A method and system for managing cages to prevent the risk of viral transmission

Abstract

The application relates to the field of animal management, and in particular to a livestock cage management method and system for preventing viral infection risks. The method comprises the following steps: acquiring a multi-source health data set of a pet individual, performing digital risk portrait construction and intelligent access triage based on the multi-source health data set, and generating a pet boarding initial information set; based on the pet boarding initial information set, indirect contact risk link modeling combined with space-time trajectory and adaptive partition scheduling for minimizing cross-infection risks based on dynamic adjustment of intelligent livestock cage unit layout are performed, and a boarding group dynamic management information set is generated; based on the boarding group dynamic management information set, automatic intervention driven by contact risk link tracing is performed, and a pet boarding whole-cycle health report is output at the end of the boarding period. In the livestock cage management process for preventing viral infection, the application can realize prospective early warning and precise intervention on cross-infection risks, and reduce animal health and public health safety hazards.

Description

Technical Field

[0001] This application relates to the field of animal management, and in particular to a method and system for managing animal cages to prevent the risk of viral transmission. Background Technology

[0002] In the field of modern high-density pet boarding and animal health management, centralized boarding has become the mainstream model to meet social needs. Its internal biosecurity control level is directly related to the health and well-being of each pet, the risk of cross-infection of epidemics, and the operational safety of service institutions. It is a key link in building a trustworthy pet service industry and animal welfare protection system.

[0003] However, the existing management methods of boarding facilities lack a mechanism for systematically and in real-time sensing and collaborative analysis of the hidden risks of individual pets, indirect contact networks among groups, and environmental media when facing complex, dynamic, and multi-source contact intensive boarding environments. This not only fails to achieve forward-looking early warning and precise intervention of cross-infection risks, but may also cause untraceable group transmission during the incubation period of diseases, posing serious risks to animal health and public health safety. Summary of the Invention

[0004] This application provides a method and system for managing livestock cages to prevent the risk of virus transmission, thereby solving the aforementioned technical problems.

[0005] Firstly, this application provides a method for managing livestock cages to prevent the risk of virus transmission, the method comprising: A multi-source health dataset of individual pets is acquired. Based on this dataset, a digital risk profile is constructed and intelligent access triage is performed to generate an initial pet boarding information set. Based on this initial pet boarding information set, indirect contact risk link modeling combining spatiotemporal trajectories and adaptive partition scheduling based on dynamic adjustment of intelligent cage unit layout to minimize cross-infection risk are performed to generate a dynamic management information set for the boarding group. Based on this dynamic management information set for the boarding group, automated intervention driven by contact risk link tracing is performed, and a full-cycle health report of the pet boarding is output at the end of the boarding period.

[0006] The above-mentioned technical solutions upgrade biosafety control from relying on human experience to data-driven and intelligent decision-making, enabling early and accurate identification, dynamic assessment and proactive blocking of infectious disease risks, significantly reducing the probability of cross-infection in groups; at the same time, through automated emergency response and full-cycle health reporting, management efficiency and service transparency are greatly improved, creating a safer boarding environment for pets and enhancing the professional reputation of institutions and customer trust.

[0007] Optionally, the process of generating the initial pet boarding information set includes: the multi-source health dataset includes: pet basic profiles, recent health history records, and on-site multimodal biosafety screening data; based on the multi-source health dataset, a digital risk profile is constructed for each boarding pet through health factor extraction and risk rule fusion to generate a dynamic risk level and individualized monitoring indicators for the pet; based on the dynamic risk level of the pet, intelligent admission triage is performed through a preset pet admission strategy to determine the initial placement area for each boarding pet; and the initial pet boarding information set is generated by combining the dynamic risk level of the pet, the individualized monitoring indicators, and the initial placement area.

[0008] Optionally, the process of acquiring the on-site multimodal biosafety screening data includes: guiding the pet through a security checkpoint equipped with non-contact sensors, and simultaneously collecting its infrared thermal imaging data and respiratory sound audio data; performing surface temperature field analysis on the infrared thermal imaging data to identify abnormally high temperature areas and their temperature gradient distribution; extracting voiceprint features and matching abnormal sound patterns on the respiratory sound audio data to identify whether there are abnormal breathing signs; and outputting the results of identifying the abnormally high temperature areas and their temperature gradient distribution and the abnormal breathing signs as the on-site multimodal biosafety screening data.

[0009] Optionally, the step of constructing a digital risk profile for each boarding pet through health factor extraction and risk rule fusion includes: extracting vaccine expiration dates, past medical history, and breed-specific susceptibility diseases as static health factors from the pet's basic file and recent health history; extracting the maximum temperature difference in the abnormal high-temperature area and the type of abnormal respiratory symptoms as dynamic health factors from the on-site multimodal biosafety screening data; inputting the static and dynamic health factors into a preset risk assessment rule, which calculates an initial risk score based on factor weights and combination logic; combining current community infectious disease pathogen information to perform environmental risk weighting correction on the initial risk score, and outputting the quantified dynamic risk level and the corresponding individualized monitoring indicators.

[0010] Optionally, the process of generating the dynamic management information set for the foster care group includes: based on the initial pet foster care information set, performing indirect contact risk link modeling by analyzing the spatiotemporal trajectories and shared facility records of all fostered pets to generate a real-time contact network graph; based on the real-time contact network graph and the pet dynamic risk level of each pet, performing adaptive partition scheduling calculation with minimizing the risk of cross-infection as the objective function to generate dynamic adjustment instructions for the layout of the intelligent cage unit; according to the dynamic adjustment instructions, controlling the intelligent cage unit to perform physical separation or regional reorganization, and updating the pet activity schedule of each pet; integrating and encapsulating the real-time contact network graph, the dynamic adjustment instructions, and the updated pet activity schedule into the dynamic management information set for the foster care group.

[0011] Optionally, the indirect contact risk link modeling includes: continuously acquiring the location tag signals worn by all boarding pets and the access control records of the smart cage unit to construct the original spatiotemporal trajectory in terms of time and location; analyzing the original spatiotemporal trajectory to identify direct contact events where pets directly coexist in the same physical space, as well as indirect contact events generated by successively using the same facility or through staff; defining a risk weight for each contact event, the risk weight being calculated based on the contact duration, contact type, and the dynamic risk level of both pets; and constructing and updating the real-time contact network graph in real time, with pets and facilities as nodes and contact events with the risk weight and timestamp as edges.

[0012] Optionally, the adaptive partitioning scheduling calculation includes: minimizing the sum of risk weights of all edges in the real-time contact network graph as the core optimization objective; using the physical layout, connection relationships, and adjustable states of all current intelligent cage units as spatial constraints; using the isolation requirements, social needs, and activity capabilities of pets with different dynamic risk levels as policy constraints; running a genetic optimization algorithm to solve for the optimal cage unit combination scheme, physical isolation state, and pet allocation scheme that satisfy the objective function under all constraints, and encoding them as the dynamic adjustment instruction.

[0013] Optionally, controlling the intelligent cage unit to perform physical separation or area reorganization includes: the central controller parsing the dynamic adjustment command to obtain the target cage unit combination scheme and physical separation status; sending control signals to the involved intelligent cage units, the control signals including: driving the mobile chassis of a specific cage unit to reposition itself, controlling the opening or closing of the liftable / sliding partition between adjacent cage units, and adjusting the opening and closing and airflow direction of the ventilation openings; during the execution of the physical separation or area reorganization, each intelligent cage unit provides real-time feedback on its position, partition status, and environmental parameters through built-in sensors, and the central controller performs closed-loop verification of the action completion; after successful verification, the global cage layout map is updated, and the actual execution results are fed back to the foster population dynamic management information set.

[0014] Optionally, the automated intervention driven by contact risk link tracing, and the output of a full-cycle health report for pet boarding at the end of the boarding period, includes: real-time monitoring of the dynamic management information set of the boarding group; when a high-risk contact event triggered by a high-risk pet appears in the real-time contact network map, automatically executing contact chain tracing and generating isolation guidance and enhanced disinfection instructions for the associated pets and facilities; throughout the entire boarding period, aggregating all health status records, behavioral data, contact history, and intervention records for each pet, and synthesizing a structured full-cycle health file according to the timeline; based on the full-cycle health file, automatically generating and outputting the full-cycle health report for pet boarding at the end of the boarding period, the full-cycle health report for pet boarding including: a health data timeline chart, contact safety proof, and a behavioral analysis summary.

[0015] Secondly, this application provides a cage management system for preventing the risk of virus transmission, the system comprising: The initial foster care information module is used to acquire multi-source health datasets of individual pets, construct digital risk profiles and perform intelligent access triage based on the multi-source health datasets, and generate an initial pet foster care information set. The cage dynamic management module is used to generate a foster care group dynamic management information set based on the initial pet foster care information set, by modeling indirect contact risk links combining spatiotemporal trajectories and adaptive partitioning scheduling based on dynamic adjustment of intelligent cage unit layout to minimize the risk of cross-infection. The health report generation module is used to perform automated intervention driven by contact risk link tracing based on the foster care group dynamic management information set, and output a full-cycle health report of pets at the end of the foster care period. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram illustrating an application scenario provided in one embodiment of this application; Figure 2 A flowchart illustrating a method for managing livestock cages to prevent the risk of virus transmission, provided as an embodiment of this application; Figure 3 This is a schematic diagram of a cage management system for preventing the risk of virus transmission, provided as an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0019] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0020] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0021] Existing management methods for animal boarding facilities lack a mechanism for systematically and collaboratively sensing and analyzing the hidden risks of individual pets, indirect contact networks within groups, and environmental media in complex, dynamic, and multi-source contact environments. This not only fails to provide proactive early warning and precise intervention for cross-infection risks but may also lead to untraceable outbreaks during the incubation period of diseases, posing serious threats to animal health and public health safety. Therefore, this application provides a method and system for managing animal cages to prevent the risk of viral transmission.

[0022] Figure 1This is a schematic diagram illustrating an application scenario provided by this application. In the process of managing livestock cages to prevent viral transmission, the method provided in this application can achieve proactive early warning and precise intervention of cross-infection risks, reducing potential threats to animal health and public health safety.

[0023] Specifically, the method of this application is applied to any server that communicates with a pet boarding management platform and obtains multi-source health datasets provided by the pet boarding management platform through the server.

[0024] For specific implementation details, please refer to the following examples.

[0025] Figure 2 This is a flowchart illustrating a method for managing livestock cages to prevent the risk of virus transmission, provided as an embodiment of this application. The method of this embodiment can be applied to servers in the above-described scenarios. Figure 2 As shown, the method includes: S201. Obtain multi-source health datasets for individual pets, construct digital risk profiles and intelligent access triage based on the multi-source health datasets, and generate initial information sets for pet boarding.

[0026] A multi-source health dataset can be a collection of raw data used to comprehensively assess the health status of pets, including pet basic profiles, recent health history records, and on-site multimodal biosafety screening data, with the data sourced from a pet boarding management platform.

[0027] Digital risk profiling can be a process of using data analysis technology to transform multi-source, heterogeneous health data into a standardized, quantifiable risk indicator model, which is used to characterize a pet's current health status, potential infectious disease risk level, and vulnerability.

[0028] Intelligent access triage can refer to the process of automatically assigning initial placement areas and corresponding monitoring levels to pets awaiting boarding based on a pre-constructed digital risk profile and according to preset, scientific access rules and triage logic.

[0029] The initial information set for pet boarding can be a structured data object that encapsulates the pet's digital risk profile (such as risk level and key monitoring indicators) and intelligent triage conclusions (initial placement area and monitoring plan), serving as the individualized basis for all subsequent group management decisions.

[0030] Specifically, biosecurity control in pet boarding services centers on the proactive management and prevention of infectious diseases. This primarily involves accurate assessment of individual health status and effective intervention in contact transmission chains within a group. The aim is to utilize digital health monitoring and intelligent environmental management systems to construct a "dynamically isolating safety network" within the boarding facility. By real-time perception and calculation of individual pet risks and group contact relationships, the risk of outbreaks of infectious diseases can be reduced, while ensuring the basic welfare and activity needs of each pet. In this process, boarding facilities can significantly improve their professional management standards and safety reputation, while pet owners can obtain a transparent and reassuring service experience, achieving a win-win situation for health, trust, and profitability. However, a significant bottleneck in existing boarding facility management technology is the low level of digitalization in health assessment and risk control processes, relying heavily on manual experience. This makes it difficult to accurately identify asymptomatic or latent infections in the early stages and to dynamically respond to the constantly changing contact transmission risks due to pet activities and the addition of new individuals, resulting in persistent health and safety risks during the boarding process. This step addresses this core pain point by acquiring and integrating multi-source health datasets such as "basic profiles," "historical records," and "on-site multimodal screening" to construct a digital risk profile that can quantitatively reflect the static vulnerability and dynamic health signs of pets. Based on this profile, intelligent admission triage is performed, generating a structured initial information set for pet boarding. This provides a unified and standardized data foundation for subsequent precise, dynamically adjustable group risk management.

[0031] S202. Based on the initial information set of pet boarding, perform indirect contact risk link modeling that combines spatiotemporal trajectories and adaptive partition scheduling that minimizes the risk of cross-infection based on the dynamic adjustment of the layout of intelligent cage units to generate a dynamic management information set for the boarding group.

[0032] Indirect contact risk link modeling can be achieved by using mathematical models such as graph theory to abstract the direct and indirect contact between pets into a network composed of "nodes" and "edges", and assigning a risk weight to each edge based on the contact duration, type, and risk level of both parties, thereby constructing a visualized and quantitative contact risk transmission network graph.

[0033] Intelligent pet cage units can refer to modular pet living units with functions such as mobility, physical separability, and independent environmental control, whose status is managed by a central control system. The units referred to in this solution emphasize being dynamically reconfigurable geospatial units, which need to integrate mobile chassis (such as AGV chassis), physical partition mechanisms that link with adjacent units (such as electric lifting platforms), and the ability to respond to central dispatch commands, so as to achieve adaptive adjustment of the overall layout based on a group risk model.

[0034] Adaptive partitioning scheduling can be a dynamic optimization process with the core objective of minimizing the total risk weight in the real-time contact network graph. It comprehensively considers multiple conditions such as physical layout constraints and pet isolation requirements, and uses optimization algorithms to calculate the optimal layout adjustment scheme in real time.

[0035] The dynamic management information set of foster care groups can be a dynamically updated core data pool that integrates real-time contact network maps reflecting the risk of group contact, dynamic adjustment instructions to guide physical space adjustments, and updated pet activity schedules, representing the risk status and space configuration plan of the entire foster care group at any time.

[0036] Specifically, infectious disease control within foster care groups is a dynamic issue of spatial management and network disruption. The main challenge lies in transforming the abstract "contact risk" into concrete "physical isolation" and "movement adjustment." Traditional models rely on fixed cages and manual patrols, which are limited by their inability to quantify the complex indirect contact relationships between pets through direct coexistence or shared facilities. Furthermore, they lack the technical means to quickly reconstruct physical space for precise isolation after risk identification. This results in often delayed and reactive risk management, with isolation only implemented after obvious symptoms appear, by which time the virus may have already formed a hidden transmission chain within the group. This step addresses this dynamic and complex challenge by continuously aggregating spatiotemporal trajectory data of all individuals based on the initial pet foster care information set. A graph model is used to model indirect contact risk links, generating a visible, quantifiable, and time-series-weighted real-time contact network graph, making the invisible transmission risk clearly visible. Then, using this graph and individual risk levels as core inputs, and minimizing the global cross-infection risk as the optimization objective, an adaptive partitioning scheduling algorithm is run, taking into account the physical constraints of the intelligent cage units, such as their mobility and separability, to generate dynamic adjustment instructions in real time. This process achieves an intelligent closed loop from "risk perception" to "spatial reshaping," enabling management strategies to adapt and adjust in response to changes in risk.

[0037] S203. Based on the dynamic management information set of the foster care group, conduct automated intervention driven by contact risk link tracing, and output a full-cycle health report of pet foster care at the end of the foster care period.

[0038] Automated intervention can be a process in which the system automatically triggers an emergency response procedure when it detects a high-risk contact event, traces back along the contact network map to find all related pets and facilities, and immediately generates targeted isolation guidance and enhanced disinfection instructions.

[0039] A comprehensive health report for a pet's boarding period can be automatically generated at the end of the boarding period. It is based on various information gathered throughout the boarding period and includes a timeline chart of health data, proof of safe contact, and a summary of behavioral analysis.

[0040] Specifically, a complete biosafety control system not only needs risk assessment and dynamic scheduling capabilities, but also a rapid emergency response mechanism based on risk, and service value output forming a data closed loop. In traditional foster care management, even if a suspected case is discovered, tracing its contact history relies on manual memory and recording, which is inefficient and prone to omissions, leading to inaccurate isolation scope and untimely disinfection. Furthermore, after the foster care service ends, what is often provided to the owners is a simple written summary, lacking a data-driven presentation of health and behavior throughout the entire foster care period, making it difficult to demonstrate the professionalism and transparency of the service. This step addresses the needs for timeliness, accuracy, and value-added services in emergency response by establishing an automated intervention mechanism driven by contact risk link tracing based on the dynamic management information set of the foster care group (especially the real-time contact network map). When the system identifies a high-risk contact event, it can automatically and instantly trace back and lock all related pets and facilities, triggering targeted isolation guidance and enhanced disinfection instructions, achieving a second-level response from risk warning to precise handling, greatly compressing the potential spread window of pathogens. Furthermore, throughout the entire boarding period, the system automatically aggregates comprehensive data for each pet, generating a structured pet boarding health report at the end of the service. This report, presented in the form of a data timeline and contact safety certificates, visually demonstrates the pet's health status and the care received, transforming the prevention and control process into a tangible trust credential.

[0041] The method provided in this embodiment first constructs a digital risk profile for each pet by integrating archives, historical records, and multimodal data from the scene, such as infrared thermal imaging and respiratory sounds. Based on this profile, intelligent admission triage is performed, generating an initial pet boarding information set that includes risk level and initial placement area. Second, the system continuously tracks the pets' spatiotemporal trajectories, modeling and generating a quantified, visual real-time contact network map. Using this map and individual risk levels as input, and aiming to minimize the risk of cross-infection, the system dynamically calculates the optimal layout scheme through optimization algorithms, driving intelligent cage units to perform physical reorganization such as movement and separation, forming a dynamic management information set for the boarding group. Finally, based on this information set, automated risk response is achieved: once a high-risk contact event is detected, it immediately and automatically traces and triggers precise isolation and disinfection commands; at the end of boarding, the system aggregates the full-cycle data to automatically generate a full-cycle health report for pet boarding, including a health timeline and proof of contact safety, and outputs it to boarding staff and pet owners. This solution upgrades biosafety control from relying on human experience to data-driven and intelligent decision-making, enabling early and accurate identification, dynamic assessment, and proactive blocking of infectious disease risks, significantly reducing the probability of cross-infection in groups. At the same time, through automated emergency response and full-cycle health reporting, it greatly improves management efficiency and service transparency, creates a safer boarding environment for pets, and enhances the professional reputation of institutions and customer trust.

[0042] In some embodiments, the multi-source health dataset includes: pet basic profiles, recent health history records, and on-site multimodal biosafety screening data; based on the multi-source health dataset, a digital risk profile is constructed for each boarding pet through health factor extraction and risk rule fusion, generating a dynamic risk level and individualized monitoring indicators for the pet; based on the dynamic risk level of the pet, intelligent admission triage is performed through a preset pet admission strategy to determine the initial placement area for each boarding pet; combining the dynamic risk level of the pet, individualized monitoring indicators, and initial placement area, an initial pet boarding information set is encapsulated and generated.

[0043] A preset pet admission policy can be a set of predefined business rules in the system. It clearly defines the standard handling procedures for pets with different "dynamic risk levels", especially the division criteria for initial placement areas (such as quarantine areas, observation areas, and general areas).

[0044] The initial placement area can be the first standardized living space automatically assigned to a pet at the beginning of its boarding period by the system based on the "pet dynamic risk level" in its digital risk profile and by executing the intelligent triage logic of the "preset pet admission strategy". This area is the basic physical unit for all subsequent dynamic scheduling and management, and usually includes different types such as independent isolation cabins, single observation rooms, and general social areas to achieve risk-level management.

[0045] Specifically, traditional boarding facilities rely heavily on visual observation and verbal inquiries by staff for health assessments and triage decisions upon pet admission. This approach suffers from inconsistent standards, strong subjectivity, low efficiency, and a lack of quantifiable records. For example, a hurried staff member might overlook a seemingly lively puppy with slightly rough breathing, mistakenly placing it in the regular group. This puppy could be a potential carrier of kennel cough. This step aims to replace this weak link with a standardized, data-driven intelligent decision-making process. By transforming scattered and heterogeneous health data into a unified "risk profile" and automatically triaging based on clear strategies, it ensures that high-risk individuals are accurately identified and isolated from the source, preventing cross-infection source management failures due to human oversight.

[0046] In the specific analysis process, health factor extraction is performed first. The system directly reads the fields "Vaccine Expiry Date" (calculating the number of days until expiry), "Past Medical History" (coded according to disease infectivity), and "Breed" (mapped to a known susceptible disease database) from the structured "Pet Basic Profile" database. From the text of "Recent Health History Records," relevant medical history descriptions are extracted using Natural Language Processing (NLP) keyword matching technology. Simultaneously, from the "On-site Multimodal Biosafety Screening Data," the "Maximum Temperature Difference in Abnormal High Temperature Areas" (numerical value) and "Type of Abnormal Respiratory Signs" (classification labels, such as moist rales, wheezing) are parsed. Subsequently, risk rule fusion is performed: the system inputs all the above factor values ​​into a preset "Risk Assessment Rule" model. This model is essentially a weighted decision tree or rule engine, and its rule base and weights are jointly defined and established by veterinary experts and epidemiological data. For example, a rule might be: "IF vaccine expired AND breed is 'French Bulldog' (susceptible to respiratory diseases) THEN risk score +20; IF dynamic factor 'abnormal respiratory sound type' is 'moist rales' THEN risk score +30." The system iterates through all relevant rules and accumulates them to obtain the "initial risk score." Next, the system calls an external public health data service via API to obtain "current community epidemic pathogen information." If there is a canine influenza outbreak, a global environmental risk weighting coefficient (e.g., multiplied by 1.2) is applied to the initial risk scores of all pets. Finally, based on the score falling into a preset range (e.g., 0-50 low risk, 51-80 medium risk, 81-100 high risk), the system outputs the "pet dynamic risk level" (e.g., M2) and generates a list of "individualized monitoring indicators" (e.g., "monitoring: respiratory rate, twice daily") based on the triggered key factors. After the pet profile is completed, the intelligent admission triage module is activated. It queries the preset "Pet Admission Strategy" rule table (e.g., "Rule: Dynamic Risk Level = H3 -> Initial Placement Area = 'Independent Isolation Chamber'; = M2 -> 'Single Room in Observation Area'; = L1 -> 'General Social Area'") and automatically determines the pet's "Initial Placement Area". Finally, the system encapsulates these three key results into a "Pet Boarding Initial Information Set" in JSON format.

[0047] In alternative or modified implementations, the "risk assessment rule" model can be replaced by a machine learning classification model (such as random forest or gradient boosting tree) trained on a large amount of historical case data, instead of using a rule base based on expert experience. This model uses health factors as features and whether the disease will eventually develop as a label, and directly outputs the probability of the risk level. The "preset pet admission strategy," in addition to being based on a static rule table, can also be designed as an optimization model dynamically calculated based on the current capacity and risk distribution of each area to achieve a balance in the initial risk load of the entire facility. For health factor extraction, if the quality of the archival data is high, a more complex NLP model can be introduced to deeply understand the free text medical records and extract more granular medical history factors.

[0048] In some embodiments, pets are guided through a security checkpoint equipped with non-contact sensors, and their infrared thermal imaging data and respiratory sound audio data are collected simultaneously. The surface temperature field of the infrared thermal imaging data is analyzed to identify abnormally high temperature areas and their temperature gradient distribution. The respiratory sound audio data is subjected to voiceprint feature extraction and abnormal sound pattern matching to identify whether there are abnormal breathing signs. The results of identifying abnormally high temperature areas and their temperature gradient distribution and abnormal breathing signs are combined as on-site multimodal biosafety screening data output.

[0049] A security checkpoint equipped with non-contact sensors can refer to a physical channel specifically designed for pets, which integrates sensor devices that can collect physiological data without physical contact with the pet. For example, an infrared thermal imager is installed at the top of the channel, and a high-sensitivity microphone array is installed on the side walls.

[0050] Specifically, traditional pet boarding admission screening relies heavily on staff touching the nose to check for dryness and listening to breathing. This method is highly subjective, inefficient, and lacks quantifiable data. For example, a stressed puppy might experience a slight increase in body temperature and rapid breathing due to stress, which could easily be misdiagnosed as pathological, or conversely, early, mild symptoms might be missed. This step aims to objectively and synchronously collect core physiological indicators (body temperature and respiratory sounds) by deploying standardized, non-contact, multimodal sensing channels, and then use algorithms for precise analysis. This allows for automated, highly sensitive initial screening of fever and respiratory abnormalities—two key indicators of infectious diseases—without stress, right from the moment the pet enters the boarding facility.

[0051] In the specific analysis process, firstly, the pets awaiting boarding are guided alone through a security checkpoint approximately 1.5 meters long and 0.6 meters wide. An infrared thermal imager (such as the FLIR Axx series) inside the checkpoint continuously captures images at a frame rate of 10Hz. Once the pet's body is fully within the field of view, the system triggers a data snapshot, acquiring a complete "infrared thermal imaging data" (e.g., a 320x240 grayscale temperature image). Simultaneously, a 4-microphone linear array deployed on both sides of the checkpoint begins recording "breathing sound audio data" for 5 seconds at a sampling rate of 44.1kHz. Parallel analysis is then performed: for the thermal imaging data, the body surface temperature field analysis module first performs image preprocessing (e.g., median filtering for noise reduction, and extraction of regions of interest based on pet contour segmentation), then calculates the average temperature and standard deviation of the entire body surface. Next, a threshold-based region growing algorithm or edge detection algorithm is used to identify connected regions with temperatures higher than "overall average temperature + 2 times the standard deviation," marking them as "abnormally high temperature regions." For each abnormal region, its centroid is calculated, and the temperature change curve radiating from the centroid outwards is analyzed to quantify its "temperature gradient distribution" (e.g., "radial gradient: -0.5°C / cm"). For respiratory sound data, the voiceprint feature extraction and abnormal sound pattern matching module first pre-emphasizes, frames, and windows the audio, then calculates 39-dimensional MFCC features (including first and second-order differences) for each frame to form a feature sequence for that respiratory sound segment. Subsequently, this feature sequence is matched against a preset "abnormal sound pattern library." This pattern library is constructed by collecting a large number of clinically diagnosed abnormal pet respiratory sound samples (such as moist rales in dogs with pneumonia and wheezing in cats with asthma), extracting their MFCC features, and then training a classification model for each type of abnormal sound using a Gaussian mixture model (GMM) or a deep neural network. The system inputs the current feature sequence into these models, calculates the probability that it belongs to each type of abnormal sound, and if the highest probability exceeds a confidence threshold (e.g., 0.7), it determines that there is an "abnormal respiratory sign" and outputs its type. Finally, the coordinates of the abnormal high-temperature areas, temperature gradient values, and the types of abnormal breathing sounds obtained from the analysis are packaged into a structured "on-site multimodal biosafety screening data" output.

[0052] In alternative or modified implementations, the infrared thermal imager can be replaced with a higher-resolution area array temperature sensor to obtain a finer distribution of body surface temperature. Breathing sound acquisition can be performed using directional microphones within the channel, or short-term acquisition using a miniature electronic stethoscope attached to the pet's collar, in addition to microphone arrays. The construction of the "abnormal sound pattern library" can utilize GMM, support vector machines (SVM), or more complex convolutional neural networks (CNN) to process the temporal spectrogram of the audio. In body surface temperature field analysis, besides region growing algorithms, deep learning-based semantic segmentation models (such as U-Net) can be used to more accurately segment specific anatomical areas such as the ears and abdomen for independent temperature analysis. The security checkpoint can also be modified to allow pets to briefly stay in specific compartments for a more comprehensive scan.

[0053] In some embodiments, the following parameters are extracted as static health factors from the pet's basic profile and recent health history: vaccine expiration date, past medical history, and breed-specific susceptibility diseases. The following parameters are extracted as dynamic health factors from on-site multimodal biosafety screening data: the maximum temperature difference in abnormally high-temperature areas and the type of abnormal respiratory symptoms. The static and dynamic health factors are input into a preset risk assessment rule, which calculates an initial risk score based on factor weights and combination logic. The initial risk score is then adjusted using environmental risk weighting, incorporating information on currently prevalent infectious pathogens in the community, and a quantified dynamic risk level and corresponding individualized monitoring indicators are output.

[0054] Static health factors can be health risk-related attributes that are relatively stable or do not change frequently during the pet's boarding period, extracted from the pet's "basic profile" and "recent health history".

[0055] Dynamic health factors can be indicators extracted from "on-site multimodal biosafety screening data" that reflect the pet's current physiological state in real time.

[0056] A preset risk assessment rule can be a risk assessment model predefined in the system. Its core is "factor weights and combination logic", which is an abstract logical framework that refers to how the system assigns importance (weights) to different health factors and how to combine multiple factors through logical operators (such as AND, OR, IF-THEN) to derive a comprehensive risk score.

[0057] Environmental risk weighting correction can be based on the "initial risk score" and then proportionally adjusted (usually amplified) according to external "current community infectious disease pathogen information" (such as regional canine influenza outbreaks) to reflect the macro-environmental factor of the overall risk level of infectious disease transmission in a specific period and region.

[0058] Specifically, traditional admission risk assessment relies on staff's memory and experience for rough judgment, failing to systematically and quantitatively combine scattered static information (such as vaccination records), dynamic vital signs (such as on-site temperature measurement), and macro-level epidemic data (such as community outbreaks). For example, which is riskier: a dog with complete vaccination records but abnormal breathing sounds on-site, or a dog with expired vaccines but normal on-site? There is a lack of unified standards. This step aims to encode veterinary epidemiological knowledge into a computable algorithm by defining clearly defined "pre-set risk assessment rules," achieving standardized, repeatable, and quantitative risk scoring for each pet. Furthermore, the scoring scale can be dynamically adjusted based on external epidemic conditions, ensuring that risk assessment conclusions are both individualized and environmentally relevant.

[0059] In the specific analysis process, the system first performs health factor extraction. For static factors: the "vaccine expiration date" is directly read from the database, and the remaining effective days are calculated; "past medical history" tags (such as "canine parvovirus infection recovery period") are extracted from "recent health history" through NLP keyword matching; and a pre-set breed susceptibility disease database is queried based on the "breed" field to obtain a list of "breed susceptibility diseases" (such as "Bulldog: Respiratory Syndrome"). For dynamic factors: the output screening data is directly parsed to extract the "maximum temperature difference in abnormal high temperature areas" (such as 2.5℃) and the "type of abnormal respiratory signs" (such as "moist rales"). Subsequently, all factors are input into a pre-set "risk assessment rule" model. This model can be a rule engine based on weighted summation and conditional logic trees. For example, its rule base might be defined as: base score = 0; factor "expired vaccine" weight = +25; "history of respiratory disease" weight = +15; "breed is susceptible" weight = +10; "maximum temperature difference > 2℃" weight = +(temperature difference value * 10); "abnormal breath sound type = moist rales" weight = +30. It also includes combinational logic: "IF ('history of respiratory disease' is true) AND ('abnormal breath sound type' is not empty) THEN +20". The system iterates through all applicable rules and accumulates to calculate the "initial risk score". Next, the system calls the public health data API to obtain "current community epidemic infectious disease pathogen information", such as "a surge in locally reported canine infectious respiratory disease cases". Based on a preset correction coefficient table (e.g., "during a respiratory disease epidemic, all pets' initial risk score × 1.3"), the system performs "environmental risk weighted correction" on the pet's initial score. Finally, the corrected final score is mapped to a preset level range (e.g., 0-40 is L1 low risk, 41-70 is M2 medium risk, and above 71 is H3 high risk), and the "Pet Dynamic Risk Level" is output. "Individualized monitoring indicators" are generated based on the key factors that have been triggered (e.g., "Daily respiratory rate monitoring" is added to the indicators because "wet rales" is triggered).

[0060] In alternative or modified implementations, the "preset risk assessment rule" model can be replaced by a machine learning model trained using historical foster care data (including health factors and whether the disease ultimately developed), such as logistic regression, random forest, or XGBoost. This model automatically learns the weights and complex interactions of each factor. The "environmental risk weighted correction" can be more than just a simple multiplication coefficient; it can be based on a spatiotemporal propagation model of an epidemic, providing more refined dynamic correction values ​​according to the specific geographical location of the foster care facility and the radius of epidemic spread. Furthermore, the analysis of "recent health history" among the static factors can employ more advanced clinical natural language processing models to extract more complex medical entities and relationships.

[0061] In some embodiments, based on the initial pet boarding information set, indirect contact risk link modeling is performed by analyzing the spatiotemporal trajectories of all boarded pets and shared facility records to generate a real-time contact network map. Based on the real-time contact network map and the dynamic risk level of each pet, adaptive partition scheduling calculation is performed with minimizing the risk of cross-infection as the objective function to generate dynamic adjustment instructions for the layout of intelligent cage units. According to the dynamic adjustment instructions, the intelligent cage units are controlled to perform physical separation or regional reorganization, and the pet activity schedule of each pet is updated. The real-time contact network map, dynamic adjustment instructions, and updated pet activity schedule are integrated and encapsulated into a dynamic management information set for the boarding group.

[0062] Real-time contact network graphs can be a model that uses graph theory data structures to represent contact risk relationships within foster care groups in real time. Nodes represent pets or public facilities (such as watering points or activity areas), edges represent contact events occurring between nodes, and the attributes attached to the edges (such as risk weights and timestamps) quantify the degree of risk associated with that contact.

[0063] Adaptive partitioning scheduling calculation can be a dynamic optimization algorithm that aims to minimize the risk of cross-infection, and uses the current physical layout of the facility and the risk status of the pets as constraints to calculate the optimal cage allocation and physical isolation scheme in real time.

[0064] Pet activity scheduling can refer to a dynamically adjusted schedule that re-plans and creates for each pet, outlining when they can enter public activity areas and what facilities they can use. Its purpose is to stagger peak times and further reduce the overlap in time and space between pets of different risk levels in public areas.

[0065] Specifically, traditional pet boarding facilities typically have static cage arrangements that rarely change once admitted, failing to respond to dynamic changes in group risk. For example, if an initially low-risk pet develops a cough midway through boarding (risk level dynamically adjusted upwards), or if two medium-risk pets are found to have shared a toy, traditional methods cannot proactively and immediately redesign the entire group's spatial layout to isolate the risk. This step aims to upgrade static cage management to dynamic, adaptive scheduling by linking the "risk map" with the "physical space" in real time, thereby proactively breaking potential indirect contact chains and minimizing the risk of cross-infection through spatial layout.

[0066] In the specific analysis process, the system first continuously receives real-time data streams from UWB positioning tags and smart access control based on the "initial pet boarding information set" of all pets, and performs "indirect contact risk link modeling" to construct and update a dynamic "real-time contact network graph". Subsequently, the "adaptive partition scheduling calculation" module is triggered (the triggering condition can be a new pet checking in, a change in the risk level of a pet, or timed optimization). This module takes the "real-time contact network graph" and the "dynamic risk level of pets" of each node as input. Its core is to solve an optimization problem with the objective function of "minimizing the sum of risk weights of all edges in the graph". The design of this objective function is essentially to transform the business goal of "reducing overall contact risk" into a computable mathematical minimization problem. The solution algorithm (such as a genetic algorithm) needs to be searched under multiple constraints: spatial constraints (such as the total number of cage units, movement range, and existing partition status); and policy constraints (such as high-risk pets must be isolated separately, and some pets need social activities). After the algorithm runs, it outputs an optimal "cage unit combination scheme" and "pet allocation scheme", which is encoded as a "dynamic adjustment instruction". The central controller parses the instruction and sends control signals via wireless network to the drive motors (mobile chassis) and partition servo motors of the relevant cage units to perform physical reorganization. After reorganization, the system generates an updated "pet activity schedule" for each pet based on the new spatial layout (e.g., low-risk pet A uses the south activity area in the morning, and medium-risk pet B uses it in the afternoon). Finally, the system encapsulates the key data generated in this process (updated maps, adjustment instructions, and new schedules) into a "foster care group dynamic management information set".

[0067] In alternative or modified implementations, the solution algorithm for "adaptive partitioned scheduling computation" can be replaced with particle swarm optimization, simulated annealing, or integer programming-based solvers to adapt to efficiency and accuracy requirements at different scales (number of pets). In addition to minimizing the total risk weight, the objective function can incorporate auxiliary objectives such as "minimizing total movement energy consumption" and "equalizing cage space areas" for multi-objective optimization. The generation of "pet activity scheduling" can be more intelligent, for example, by combining historical pet behavior data (active time preferences) for personalized arrangements, rather than simply rotating and staggering activity times.

[0068] In some embodiments, the system continuously acquires the location tag signals worn by all boarding pets and the access control records of the smart cage unit to construct the original spatiotemporal trajectory in terms of time and location. The system analyzes the original spatiotemporal trajectory to identify direct contact events where pets directly coexist in the same physical space, as well as indirect contact events generated by sequential use of the same facility or through staff. A risk weight is defined for each contact event, which is calculated based on the contact duration, contact type, and the dynamic risk level of both pets. A real-time contact network graph is constructed and updated in real time, with pets and facilities as nodes and contact events with risk weights and timestamps as edges.

[0069] The original spatiotemporal trajectory can be a sequence of raw data containing timestamps and spatial location information continuously generated by the positioning tag (signal source) worn by the pet and the access control system (event recorder) of the smart cage unit; its source is the UWB (ultra-wideband) or RFID indoor positioning system deployed in the foster care environment, as well as the electronic door lock logs of each cage door and activity area door.

[0070] Risk weight can be a quantifiable value used to characterize the potential risk of pathogen transmission from a single "contact event." Its calculation is based on a predefined mathematical model whose input variables include: contact duration (T, in minutes), contact type (K, with higher values ​​for direct contact, such as 1.0; lower values ​​for indirect contact, such as 0.3), and the dynamic risk levels (R1, R2) of the pets involved in the contact. An exemplary formula is: Risk Weight W = K * T * (R1 + R2).

[0071] Specifically, traditional pet boarding facilities' understanding of pet contact risks is limited to visible "fighting" or "cohabitation," completely failing to track and quantify more hidden and common indirect contact risks (such as sharing food and water, or transmission through staff hands). For example, a seemingly healthy carrier and a puppy licking the same toy in succession—this high-risk indirect contact would be completely overlooked in traditional management. This step aims to systematically analyze the spatiotemporal data of all pets, like an epidemiological survey, to fully reconstruct all direct and indirect contact chains within the group, and assign a quantified risk value to each contact event, thereby providing an accurate and comprehensive data foundation for subsequent dynamic zoning and scheduling.

[0072] During the specific analysis process, a data stream processing engine continuously runs in the system background. This engine consumes information from two data sources in real time: first, the coordinate data reported every second by the UWB positioning tags worn by all pets; and second, the opening and closing records (including pet IDs and timestamps) of all smart cage unit doors and public area doors. The engine first performs time alignment and cleaning on these data streams to construct a continuous "original spatiotemporal trajectory" for each pet. Then, the contact recognition algorithm begins to work: for direct contact events, the algorithm divides the space into a grid. If two pets spend more than 80% of the time in the same grid within the same time window (e.g., within 30 consecutive seconds), it is determined to be direct contact. For indirect contact events such as "using the same facility sequentially," the algorithm maintains a usage queue for each public facility (e.g., water fountain #1). When pet B's usage record immediately follows pet A's (the time interval is less than the preset "facility residual risk window," e.g., 10 minutes), an indirect contact edge is established from A to B. For events occurring "via staff intermediaries," the system also tracks the movements of staff wearing name tags. If their movements intersect with those of pet A and pet B within a short period (e.g., within 5 minutes) (e.g., they stop in front of pet A's cage and then go to pet B's cage), an indirect contact edge is inferred. For each identified event, the system immediately calculates a specific risk weight (W value) based on the preset risk weight calculation formula, using the "dynamic risk level" (R value) of both parties involved. Ultimately, the system constructs and maintains a dynamic, weighted "real-time contact network graph" in memory, using pets and facilities as nodes and these contact events with timestamps and weights (W) as directed edges. This graph continuously adds, deletes, and updates edge weights as new data flows in.

[0073] In alternative or modified implementations, the positioning technology can be replaced with RSSI (Received Signal Strength Indication) positioning based on Bluetooth beacons, or video tracking based on computer vision. The "facility residual risk window" for identifying "sequential use of the same facility" can be differentiated based on the in vitro survival time of different pathogens; for example, for canine parvovirus (long survival time), the window can be extended to several hours. The mathematical model used to calculate risk weights can be more complex, for example, by incorporating ambient temperature and humidity as influencing factors, or by using a machine learning model trained on historical transmission data instead of a fixed formula. For inferences about "worker-mediated" exposure, sensor data from hand sanitizer dispensers installed in key areas can be used to increase or decrease the confidence level of the inference.

[0074] In some embodiments, the core optimization objective is to minimize the sum of risk weights of all edges in the real-time contact network graph; the physical layout, connection relationships, and adjustable states of all current smart cage units are used as spatial constraints; the isolation requirements, social needs, and activity capabilities of pets with different dynamic risk levels are used as policy constraints; a genetic optimization algorithm is run to solve for the optimal cage unit combination scheme, physical isolation state, and pet allocation scheme that satisfy all constraints, and these are encoded as dynamic adjustment instructions.

[0075] Spatial constraints can be restrictions on the physical attributes and layout of the "intelligent cage unit" that must be followed in optimization calculations; their source is the system's real-time equipment status database, specifically including: the physical layout of all cage units (such as a two-dimensional coordinate map), connection relationships (which units are adjacent and can be merged or isolated), and adjustable states (which unit chassis can be moved, and which partitions can be raised and lowered).

[0076] Strategy constraints can be soft restrictions based on management rules and pet welfare that must be followed in optimization calculations; their sources are a pre-set management strategy library and pet profiles, specifically including: isolation requirements (e.g., high-risk pets should not share space with any other pets), social needs (e.g., certain pet pairs that have established friendly relationships should be arranged to be adjacent to each other as much as possible under controllable risk), and activity levels (e.g., the cages of elderly or disabled pets should not be frequently moved to the far end).

[0077] Genetic optimization algorithms can be specific algorithms used to solve complex combinatorial optimization problems, and belong to evolutionary computation.

[0078] Specifically, traditional boarding facilities rely entirely on the experience and intuition of administrators for zoning and scheduling. Faced with numerous pets, diverse risk levels, and complex physical limitations of cages, manual solutions often fall short, either failing to achieve optimal risk isolation or severely sacrificing space utilization and pet welfare. This step aims to formalize the scheduling problem into a mathematical optimization model with multiple constraints and use the powerful search tool of "genetic optimization algorithm" for automated solution, thereby outputting a mathematically provable, low-risk (or near-low-risk) optimal scheduling solution within seconds that satisfies all hard constraints.

[0079] In the specific analysis process, after the scheduling calculation is triggered, the algorithm module first structures all input parameters. The core optimization objective is defined as the fitness function Fitness = 1 / (ΣW + ε), where ΣW is the total weight of contact edges predicted based on the new layout and activity schedule under a given scheduling scheme, and ε is a small constant to prevent division by zero. Spatial constraints are encoded as verification rules during chromosome decoding: for example, a gene string representing cage allocation, if it causes two cell positions to overlap when decoded into physical coordinates, is judged as illegal, and its fitness is set to zero. Policy constraints are partly implemented as hard constraints (such as isolation requirements) through decoding verification; partly as soft constraints (such as social needs) are incorporated into the fitness calculation through a penalty function, that is, if not satisfied, a large penalty value is added to ΣW. Next, the algorithm is initialized in the computing environment (such as using Python's DEAP library): a batch of "chromosomes" are randomly generated, each chromosome encoding a complete "cage cell combination scheme", "physical isolation state", and "pet allocation scheme". For example, an integer array of length (number of units + number of pets) is used. The first half represents the state of each unit (0: closed, 1: merged with the right, 2: merged with the bottom, etc.), and the second half represents the unit ID assigned to each pet. In each generation of evolution, the algorithm calculates the fitness of each chromosome, retains superior individuals through a "roulette wheel selection" method, and performs single-point crossover (exchanging some gene segments) and uniform mutation (randomly changing a gene value) on the selected individuals with a certain probability to produce offspring. After this process iterates for hundreds to thousands of generations, the scheme represented by the chromosome with the highest fitness is encoded as the final "dynamic adjustment instruction".

[0080] In alternative or modified implementations, the "genetic optimization algorithm" can be replaced with other metaheuristic algorithms, such as simulated annealing, particle swarm optimization, or, when the problem size is small, an exact solver (such as the mixed-integer programming solver Gurobi) can be used. The fitness function can be designed to be more complex, for example, by incorporating considerations of total movement cost and spatial equilibrium, forming a multi-objective optimization. For handling soft "policy constraints," constrained multi-objective evolutionary algorithms can also be used to directly find the Pareto optimal solution set.

[0081] In some embodiments, the central controller parses the dynamic adjustment instructions to obtain the target cage unit combination scheme and physical partition status; it sends control signals to the relevant smart cage units, including: driving the mobile chassis of a specific cage unit to reposition itself, controlling the opening or closing of the liftable / sliding partition between adjacent cage units, and adjusting the opening and closing and airflow direction of the ventilation openings; during the physical separation or area reorganization process, each smart cage unit provides real-time feedback on its position, partition status, and environmental parameters through built-in sensors, and the central controller performs closed-loop verification of the action completion; after successful verification, it updates the global cage layout map and feeds back the actual execution results to the foster population dynamic management information set.

[0082] The central controller can refer to the core computing and instruction distribution device that runs the core scheduling and control algorithm and is responsible for coordinating the collaborative actions of all "intelligent cage units". It is usually an industrial PLC, edge computing server or high-performance embedded industrial control computer.

[0083] Physical separation can refer to the operation of physically dividing a originally connected space into independent areas by controlling the liftable / sliding partition between adjacent units without changing the physical location of the "intelligent cage unit".

[0084] Regional reorganization can refer to the operation of changing the spatial position of a "smart cage unit" by driving the mobile chassis at the bottom and combining it with partition control to reassemble it into a new large space (such as merging multiple units into an activity area) or a new layout that meets scheduling requirements.

[0085] A mobile chassis can refer to a motorized platform integrated into the bottom of each "intelligent cage unit" and containing a built-in drive motor and navigation module (such as an encoder and magnetic guidance sensor). It is the core execution component that enables autonomous repositioning.

[0086] A liftable / sliding partition can refer to a movable partition installed between the side walls of adjacent "intelligent livestock cage units" and driven by an electric push rod or servo motor; its "open" state connects the two unit spaces, while its "closed" state forms a physical barrier.

[0087] The opening and closing of the vents and the direction of airflow can refer to the controllable ventilation components integrated into the "intelligent cage unit," including switchable louvers and air guide vanes, used to adjust the ventilation path of each independent space after reorganization to avoid cross-flow of air.

[0088] A global cage layout map can refer to a digital twin map maintained in the memory of a central controller that reflects in real time the precise location, orientation, partition status, and connection relationships of all "intelligent cage units".

[0089] Specifically, traditional foster care facilities rely entirely on manual handling and manual insertion / removal of partitions to adjust cage layouts. This is not only inefficient and labor-intensive, but also difficult to execute complex spatial reconfiguration schemes accurately, making it prone to errors and preventing virtual risk management strategies from being accurately implemented in the physical world. This step aims to achieve seamless, precise, and automated execution of scheduling commands to physical space changes through precise driving and closed-loop verification of the mechatronic "intelligent cage unit" by a central controller, ensuring that the optimal risk isolation scheme is faithfully reproduced.

[0090] In the specific analysis process, the central controller (such as an industrial PLC or edge server) first parses the "dynamic adjustment instruction" calculated by the "genetic optimization algorithm" after receiving it. This instruction is essentially a structured task list, for example: {Unit A: Move to coordinates (X1, Y1); Units B and C: Lower the partition plate; Unit D area: Close the east-facing ventilation vent}. The controller plans a collision-free movement path based on the "global cage layout map," and then sends specific "control signals" to the relevant units sequentially via a wireless communication network (such as Wi-Fi or Zigbee). For units that need to move, the signal triggers the drive motor of its "moving chassis," which autonomously moves to the target coordinates based on a preset magnetic strip track or real-time positioning and navigation via laser SLAM, with the encoder providing real-time feedback on the travel distance. For units that need to operate the partition plate, the signal controls the electric push rod to move until the limit switch at the top of the partition plate is triggered, providing feedback of "closed" or "open" signals. Simultaneously, the controller sends instructions to the ventilation actuators of the relevant units to adjust the opening and closing of the louvers and the angle of the guide vanes. Throughout the execution process, the built-in sensors in each unit continuously report real-time feedback data streams (such as location coordinates, partition status codes, and vent angle values) to the central controller. The controller performs a closed-loop verification of this feedback data against the expected command values, for example, determining whether the actual coordinates of unit A are within the ±5cm tolerance range of the target coordinates. Only when all subtasks in the list pass the verification is the entire scheduling considered successful. Subsequently, the controller immediately updates the global cage layout map with the latest unit status. Finally, the complete execution log of this "physical separation or area reorganization" (including commands, feedback data, and verification results) is packaged and "feedback" to the "fostered population dynamic management information set" to form a traceable record.

[0091] In alternative or modified implementations, the "mobile chassis" can use UWB ultra-wideband positioning for precise positioning instead of magnetic strip tracks. The "liftable / sliding partition" can be made of flexible, transparent material in a roller blind style to save space and maintain visual openness. Ventilation control can be upgraded to an independent fresh air system with HEPA filtration and directional airflow adjustment for stricter air isolation. In addition to basic position and limit switches, the "closed-loop verification" sensors can include force sensors to prevent pinching, or visual sensors for secondary confirmation. Updates to the "global cage layout map" can be event-driven, rather than only occurring after all actions are completed, to achieve near real-time map status synchronization.

[0092] In some embodiments, the system monitors the dynamic management information set of the foster care group in real time. When a high-risk contact event triggered by a high-risk pet appears in the real-time contact network graph, it automatically executes contact chain tracing and generates isolation guidance and enhanced disinfection instructions for the associated pets and facilities. Throughout the foster care period, it aggregates all health status records, behavioral data, contact history, and intervention records for each pet and synthesizes a structured full-cycle health record in chronological order. Based on the full-cycle health record, at the end of the foster care period, it automatically generates and outputs a full-cycle health report for pets. The full-cycle health report for pets includes: a health data timeline chart, contact safety certificate, and a behavioral analysis summary.

[0093] High-risk pets are those whose "dynamic risk level" is determined to be the highest level after being quantitatively calculated using the "preset risk assessment rules".

[0094] A high-risk contact event can be defined as a newly generated or updated edge (contact event) in the "real-time contact network graph" whose "risk weight" value exceeds the system's preset alarm threshold.

[0095] The isolation guidance and enhanced disinfection instructions can be two types of structured operation commands automatically generated by the system based on the "contact chain tracing" results. The isolation guidance instruction is sent to the controller of the relevant pet's "intelligent cage unit" to trigger it to enter the isolation mode (such as lighting up the warning light, locking the cage door, and closing the public passage). The enhanced disinfection instructions are sent to the facility management system or robot to specify the level of cleaning procedure (such as using disinfectant or extending the disinfection time) for the relevant facilities (such as drinking points and toys).

[0096] Specifically, traditional foster care management often relies on manual memory and scattered records for rudimentary isolation and disinfection when potential infectious events occur. This approach is slow to respond and prone to missing related individuals. Furthermore, it cannot provide clients with a systematic and reliable health history and safety documentation afterward. This step aims to automate and refine emergency response to risk events, and to consolidate all data throughout the foster care period into an authoritative, intuitive, and verifiable "digital health passport," significantly enhancing the professional transparency of management and increasing client trust.

[0097] During the specific analysis process, a real-time monitoring process runs in the background of the system, continuously scanning the "real-time contact network graph." It defines a preset trigger threshold: when the risk weight of a newly generated edge (contact event) exceeds the threshold (e.g., W>15), and the "dynamic risk level" of the pet initiating the event is high risk, it is determined to be a "high-risk contact event." Once triggered, the system immediately starts the "contact chain tracing" module, starting with the pets of both parties involved in the event, and performs a breadth-first search (BFS) in the graph to find all nodes (pets, facilities) that have had direct or indirect contact with them within the past 24 hours (this time window is a preset reference value for the pathogen incubation period). The tracing results generate a list of associated objects. Subsequently, the system automatically generates instructions based on a preset rule base: for the associated pets in the list, an "isolation guidance instruction" is generated and issued to their respective cage unit; for the associated facilities in the list, an "enhanced disinfection instruction" is generated and added to the facility management queue. Simultaneously, from the start, the system maintains a "structured full-cycle health record" for each pet. This record's data model is time-based, with each record containing a timestamp, data type (such as "temperature monitoring," "risk level update," "contact event," and "intervention instructions"), and specific values / content. All data streams are appended in real-time according to this model. At the end of the boarding period, the report generation module is triggered: it calls a chart library to extract time-series data such as temperature and risk level from the record, generating a "health data timeline chart"; it runs a logic-based judgment program to check all contact records in the record; if none meet the high-risk criteria, it generates a "contact safety certificate"; it performs statistical analysis on behavioral data in the record, such as activity levels and number of times the pet enters and leaves its cage (e.g., calculating average daily activity and most frequently used facilities), generating a "behavioral analysis summary." Finally, these contents are automatically formatted and combined into a complete "pet boarding full-cycle health report" (such as a PDF document), which is output through the client terminal.

[0098] In alternative or modified implementations, "contact chain tracing" can employ a depth-first search (DFS) algorithm or use a more efficient community detection algorithm for approximate tracing within a large-scale data map. The threshold for triggering a "high-risk contact event" can be dynamically adjusted, for example, automatically lowered based on an increase in the current intensity of the community's infectious disease epidemic. "Enhanced infectious virus disinfection instructions" can be specific, such as dispatching a particular UV disinfection robot to designated coordinates to perform a task. The data structure of the "full-cycle health record" can adopt a document model from a NoSQL database (such as MongoDB) to better accommodate semi-structured and time-series data. The "behavioral analysis summary" in the report can incorporate a simple machine learning model to cluster behavioral patterns to identify behavioral labels such as "anxious" and "well-adjusted."

[0099] Figure 3 A schematic diagram of a cage management system for preventing the risk of virus transmission, provided in an embodiment of this application, is shown below. Figure 3 As shown, the livestock cage management system 300 for preventing the risk of virus transmission in this embodiment includes: an initial foster care information module 301, a livestock cage dynamic management module 302, and a health report generation module 303.

[0100] The initial boarding information module 301 is used to acquire multi-source health datasets of individual pets, construct digital risk profiles and perform intelligent access triage based on the multi-source health datasets, and generate an initial pet boarding information set. The cage dynamic management module 302 is used to generate a foster group dynamic management information set based on the initial pet fostering information set, by combining indirect contact risk link modeling with spatiotemporal trajectory and adaptive partition scheduling to minimize cross-infection risk based on dynamic adjustment of intelligent cage unit layout. The health report generation module 303 is used to perform automated intervention driven by contact risk link tracing based on the dynamic management information set of the fostering group, and output a full-cycle health report of pet fostering at the end of the fostering period.

[0101] The system in this embodiment can be used to execute the methods of any of the above embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.

Claims

1. A method for managing livestock cages to prevent the risk of viral transmission, characterized in that, include: Obtain multi-source health datasets of individual pets, construct digital risk profiles and intelligent access triage based on the multi-source health datasets, and generate initial information sets for pet boarding. Based on the initial pet boarding information set, an indirect contact risk link modeling combining spatiotemporal trajectories is performed, along with adaptive partition scheduling based on dynamic adjustment of intelligent cage unit layout to minimize the risk of cross-infection, thereby generating a dynamic management information set for the boarding group. Based on the dynamic management information set of the foster care group, automated intervention driven by contact risk tracing is carried out, and a full-cycle health report of the pet foster care is output at the end of the foster care period.

2. The method according to claim 1, characterized in that, The process of generating the initial pet boarding information set includes: The multi-source health dataset includes: pet basic profiles, recent health history records, and on-site multimodal biosafety screening data; Based on the multi-source health dataset, a digital risk profile is constructed for each boarding pet by extracting health factors and fusing risk rules, generating a dynamic risk level and individualized monitoring indicators for the pet. Based on the dynamic risk level of the pets, intelligent admission triage is performed through a preset pet admission strategy to determine the initial placement area for each boarding pet. The initial pet boarding information set is generated by combining the pet's dynamic risk level, the individualized monitoring indicators, and the initial placement area.

3. The method according to claim 2, characterized in that, The process of acquiring the on-site multimodal biosafety screening data includes: Guide pets through security checkpoints equipped with non-contact sensors, simultaneously collecting their infrared thermal imaging data and respiratory audio data; The infrared thermal imaging data is analyzed to identify abnormally high temperature areas and their temperature gradient distribution. The audio data of the breathing sounds are subjected to voiceprint feature extraction and abnormal sound pattern matching to identify whether there are abnormal breathing signs. The abnormally high temperature area and its temperature gradient distribution, together with the identification results of the abnormal respiratory symptoms, are used as the output of the on-site multimodal biosafety screening data.

4. The method according to claim 3, characterized in that, The process of constructing a digital risk profile for each boarding pet through the extraction of health factors and the fusion of risk rules includes: From the pet's basic profile and recent health history, vaccine expiration date, past medical history, and breed-specific susceptibility diseases are extracted as static health factors. From the on-site multimodal biosafety screening data, the maximum temperature difference value of the abnormal high temperature area and the type of the abnormal respiratory symptoms are extracted as dynamic health factors; The static health factors and the dynamic health factors are input into a preset risk assessment rule, and the risk assessment rule calculates an initial risk score based on factor weights and combination logic. Based on the current information on infectious pathogens prevalent in the community, the initial risk score is adjusted by environmental risk weighting, and the quantified dynamic risk level and the corresponding individualized monitoring indicators are output.

5. The method according to claim 2, characterized in that, The process of generating the dynamic management information set of the foster population includes: Based on the initial pet boarding information set, by analyzing the spatiotemporal trajectories of all boarded pets and records of shared facilities, indirect contact risk links are modeled to generate a real-time contact network map. Based on the real-time contact network map and the dynamic risk level of each pet, with the goal of minimizing the risk of cross-infection, adaptive partition scheduling calculation is performed to generate dynamic adjustment instructions for the layout of the intelligent animal cage unit. According to the dynamic adjustment instructions, the intelligent cage unit is controlled to perform physical separation or area reorganization, and the pet activity schedule for each pet is updated. The real-time contact network graph, the dynamic adjustment instructions, and the updated pet activity schedule are integrated and packaged into the dynamic management information set of the foster care group.

6. The method according to claim 5, characterized in that, The indirect contact risk link modeling includes: Continuously acquire the location tag signals worn by all boarding pets and the access control records of the smart cage unit to construct the original spatiotemporal trajectory in terms of time and location; Analyze the original spatiotemporal trajectories to identify direct contact events where pets directly coexist in the same physical space, as well as indirect contact events that occur through the sequential use of the same facility or through staff intermediaries. A risk weight is defined for each contact event, which is calculated based on the contact duration, contact type, and the dynamic risk level of both pets; Using pets and facilities as nodes and contact events with the aforementioned risk weights and timestamps as edges, a real-time contact network graph is constructed and updated in real time.

7. The method according to claim 5, characterized in that, The adaptive partition scheduling calculation includes: The core optimization objective is to minimize the sum of risk weights of all edges in the real-time contact network graph. The physical layout, connection relationship, and adjustable state of all the current intelligent cage units are used as spatial constraints. The isolation requirements, social needs, and activity levels of pets with different dynamic risk levels are used as policy constraints. Run the genetic optimization algorithm to find the optimal cage unit combination scheme, physical partition state and pet allocation scheme that satisfy all constraints, and encode them as the dynamic adjustment command.

8. The method according to claim 7, characterized in that, The control of the intelligent cage unit to perform physical separation or area reorganization includes: The central controller parses the dynamic adjustment instructions to obtain the target cage unit combination scheme and physical isolation status; Send control signals to the intelligent livestock cage unit involved, the control signals including: driving the mobile chassis of a specific livestock cage unit to reposition itself, controlling the opening or closing of the liftable / sliding partition between adjacent livestock cage units, and adjusting the opening and closing and airflow direction of the ventilation openings. During the physical separation or area reorganization process, each of the intelligent cage units provides real-time feedback on its position, separation status, and environmental parameters through built-in sensors, and the central controller performs closed-loop verification of the completed action. After successful verification, the global cage layout map is updated, and the actual execution results are fed back to the foster population dynamic management information set.

9. The method according to claim 5, characterized in that, The automated intervention driven by contact risk tracing, and the output of a full-cycle health report for pets at the end of the boarding period, include: The system monitors the dynamic management information set of the foster care group in real time. When a high-risk contact event triggered by a high-risk pet appears in the real-time contact network map, it automatically executes contact chain tracing and generates isolation guidance and enhanced disinfection instructions for the associated pets and facilities. Throughout the entire foster care period, all health status records, behavioral data, contact history, and intervention records of each pet are aggregated and synthesized into a structured full-cycle health record in chronological order. Based on the full-cycle health record, at the end of the boarding period, the pet boarding full-cycle health report is automatically generated and output. The pet boarding full-cycle health report includes: a health data timeline chart, contact safety certificate, and behavior analysis summary.

10. A livestock cage management system for preventing the risk of virus transmission, characterized in that, The method applied to any one of claims 1-9 includes: The initial boarding information module is used to obtain multi-source health datasets of individual pets, construct digital risk profiles and perform intelligent access triage based on the multi-source health datasets, and generate an initial pet boarding information set; The cage dynamic management module is used to generate a dynamic management information set for the fostering group by combining indirect contact risk link modeling based on the initial pet fostering information set with spatiotemporal trajectory and adaptive partition scheduling based on dynamic adjustment of intelligent cage unit layout to minimize cross-infection risk. The health report generation module is used to perform automated intervention driven by contact risk tracing based on the dynamic management information set of the foster care group, and output a full-cycle health report of the pet foster care at the end of the foster care period.

Images