Intelligent monitoring and early warning device for mouse-borne diseases based on multi-source data fusion

Abstract

The invention relates to the technical field of preventive medicine, and discloses a mouse-borne disease intelligent monitoring and early warning device based on multi-source data fusion, which comprises an environment sensor module, a mouse activity monitoring module, a pathogen detection module, a human case data acquisition module, a data fusion processing unit and an early warning and decision support unit, the data fusion processing unit is used for integrating and analyzing multi-source heterogeneous data from the above modules, and the early warning and decision support unit generates early warning information and prevention and control suggestions based on an analysis result. According to the method, the environment data, the mouse activity data, the pathogen detection data and the human case data are integrated, a multi-dimensional monitoring system is constructed, space-time correlation analysis is performed in combination with a deep learning algorithm, the prediction accuracy of disease outbreak is remarkably improved, the mouse recognition accuracy is high, and the pathogen detection specificity is excellent.

Description

Technical Field

[0001] The present invention relates to the technical field of preventive medicine, and in particular to an intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion. Background Art

[0002] In the field of preventive medicine, the monitoring and prevention of rodent-borne diseases (such as plague, hantavirus, and leptospirosis) are crucial components of public health and safety. The transmission mechanisms of these diseases involve complex interactions among multiple factors, including rodent population dynamics, environmental microecology, pathogen mutations, and human exposure risk. Therefore, building intelligent monitoring systems that encompass multi-source data has become a key technical direction in the industry.

[0003] The current mainstream technical solutions for rodent-borne disease monitoring are mainly based on single-dimensional data and traditional analysis methods. They rely on manual rodent density surveys (such as the night trap method and powder trace method) to obtain rodent population data, passively collect human case reports through public health agencies, and mostly use fixed-point temperature and humidity sensors for environmental parameter monitoring. However, there is a lack of linkage collection with geographic information, air quality and other data.

[0004] The inventors of the present application discovered that the existing technical solutions solely rely on mouse density or case data, resulting in the inability to capture key risk factors. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides an intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion, which solves the problem in the existing technical solutions that rely solely on rodent density or case data, resulting in the inability to capture key risk factors.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion, including an environmental sensor module, a rodent activity monitoring module, a pathogen detection module, a human case data acquisition module, a data fusion processing unit and an early warning and decision support unit; the data fusion processing unit is used to integrate and analyze multi-source heterogeneous data from the above modules, and the early warning and decision support unit generates early warning information and prevention and control recommendations based on the analysis results.

[0007] By adopting the above technical solutions, multi-dimensional environmental parameters such as temperature and humidity, air quality, and geographic information are collected in real time through the environmental sensor module. The intelligent infrared sensing and RFID tag tracking technologies of the rodent activity monitoring module are used to obtain the rodent density, activity trajectory and population dynamic characteristics. The real-time fluorescence quantitative PCR and other technologies of the pathogen detection module are used to achieve advance warning of pathogen mutations. The human case data collection module is connected to the public health system to realize real-time coupling analysis of case spatiotemporal data. The machine learning algorithm of the data fusion processing unit is used to perform feature correlation analysis on multi-source heterogeneous data and construct a risk prediction model. Finally, the early warning and decision support unit generates multi-level early warning information, visual risk heat maps and personalized prevention and control recommendations, and constructs a monitoring system covering the entire chain of "environment-rodent population-pathogen-human population". It realizes the early identification and dynamic tracking of key risk factors such as pathogen mutation, abnormal migration of rodent population, and environmental stress factors, breaking through the prediction limitations of a single data dimension and providing intelligent decision-making support for the precise prevention and control of rodent-borne diseases.

[0008] Preferably, the environmental sensor module includes a temperature and humidity sensor, an air quality sensor and a geographic information acquisition unit, which are used to collect environmental data in real time and dynamically compare it with preset thresholds.

[0009] Preferably, the rodent activity monitoring module uses infrared sensing technology and image recognition algorithm to monitor the activity trajectory, density and distribution range of rodents in real time, and transmits data to the data fusion processing unit through a wireless communication module.

[0010] Preferably, the pathogen detection module integrates a biosensor and a portable gene sequencing device to detect the types, concentrations and mutation characteristics of pathogens carried by rodents, and conducts real-time correlation analysis with the pathogen database.

[0011] Preferably, the human case data collection module is interconnected with the public health information system through an API interface to obtain and integrate the geographic location, symptom descriptions and epidemiological history data of confirmed and suspected cases in the region in real time.

[0012] Preferably, the data fusion processing unit uses a deep learning algorithm to construct a dynamic risk assessment model, performs spatiotemporal correlation analysis on multi-source heterogeneous data, and outputs a probability prediction of disease outbreaks and a heat map of high-risk areas.

[0013] Preferably, the dynamic risk assessment model is adaptive, automatically optimizes model parameters through real-time data feedback, and introduces transfer learning technology to adapt to different geographical environments and epidemic scenarios.

[0014] Preferably, the early warning and decision support unit includes a risk level classification module and a prevention and control strategy generation module; the risk level classification module divides the risk levels into low, medium and high levels according to the predicted probability, and the prevention and control strategy generation module recommends differentiated prevention and control measures based on the risk levels.

[0015] Preferably, differentiated prevention and control measures include targeted rodent control programs, environmental disinfection route planning, and vaccination priority allocation, and the simulation of the implementation effect of prevention and control measures is dynamically displayed through a visual interface.

[0016] Preferably, the device is also provided with edge computing nodes and a cloud collaborative processing architecture, the edge computing nodes are used for local real-time data processing, and the cloud collaborative processing architecture is used for global data modeling and cross-regional risk association analysis.

[0017] The present invention provides an intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion. It has the following beneficial effects:

[0018] 1. The present invention integrates environmental data, rodent activity data, pathogen detection data and human case data to build a multi-dimensional monitoring system, and combines deep learning algorithms to perform spatiotemporal correlation analysis, which significantly improves the accuracy of disease outbreak prediction, has high rodent identification accuracy, and excellent pathogen detection specificity.

[0019] 2. This invention optimizes parameters through an incremental learning framework and real-time data feedback, introducing transfer learning technology to adapt to different geographical environments and epidemic scenarios, thereby improving the robustness of the system. The target domain has low data requirements and high accuracy, significantly reducing cross-domain prediction errors.

[0020] 3. The present invention generates differentiated prevention and control measures based on risk level classification, and intuitively displays the execution effect through a three-dimensional heat map and a dynamic simulation interface. The activity of rodent control has decreased significantly, and the vaccination rate has increased, effectively suppressing the infection peak.

[0021] 4. This invention uses multiple encryption technologies to ensure data transmission and storage security, and protects sensitive information through permission control and de-identification. The system has high availability and data sharding is safe and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a system architecture diagram of the intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion of the present invention. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] Please see the attached Figure 1 An embodiment of the present invention provides an intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion, including an environmental sensor module, a rodent activity monitoring module, a pathogen detection module, a human case data collection module, a data fusion processing unit, and an early warning and decision support unit; the data fusion processing unit is used to integrate and analyze multi-source heterogeneous data from the above modules, and the early warning and decision support unit generates early warning information and prevention and control recommendations based on the analysis results.

[0025] Specifically, a monitoring and early warning system was constructed by integrating an environmental sensor module, a rodent activity monitoring module, a pathogen detection module, and a human case data collection module. The environmental sensor module is equipped with temperature and humidity sensors, air quality sensors, and a geographic information collection unit to monitor environmental parameters in real time and compare them with thresholds. The rodent activity monitoring module uses infrared sensing and image recognition technology to track rodent activity and density distribution. The pathogen detection module uses biosensors and portable gene sequencing devices to analyze the types and mutation characteristics of pathogens carried by rodents. The human case data collection module connects to the public health system to obtain data on the spatiotemporal distribution of cases. The multi-source heterogeneous data collected by these modules is uniformly transmitted to the data fusion processing unit for in-depth analysis.

[0026] The data fusion processing unit is the core of the device. It uses deep learning algorithms to construct a dynamic risk assessment model. By analyzing the spatiotemporal correlations of environmental data, rodent activity data, pathogen detection results, and human case data, it generates probability predictions for disease outbreaks and heat maps of high-risk areas. The model is adaptive, optimizing parameters through real-time data feedback and using transfer learning techniques to adapt to the characteristics of epidemics in different regions. After the analysis results are transmitted to the early warning and decision support unit, the system automatically divides the risk into three levels: low, medium, and high, based on the risk probability, and generates differentiated prevention and control recommendations: low-risk areas are recommended to strengthen environmental cleaning, medium-risk areas to initiate targeted rodent control and local disinfection, and high-risk areas to implement regional rodent control and vaccination priority management. All prevention and control measures are displayed through a visual interface to demonstrate the execution path and simulation results.

[0027] The architecture design adopts a collaborative edge computing and cloud computing approach. Edge nodes are responsible for real-time processing of local data to reduce response delays; the cloud performs global data modeling and cross-regional risk analysis to ensure the comprehensiveness and scalability of the early warning system. Multi-source data fusion improves early warning accuracy, adaptive models enhance scenario adaptability, and edge-cloud collaboration optimizes computing efficiency. Typical application scenarios include urban community rodent pest monitoring and rural epidemic early warning. For example, when the system detects that the humidity in a certain area exceeds the standard and the rodent density increases sharply, it will automatically trigger an early warning based on the pathogen detection results, and mark the high-risk range on the map to provide accurate decision-making support for the disease control department. The attached figure shows in detail the key designs such as the device architecture, risk assessment heat map and prevention and control strategy visualization interface, and fully presents the closed-loop management of the entire process from data collection to early warning output.

[0028] The environmental sensor module includes a temperature and humidity sensor, an air quality sensor, and a geographic information acquisition unit, which is used to collect environmental data in real time and dynamically compare it with preset thresholds.

[0029] Specifically, a three-dimensional monitoring network is formed by temperature and humidity sensors (DHT22 series, measuring range -20°C to 60°C, accuracy ±0.5°C), multi-component air quality sensors (capable of detecting harmful gases such as PM2.5 / PM10, CO2, and NH3 / H2S), and a multi-functional geographic information acquisition unit (GPS + Beidou dual-mode positioning, accuracy ±1.5m). These sensors transmit data in real time via a hybrid I2C / LoRa communication protocol, triggering graded alarms when persistent high humidity (>85%) or abnormal temperature fluctuations are detected.

[0030] The air quality monitoring subsystem utilizes a laser scattering (SDS011) sensor in conjunction with an electrochemical sensor. This sensor not only captures real-time changes in particulate matter concentrations (ranging from 0 to 1000 μg / m³) in high-risk areas such as garbage stations and sewers, but also identifies rodent concentrations through the presence of characteristic gases (such as NH3). All environmental data is deeply integrated with the GIS (Geographic Information System). Using RTK positioning technology and IMU inertial compensation, the system accurately annotates the three-dimensional geographic features (latitude and longitude, altitude, and terrain type) of monitoring points and automatically links them to spatial risk models derived from historical epidemic data. The system's built-in adaptive algorithm dynamically optimizes parameter thresholds based on machine learning, for example, automatically lowering humidity sensitivity thresholds during the rainy season or enhancing rodent migration warnings in low-temperature winter environments.

[0031] When the module is deployed at a city sewer monitoring point, temperature and humidity sensors track real-time changes in the pipeline microenvironment, while air quality sensors simultaneously detect sudden increases in NH3 concentrations. Combined with the "underground confined space" attribute annotated by the GIS system, the data fusion unit completes a risk assessment within 10 seconds. If humidity levels exceed 90% and NH3 levels exceed 50 ppm for 72 consecutive hours, the system automatically generates a "high risk of rodent infestation" alert and identifies key control areas within a 500-meter radius using a three-dimensional heat map. This multi-source data collaborative analysis mechanism improves environmental monitoring accuracy by over 40% compared to traditional methods, providing critical spatial and temporal data support for subsequent pathogen transmission predictions.

[0032] The rodent activity monitoring module uses infrared sensing technology and image recognition algorithms to monitor the trajectory, density and distribution range of rodent activity in real time, and transmits data to the data fusion processing unit through the wireless communication module.

[0033] Specifically, the infrared sensing array uses a third-generation pyroelectric sensor (PIR) combined with a Fresnel lens, which can detect mouse motion within a radius of 15 meters (sensitivity 0.3-3m / s), and effectively suppress environmental thermal noise through differential signal processing technology; a 2-megapixel starlight-level camera (0.001Lux) combined with an 850nm infrared fill light ensures the quality of night monitoring images, and is equipped with a dedicated DSP chip to achieve H.265 efficient encoding; the intelligent analysis unit is based on the deeply optimized YOLOv5s algorithm (model size <8MB), and completes mouse recognition (accuracy ≥95%), multi-target trajectory tracking and behavior pattern analysis at the edge.

[0034] The system utilizes LoRa and 4G dual-mode wireless transmission, enabling data transmission within a 3km radius in urban environments. Built-in encryption and compression algorithms ensure secure and efficient data transmission. When the infrared array triggers motion detection, the vision system captures high-definition images (1920×1080@30fps). The edge computing unit completes target recognition within 200ms and generates a data packet containing elements such as population counts, activity heat maps, and spatiotemporal distribution matrices. This data is ultimately uploaded to the data center using adaptive frequency hopping technology. The module has been specifically optimized for anti-interference capabilities and can operate continuously for 30 days in ambient temperatures of -20°C to 60°C (IP67 protection rating).

[0035] The pathogen detection module integrates biosensors and portable gene sequencing devices to detect the types, concentrations and mutation characteristics of pathogens carried by rodents, and conducts real-time correlation analysis with the pathogen database.

[0036] Specifically, the pathogen detection module integrates biosensors and portable gene sequencing devices, using a dual-mode design of immunomagnetic bead sensors (based on the principle of antigen-antibody specific binding, which can quickly capture characteristic proteins of pathogens such as Yersinia pestis and Hantavirus, with a detection limit of 10³CFU / mL and a detection time of less than 15 minutes) and nucleic acid aptamer sensors (using chemically modified aptamers to identify pathogen nucleic acids, combined with electrochemiluminescence technology to achieve quantitative detection of multi-pathogen complex infections, with a specificity of 99.7%). Rodent samples are collected through a portable sampler and automatically lysed and purified by a microfluidic chip before being introduced into the detection chamber; at the same time, the portable gene sequencing device equipped with nanopore sequencing technology supports single-molecule real-time long fragment sequencing, without the need for PCR amplification, and can complete the identification of 23 major rodent-borne pathogens listed by the WHO within 1 hour. Combined with the digital PCR module, it can achieve 10²-10 8 It can achieve absolute quantification of nucleic acids with a precision error of <5% in copies / μL, and can compare in real time with the cloud-based pathogen database (supports local deployment or docking with public databases, updated ≥2 times a day, including drug-resistant gene databases and phylogenetic analysis tools), completing triple analysis of pathogen types, concentrations, and variation characteristics (such as SNPs, InDels, drug-resistant mutations, and antigenic drift); detection data is transmitted in real time via 5G satellite communications, and combined with temperature, humidity, air quality data and rodent activity trajectory data from environmental sensors to establish an environment-pathogen-rodent dynamic coupling model. This can provide full-chain data support from rapid screening to in-depth genetic analysis for early warning and precise prevention and control of rodent-borne diseases in scenarios such as field epidemic monitoring (such as 4-hour full-process detection of natural plague foci, which is 80% faster than traditional methods) and urban emergency response (such as dynamic monitoring of pathogens around garbage treatment plants).

[0037] The human case data collection module is connected to the public health information system through the API interface to obtain and integrate the geographic location, symptom descriptions and epidemiological history data of confirmed and suspected cases in the region in real time.

[0038] Specifically, the human case data collection module is interconnected with the public health system through the HTTPS security interface using the RESTfulAPI architecture, and authorized access is ensured through OAuth2.0 authentication. At the same time, it uses data adapters to be compatible with multiple source systems such as hospital HIS and disease control databases, and supports two data acquisition modes: Webhook real-time push and timed polling with an interval of no more than 5 minutes.

[0039] The collected data covers multi-dimensional information such as the longitude and latitude coordinates of the cases (corrected by Beidou / GPS to an accuracy of no more than 10 meters), symptoms coded in ICD-11 (such as fever, bleeding, etc.), and epidemiological history (including frequency of rodent contact, travel history to epidemic areas, and clustering characteristics).

[0040] In terms of data processing, AES-256 encryption is used to de-identify sensitive information, retaining only anonymous IDs for data association. In accordance with the principle of data minimization, irrelevant fields are eliminated. Data cleaning and semantic unification are also achieved through logical verification (such as geographic location validity and timeline consistency checks), marking of missing values for key fields with a missing rate of more than 30% for review, and use of dictionaries to convert unstructured text into standardized terms.

[0041] This module will spatiotemporally overlay case data with environmental temperature and humidity, and rodent density heat maps, and use kernel density estimation (KDE) to generate a "case-environment-rodent population" coupled risk layer. With the help of a case-control algorithm, it will identify risk factors such as a 5-fold increase in infection risk when contact with rodents more than 3 times a week, and will correlate pathogen resistance mutations with treatment failure cases in real time, constructing a "resistant strain-case" map to assist in adjusting prevention and control and clinical strategies.

[0042] The data fusion processing unit uses deep learning algorithms to build a dynamic risk assessment model, conducts spatiotemporal correlation analysis on multi-source heterogeneous data, and outputs probability predictions of disease outbreaks and heat maps of high-risk areas.

[0043] Specifically, the data fusion processing unit uses deep learning algorithms to build a dynamic risk assessment model, conducts comprehensive and in-depth spatiotemporal correlation analysis on multi-source heterogeneous data, and ultimately outputs accurate disease outbreak probability predictions and heat maps of high-risk areas.

[0044] The working mechanism is as follows: first, multi-source heterogeneous data is integrated, covering human case data (such as geographic location accurate to longitude and latitude, structured symptom descriptions, and detailed epidemiological history), pathogen detection data (pathogen type, concentration, and variation characteristics), environmental data (temperature, humidity, air quality), and rodent activity data (activity trajectory, density). This data is then preprocessed through cleaning, standardization, and feature extraction. In terms of algorithm and model construction, recurrent neural networks such as long short-term memory (LSTM) and gated recurrent units (GRU) are used to process time series data, combined with convolutional neural networks (CNN) to process spatial data. An attention mechanism is also used to enhance focus on important features. Model training and tuning are accomplished by partitioning the dataset into training, validation, and test sets, and defining appropriate loss functions and optimizers.

[0045] Spatiotemporal correlation analysis analyzes data trends and the chronological relationships between events at different time scales. Spatial correlations are calculated across regions to identify spatial clusters. The model outputs the probability of disease outbreaks in different regions over the next period of time, along with confidence intervals. Furthermore, the model integrates with geographic information systems to generate heat maps of high-risk areas, using color saturation to indicate risk.

[0046] The dynamic risk assessment model is adaptive, automatically optimizing model parameters through real-time data feedback, and introducing transfer learning technology to adapt to different geographical environments and epidemic scenarios.

[0047] Specifically, the dynamic risk assessment model constructs an online learning architecture through an incremental learning framework, uses a real-time data buffer pool to update parameters using mini-batch gradient descent, triggers adaptive retraining through an error monitoring loop, combines an attention mechanism with geographically weighted regression to achieve dynamic adjustment of spatiotemporal weights, and sorts feature importance in real time based on SHAP values to eliminate redundant features. At the same time, it adopts a layered migration architecture, freezes the underlying CNN / RNN layers to reuse common features, fine-tunes top-level parameters to adapt to localized risk factors, and combines adversarial transfer learning (minimizing cross-domain distribution differences through domain discriminators and maximum mean differences) with meta-learning (MAML algorithm pre-training for fast convergence). Realize cross-domain knowledge reuse, dynamically adjust temperature and humidity weight coefficients in the dry north / humid south, and call historical similar event feature representations in the scenario of newly emerging recombinant pathogens; relying on the edge-cloud collaborative architecture, the edge node is responsible for local data preprocessing and lightweight parameter updates (delay <5 minutes), and the cloud generates a global optimized version by integrating cross-regional models through federated learning; in terms of performance, the parameter convergence of the data distribution mutation scenario is ≤12 hours (error reduction >30%), the accuracy rate is ≥85% (traditional model 60%) with 10% data volume in the target field, and the cross-domain prediction RMSE is <0.2 (58% lower than the non-migration model), which significantly improves the robustness and applicability of early warning in complex scenarios.

[0048] The early warning and decision support unit includes a risk level classification module and a prevention and control strategy generation module; the risk level classification module divides the risk levels into low, medium and high levels according to the predicted probability, and the prevention and control strategy generation module recommends differentiated prevention and control measures based on the risk level.

[0049] Specifically, the early warning and decision support unit includes a risk level classification module and a prevention and control strategy generation module. The risk level classification module takes the disease outbreak prediction probability output by the dynamic risk assessment model as the core, and combines regional disease historical data and the experience of public health experts to set differentiated probability thresholds (such as low risk 5%-20%, medium risk 20%-50%, high risk >50%) to divide the risk levels into low, medium and high. It also uses map visualization technology to mark different risk areas with green, yellow and red colors to intuitively present the risk distribution and predicted probability. The prevention and control strategy generation module follows the principles of differentiation, comprehensive measures and feasibility, and formulates measures to strengthen monitoring and early warning, health education and environmental sanitation improvement for low-risk areas. Professional rodent control, vaccination of key populations and health monitoring of key populations are implemented in medium-risk areas. High-risk areas adopt strategies such as personnel and material control, full-area disinfection and sterilization, and concentrated medical resources for treatment. All prevention and control measures are dynamically adjusted with the risk level. For example, after the risk is downgraded, the control scope is gradually reduced and the frequency of disinfection is reduced, thereby providing accurate and dynamic decision support for the prevention and control of rodent-borne diseases.

[0050] Differentiated prevention and control measures include targeted rodent control programs, environmental disinfection route planning, and vaccination priority allocation, and the implementation effect simulation of prevention and control measures is dynamically displayed through a visual interface.

[0051] Specifically, the targeted rodent control plan: Based on the rodent activity trajectory and pathogen detection data, spatial autocorrelation analysis is used to identify rodent gathering hotspots (such as areas with rat density >8 rats / hectare and virus infection rate >30%), and a 50m×50m rodent control priority grid map is generated; a "bait station + smart trap" combination is used in high-risk areas (the bait contains pheromone attractants, the traps report capture data in real time, and the capture rate is ≥92%), silent mousetrap cages are used in residential areas, bromadiolone is placed in closed areas based on meteorological conditions, and owl roosting boxes are introduced into farmland for ecological rodent control; within 48 hours after rodent control, the effect is evaluated using the dual indicators of rodent activity (a decrease of >60%) and virus infection rate (a decrease of >50%), and a report is automatically generated.

[0052] Environmental disinfection route planning: According to the risk level and environmental characteristics, three types of disinfection areas are divided into high (100-meter radius of the case's residence), medium (community main roads), and low (parks and green spaces). Atomizing disinfection vehicles (twice a day, effective chlorine 1000-2000 mg / L), backpack sprayers (once a day, concentration reduced by half), and preventive disinfection (once a week) are used respectively; the disinfection route is optimized based on the VRP algorithm, and parameters such as geographical accessibility, residents' travel peak, and agent volatilization conditions are input to complete the coordinated operation of multiple vehicles within 6 hours; the visual interface displays the trajectory and coverage of the disinfection vehicle in real time, and the change of NH3 concentration is superimposed to verify the effect. A "disinfection-pathogen attenuation model" is established to simulate the disinfection time.

[0053] Vaccination priority allocation: A two-dimensional "risk-vulnerability" assessment model is constructed. People with a combined score of ≥120 for infection risk factors (such as 100 points for being less than 50 meters away from high-risk areas and 80 points for daily contact with rodents) and health vulnerability factors (such as 50 points for age >65 and 40 points for diabetes) are classified as first-level priority groups; mobile vaccination units are set up in high-risk areas for "door-to-door service", and medium-risk areas are guided by smart appointments in different time periods (20% of the numbers are reserved), and vaccine transportation uses blockchain traceability cold chain temperature (2-8°C); based on the SEIR model to simulate the vaccination effect, a vaccination rate of >90% for the first-level population can reduce the infection peak by 75%, and the dynamic curve shows the correlation between vaccination progress and risk reduction (R²=0.91).

[0054] Dynamic display of the visualization interface: Based on CityEngine, a city-level 3D model is generated, integrating multi-source data such as rodent heat maps, case scatter points, and disinfection vehicle trajectories, and supports zooming to individual buildings; simulation functions include overlaying heat maps of density reduction before and after rodent control, disinfection path playback (displaying disinfectant deposition and inactivation rate), and vaccination progress Sankey diagram (calculating the rate of achievement of the herd immunity threshold); it supports manual adjustment of prevention and control parameters (such as disinfection frequency) and real-time simulation of risk changes. It has a built-in library of 12 scenario plans (such as rodent migration after heavy rain), and you can load preset measures combinations with a click.

[0055] The device also has edge computing nodes and cloud-based collaborative processing architecture. The edge computing nodes are used for local real-time data processing, and the cloud-based collaborative processing architecture is used for global data modeling and cross-regional risk correlation analysis.

[0056] Specifically, the edge computing node is a localized real-time data processing center that uses embedded hardware (such as NVIDIA Jetson Nano, equipped with a quad-core Cortex-A57 CPU, a 128-core Maxwell GPU, supporting a wide operating temperature range of -20°C to 60°C and IP67 protection) and the Ubuntu IoT + ROS 2 software stack. The TensorFlow Lite model is deployed through the Edge Impulse framework to achieve 200ms-level data preprocessing; sliding filtering and aggregation of high-frequency data such as temperature and humidity, differential threshold detection (compression ratio 8:1) is used for infrared data of rodent activity, and the YOLOv5s-tiny model (accuracy ≥95%) and the Isolation Forest algorithm are deployed to identify anomalies; local rodent traps / disinfection equipment can be directly controlled (response delay <5 minutes), 72 hours of data (about 5GB) can be cached when the network is interrupted, and breakpoint resuming is supported.

[0057] Cloud-based collaborative processing architecture: Global modeling and cross-regional analysis core, the cloud is deployed based on Kubernetes containers, and the underlying layer uses Ceph distributed storage and NVIDIA A100 GPU clusters (peak 10PFLOPS computing power) to build a rodent-borne disease data lake (supporting PB-level storage and second-level queries); uses the spatial Doberman model to analyze cross-regional risk spillover effects (such as the impact coefficient of rat density in City A on the surrounding City B 200 kilometers away), and uses the Transformer model to input 10 years of historical data to predict quarterly epidemic trends; uses federated learning to jointly train edge nodes for global model training, and distributes model updates via encrypted OTA every week (deployment delay <30 minutes).

[0058] With collaborative workflow and technical advantages, the edge layer completes 90% of real-time data filtering and simple analysis (only key features are uploaded, and bandwidth consumption is reduced by 70%), and the cloud is responsible for cross-regional modeling and trend prediction (such as data modeling response across 5 provinces is less than 2 hours); in earthquake-stricken areas, edge nodes are disconnected from the network and operate independently. After being connected to the network, the cloud aggregates data to identify the homology of strains (such as 98% homology); during the migratory season of migratory birds, monitoring along the way is triggered through spatiotemporal intersection analysis (such as setting up infrared points within 50 kilometers of migratory bird stopovers); the system supports access to 100,000 edge nodes (throughput 10TB / day), edge decision delay is less than 500ms, and cloud availability reaches 99.99%.

[0059] For data security and privacy protection, edge and cloud communications use TLS1.3+National Secret SM4 encryption (data sharding <1KB), and sensitive data is ciphertext modeled through homomorphic encryption; fine-grained permissions are allocated based on the RBAC system (such as edge operations only viewing local data), edge nodes have dual power redundancy, and cloud multi-active data centers ensure service continuity, ensuring the security and compliance of the entire process of data transmission, storage, and access.

[0060] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion, characterized in that: It includes an environmental sensor module, a rodent activity monitoring module, a pathogen detection module, a human case data collection module, a data fusion processing unit, and an early warning and decision support unit; the data fusion processing unit is used to integrate and analyze multi-source heterogeneous data from the above modules, and the early warning and decision support unit generates early warning information and prevention and control recommendations based on the analysis results.

2. The intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion according to claim 1 is characterized in that: The environmental sensor module includes a temperature and humidity sensor, an air quality sensor and a geographic information acquisition unit, which is used to collect environmental data in real time and dynamically compare it with preset thresholds.

3. The intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion according to claim 1 is characterized in that: The rodent activity monitoring module uses infrared sensing technology and image recognition algorithm to monitor the activity trajectory, density and distribution range of rodents in real time, and transmits data to the data fusion processing unit through the wireless communication module.

4. The intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion according to claim 1 is characterized in that: The pathogen detection module integrates a biosensor and a portable gene sequencing device to detect the types, concentrations and mutation characteristics of pathogens carried by rodents, and conducts real-time correlation analysis with the pathogen database.

5. The intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion according to claim 1 is characterized in that: The human case data collection module is interconnected with the public health information system through an API interface to obtain and integrate the geographic location, symptom descriptions and epidemiological history data of confirmed and suspected cases in the region in real time.

6. The intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion according to claim 1 is characterized in that: The data fusion processing unit uses a deep learning algorithm to build a dynamic risk assessment model, performs spatiotemporal correlation analysis on multi-source heterogeneous data, and outputs a probability prediction of disease outbreaks and a heat map of high-risk areas.

7. The intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion according to claim 1 is characterized in that: The dynamic risk assessment model is adaptive, automatically optimizing model parameters through real-time data feedback, and introducing transfer learning technology to adapt to different geographical environments and epidemic scenarios.

8. The intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion according to claim 1 is characterized in that: The early warning and decision support unit includes a risk level classification module and a prevention and control strategy generation module; the risk level classification module divides the risk levels into low, medium and high levels according to the predicted probability, and the prevention and control strategy generation module recommends differentiated prevention and control measures based on the risk levels.

9. The intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion according to claim 1 is characterized in that: Differentiated prevention and control measures include targeted rodent control programs, environmental disinfection route planning, and vaccination priority allocation, and the implementation effect simulation of prevention and control measures is dynamically displayed through a visual interface.

10. The intelligent monitoring and early warning device for rodent-borne diseases based on multi-source data fusion according to claim 1 is characterized in that: The device is also equipped with edge computing nodes and a cloud-based collaborative processing architecture. The edge computing nodes are used for local real-time data processing, and the cloud-based collaborative processing architecture is used for global data modeling and cross-regional risk association analysis.

Images