g. 10% higher heart rate);

[0246] Warning level: Significant deviation from normal range (e.g. 20% higher heart rate).

[0247] Emergency level: Severe deviation from normal range (e.g. 30% or more higher heart rate).

[0248] Threshold-based alarm processing flow: Sensors continuously collect vital sign data, edge computing unit performs data preprocessing and parameter extraction, system obtains current animal activity state, selects corresponding threshold set, compares processed physiological parameters with corresponding thresholds, triggers different levels of alarms according to deviation degree:

[0249] Parameter exceeds warning level threshold: System interface displays yellow prompt, records abnormal event

[0250] Parameter exceeds attention level threshold: System interface displays orange warning, sends notification to breeder

[0251] Parameter exceeds warning level threshold: System interface displays red alarm, sends notification to breeder and veterinarian

[0252] Parameter exceeds emergency level threshold: System interface displays black alarm, sends notification to breeder and veterinarian, and synchronously notifies staff to receive alarm and perform on-site confirmation.

[0253] Record processing measures and results: Mark the effectiveness of the alarm (real abnormality / false alarm).

[0254] Optimization phase based on machine learning:

[0255] As data accumulates, the system gradually introduces machine learning algorithms to improve monitoring accuracy:

[0256] Data accumulation and preprocessing: The system continuously collects and stores animal vital sign data, cleans, standardizes and extracts features from the data, and classifies and organizes them according to animal individuals, activity states, environmental conditions, etc.

[0257] Individualized model training: Use the historical data of each animal to train individualized health models, apply supervised learning methods: use veterinarian-labeled health / abnormal states as training labels, apply unsupervised learning methods: identify natural clusters and abnormal patterns in the data;

[0258] Model types include: Time series anomaly detection model, multi-parameter correlation analysis model, health state classification model, trend prediction model and growth and development evaluation model

[0259] Model deployment and application: the trained model is deployed to edge computing units and central servers. Edge computing units run lightweight models for preliminary anomaly detection, and central servers run complex models for deep analysis. Model output and threshold judgment results are combined to determine whether to trigger an alarm.

[0260] Threshold and model cooperative working mechanism:

[0261] Initial system deployment: 100% dependent on threshold judgment;

[0262] After 30 days of data accumulation: 70% threshold judgment + 30% model judgment;

[0263] After 90 days of data accumulation: 50% threshold judgment + 50% model judgment;

[0264] After 180 days of data accumulation: 30% threshold judgment + 70% model judgment.

[0265] Regardless of how the machine learning model is optimized, the basic threshold always serves as a safety boundary. If the parameter exceeds the safety threshold, an alarm will be triggered regardless of the model judgment;

[0266] Model continuous optimization: regularly update the model using newly collected data, adjust model parameters based on alarm feedback, analyze false positive and false negative cases, improve algorithms, introduce new features and algorithms, and improve detection accuracy.

[0267] The second aspect of the application proposes a health monitoring and alarm method for individual animals in a zoo, as shown in Figure 2 The method comprises the following steps:

[0268] S100: continuously collect animal vital sign data through a perception layer sensor;

[0269] S200: the edge computing layer pre-processes the data and extracts heart rate, respiratory rate, body temperature and activity parameters;

[0270] S300: preliminary anomaly detection based on a preset threshold triggers a local alarm;

[0271] S400: transmit the processed data to the cloud analysis layer for deep analysis combined with a machine learning model;

[0272] S500: generate a health report based on the anomaly detection result and trigger an alarm based on the grading rules.

[0273] The third aspect of the application proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above method.

[0274] The fourth aspect provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the steps of the above method.

[0275] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the above described functions. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or software. In addition, the specific names of each functional unit and module are only for easy distinction, and do not limit the protection scope of the present application. The specific working process of the units and modules in the above system can refer to the corresponding process in the above method embodiments, which will not be repeated here.

[0276] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can be referred to the related description of other embodiments.

[0277] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present disclosure.

[0278] In the embodiments provided in the present disclosure, it should be understood that the disclosed devices / computer equipment and methods can be implemented in other ways. For example, the device / computer equipment embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division, and actual implementation can have another division manner. Multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0279] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e. may be located in one place, or may be distributed on multiple network units. Part or all of the units may be selected according to actual needs to achieve the purpose of the embodiment.

[0280] In addition, each functional unit in various embodiments of the present disclosure can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0281] If the integrated module / unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program can be executed by a processor to implement the steps of each method embodiment. The computer program can include computer program code, which can be in the form of source code, object code, executable file or some intermediate form. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier wave signal, telecommunication signal and software distribution medium, etc. It should be noted that the content included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electric carrier wave signal and telecommunication signal.

[0282] The above is only the preferred embodiment of the present application, and it should be noted that the technical solutions of several modifications and improvements made by those skilled in the art without departing from the present technical solutions should also be considered to fall within the scope of protection of the present application.

Claims

1. A health monitoring and alarm system for individual zoo animals, characterized by: Including perception layer, edge computing layer, cloud analysis layer, communication network and user terminals: The perception layer is composed of a variety of non-invasive sensors, including magnetic array sensors, infrared thermal imaging cameras, sound collection devices, behavior monitoring cameras and weight monitoring platforms, which are used to collect animal vital signs data; The edge computing layer is used to perform preliminary anomaly detection analysis and anomaly detection on the animal vital sign data collected by multiple non-invasive sensors using multiple animal health analysis algorithms combined with a basic threshold database, and trigger a local alarm when necessary; The cloud analysis layer is used to perform in-depth analysis, comprehensive health assessment, and hierarchical alarm decision-making on the data transmitted by the edge computing layer, including establishing an individualized health baseline parameter set based on the physiological parameter time series data and basic threshold data in the edge computing layer, performing anomaly detection on the individualized health baseline parameter set through an anomaly detection algorithm, and triggering an alarm based on the anomaly detection results and hierarchical alarm decision; The communication network is used to connect the perception layer, edge computing layer and cloud analysis layer, supports wired and wireless backup, and has a network security protection mechanism; The user terminal includes a web application, a mobile application, and a large-screen display for receiving alarm information, viewing health reports, and performing system management; The perception layer, edge computing layer and cloud analysis layer together form a three-layer architecture for data collection, local processing and cloud collaborative analysis, and perform long-term health monitoring based on a dynamic optimization mechanism combining preset thresholds and machine learning models.

2. The system according to claim 1, wherein: The edge computing layer includes a heart rate data analysis module and a temperature data analysis module; The heart rate data analysis module is used to input the original data of the magnetic array sensor, calculate it through the multi-channel adaptive magnetic field heart rate detection algorithm, and output the heart rate value, heart rate variability and signal quality index; The temperature data analysis module is used to input infrared thermal imaging video stream, calculate through multi-region thermal imaging body temperature estimation algorithm, and output animal core body temperature, surface temperature distribution characteristics and temperature abnormality areas.

3. The system according to claim 2, characterized in that: The heart rate data analysis module includes a signal preprocessing unit, a heart rate detection and calculation unit, a quality assessment unit and a basic threshold comparison unit; The signal preprocessing unit is configured to apply a notch filter to remove power frequency interference, apply a bandpass filter to extract the cardiac magnetic field frequency band, apply a wavelet denoising algorithm to each sensor channel to calculate the signal-to-noise ratio of each channel, select the top 25% channels with the highest signal-to-noise ratio, apply independent component analysis to separate the cardiac signal from the selected channels, and select the independent component that best matches the cardiac template as the cardiac signal; The heart rate detection and calculation unit is used to apply the Pan-Tompkins algorithm to detect the R wave peak, calculate the interval between adjacent R waves, apply an adaptive threshold to filter abnormal RR intervals, calculate the average RR interval within a preset time window, heart rate = 60 / average RR interval, and calculate the standard deviation of the RR interval as a heart rate variability indicator; The quality evaluation unit is used to calculate a signal quality index. The signal quality index is based on the signal-to-noise ratio and the consistency of R-wave detection. The signal quality index calculation formula is: in, represents the signal-to-noise ratio, which measures the strength of the cardiac signal relative to the background noise. and represents their respective weight coefficients, Indicates the current RR interval, which is the time between adjacent heartbeats. represents the average RR interval, When SQI < 0.6, it is marked as low quality and the heart rate value is unreliable. When 0.6 ≤ SQI < 0.8, it is marked as medium quality and the heart rate value can be used but with caution. When SQI ≥ 0.8, it is marked as high quality and the heart rate value is reliable. The basic threshold comparison unit is used to obtain the current animal's heart rate basic threshold range [HR_min, HR_max] from the basic threshold database, calculate the deviation percentage between the current heart rate and the threshold, and mark it as a potential abnormality when the deviation exceeds ±15%; when the deviation exceeds ±30%, mark it as a clear abnormality.

4. The system according to claim 3, wherein: The temperature data analysis module includes an image preprocessing unit, an image detection and segmentation unit, a key area identification unit and a temperature analysis unit; The image preprocessing unit is used to apply Gaussian filtering to reduce thermal noise, perform geometric correction, compensate for lens distortion, apply radiometric correction, and compensate for environmental reflections; The image detection and segmentation unit is used to identify animals in the image using an object detection algorithm, apply a semantic segmentation network to separate the animal outline from the background, and generate a heat map mask of the animal's body; The key area recognition unit is used to recognize the inner corner area of ​​the eye, the base area of ​​the ear and the nose area; The temperature analysis unit is used to extract the corresponding temperature in key areas. The body temperature in the inner corner of the eye is the core body temperature and has the highest correlation. The temperature in the ear base area is less correlated. The temperature in the nose area is greatly affected by the ambient temperature and is used for temperature calibration. The temperature extraction formula is: in, Indicates the temperature of the inner corner of the eye. Indicates the temperature of the ear base area. Indicates the weight coefficient of the temperature of the inner corner of the eye. represents the weight coefficient of the temperature in the base area of ​​the ear, represents the bias term, which is determined based on experimental data; The ambient temperature compensation formula is: in, Indicates the current ambient temperature. Indicates standard ambient temperature, Represents the environmental compensation coefficient.

5. The system according to claim 4, characterized in that: The cloud analysis layer includes an individualized health baseline establishment module, an anomaly detection module, and a hierarchical alarm decision module; The individualized health baseline establishment module is used to input animal ID, physiological parameter time series data, expert-marked health status and basic threshold data, calculate through the adaptive health baseline generation algorithm, and output the individual health baseline parameter set; The anomaly detection module is used to input an individual health baseline parameter set and basic threshold data, perform calculations using an anomaly detection algorithm, and output an anomaly detection result; The hierarchical alarm decision module is used to input anomaly detection results, historical alarm records, user feedback data, time environment information and basic threshold deviation, calculate through the context-aware hierarchical alarm decision algorithm, and output the alarm level, notification object and alarm content.

6. The system according to claim 5, characterized in that: The individualized health baseline establishment module includes a basic threshold initialization unit, a data segmentation filtering unit, a basic statistical feature extraction unit, a time series feature extraction unit, an environmental factor association analysis unit and a baseline parameter set generation unit; The basic threshold initialization unit is used to obtain an initial threshold from a basic threshold database based on the animal species, age, and sex, and to generate an initial baseline parameter set based on the animal-specific adjustment factor; The data segmentation and filtering unit is used to determine the animal's movement speed and posture based on the activity status segmentation, wherein the activity status includes rest, light activity, moderate activity and high activity, and filter out the time period data marked as unhealthy by experts; The basic statistical feature extraction unit is used to calculate the mean and standard deviation of each physiological parameter in each activity state, divide 24 hours into multiple time periods, and calculate the mean, maximum, minimum and coefficient of variation of each time period. The mean and standard deviation calculation formula is: The coefficient of variation calculation formula is: in, represents the measured value of the i-th physiological parameter, represents the number of samples, represents the mean, represents the standard deviation; is the coefficient of variation, which represents the percentage of the standard deviation relative to the mean; The time series feature extraction unit is used to extract time series features by applying an autoregressive integrated moving average model, including calculating short-term, medium-term and long-term trend features to obtain periodic features; The environmental factor correlation analysis unit is used to calculate the correlation coefficient between physiological parameters and ambient temperature and establish a linear regression model: in, Indicates the physiological parameter value after environmental adjustment, Represents the original measured physiological parameter value, represents the environmental adjustment coefficient, Indicates the current ambient temperature. Indicates the reference ambient temperature, represents the i-th ambient temperature sample, represents the average ambient temperature, represents the i-th physiological parameter sample, represents the average physiological parameter value; The baseline parameter set generation unit is used to generate a normal range for each activity state and each physiological parameter, ensuring that the generated normal range does not exceed the safety boundary of the basic threshold. When individual data is insufficient, the basic threshold is used as the basis. As data accumulates, the weight of the individualized data is adaptively increased to obtain the final baseline parameter. The calculation formula of the normal range is: The calculation formula for weight adaptive adjustment is: The final baseline parameter calculation formula is: in, represents the weight of individualized data, represents the weight of the basic threshold data, Indicates the amount of accumulated individual data, Indicates the threshold value of completely individualized data, represents the final baseline parameter value, represents the parameter value calculated from individual data, Indicates the parameter value in the basic threshold.

7. The system according to claim 6, characterized in that: The anomaly detection module includes a single parameter anomaly detection unit, a multi-parameter correlation analysis unit, a time series pattern analysis unit, an anomaly rating unit, and an anomaly rating and report generation unit; The single parameter anomaly detection unit is used to obtain the individualized baseline parameters under the current activity state and calculate the Z score of each physiological parameter. The formula is: Applying basic thresholds as safety margins, marking parameters that exceed individualized baselines or basic thresholds; The multi-parameter association analysis is used to apply the Bayesian network model to analyze the conditional dependency relationship between physiological parameters, detect abnormal combinations of physiological parameters and calculate abnormal scores of multiple physiological parameters; The time series pattern analysis unit is used to use a sliding window to analyze parameter change trends and detect abnormal trends, such as continuous rise / fall, sudden change, and increased fluctuation, and compare with historical time series patterns to identify deviations from normal patterns; The anomaly rating unit is used to calculate the basic anomaly score by combining single parameter anomalies, multi-parameter correlation anomalies, and time series anomalies. The formula is: in, is the Z score of the i-th parameter, which indicates the number of standard deviations that the observation deviates from the mean. represents the observed value of the i-th parameter, represents the basic anomaly score, represents the weight of the i-th parameter, Represents the time factor, reflecting the impact of abnormal duration, Indicates the rate of change factor, reflecting the influence of the parameter change speed, Indicates the duration of the abnormality, Indicates the base time, represents the rate of change of the Z score; The abnormality level is determined according to the abnormality score and duration, including mild abnormality, moderate abnormality, severe abnormality and urgent abnormality; The anomaly rating and report generation unit is used to assess the anomaly level based on the number of abnormal parameters, degree of deviation and duration, generate an anomaly description including start time, impact parameters and severity, and generate a visual report including an abnormal parameter trend graph and comparative data.

8. The system according to claim 7, characterized in that: The hierarchical alarm decision module includes a final anomaly score calculation unit, an alarm level determination unit, a notification object determination unit, and an alarm content generation unit; The final anomaly score calculation unit is used to calculate a context adjustment coefficient based on time factors, historical factors, and animal-specific factors, and calculate a final anomaly score based on the calculated context adjustment coefficient. The formula is: in, represents the final anomaly score, represents the context adjustment coefficient, represents the time-dependent adjustment factor, Historically relevant adjustment factors, Indicates the special state adjustment factor, Indicates the maximum absolute value of the Z score among all parameters, represents the time factor weight of the contextual factor, represents the historical factor weights of contextual factors, Special state weights representing contextual factors, represents the urgency factor, which is associated with the largest Z score; The alarm level determination unit is used to determine the alarm level using fuzzy logic rules, including when <3.0 or the deviation of a single parameter exceeds 10% of the basic threshold, it is determined to be a warning level alarm. <3.0 or the deviation of a single parameter exceeds 10% to 20% of the basic threshold, it is determined to be a concern level alarm. <9.0 or the deviation of a single parameter exceeds 20% to 30% of the basic threshold, it is determined to be a warning level alarm. ≥9.0 or the deviation of a single parameter exceeds 30% of the basic threshold, it is determined to be an emergency alarm; The notification object determination unit is used to determine the notification object according to the alarm level. When it is a prompt-level alarm, the record is displayed in the system; when it is a concern-level alarm, the record is displayed in the system and the breeder is notified; when it is a warning-level alarm, the record is displayed in the system and the breeder and the on-duty veterinarian are notified; when it is an emergency-level alarm, the record is displayed in the system and the breeder, on-duty veterinarian and management personnel are notified; The alarm content generation unit is used to generate a concise alarm title including the animal ID and the abnormality type, generate detailed alarm content including abnormal parameters, values ​​and duration, add historical comparison information, and display parameter change trends.

9. A method for monitoring and alarming the health of individual animals in a zoo, characterized by: The following steps are involved: Continuously collect animal vital sign data through perception layer sensors; The edge computing layer pre-processes the data and extracts parameters such as heart rate, respiratory rate, body temperature, and activity level; Perform preliminary anomaly detection based on preset thresholds and trigger local alarms; The processed data is transmitted to the cloud analysis layer and combined with machine learning models for in-depth analysis; Generate health reports based on anomaly detection results and trigger alarms based on classification rules.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to claim 9 are implemented.

Citation Information

Patent Citations

  • Pig behavior and health method based on deep learning and physiological parameter dynamic monitoring

    CN118986303A

  • Non-contact vital sign data monitoring method and system

    CN119073946A

  • Animal body state monitoring method and system based on implantable chip

    CN119097288A

  • Livestock disease prevention and control monitoring and early warning system based on Internet of Things

    CN119905255A

Cited By

  • Animal health prediction and intervention system and method based on digital twin

    CN121457326A