Pig individual identification and health monitoring system based on biometrics

Abstract

The present invention discloses a biometric-based pig individual identification and health monitoring system, which relates to the field of medical technology. The system includes a multimodal biometric acquisition module for acquiring a pig's 3D facial features, ear vascular distribution characteristics, gait dynamic characteristics, and physiological data; a biometric fusion recognition module that fuses multimodal features based on a Transformer architecture to achieve individual uniqueness identification and dynamic updating of feature templates; and a digital twin health monitoring module that constructs a digital twin model of the pig's entire life cycle and combines it with an LSTM neural network to achieve health status assessment and early disease warning. This biometric-based pig individual identification and health monitoring system uses multiple devices, including an RGB-D camera, a near-infrared camera, and a binocular vision sensor, to construct a 1024-dimensional standardized feature vector. Combined with a Transformer network to dynamically assign modal weights, it achieves high-precision recognition in complex scenarios.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a pig individual identification and health monitoring system based on biological characteristics. Background Art

[0002] Research on group-reared animal behavior has shown that healthy pigs have specific habits, such as excretion and feeding. Observing these behaviors can promptly identify suspected pig diseases and reduce the likelihood of swine epidemics. To accurately and promptly assess the health of pigs, breeders need to monitor their feeding, drinking, and excretion behaviors. Pigs' behavioral, social, and psychological activities are often intertwined and can affect their health. Therefore, breeders must promptly detect and diagnose abnormal behaviors during the pig breeding process. For large and medium-sized pig farms, relying on individual breeders to monitor behavior is inefficient. Furthermore, large-scale pig farms have too many individuals and breeders have limited resources to monitor every pig. This makes it difficult to monitor every pig, leading to missed screenings. If diseased pigs are not promptly identified and treated, the disease can spread, affecting other healthy pigs and causing immeasurable financial losses.

[0003] Patent number CN115777571A discloses a monitoring system and method for identifying pig behavior. The system includes a camera for capturing real-time visual information within a pigpen. The system also includes an audio acquisition module for acquiring the pig's voice information; an infrared temperature detector for tracking and acquiring the body temperature of a specific pig; and a data processing module for acquiring visual information, audio information, and the pig's body temperature. The method includes the following steps: S1: acquiring image information of individual pigs within the pigpen; S2: acquiring and storing the appearance, voice, and body temperature characteristics of each individual pig; and uniquely numbering each pig.

[0004] However, existing monitoring systems for identifying pig behavior rely on a single sensor and are unable to integrate multi-dimensional biometric features such as sound, body temperature, and gait. This makes individual identification susceptible to occlusion and lighting, and makes it difficult to distinguish individuals with similar appearances. In addition, traditional methods have not established a dynamic feature template library, and have poor adaptability to changes in the appearance of pigs during growth. Frequent manual recalibration is required, which is time-consuming and labor-intensive. Moreover, existing systems only assess health status through a few indicators such as body temperature and food intake, and lack real-time analysis of key features such as gait stability and surface temperature distribution. This leads to delayed disease warnings, the failure to build an individualized digital twin model, the inability to dynamically fit the growth curve, and the fixed warning threshold. It is not dynamically adjusted according to the age and breed of the pig, making it difficult to achieve precise health management. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a pig individual identification and health monitoring system based on biological characteristics, which solves the existing problems.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a pig individual identification and health monitoring system based on biometrics, comprising:

[0007] Multimodal biometric acquisition module, used to obtain pigs' 3D facial features, ear vascular distribution characteristics, gait dynamic characteristics and physiological data;

[0008] The biometric fusion recognition module integrates multimodal features based on the Transformer architecture to achieve individual unique identification and dynamic update of feature templates;

[0009] The digital twin health monitoring module builds a digital twin model of the pig's entire life cycle and combines it with the LSTM neural network to achieve health status assessment and early disease warning.

[0010] Preferably, the 3D facial features include facial contour point cloud and nose spot distribution, the ear blood vessel features include blood vessel curvature and branching angle, and the gait dynamic features include step length and forelimb swing angle.

[0011] Preferably, the loss function of the biometric fusion recognition module is:

[0012] ;

[0013] Among them, is the dynamic weight coefficient, 、 、 are the facial, ear and gait feature contrast loss functions respectively, and .

[0014] Preferably, the health status assessment formula of the digital twin health monitoring module is:

[0015] ;

[0016] in, Score the activity level. Score weight gain, Score abnormal skin temperature. 、 、 Dynamic adjustment based on pig age and breed.

[0017] Preferably, the multimodal biometric feature acquisition module includes:

[0018] Multi-angle RGB-D camera for collecting 3D facial point clouds and contour features;

[0019] Near-infrared camera (wavelength 850nm), used to collect images of ear blood vessels;

[0020] Binocular vision sensor (baseline distance 120mm) is used to capture gait sequence video (frame rate 30fps).

[0021] Preferably, the biometric fusion recognition module includes an online learning unit, which updates the individual feature template every 30 minutes based on the latest collected data to adapt to the changes in the characteristics of the pigs during their growth.

[0022] Preferably, the digital twin health monitoring module has a built-in disease prediction model, which analyzes growth curves, gait stability index (CSI) and body surface temperature distribution based on the LSTM neural network, and identifies limb diseases or growth abnormalities 72 hours in advance.

[0023] Preferably, the multimodal biometric feature acquisition module further includes:

[0024] Weight sensor, used to obtain pig weight data in real time;

[0025] Infrared thermal imager, used to collect body surface temperature distribution and identify fever areas.

[0026] Preferably, the digital twin health monitoring module supports a hierarchical early warning mechanism:

[0027] Yellow warning (health score H < 60): triggers an alarm for abnormal activity or weight gain;

[0028] Red alert (health score H < 40): triggers a suspected disease alert and automatically associates historical health data to generate diagnostic recommendations.

[0029] The present invention also discloses a health monitoring method of a pig individual identification and health monitoring system based on biological characteristics, comprising the following steps:

[0030] Step 1: Data Collection: Multi-angle RGB-D cameras, near-infrared cameras, and binocular vision sensors collect real-time 3D facial, ear vascular, and gait data of pigs. Weight sensors and infrared thermal imagers simultaneously collect weight and temperature data.

[0031] Step 2: Preprocessing: The edge computing node performs point cloud denoising and image enhancement on the raw data to generate a 1024-dimensional standardized feature vector;

[0032] Step 3: Individual identification: The biometric fusion recognition module fuses multimodal features through the Transformer architecture, outputs an individual number, and updates the individual feature template in the digital twin model;

[0033] Step 4: Health Assessment: The digital twin platform calculates the health score H based on the health status assessment formula and uses the LSTM neural network to analyze growth curves and gait data to identify abnormal conditions.

[0034] Step 5. Graded warning: When H < 60 or H < 40, a yellow or red warning is triggered respectively, which is synchronized to the breeding management system and generates treatment suggestions (such as adjusting feed formula, isolation and observation), and records the treatment effects to optimize the warning model.

[0035] Beneficial effects

[0036] The present invention provides a pig individual identification and health monitoring system based on biometrics. Compared with the existing technology, it has the following advantages:

[0037] 1. This biometric-based pig individual identification and health monitoring system uses the collaboration of multiple devices such as RGB-D cameras, near-infrared cameras, and binocular vision sensors to construct a 1024-dimensional standardized feature vector. Combined with the Transformer network, it dynamically allocates modal weights to achieve high-precision identification in complex scenarios and supports rapid cross-breed adaptation. The dynamic template update mechanism can adapt to the appearance changes of pigs during their growth, solving the occlusion and light dependence problems of traditional single-modality recognition, and providing a data foundation for precise individual management.

[0038] 2. This biometric-based pig individual identification and health monitoring system builds a digital twin health profile based on the Gompertz growth model and LSTM neural network, fits parameters such as daily weight gain and activity level in real time, and realizes multi-dimensional health assessment through gait stability index and body surface temperature abnormality score. The graded early warning mechanism significantly improves the speed of disease detection, early identification of fever symptoms, early warning of limb diseases, and reduced missed detection rate. Combined with dynamic weight adjustment and automated processing suggestions, it shortens the average time to market and reduces the use of veterinary drugs, thereby reducing breeding risks from the source and realizing precise and intelligent health management. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the system architecture of the present invention;

[0040] Figure 2 This is a schematic diagram of the biometric fusion algorithm flow of the present invention;

[0041] Figure 3 This is a schematic diagram of the health monitoring and early warning logic of the present invention. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0043] See Figure 1-3 , the present invention provides a variety of technical solutions:

[0044] The first implementation method: a pig individual identification and health monitoring system based on biometrics, including:

[0045] A multimodal biometric acquisition module is used to acquire the pig's 3D facial features, ear vascular distribution characteristics, gait dynamic characteristics, and physiological data. The multimodal biometric acquisition module includes:

[0046] Multi-angle RGB-D camera for collecting 3D facial point clouds and contour features;

[0047] Near-infrared camera (wavelength 850nm), used to collect images of ear blood vessels;

[0048] Binocular vision sensor (baseline distance 120mm) is used to capture gait sequence video (frame rate 30fps).

[0049] The 3D facial features include facial contour point cloud and nose spot distribution, the ear blood vessel features include blood vessel curvature and branching angle, and the gait dynamic features include step length and forelimb swing angle.

[0050] The multimodal biometric feature acquisition module further includes:

[0051] Weight sensor, used to obtain pig weight data in real time;

[0052] Infrared thermal imager, used to collect body surface temperature distribution and identify fever areas.

[0053] The biometric fusion recognition module fuses multimodal features based on the Transformer architecture to achieve individual unique identification and dynamic update of feature templates. The loss function of the biometric fusion recognition module is:

[0054] ;

[0055] Among them, is the dynamic weight coefficient, 、 、 are facial, ear, and gait feature contrast loss functions, respectively, and .

[0056] The biometric fusion recognition module includes an online learning unit that updates the individual feature template every 30 minutes based on the latest collected data to adapt to feature changes during the growth process of pigs.

[0057] The digital twin health monitoring module builds a digital twin model of the pig's entire life cycle and combines it with the LSTM neural network to achieve health status assessment and early disease warning. The health status assessment formula of the digital twin health monitoring module is:

[0058] ;

[0059] in, Score the activity level. Score weight gain, Score abnormal skin temperature. 、 、 Dynamic adjustment based on pig age and breed.

[0060] The digital twin health monitoring module has a built-in disease prediction model that analyzes growth curves, gait stability index (CSI) and body surface temperature distribution based on an LSTM neural network, identifying limb diseases or growth abnormalities 72 hours in advance.

[0061] The digital twin health monitoring module supports a hierarchical early warning mechanism:

[0062] Yellow warning (health score H < 60): triggers an alarm for abnormal activity or weight gain;

[0063] Red alert (health score H < 40): triggers a suspected disease alert and automatically associates historical health data to generate diagnostic recommendations.

[0064] The second embodiment: a health monitoring method of a pig individual identification and health monitoring system based on biometrics, comprising the following steps:

[0065] Step 1: Data Collection: Multi-angle RGB-D cameras, near-infrared cameras, and binocular vision sensors collect real-time 3D facial, ear vascular, and gait data of pigs. Weight sensors and infrared thermal imagers simultaneously collect weight and temperature data.

[0066] Step 2: Preprocessing: The edge computing node performs point cloud denoising and image enhancement on the raw data to generate a 1024-dimensional standardized feature vector;

[0067] Step 3: Individual identification: The biometric fusion recognition module fuses multimodal features through the Transformer architecture, outputs an individual number, and updates the individual feature template in the digital twin model;

[0068] Step 4: Health Assessment: The digital twin platform calculates the health score H based on the health status assessment formula and uses the LSTM neural network to analyze growth curves and gait data to identify abnormal conditions.

[0069] Step 5. Graded warning: When H < 60 or H < 40, a yellow or red warning is triggered respectively, which is synchronized to the breeding management system and generates treatment suggestions (such as adjusting feed formula, isolation and observation), and records the treatment effects to optimize the warning model.

[0070] The third implementation method: system deployment

[0071] 1. System Hardware and Data Acquisition

[0072] Multimodal biometric acquisition module

[0073] 3D facial and ear feature collection

[0074] RGB-D camera: Deploy 2-4 depth cameras (such as Intel RealSense D435i) on the ceilings of the feeding and resting areas of the pig house. The installation height is 2.5-3m, and the coverage area is 5m×5m. The front and side facial point cloud data of the pigs are collected every 5 minutes, and RGB images are obtained simultaneously for nose spot feature extraction.

[0075] Near-infrared camera: Install a near-infrared imaging device (wavelength 850nm, such as FLIRA35) 30 cm above the feeding trough in the feeding area, focus on the pig's ear area, and collect blood vessel distribution images. The collection frequency should be ≥ 20 times per day, covering different lighting conditions (automatically switching between day and night modes).

[0076] Gait and physiological data collection

[0077] Binocular vision sensor: A binocular camera (baseline distance 120mm, such as ZED2i) is installed 1.8m above the pig house passage to capture pig walking videos and simultaneously record dynamic characteristics such as stride length and forelimb swing angle. ≥10 gait cycles of a single pig walking in a single time are collected.

[0078] Weight and temperature sensors: Four sets of pressure sensors (such as Honeywell weighing modules) are embedded in the ground of the feeding area to automatically obtain weight data when the pigs are feeding; an infrared thermal imager (such as NECR300) is installed on the top of the pig house, scanning the entire area every 10 minutes to identify abnormal temperatures in areas such as the ears and back (such as local temperature rise ≥1°C).

[0079] (2) Edge computing node preprocessing process

[0080] Point cloud data processing

[0081] Voxel filtering (voxel size 5 mm) was used to remove noise from the 3D facial point cloud. The pig's face was separated from the background using the RANAC planar segmentation algorithm, and valid point cloud data (≥5000 points) were retained.

[0082] Extract facial contour key points (15 feature points such as the nose tip and ear root) and calculate 50-dimensional geometric features such as contour curvature and symmetry.

[0083] Image and video processing

[0084] Adaptive histogram equalization was performed on near-infrared ear images to enhance vascular contrast. The U-Net network was used to segment the vascular region and extract 20-dimensional features such as vascular branching angle and curvature radius.

[0085] The gait video uses the optical flow method to track the motion trajectory of the joints, combines the Kalman filter to smooth the trajectory, and calculates 12-dimensional dynamic features such as step length, step frequency, and center of gravity offset.

[0086] Data fusion

[0087] 3D facial features (4096 dimensions), ear vascular features (512 dimensions), gait features (128 dimensions), weight (1 dimension), and body temperature (1 dimension) are spliced into a 1024-dimensional standardized feature vector and uploaded to the central server via the MQTT protocol.

[0088] 2. Biometric Fusion Recognition Module

[0089] Transformer Network Architecture

[0090] Input layer: Position encoding of different modal features (facial feature encoding length 50, gait sequence encoding length 10) is performed to generate feature vectors with position information.

[0091] Multi-head self-attention layer: 8 attention heads learn the interaction between different modalities, such as the association weights between facial features and ear features (formula: ), dynamically assign weights (e.g., the weight of ear features in night scenes is increased by 30%).

[0092] Fusion layer: The attention output is mapped to the individual recognition space through the fully connected layer, and a 128-dimensional feature vector is output for individual matching.

[0093] Dynamic template update mechanism

[0094] An individual feature template library was established, and the initial template was generated by averaging the data collected three times at birth.

[0095] Online learning algorithm: When the matching degree between the newly acquired features and the template is less than 95%, the template is updated and the exponential moving average (EMA) is used to fuse the new features (update formula: ), ensure that the template update error is ≤2%.

[0096] (2) Individual Identification Process

[0097] Real-time matching: When a pig enters the monitoring area, the edge computing node generates a feature vector and uploads it. The central server matches the individual number using the nearest neighbor algorithm (cosine distance threshold 0.8), and the matching time is ≤ 200ms.

[0098] Cross-breed adaptability: For different breeds such as Landrace and Duroc, model parameters are pre-trained using a meta-learning algorithm (MAML), allowing adaptation to new pig herds with just 50 samples.

[0099] 3. Digital Twin Health Monitoring Module

[0100] 1. Construction of individualized health records

[0101] Digital twin model parameters

[0102] Growth curve: Fitting daily weight gain data based on the Gompertz growth model (Formula: ,in For mature weight, 、 is the variety parameter), the prediction error is ≤3%.

[0103] Activity patterns: Calculate daily activity duration through gait data (threshold: ≥8 hours for piglets, ≥5 hours for fattening pigs), and combine it with feeding frequency (≥3 times / day) to construct an activity score Sa (0-100 points, with a score below 60 triggering an alert).

[0104] Health status assessment model

[0105] Dynamic weight adjustment: Automatically adjust weight according to pig age (e.g. 1-20 days old: ; 60 days and older: ), ensuring that the assessment focus is different at different growth stages.

[0106] Abnormal body temperature score: The temperature of 5 key areas such as the ear and abdomen is identified through thermal imaging data. When the temperature of ≥3 areas exceeds the mean +1°C, point.

[0107] 2. Early warning mechanism for diseases

[0108] LSTM prediction model

[0109] Input features: daily weight gain, activity duration, and body temperature data from the past 7 days (21 dimensions in total), and output growth trend prediction for the next 3 days.

[0110] Abnormal judgment: When the daily weight gain decreases by ≥10% for three consecutive days and the predicted trend continues to deteriorate, a growth abnormality warning is triggered.

[0111] Gait stability index (CSI)

[0112] Calculation method: ,when If the weight-bearing time of one limb is greater than 15 minutes, it is judged as limb abnormality, and the disease probability is output in combination with historical health data (such as arthritis probability ≥ 70%).

[0113] Graded warning response

[0114] Yellow warning (H<60): Push abnormal information (such as "Pig No. 0012 is not active enough") through the breeding management system, and automatically adjust the feed formula (increase the energy feed by 10%).

[0115] Red alert (H < 40): Triggers an audible and visual alarm and generates an isolation recommendation. Simultaneously retrieves historical data to generate a diagnostic report (such as "elevated body temperature + abnormal gait in the past three days, suspected of Streptococcus suis infection"). The response time for handling the recommendation is ≤ 5 minutes.

[0116] Fourth implementation method: Example, large-scale pig farm application verification

[0117] (1) Scenario Configuration

[0118] Pig farm scale: 1,000 fattening pigs, divided into 10 pig houses, 100 pigs in each house, and a density of 1.2 m2 / pig.

[0119] Hardware deployment: Each dormitory is equipped with 4 RGB-D cameras, 2 near-infrared cameras, 1 set of binocular vision sensors, and 8 sets of weight sensors.

[0120] Algorithm parameters: Transformer model training data: 5,000 pigs (covering 3 breeds), LSTM prediction window: 7 days, online learning update interval: 30 minutes.

[0121] (2) Operational results

[0122] Individual recognition performance

[0123] The recognition accuracy rate is 99.2% under normal lighting, 98.5% at night (infrared fill light), and 97.3% in scenarios where the ear tag is blocked.

[0124] The deployment time for a new herd (Duroc pigs) is 24 hours, while traditional methods take 2 weeks, and the cross-breed identification accuracy is 95%.

[0125] Health monitoring effect

[0126] Early disease warning time: fever symptoms can be identified 48 hours in advance, limb diseases can be warned 72 hours in advance, and the missed detection rate has been reduced from 15% to 3%.

[0127] Improved breeding efficiency: average time to market shortened by 7 days, veterinary drug usage reduced by 30%, and manual inspection intensity reduced by 50%.

[0128] (3) Exception handling example

[0129] Case 1: Early warning of piglet diarrhea

[0130] Piglet No. 0035 experienced a 15% decrease in daily weight gain for two consecutive days, with a CSI of 0.78. Its surface temperature remained normal, but its activity level decreased. The system triggered a yellow alert, stating "Abnormal Growth, Fecal Testing Recommended." Veterinary testing confirmed mild diarrhea, and the patient recovered within 48 hours after prompt medication, preventing widespread infection.

[0131] Case 2: Early warning of arthritis in adult pigs

[0132] Adult pig No. 1021 had a gait CSI of 0.65, an ear temperature elevated by 1.2°C, and a health score of 35. The system triggered a red alert and recommended isolation. An X-ray confirmed knee joint inflammation, shortening treatment by five days to prevent further damage.

[0133] Through the above-mentioned specific implementation methods, the present invention realizes the full process from hardware deployment, data processing to intelligent decision-making, verifies the significant advantages of multimodal biometric fusion and digital twin technology in individual pig identification and health monitoring, meets the needs of large-scale pig farms for precise breeding and intelligent management, and has significant practical value.

[0134] At the same time, the contents not described in detail in this specification belong to the existing technology known to those skilled in the art, and the model parameters of each electrical appliance are not specifically limited, and conventional equipment can be used.

[0135] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0136] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A pig individual identification and health monitoring system based on biometrics, characterized by: include: Multimodal biometric acquisition module, used to obtain pigs' 3D facial features, ear vascular distribution characteristics, gait dynamic characteristics and physiological data; The biometric fusion recognition module integrates multimodal features based on the Transformer architecture to achieve individual unique identification and dynamic update of feature templates; The digital twin health monitoring module builds a digital twin model of the pig's entire life cycle and combines it with the LSTM neural network to achieve health status assessment and early disease warning. The biometric fusion recognition module includes an online learning unit that updates the individual feature template every 30 minutes based on the latest collected data to adapt to the changes in the characteristics of the pig during its growth process; The digital twin health monitoring module has a built-in disease prediction model that uses an LSTM neural network to analyze growth curves, gait stability index, and body surface temperature distribution to identify limb diseases or growth abnormalities in advance. The 3D facial features include facial contour point cloud and nose spot distribution, the ear blood vessel distribution features include blood vessel curvature and branching angle, and the gait dynamic features include step length and forelimb swing angle; The loss function of the biometric fusion recognition module is: ； Among them, is the dynamic weight coefficient, 、 、 are the facial, ear and gait feature contrast loss functions respectively, and .

2. The pig individual identification and health monitoring system based on biometrics according to claim 1, characterized in that: The health status evaluation formula of the digital twin health monitoring module is: ； in, Score the activity level. Score weight gain, Score abnormal skin temperature. 、 、 Dynamic adjustment based on pig age and breed.

3. The pig individual identification and health monitoring system based on biometrics according to claim 1, characterized in that: The multimodal biometric feature acquisition module includes: Multi-angle RGB-D camera for collecting 3D facial point clouds and contour features; A near-infrared camera is used to collect images of the blood vessels in the ear; Binocular vision sensor, used to collect gait sequence videos.

4. The pig individual identification and health monitoring system based on biometrics according to claim 1, characterized in that: The multimodal biometric feature acquisition module further includes: Weight sensor, used to obtain pig weight data in real time; Infrared thermal imager, used to collect body surface temperature distribution and identify fever areas.

5. The pig individual identification and health monitoring system based on biometrics according to claim 1, characterized in that: The digital twin health monitoring module supports a hierarchical early warning mechanism: Yellow warning: triggers an alarm for abnormal activity or weight gain; Red alert: Triggers a suspected disease alert and automatically associates historical health data to generate diagnostic recommendations.

6. A health monitoring method based on the biometric-based pig individual identification and health monitoring system according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Data Collection: Multi-angle RGB-D cameras, near-infrared cameras, and binocular vision sensors collect real-time 3D facial, ear vascular, and gait data of pigs. Weight sensors and infrared thermal imagers simultaneously collect weight and temperature data. Step 2: Preprocessing: The edge computing node performs point cloud denoising and image enhancement on the raw data to generate a 1024-dimensional standardized feature vector; Step 3: Individual identification: The biometric fusion recognition module fuses multimodal features through the Transformer architecture, outputs an individual number, and updates the individual feature template in the digital twin model; Step 4: Health Assessment: The digital twin platform calculates the health score H based on the health status assessment formula and uses the LSTM neural network to analyze growth curves and gait data to identify abnormal conditions. Step 5. Gradual warning: When H < 40, a red warning is triggered; when 40 ≤ H < 60, a yellow warning is triggered. The warning is synchronized to the breeding management system and treatment suggestions are generated. The treatment effects are recorded to optimize the warning model.

Images