Animal individual digital twinning and self-learning label generation system and method

Abstract

The invention discloses an individual animal digital twinning and self-learning label generation system and method, and the system comprises a perception collection layer which is used for capturing the visual information, sound, motion track, physiological signal and surrounding environment data of an animal in a real-time and multi-modal manner; the AI cognitive layer is used for carrying out preprocessing and feature extraction processing on data acquired by the perception acquisition layer by utilizing a deep learning model, and identifying animal posture, behavior, sound mode and physiological index abnormity; the digital twinning and knowledge reasoning layer is used for generating a globally unique identifier for each animal individual and carrying out data integration to generate a knowledge graph; the federation collaboration layer is used for realizing multi-park data sharing and model training; and the application service layer is used for providing an application interface and realizing man-machine interaction. According to the method, global unique identification of animal individual identities can be realized, the ID conflict problem is thoroughly solved, and a solid foundation is laid for animal information management, tracking and cross-mechanism cooperation.

Description

Technical Field

[0001] This invention relates to the field of animal information management technology, and in particular to a system and method for generating digital twins and self-learning tags for individual animals. Background Technology

[0002] Currently, zoos, research institutions, and nature reserves generally face challenges in managing individual animal information. Existing management systems are mostly based on traditional databases or simple spreadsheets, and have the following significant shortcomings: 1. Inconsistent and conflict-prone ID systems: Animal ID generation rules vary across different organizations, and even between different projects within the same organization, lacking global uniqueness. This makes data integration, sharing, and cross-organizational collaboration extremely difficult. For example, an animal may have the same ID in different organizations, or multiple IDs may be assigned to the same animal within a single organization.

[0003] 2. Fragmented and Delayed Information Management: Animal physiological, behavioral, and health information is scattered across different recording systems or paper documents, making it difficult to create complete individual profiles. Information updates often rely on manual entry, resulting in poor real-time performance and a lack of in-depth correlation analysis.

[0004] 3. Label generation relies on human experience or a single model: Existing labels (such as behavioral classification and health indicators) mostly rely on expert experience or are based on pre-trained models of a single modality (such as vision), which makes it difficult to capture the complex, multimodal behavioral patterns and subtle health changes of animals, and the dynamic updating and adaptive capabilities of the labels are insufficient.

[0005] 4. Lack of digital twin concept: There is a lack of dynamic, real-time virtual mapping (digital twin) of individual animals, making it impossible to conduct effective behavior simulation, health status prediction, risk warning and causal analysis.

[0006] 5. Weak cross-institutional collaboration: Due to inconsistent IDs, incompatible data formats, and a lack of standardized information sharing mechanisms, collaboration between different institutions in areas such as scientific research, disease control, and species conservation is inefficient, making it difficult to maximize the value of data.

[0007] Therefore, there is an urgent need for an innovative system and method that can break down information silos, achieve globally unique identification of individual animals, integrate multimodal data for intelligent analysis and tag generation, construct dynamic digital twins, and support secure cross-institutional collaboration. Summary of the Invention

[0008] The purpose of this invention is to provide a digital twin and self-learning tag generation system and method for individual animals, in order to solve the technical problems in the prior art such as easy conflict of individual animal IDs, fragmented information management, low tag generation efficiency and lack of adaptability, lack of digital twin capabilities, and difficulty in cross-institutional collaboration.

[0009] This invention is achieved using the following technical solution: a digital twin and self-learning tag generation system for individual animals, comprising: The perception and acquisition layer is used to capture the animal's visual information, sound, movement trajectory, physiological signals, and surrounding environmental data in real time and in multiple modalities. The AI ​​cognitive layer uses deep learning models to preprocess and extract features from the data collected by the perception and acquisition layer, and to identify abnormalities in animal posture, behavior, sound patterns and physiological indicators. The digital twin and knowledge reasoning layer is used to generate a globally unique identifier for each individual animal and to integrate data to generate a knowledge graph. The federal collaboration layer is used to enable data sharing and model training across multiple campuses; The application service layer provides the application interface and enables human-computer interaction.

[0010] Furthermore, the sensing and acquisition layer includes a variety of sensors deployed in animal habitats or research areas, including one or more of high-definition cameras, infrared thermal imagers, microphone arrays, wearable biosensors, and environmental sensors.

[0011] Furthermore, the deep learning model includes one or more of CNN, RNN, Transformer, and GNN.

[0012] Furthermore, the digital twin and knowledge reasoning layer includes a globally unique ID generation module and a multimodal data fusion module. The globally unique ID generation module adopts a distributed, AI-driven conflict prediction and detection mechanism to generate a globally unique identifier for each individual animal. The multimodal data fusion module is used to combine the features output by the AI ​​cognitive layer with the original perceptual data to form a more comprehensive representation of the animal's state.

[0013] Furthermore, the digital twin and knowledge reasoning layer also includes a self-learning tag generation module. This module is based on an AI model and automatically discovers new behavioral patterns and health status characteristics to generate dynamically updated animal individual tags containing predictive information.

[0014] Furthermore, the digital twin and knowledge reasoning layer also includes a digital twin construction module, which integrates the generated unique ID, collected multimodal data, AI cognitive results and self-learning tags to construct a dynamic digital twin of an individual animal.

[0015] Furthermore, the digital twin and knowledge reasoning layer also includes a knowledge graph and a reasoning engine. The knowledge graph and reasoning engine are used to construct a knowledge graph containing information related to individual animals, species, behaviors, health, and environment, and integrate the reasoning engine to support multi-dimensional queries, causal analysis, correlation mining, and scientific research-assisted decision-making.

[0016] Furthermore, the federated collaboration layer employs federated learning technology to achieve model parameter sharing and collaborative updates while protecting the privacy of the original data of each institution. It also achieves model integration and global knowledge enhancement across institutions and heterogeneous systems through knowledge distillation and heterogeneous label mapping technologies.

[0017] Furthermore, the application service layer provides a user application interface, supports individual information retrieval, behavior analysis, health warning, scientific research report generation and permission management functions, and integrates with external systems through standard API interfaces.

[0018] A method for generating digital twins and self-learning tags for individual animals, based on the aforementioned system for generating digital twins and self-learning tags for individual animals, includes the following steps: S1: Generate a globally unique ID; S2: Multimodal data acquisition and preprocessing; S3: AI Cognition and Feature Extraction; S4: Self-learning tag generation; S5: Digital Twin Construction; S6: Knowledge Graphs and Reasoning; S7: Federal Collaboration and Privacy Protection; S8: Application services.

[0019] The beneficial effects of this invention are as follows: This invention can achieve globally unique identification of individual animals. It completely solves the ID conflict problem, laying a solid foundation for animal information management, tracking, and cross-agency collaboration; it enhances the intelligence and real-time performance of information management, and through multimodal data fusion and AI analysis, it achieves comprehensive, real-time perception and deep understanding of animal status.

[0020] This invention generates dynamic, adaptive intelligent tags. The AI-driven self-learning tag generation can automatically discover new features, capture complex behaviors, and provide predictive information, greatly improving the accuracy and application value of tags.

[0021] This invention constructs comprehensive digital twins of individual animals, enabling the simulation of animal behavior and prediction of health risks, supporting refined management, early intervention, and scientific research. It promotes efficient cross-institutional collaboration and knowledge sharing; the federated learning mechanism, while ensuring data privacy, achieves federated growth of models and knowledge, improving overall research and management levels. Furthermore, the invention's knowledge graph and inference engine provide powerful analytical tools, supporting researchers and managers in making more informed decisions; and it enhances data security and traceability, with strict access control and operational auditing mechanisms ensuring data security and compliance. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0023] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0025] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0026] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0027] See Figure 1 A digital twin and self-learning tag generation system for individual animals, comprising: The perception and acquisition layer is used to capture the animal's visual information, sound, movement trajectory, physiological signals, and surrounding environmental data in real time and in multiple modalities. The AI ​​cognitive layer uses deep learning models to preprocess and extract features from the data collected by the perception and acquisition layer, and to identify abnormalities in animal posture, behavior, sound patterns and physiological indicators. The digital twin and knowledge reasoning layer is used to generate a globally unique identifier for each individual animal and to integrate data to generate a knowledge graph. The federal collaboration layer is used to enable data sharing and model training across multiple campuses; The application service layer provides the application interface and enables human-computer interaction.

[0028] In this embodiment, the perception and acquisition layer integrates multiple sensors (such as high-definition cameras, infrared cameras, microphone arrays, biosensors, and environmental sensors) to collect real-time visual, auditory, movement, physiological parameters, and environmental information of the animal. Specifically, sensors can be deployed in animal habitats or research areas, including but not limited to high-definition cameras, infrared thermal imagers, microphone arrays, wearable biosensors (such as heart rate and body temperature monitoring), and environmental sensors (temperature, humidity, light intensity, ammonia concentration, etc.). These devices are responsible for capturing the animal's visual information, sound, movement trajectory, physiological signals, and surrounding environmental data in real-time and multimodally. Furthermore, some AI inference capabilities (such as behavior recognition and simple anomaly detection) can be deployed in the perception and acquisition layer to reduce data transmission volume and improve real-time response speed.

[0029] In this embodiment, the AI ​​cognitive layer primarily utilizes deep learning models (such as CNN, RNN, Transformer, GNN, etc.) to preprocess and extract features from multimodal acquired data, identifying animal posture, behavior, vocal patterns, and abnormal physiological indicators. Further, this AI cognitive layer receives data from the perception acquisition layer. The data preprocessing unit performs denoising, calibration, and format unification on the raw data. The feature extraction unit uses deep learning models (such as CNN for images, MFCC for sound, and LSTM / Transformer for time-series data) to extract key features. Behavior recognition models (such as posture estimation-based models and action recognition models) identify the animal's activity types (such as eating, resting, socializing, and abnormal movements). Health status assessment models (such as those based on body temperature, heart rate, respiratory rate, and activity level) assess the animal's physiological health indicators. An anomaly detection model focuses on identifying any signals deviating from normal patterns.

[0030] In this embodiment, the digital twin and knowledge reasoning layer mainly includes the following aspects: 1. A distributed, AI-driven conflict prediction and detection mechanism is adopted to generate a globally unique identifier for each animal individual. This ID contains structured information and can be associated with metadata such as species, individual characteristics, occurrence time, and affiliated institution. 2. Based on AI models (such as contrastive learning, self-supervised learning, clustering algorithms, and weakly supervised learning), new behavioral patterns and health status characteristics are automatically discovered, generating dynamically updated animal individual tags containing predictive information, supporting iterative optimization of tags with minimal human confirmation. 3. The generated unique IDs, collected multimodal data, AI cognitive results, and self-learning tags are integrated to construct a dynamic digital twin of the animal individual. The digital twin not only contains historical and real-time data but can also predict future behavioral patterns, health trends, and potential risks through AI models (such as time-series prediction and graph neural networks). 4. A knowledge graph containing related information such as animal individuals, species, behavior, health, and environment is constructed, and an inference engine is integrated to support multi-dimensional queries, causal analysis, correlation mining, and scientific research-assisted decision-making.

[0031] The digital twin and knowledge reasoning layer includes a globally unique ID generation module, a multimodal data fusion module, a self-learning tag generation module, a digital twin construction module, and a knowledge graph and reasoning engine.

[0032] Globally Unique ID Generation Module: Activated during system initialization or when a new animal enters the park. It employs a distributed algorithm, combined with AI prediction technology (such as predicting potential conflicting ID segments based on the distribution patterns of existing IDs), and works in conjunction with a cross-park synchronization mechanism to ensure that the generated ID (e.g., format: `GSSPP-YYYY-NNNN-MID-SEQ-C`, where GSSPP is the species / subspecies code, YYYY is the year of entry / birth, NNNN is the park / equipment identifier, MID is the individual identifier, SEQ is the serial number, and C is the check digit) is globally unique and cannot be reused. This ID is bound to the animal entity, collected multimodal data, AI cognitive results, and subsequently generated digital twins.

[0033] Multimodal data fusion module: Combines the features output by the AI ​​cognitive layer with the raw perception data to form a more comprehensive animal state representation, providing a comprehensive and detailed perception of the animal state.

[0034] Self-learning label generation module: This is the key innovation of this invention. It does not rely on a pre-set fixed label set, but instead utilizes contrastive learning (learning the similarity and differences between samples), cluster analysis (grouping similar behaviors or states), self-supervised learning (learning representations by predicting a portion of the data itself), and minimal human verification (after initial classification by the AI ​​model, only a small amount of expert verification is needed for the model to iteratively optimize and discover new, unknown behavioral patterns or health indicators). For example, the model may automatically discover that a specific combination of movement patterns and sound signals indicates early signs of a disease and generate new predictive labels. The generated labels are dynamically updated and bound to a digital twin and a global ID. This module achieves dynamic label updates and automated discovery of new features, eliminating dependence on a fixed label set and greatly enhancing the generalization ability and application value of AI.

[0035] Digital Twin Building Module: Based on the animal's globally unique ID, this module aggregates all relevant historical multimodal data, AI cognitive results, and self-learning tags to construct a dynamic digital twin of the individual animal. It utilizes Graph Neural Networks (GNNs) to model social relationships between individuals or connections within an individual's physiological systems, and combines time-series prediction models (such as LSTM, Transformer, and Prophet) to predict future behavioral trends, changes in health status, and potential risks (such as disease incidence probability and decreased activity levels). This module can simulate and predict animal behavior and health, enabling proactive management. Furthermore, it allows users to explore the digital twin's responses through virtual interactions (such as the impact of simulated feeding on behavior); and overlays digital twin information onto the AR / VR view of the real animal, providing a more intuitive management and display experience.

[0036] Knowledge Graph and Inference Engine: All structured data (IDs, tags, prediction results) and unstructured data (meta-information from images and sounds) are combined to construct a knowledge graph encompassing individual animals, species, behaviors, health, and environmental factors. The inference engine (e.g., rule-based reasoning, graph embedding-based reasoning) enables complex queries (e.g., "find all animals with similar behaviors to the target individual"), causal analysis (e.g., "analyze the impact of environmental changes on animal activity levels"), and correlation mining (e.g., "discover the association between specific behavioral patterns and a decline in a certain health indicator"), providing deep insights for research and management. Furthermore, the knowledge inference layer can be further strengthened by introducing causal inference models to more accurately reveal "what caused what."

[0037] For ID generation strategies, blockchain technology can be used to achieve more decentralized ID management, depending on the specific application scenario. For AI cognitive models, different model architectures can be selected based on animal species and data characteristics; for example, lighter models can be used for edge devices, while more complex models can be used for centralized servers. For self-learning tag generation algorithms, more advanced AI technologies such as reinforcement learning and meta-learning can be explored to assist in tag discovery and optimization. For digital twin modeling, for specific animals or scenarios, a combination of physical models (such as dynamic models) and AI models can be used to construct digital twins. For knowledge graph construction, different graph databases (such as Neo4j and TigerGraph) or different knowledge representation methods can be used.

[0038] In this embodiment, the federated collaboration layer aims to address privacy and compatibility issues in data sharing and model training across multiple campuses. This layer employs federated learning technology to achieve model parameter sharing and collaborative updates while protecting the privacy of the original data from each institution. Through techniques such as knowledge distillation and heterogeneous label mapping, it enables cross-institutional and cross-heterogeneous system model integration and global knowledge enhancement. Achieving cross-institutional collaboration and knowledge sharing while protecting privacy is key to solving multi-campus collaboration problems. Furthermore, for knowledge graph construction, different graph databases (such as Neo4j and TigerGraph) or different knowledge representation methods can be used.

[0039] Specifically, the federated collaboration layer includes: Federated Learning Training Nodes: Each participating institution deploys this node locally, using local animal data to train AI models (such as behavior recognition and health assessment models) without uploading the original data. Parameter Server / Aggregator: Receives (encrypted) model parameters or gradients uploaded by each node, aggregates them, and generates a globally optimal model. Knowledge Distillation Module: Distills the knowledge from the globally aggregated "teacher model" into each institution's local "student model," enabling it to achieve performance improvements from the global model without directly accessing the global data. Heterogeneous Label Mapping Module: Considering that different institutions may use slightly different labeling systems, this module maps the labels generated by the global model to each institution's local labeling system, ensuring the compatibility of model output. Privacy Protection Module: Employs differential privacy, homomorphic encryption, and secure multi-party computation technologies to ensure that the privacy of the original data is not leaked during model parameter exchange.

[0040] In this embodiment, the application service layer is directed towards the end user, providing a user-friendly application interface that supports functions such as individual information retrieval, behavior analysis, health alerts, scientific research report generation, and access control, and integrates with external systems through standard API interfaces.

[0041] Specifically, the application service layer includes: a user interface module: providing an intuitive web or app interface to display animal information, digital twin status, real-time alarms, predicted trends, and analysis reports; a retrieval and analysis module: supporting multi-dimensional retrieval and analysis of animal information based on globally unique IDs, species, behavioral patterns, health indicators, time periods, etc.; an alarm and early warning module: issuing timely alarms for health risks, behavioral abnormalities, and environmental risks based on the prediction results of the digital twin and anomaly detection models; a report generation module: automatically generating individual behavior analysis reports, health assessment reports, and scientific research data summary reports; a permission management module: implementing role-based access control, setting different data access and operation permissions for different users (administrators, researchers, keepers), and recording detailed operation audit logs to ensure data security and compliance; and an API interface module: providing a standardized RESTful API, allowing external systems (such as other animal management software and scientific research platforms) to access this system and achieve data interconnection. Furthermore, it should be noted that the use of standardized data interfaces should follow internationally accepted animal information data exchange standards (such as ZooStandard and Species360 standards) to further improve interoperability.

[0042] See Figure 2 A method for generating digital twins and self-learning labels for individual animals, based on the aforementioned system for generating digital twins and self-learning labels for individual animals, includes the following steps: S1: Generate a globally unique ID; S2: Multimodal data acquisition and preprocessing; S3: AI Cognition and Feature Extraction; S4: Self-learning tag generation; S5: Digital Twin Construction; S6: Knowledge Graphs and Reasoning; S7: Federal Collaboration and Privacy Protection; S8: Application services.

[0043] Specifically, step S1 includes: During system initialization, S101 sets the ID generation strategy. S102 initiates a real-time prediction and detection mechanism for cross-organization / equipment ID conflicts. If a potential conflict is detected, S103 uses an AI model to predict the conflict and automatically generates a new, conflict-free ID for repair; if there is no conflict, S104 generates a globally unique ID for newly admitted or newly discovered animal individuals, which will persist throughout their entire lifespan.

[0044] Specifically, step S2 includes: S201 Deploying and activating the sensing and acquisition layer device. S202 Starting to collect various types of animal data in real time. S203 Transmitting the collected raw data to the AI ​​cognitive layer for preprocessing, including noise reduction, calibration, and format unification.

[0045] Specifically, step S3 includes: S301 Extracting key features from multimodal data using various deep learning models. S302 Fusing the extracted features. S303 Performing cognitive analysis of the animal's activities and physiological condition using AI models such as behavior recognition and health status assessment. S304 Running an anomaly detection model to identify any signals deviating from the normal pattern.

[0046] Specifically, step S4 includes: S401. Based on the fused features, self-learning algorithms such as contrastive learning and clustering are applied to explore potential patterns in the data. S402. The system prompts the user to perform minimal manual confirmation on potential new features or patterns discovered by the AI ​​(for example, the AI ​​identifies a certain type of behavior and provides a preliminary classification; the expert only needs to confirm whether it is a new behavior or belongs to a known category). S403. Based on the confirmation results, the AI ​​model automatically generates or updates intelligent tags for individual animals. These tags may contain predictive information, and the tags are bound to the animal's globally unique ID.

[0047] Specifically, step S5 includes: S501 binding the collected multimodal data with the animal's globally unique ID. S502 binding the self-learned intelligent tags with the ID. S503 binding the recognition results (such as behavior and health status) output by the AI ​​cognitive layer with the ID. S504 aggregating all the bound information and using time-series prediction and GNN models to construct a dynamic digital twin of the individual animal, which includes historical data, real-time status, and future behavior and health predictions.

[0048] Specifically, step S6 includes: S601 persistently storing the data of the digital twin. S602 constructing a comprehensive knowledge graph based on the stored data, connecting entities such as individuals, species, behaviors, health, and the environment. S603 launching the knowledge reasoning engine to query, analyze, and predict the knowledge graph, providing support for scientific research and decision-making.

[0049] Specifically, step S7 includes: S701 initiating the federated learning process. Each participating institution's local node S702 trains the model using local data. S703 uploads the trained model parameters (encrypted or differentially privacy-preserving) to the aggregator in S704. The aggregator S704 aggregates the parameters to generate a global model. S705 transfers the global model knowledge to the local models through knowledge distillation and uses a heterogeneous label mapping module to ensure the compatibility of the model output, completing a collaborative update. This process is performed under the privacy protection mechanism of S705 and repeats in S701.

[0050] Specifically, step S8 includes: S801 activating the application service layer; S802 providing a user-friendly interface, allowing users to perform operations such as information retrieval, behavior analysis, health alerts, and report generation through the analysis results of digital twins and knowledge graphs, self-learning tags, and AI prediction information; S803 integrating with other systems through API interfaces; and S804 ensuring that all operations comply with permission settings and generating audit logs.

[0051] For the foregoing embodiments, in order to simplify the description, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to this application.

[0052] The above embodiments describe the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Modifications and variations made by those skilled in the art without departing from the spirit and scope of the invention should be within the protection scope of the appended claims.

Claims

1. An animal individual digital twin and self-learning label generation system, characterized in that, Comprise: a perception collection layer for capturing visual information, sound, motion trajectory, physiological signals, and surrounding environment data of animals in real time and in multiple modalities; an AI cognition layer for pre-processing and feature extraction of the data collected by the perception collection layer using deep learning models to identify animal posture, behavior, sound pattern, and physiological index abnormalities; a digital twin and knowledge reasoning layer for generating a globally unique identifier for each animal individual, integrating data, and generating a knowledge graph; a federal collaboration layer for realizing multi-park data sharing and model training; an application service layer for providing an application interface to realize human-computer interaction.

2. The animal individual digital twin and self-learning tag generation system of claim 1, wherein, The perception collection layer comprises various sensors, which are deployed in animal habitats or research areas, including one or more of high-definition cameras, infrared thermal imagers, microphone arrays, wearable biosensors, and environmental sensors.

3. The animal individual digital twin and self-learning tag generation system of claim 1, wherein, The deep learning model comprises one or more of CNN, RNN, Transformer, and GNN.

4. The animal individual digital twin and self-learning tag generation system of claim 1, wherein, The digital twin and knowledge reasoning layer comprises a globally unique ID generation module and a multi-modal data fusion module, wherein the globally unique ID generation module generates a globally unique identifier for each animal individual using a distributed, AI-driven conflict prediction and detection mechanism; the multi-modal data fusion module combines the features output by the AI cognition layer with the original perception data to form a more comprehensive representation of the animal state.

5. The animal individual digital twin and self-learning tag generation system of claim 4, wherein, The digital twin and knowledge reasoning layer further comprises a self-learning label generation module that automatically discovers new behavior patterns and health status features based on AI models to generate dynamically updated animal individual labels containing predictive information.

6. The animal individual digital twin and self-learning tag generation system of claim 5, wherein, The digital twin and knowledge reasoning layer further comprises a digital twin construction module that integrates the generated unique ID, collected multi-modal data, AI cognition results, and self-learning labels to construct a dynamic digital twin of the animal individual.

7. The animal individual digital twin and self-learning tag generation system of claim 6, wherein, The digital twin and knowledge reasoning layer further comprises a knowledge graph and reasoning engine that constructs a knowledge graph containing information about animal individuals, species, behavior, health, and environmental correlations, and integrates a reasoning engine to support multi-dimensional queries, causal analysis, correlation mining, and scientific research decision support.

8. The animal individual digital twin and self-learning tag generation system of claim 1, wherein, The federal collaboration layer uses federated learning technology to share and collaboratively update model parameters while protecting the privacy of original data from each institution, and through knowledge distillation and heterogeneous label mapping technology, realizes model integration and global knowledge improvement across institutions and heterogeneous systems.

9. The animal individual digital twin and self-learning tag generation system of claim 1, wherein, The application service layer provides a user application interface to support individual information retrieval, behavior analysis, health warning, scientific research report generation, and permission management functions, and integrates with external systems through standard API interfaces.

10. An animal individual digital twin and self-learning tag generation method, based on the animal individual digital twin and self-learning tag generation system of any one of claims 1-9, wherein Comprise the following steps: S1: Globally unique ID generation; S2: Multi-modal data collection and pre-processing; S3: AI cognition and feature extraction; S4: Self-learning label generation; S5: Digital twin construction; S6: Knowledge graph and reasoning; S7: Federal collaboration and privacy protection; S8: Application service.

Images