Fire monitoring method and system for scenic spot based on internet of things big data

Abstract

The application discloses a scenic spot fire monitoring method and system based on Internet of Things big data, and relates to the technical field of public safety.The method comprises the following steps: acquiring real-time multi-source heterogeneous data in a scenic spot, and constructing a dynamic digital twin model; performing predictive deduction to generate a scenic spot risk evolution heat map in a future time period; performing multi-strategy simulation scheduling evaluation to determine an optimal monitoring resource scheduling instruction set; synchronously collecting scenic spot field feedback data collected by the scenic spot monitoring resources after executing the optimal scheduling instruction set; and iteratively optimizing the tourist behavior rule base and risk propagation model using a deviation value.The application solves the problems of static rigidity of monitoring resource configuration, difficulty in coping with dynamic risk changes, existence of monitoring blind spots and lag of emergency resource scheduling in the prior art by constructing a dynamic digital twin model for risk prediction, performing multi-strategy resource scheduling evaluation and establishing a closed-loop optimization mechanism.

Description

Technical Field

[0001] This invention relates to the field of public safety technology, and in particular to a method and system for fire monitoring in scenic areas based on Internet of Things big data. Background Technology

[0002] With the deepening of smart tourism development and the widespread application of IoT technology, scenic area safety management is moving towards digitalization and intelligence. In particular, fire safety management in scenic areas, as a crucial link in ensuring the safety of tourists' lives and property, is receiving increasing attention. By deploying various sensors and monitoring equipment to build an IoT system for scenic areas, real-time collection of environmental data and tourist behavior is achieved, providing a data foundation for intelligent fire monitoring.

[0003] Existing fire monitoring systems in scenic areas typically monitor fire conditions by deploying fire detectors and surveillance cameras at fixed locations, combined with manual inspections. These systems mainly rely on a trigger-based alarm mechanism based on preset thresholds, issuing an alarm when sensors detect abnormal temperatures or excessive smoke concentrations.

[0004] However, existing static monitoring systems have significant limitations. First, because the locations and ranges of monitoring points are fixed, the system struggles to adapt to dynamic changes in fire risk caused by factors such as tourist movement and seasonal variations, resulting in a mismatch between static resource allocation and dynamic risk evolution. Second, when a sudden event causes a surge in risk in a certain area, fixed monitoring equipment cannot autonomously adjust its coverage, creating monitoring blind spots and ineffective resource coverage. More importantly, traditional resource allocation relies on manual decision-making, making it difficult to respond promptly to rapidly changing risk situations, leading to a disconnect between emergency resource allocation and real-time risk conditions. These problems make it difficult for existing systems to achieve accurate and efficient fire risk prevention and control in the face of complex and ever-changing scenic environments. Summary of the Invention

[0005] This application provides a scenic area fire monitoring method and system based on Internet of Things big data. It solves the problems in the prior art where static monitoring systems are difficult to adapt to dynamic risk changes, cannot eliminate monitoring blind spots, and are disconnected from real-time risk status. It achieves the technical effect of building an adaptive and fully covered scenic area fire monitoring system through dynamic risk prediction and intelligent resource scheduling.

[0006] This application provides a method for fire monitoring in scenic areas based on Internet of Things (IoT) big data. The method is applied to a fire monitoring system for scenic areas based on IoT big data, including: acquiring real-time multi-source heterogeneous data within the scenic area, and constructing a dynamic digital twin model that is synchronized with the spatiotemporal state of the scenic area based on the real-time multi-source heterogeneous data. Based on the dynamic digital twin model, predictive simulations are performed to generate a heat map of the scenic area's risk evolution over future time periods. Based on the aforementioned risk evolution heatmap, perform multi-strategy simulation scheduling evaluation to determine the optimal set of monitoring resource scheduling instructions; The optimal scheduling instruction set is executed to drive the deployment of scenic area monitoring resources, and the on-site feedback data of the scenic area monitoring resources collected after the optimal scheduling instruction set is executed is collected simultaneously. The collected on-site feedback data of the scenic area is compared with the scenic area risk evolution heat map to generate a prediction-on-site deviation value, and the deviation value is used to iteratively optimize the tourist behavior rule base and risk propagation model.

[0007] Furthermore, the steps for constructing a dynamic digital twin model synchronized with the spatiotemporal state of the scenic area based on the aforementioned real-time multi-source heterogeneous data include: Obtain multi-source numerical sequences within the scenic area and preprocess the multi-source numerical sequences; The preprocessed multi-source numerical sequence is input into 3D modeling software to generate a scenic area terrain mesh model, and a real-time data layer is superimposed to form a spatiotemporally synchronized representation, resulting in a dynamic digital twin model. The spatiotemporal synchronization representation is updated using a particle filter algorithm.

[0008] Furthermore, the steps for performing predictive simulations and generating a heatmap of scenic area risk evolution over future time periods include: Extract the current spatiotemporal state data structure from the dynamic digital twin model; Load the pre-defined tourist behavior rules library and risk propagation model; By combining the current spatiotemporal state data structure with the tourist behavior rule base, the future tourist location distribution sequence is simulated and calculated. The future tourist location distribution sequence is input into the risk propagation model, the fire source diffusion equation is solved to generate the fire source propagation path sequence, and the smoke propagation equation is solved to generate the smoke concentration change sequence. By integrating the fire source propagation path sequence, smoke concentration change sequence, and future tourist location distribution sequence, a risk probability grid is calculated. The risk probability grid is extended over time to generate a risk evolution sequence covering future time periods; The risk evolution sequence is converted into a heatmap representation, where the color gradient of the heatmap corresponds to the risk value level.

[0009] Furthermore, the steps for performing multi-strategy simulation scheduling evaluation to determine the optimal set of monitoring resource scheduling instructions include: Extract the future risk distribution matrix from the aforementioned risk evolution heatmap; Query the state vectors of fixed monitoring resources and mobile monitoring resources within the scenic area; Generate multiple sets of scheduling strategies, simulate the execution of each set of scheduling strategies, and calculate the coverage and response time metrics after scheduling. Compare the coverage metric and the response time metric, and select the scheduling strategy set with the largest sum of the metrics as the candidate instruction set; The candidate instruction set is subjected to constraint checks, and the candidate instruction set is adjusted until all constraints are satisfied, thereby generating the final optimal monitoring resource scheduling instruction set.

[0010] Furthermore, the step of synchronously collecting the on-site feedback data of the scenic area monitoring resources after executing the optimal scheduling instruction set includes: Send reconfiguration commands to fixed monitoring resources and path planning commands to mobile monitoring resources; Monitor the status of the commands executed by the fixed and mobile monitoring resources; Collect and execute multi-source numerical sequences after executing the optimal scheduling instruction set; The multi-source numerical sequence after executing the optimal scheduling instruction set is preprocessed to obtain the preprocessed multi-source numerical sequence. The preprocessed multi-source numerical sequences are aggregated to form a scenic area on-site feedback dataset.

[0011] Further steps in generating the predicted-site deviation value include: Extract the scenic area on-site data structure from the scenic area on-site feedback dataset; Extract the predicted spatiotemporal state vector for the corresponding time point from the aforementioned scenic area risk evolution heat map; The temperature deviation component, humidity deviation component, smoke deviation component, gas deviation component, and tourist deviation component are used to generate the overall prediction-site deviation value.

[0012] Furthermore, the step of iteratively optimizing the tourist behavior rule base and risk propagation model using the deviation value includes: Load the current tourist behavior rule base and the current risk propagation model; The prediction-reality deviation value is decomposed into a behavioral deviation component and a propagation deviation component; For the aforementioned behavioral deviation component, the gradient descent algorithm is applied to adjust the element values ​​of the tourist movement probability transition matrix and the element values ​​of the tourist aggregation threshold matrix. For the propagation deviation component, the least squares method is applied to update the coefficient values ​​of the fire source diffusion equation parameter set and the smoke propagation equation parameter set. Verify the adjusted and updated tourist behavior rule base and risk propagation model.

[0013] Furthermore, the steps for simulating and calculating the future tourist location distribution sequence include: Obtain the current tourist distribution state tensor from the dynamic digital twin model; Load the spatiotemporal transition probability tensor from the tourist behavior rule base; Establish an improved Markov prediction equation: ; In the formula, This represents the tourist distribution vector at the current moment. Indicates the future The predicted tourist distribution vector at any given time. This represents the transition probability matrix that changes over time. This represents the field vector representing the tourist's movement velocity. Represents the diffusion coefficient matrix. Represents the tourist density gradient vector. This indicates normalization processing.

[0014] Furthermore, the steps for calculating the risk probability grid include: The flame intensity and thermal radiation flux of each grid cell are extracted from the fire source propagation path sequence. The visibility impact index and toxic gas concentration index of each grid cell were extracted from the smoke concentration change sequence; Extract the population density index and evacuation difficulty index of each grid cell from the future tourist location distribution sequence; Establish a multi-factor risk fusion equation: ; In the formula, where This represents the final risk value of the grid cell. Indicates the flame intensity index. Indicates the distance to the fire source. Indicates the attenuation coefficient. Indicates smoke concentration index, Indicates visibility index, Indicates the baseline visibility. Indicators representing the concentration of toxic gases Indicates the baseline concentration. Indicators representing population density Indicates the evacuation difficulty coefficient. , , , This represents the weighting coefficient of each indicator. It is a natural constant.

[0015] This application provides a scenic area fire monitoring system based on Internet of Things (IoT) big data, which is used to implement a scenic area fire monitoring method based on IoT big data, including: Model building module, heatmap generation module, instruction set determination module, resource deployment module, iterative optimization module; The model building module is used to acquire real-time multi-source heterogeneous data within the scenic area and to build a dynamic digital twin model that is synchronized with the spatiotemporal state of the scenic area based on the real-time multi-source heterogeneous data. The heat map generation module is used to perform predictive inference based on the dynamic digital twin model and generate a heat map of the scenic area risk evolution over a future time period. The instruction set determination module is used to perform multi-strategy simulation scheduling evaluation based on the risk evolution heatmap to determine the optimal monitoring resource scheduling instruction set; The resource deployment module is used to execute the optimal scheduling instruction set to drive the deployment of scenic area monitoring resources, and simultaneously collect the on-site feedback data of the scenic area monitoring resources after the execution of the optimal scheduling instruction set. The iterative optimization module is used to compare the collected on-site feedback data of the scenic area with the risk evolution heat map of the scenic area, generate a prediction-on-site deviation value, and use the deviation value to iteratively optimize the tourist behavior rule base and risk propagation model.

[0016] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: By sensing factors such as tourist distribution and environmental changes in real time, and predicting future fire risk areas, the monitoring work has shifted from passive response to proactive early warning, significantly improving the timeliness and accuracy of risk perception. Furthermore, based on risk prediction, through multi-strategy simulation and scheduling evaluation, the optimal monitoring resource scheduling plan is automatically generated. This plan directs fixed cameras to adjust their viewing angles and mobilizes mobile monitoring resources to cover high-risk areas, proactively filling monitoring efficiency gaps caused by changes in risk. This upgrades the scenic area monitoring network from static deployment to dynamic optimization. Moreover, during the execution of the scheduling plan, on-site feedback data is continuously collected and compared with prediction results. Deviation values ​​are used to continuously optimize behavioral rules and risk models. This iterative optimization mechanism ensures continuous adaptation to changes in the scenic area environment, gradually improving the intelligence level and long-term reliability of the entire fire monitoring system. Attached Figure Description

[0017] Figure 1 A flowchart of a scenic area fire monitoring method based on Internet of Things big data provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a scenic area fire monitoring system based on Internet of Things big data, provided in an embodiment of this application. Detailed Implementation

[0018] This application provides a method and system for fire monitoring in scenic areas based on IoT big data. It solves the problems in the prior art such as the mismatch between the static configuration of monitoring resources and the dynamic evolution of risks, the existence of monitoring efficiency gaps, and the conflict between the pre-positioning of emergency resources and real-time scheduling. By constructing a dynamic digital twin model and performing predictive simulations, executing multi-strategy simulation scheduling evaluation and dynamic resource deployment, and establishing a closed-loop iterative optimization mechanism based on feedback data, it achieves accurate prediction of fire risks in scenic areas, intelligent scheduling and adaptive optimization of monitoring resources, and improves the initiative and reliability of fire monitoring in scenic areas.

[0019] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0020] like Figure 1 As shown, this application provides a scenic area fire monitoring method based on Internet of Things (IoT) big data. The method is applied to a scenic area fire monitoring system based on IoT big data, including: acquiring real-time multi-source heterogeneous data generated by IoT sensors, video streams and ticketing systems in the scenic area, and constructing a dynamic digital twin model that is synchronized with the spatiotemporal state of the scenic area based on the real-time multi-source heterogeneous data. Real-time multi-source heterogeneous data includes temperature and humidity data collected by temperature and humidity sensors, smoke concentration data collected by smoke sensors, gas composition data collected by gas sensors (such as those detecting carbon monoxide and carbon dioxide), video and audio data collected by surveillance cameras deployed in various areas of the scenic area, and visitor entry timestamp sequences, visitor count sequences, and visitor location sequences obtained through ticketing systems and entrance gates.

[0021] Based on the dynamic digital twin model, a pre-set tourist behavior rule base and risk propagation model are integrated to perform predictive simulations and generate a heat map of scenic area risk evolution over future time periods. The pre-defined tourist behavior rule base was established by analyzing historical monitoring data and ticketing data of the scenic area. The tourist movement probability transfer matrix was obtained by statistical analysis of historical tourist trajectory data, and the tourist gathering threshold matrix was set based on historical population density data and safety management regulations for each area. The risk propagation model was established based on the principles of fire science. The fire source diffusion equation adopted the classic forest fire spread model of Rothermel et al., and the smoke propagation equation adopted the convection-diffusion equation in fluid dynamics. Its parameters were determined according to the vegetation type, building layout and local meteorological data of the scenic area.

[0022] Based on the future risk distribution revealed by the risk evolution heat map, and combined with the real-time status of fixed monitoring resources and the real-time status and mobility of mobile monitoring resources within the scenic area, a multi-strategy simulation scheduling evaluation is performed to determine the optimal set of monitoring resource scheduling instructions. The real-time status of fixed monitoring resources includes: the location coordinates of the fixed monitoring resources: manually entered or calibrated based on a geographic information system (GIS); the coverage area of ​​the fixed monitoring resources: preset according to the model and performance parameters (such as detection distance, lens focal length and viewing angle) of the sensors and cameras; and the activation status of the fixed monitoring resources: determined by the heartbeat signal or polling mechanism through regular communication between the equipment and the monitoring center to determine its online / offline, working / fault status. The real-time status and mobility of mobile monitoring resources include: the current location coordinates of the mobile monitoring resources: obtained in real time through the onboard GPS or Beidou positioning module; the current velocity vector of the mobile monitoring resources: calculated through its own inertial measurement unit (IMU) or in combination with positioning system data; and the available capacity of the mobile monitoring resources (such as remaining fire extinguishing agent and power): measured and transmitted back through the built-in sensors of the equipment.

[0023] The optimal scheduling instruction set is executed to drive the deployment of scenic area monitoring resources, and the scenic area monitoring resources, namely fixed monitoring resources and mobile monitoring resources, collect on-site feedback data of the scenic area after the optimal scheduling instruction set is executed. By comparing the collected on-site feedback data of the scenic area with the scenic area risk evolution heat map, a prediction-on-site deviation value is generated. The deviation value is then used to iteratively optimize the tourist behavior rule base and risk propagation model to improve the accuracy of the next round of predictive simulation.

[0024] Furthermore, the steps for constructing a dynamic digital twin model synchronized with the spatiotemporal state of the scenic area based on the aforementioned real-time multi-source heterogeneous data include: The system acquires multi-source numerical sequences within the scenic area. These sequences include temperature, humidity, smoke concentration, and gas composition sequences generated by IoT sensors, image frame sequences and audio signal sequences generated by video streams, and visitor entry timestamp sequences, visitor count sequences, and visitor location sequences generated by the ticketing system. The system then preprocesses these multi-source numerical sequences, including time-series alignment of the temperature, humidity, smoke concentration, and gas composition sequences, filling in missing numerical sequences using linear interpolation, and applying a low-pass filter to remove noise interference. The image frame sequence is subjected to inter-frame difference calculation to extract the contour of moving objects, and the object trajectory is tracked using an optical flow algorithm. The audio signal sequence is subjected to Fourier transform to extract frequency spectrum features. The steps for performing inter-frame difference calculation to extract the contours of moving objects include: Inter-frame difference calculation: For consecutive video image frames, calculate the difference in grayscale values ​​of pixels between adjacent frames. If the difference exceeds a set threshold, the pixel is determined to belong to a moving object. The formula is expressed as: ; In the formula, Represents the pixel at time t grayscale value, This is the difference result. Through the... Binarization and morphological processing can be used to obtain the outline of a moving object.

[0025] The visitor entry timestamp sequence, visitor count sequence, and visitor location sequence are aggregated and statistically analyzed to calculate the visitor density distribution grid. The steps for calculating the tourist density distribution grid include: Divide the scenic area map into a regular spatial grid (e.g., 10m×10m). Based on the obtained sequence of tourist location coordinates (such as those obtained from camera analysis or ticketing systems), the number of tourists within a specific time period in each grid is counted. Tourist density The calculation formula is: ; in, It is a grid The number of tourists inside, It is the area of ​​the grid.

[0026] The preprocessed multi-source numerical sequence is input into the 3D modeling software. The 3D modeling software uses a meshing algorithm to generate a scenic area terrain mesh model and overlays a real-time data layer to form a spatiotemporally synchronized representation, thus obtaining a dynamic digital twin model. The spatiotemporal synchronization representation is updated using a particle filtering algorithm to ensure that the dynamic digital twin model remains synchronized with the actual state of the scenic area.

[0027] Furthermore, the steps for performing predictive simulations and generating a heatmap of scenic area risk evolution over future time periods include: The current spatiotemporal state data structure is extracted from the dynamic digital twin model, including temperature distribution matrix, humidity distribution matrix, smoke concentration distribution matrix, gas composition distribution matrix, tourist density distribution matrix, and object trajectory vector; Load the preset tourist behavior rule base, which includes a tourist movement probability transition matrix and a tourist aggregation threshold matrix, as well as a risk propagation model, which includes a fire source diffusion equation and a smoke propagation equation; The current spatiotemporal state data structure is combined with the tourist behavior rule base, and the future tourist location distribution sequence is calculated by Markov chain simulation. The future tourist location distribution sequence is input into the risk propagation model. The finite difference method is used to solve the fire source diffusion equation to generate the fire source propagation path sequence, and the smoke propagation equation is solved to generate the smoke concentration change sequence. The steps to solve the fire source diffusion equation to generate the fire source propagation path sequence include: Step 1. Discretize the scenic area into a grid.

[0028] Step 2. At each time step, based on parameters such as the current fire source location, wind speed, wind direction, fuel (vegetation) type, and humidity, calculate the spread rate and direction of the fire line at each grid point using the fire source diffusion equation (such as the Rothermel model).

[0029] Step 3. Update the fire source boundary based on the calculated spread rate and direction, and generate a new fire source location.

[0030] Step 4. Repeat steps 2 and 3 to obtain the sequence of fire source propagation paths over a future period of time.

[0031] The steps to solve the smoke propagation equation and generate a smoke concentration change sequence include: Step 1. Discretize the scenic area into a grid.

[0032] Step 2. At each time step, based on parameters such as fire source location, smoke generation rate, wind speed, wind direction, and turbulent diffusion coefficient, solve the smoke transport equation (usually a simplified convection-diffusion equation).

[0033] Step 3. Use the finite difference method or finite volume method to perform numerical solutions to obtain the smoke concentration of each grid at future times, forming a smoke concentration change sequence.

[0034] By integrating the fire source propagation path sequence, smoke concentration change sequence, and future tourist location distribution sequence, a risk probability grid is calculated, where the risk value of each grid cell is calculated using a weighted summation formula. The risk probability grid is extended over time to generate a risk evolution sequence covering future time periods; The specific steps are as follows: Define the time series to be predicted in the future, for example .

[0035] For each future point in time Obtain the fire source propagation path sequence and smoke concentration change sequence corresponding to that moment.

[0036] At the same time, obtain this time point The future tourist location distribution sequence (predicted by Markov chain).

[0037] For each time point The calculation process of the risk probability grid is repeated.

[0038] Each time point The resulting risk probability grid is arranged in chronological order, thus generating a risk evolution sequence covering future time periods.

[0039] The risk evolution sequence is converted into a heatmap representation using the Matplotlib visualization library, where the color gradient of the heatmap corresponds to the risk value level.

[0040] Furthermore, the steps for performing multi-strategy simulation scheduling evaluation to determine the optimal set of monitoring resource scheduling instructions include: Extract the future risk distribution matrix from the risk evolution heatmap, including the coordinate set of high-risk areas and the time series of risk peaks; The IoT platform queries the status vectors of fixed monitoring resources within the scenic area, including the location coordinates of fixed sensors, the coverage area of ​​fixed cameras, and the activation status of fixed alarms, as well as the status vectors of mobile monitoring resources, including the current location coordinates of mobile drones, the speed vector of mobile inspection vehicles, and the available capacity of mobile fire extinguishers. Multiple sets of scheduling strategies are generated, each set of strategies containing fixed monitoring resource reconfiguration instructions and mobile monitoring resource path planning instructions; The command for reconfiguring fixed monitoring resources includes adjusting the camera pan-tilt angle, switching preset positions, and changing the sensor sampling frequency. Different combinations are generated by changing these parameters.

[0041] Mobile monitoring resource path planning instructions include: generating multiple feasible paths from the current location to the target risk area for drones or inspection vehicles through algorithms (such as Dijkstra's algorithm and genetic algorithm), thereby forming different path planning instruction sets.

[0042] Simulate the execution of each set of scheduling strategies and use the Monte Carlo method to calculate the coverage and response time metrics after scheduling. The coverage rate after scheduling is the total area of ​​high-risk areas covered by the scheduled monitoring resources (camera field of view, sensor detection range, and drone inspection path coverage area). Total area of ​​high-risk areas The ratio, i.e. In Monte Carlo simulations, coverage can be estimated by randomly scattering points within a risk area and counting the proportion of points that fall within the monitoring range.

[0043] Response time refers to the time required for mobile monitoring resources to reach a designated high-risk area. The calculation formula is as follows: , It is the shortest path distance from the current location of the moving resource to the edge of the target area, which can be obtained through a path planning algorithm; This is the average speed of the moving resource, which is either a known parameter of the equipment or an estimate based on historical data. Monte Carlo simulations can account for factors such as speed fluctuations and path uncertainties.

[0044] Compare the coverage metric and the response time metric, and select the scheduling strategy set with the largest sum of the metrics as the candidate instruction set; The candidate instruction set is subjected to constraint checks, including resource mobility limitations and path reachability verification; the candidate instruction set is adjusted until all constraints are met, and the final optimal monitoring resource scheduling instruction set is generated.

[0045] Constraint checks include resource mobility limitations: checking whether commands exceed the physical limits of the equipment. For example, whether the turning radius of the drone's flight path is smaller than its minimum turning radius; whether the commanded speed exceeds its maximum speed; and whether the total mission duration exceeds its endurance.

[0046] Path reachability verification: This uses path planning algorithms to verify the feasibility of the path in the instruction. For example, it checks whether the path crosses no-fly zones or buildings; for inspection vehicles, it checks whether the path is a drivable road and whether there are any obstacles.

[0047] If the candidate instruction set does not meet the constraints, adjustments are made. For example, for unreachable paths, a path planning algorithm is used to replan the path; for instructions that exceed the limits, the parameters are modified to be within the allowable range.

[0048] After adjustment, constraint checks are performed again, and the process is iterated until the instruction set meets all preset constraints, thus generating the final optimal monitoring resource scheduling instruction set.

[0049] Furthermore, the step of synchronously collecting the on-site feedback data of the scenic area monitoring resources after executing the optimal scheduling instruction set includes: Send reconfiguration commands to fixed monitoring resources, including adjusting the sampling frequency of fixed sensors and the focal length parameters of fixed cameras, and send path planning commands to mobile monitoring resources, including specifying the flight trajectory of mobile drones and the driving route of mobile inspection vehicles; Monitor the status of the fixed and mobile monitoring resources executing commands, and confirm deployment completion via heartbeat signals; The status of fixed monitoring resources mainly includes: confirmation of successful / failed command reception, pan-tilt rotation angle, focal length adjustment value, and current sensor sampling frequency.

[0050] The status of mobile monitoring resources executing commands mainly includes: the current real-time location, altitude, speed, heading, battery level, flight mode (e.g., automatic, manual), and mission progress of the UAV; and the current location, speed, direction of travel, battery / fuel level, and mission execution status of the mobile inspection vehicle.

[0051] Collect multi-source numerical sequences after executing the optimal scheduling instruction set; the multi-source numerical sequences include temperature numerical sequences, humidity numerical sequences, smoke concentration numerical sequences, and gas composition numerical sequences collected from fixed sensors after executing the optimal scheduling instructions; Image frame sequences and video clip sequences were collected from fixed cameras and mobile drones, and location coordinate sequences and environmental audio signal sequences were collected from mobile inspection vehicles; After executing the optimal scheduling instruction set, the multi-source numerical sequence is preprocessed. The preprocessing includes timestamp synchronization and alignment, and using a network time protocol to ensure data consistency, resulting in a preprocessed multi-source numerical sequence. The preprocessed multi-source numerical sequences are aggregated to form a scenic area on-site feedback dataset, which includes multimodal feature vectors.

[0052] Further steps in generating the predicted-site deviation value include: Extract the scenic area on-site data structure from the on-site feedback dataset, including the on-site temperature distribution matrix, on-site humidity distribution matrix, on-site smoke concentration distribution matrix, on-site gas composition distribution matrix, and on-site tourist density distribution matrix; The predicted spatiotemporal state vectors for corresponding time points are extracted from the scenic area risk evolution heat map, including the predicted temperature distribution matrix, the predicted humidity distribution matrix, the predicted smoke concentration distribution matrix, the predicted gas composition distribution matrix, and the predicted tourist density distribution matrix. The Euclidean distance between the on-site temperature distribution matrix and the predicted temperature distribution matrix is ​​calculated as the temperature deviation component; the calculation formula is: ; In the formula, Location in the on-site temperature distribution matrix The temperature values ​​are derived from on-site sensors and video inversion. To predict the location in the temperature distribution matrix The temperature values ​​are derived from the underlying data of the risk evolution heatmap. , The number of rows and columns in each grid after the scenic area map is gridded is determined by pre-defined grid division rules. , These are the row and column indices of the grid cell in the two-dimensional grid, respectively.

[0053] The Euclidean distance between the on-site humidity distribution matrix and the predicted humidity distribution matrix is ​​calculated as the humidity deviation component; the calculation formula is: ; In the formula, Location in the on-site humidity distribution matrix The humidity value was obtained through on-site sensors. To predict the location in the humidity distribution matrix The humidity value is derived from the underlying data of the risk evolution heatmap.

[0054] The Euclidean distance between the on-site smoke concentration distribution matrix and the predicted smoke concentration distribution matrix is ​​calculated as the smoke deviation component; the calculation formula is: ; In the formula, Location in the on-site smoke concentration distribution matrix The concentration value is obtained from the on-site smoke sensor. To predict the location in the smoke concentration distribution matrix The concentration values ​​are derived from the underlying data of the risk evolution heatmap.

[0055] The Euclidean distance between the on-site gas composition distribution matrix and the predicted gas composition distribution matrix is ​​calculated as the gas deviation component; the calculation formula is: ; In the formula, Position in the on-site gas composition distribution matrix The concentration value is obtained through on-site gas sensors. To predict the position in the gas composition distribution matrix The concentration values ​​are derived from the underlying data of the risk evolution heatmap. The Euclidean distance between the on-site tourist density distribution matrix and the predicted tourist density distribution matrix is ​​calculated as the tourist deviation component. The calculation formula is: ; In the formula, Location in the on-site tourist density distribution matrix The density values ​​were obtained through on-site cameras and the ticketing system. To predict the location in the tourist density distribution matrix The density value is derived from the underlying data of the risk evolution heatmap.

[0056] The overall predicted-site deviation value is generated by averaging the temperature deviation component, humidity deviation component, smoke deviation component, gas deviation component, and tourist deviation component.

[0057] The calculation formula is: .

[0058] Furthermore, the step of iteratively optimizing the tourist behavior rule base and risk propagation model using the deviation value includes: The system loads the current tourist behavior rule base and the current risk propagation model. The current tourist behavior rule base includes a tourist movement probability transition matrix and a tourist aggregation threshold matrix, which are obtained through historical data statistical analysis and expert experience presets. The current risk propagation model includes a fire source diffusion equation parameter set and a smoke propagation equation parameter set, based on fire science theory and scenic area environmental parameters presets. During iterative optimization, "current" refers to the version saved in the database after the end of the previous optimization cycle. The prediction-reality deviation value is decomposed into behavioral deviation component and propagation deviation component, and the contribution is separated by principal component analysis. For the aforementioned behavioral deviation component, the gradient descent algorithm is applied to adjust the element values ​​of the tourist movement probability transition matrix to converge in the direction of reducing deviation, and the element values ​​of the tourist aggregation threshold matrix are also adjusted to converge in the direction of reducing deviation. For the propagation deviation component, the least squares method is applied to update the coefficient values ​​of the fire source diffusion equation parameter set and the smoke propagation equation parameter set. Verify the updated tourist behavior rule base and risk propagation model by replaying historical data to determine whether the new deviation value has decreased.

[0059] The specific steps are as follows: Extract a set of historical data from the database as a validation set, which was not involved in the previous round of model optimization; Using the model before the update to predict this set of historical data, we obtain an old bias value; Using the adjusted and updated model to predict the same set of historical data yields a new bias value; Compare the new deviation value with the old deviation value.

[0060] Furthermore, the steps for simulating and calculating the future tourist location distribution sequence include: Obtain the current tourist distribution state tensor from the dynamic digital twin model. This tensor contains the number of tourists, tourist movement speed, and tourist aggregation degree in each grid cell. Load the spatiotemporal transition probability tensor from the tourist behavior rule base. This tensor contains inter-grid transition weights adjusted based on time factors and transition correction coefficients based on density factors. Establish an improved Markov prediction equation: ; In the formula, The vector representing the tourist distribution at the current moment is obtained in real time from the dynamic digital twin model. Indicates the future The predicted tourist distribution vector for each time period is loaded from a tourist behavior rule base. This matrix can be preset with different values ​​based on the time period (e.g., morning, afternoon). The transition probability matrix, representing the change over time, is obtained by differential calculation of tourist location data at consecutive time points in the dynamic digital twin model. This represents the field vector representing the tourist's movement velocity. This represents the diffusion coefficient matrix, obtained through statistical analysis of historical tourist movement data, reflecting the tendency of tourists to spread randomly in space. This represents the tourist density gradient vector, obtained by spatial difference calculation of the current tourist distribution vector. This indicates normalization processing.

[0061] Furthermore, the steps for calculating the risk probability grid, where the risk value of each grid cell is calculated using a weighted summation formula, include: The flame intensity and thermal radiation flux of each grid cell are extracted from the fire source propagation path sequence. The visibility impact index and toxic gas concentration index of each grid cell were extracted from the smoke concentration change sequence; Extract the population density index and evacuation difficulty index of each grid cell from the future tourist location distribution sequence; Establish a multi-factor risk fusion equation: ; In the formula, where This represents the final risk value of the grid cell. The flame intensity index is extracted from the calculation results of the fire source diffusion equation. This represents the distance to the fire source, which is the Euclidean distance between the coordinates of the grid cell center and the predicted fire source location coordinates in the fire propagation path sequence. This represents the attenuation coefficient, which is preset based on the blocking and absorption characteristics of heat radiation by the vegetation type, building layout, and materials of the scenic area. The smoke concentration index is extracted from the calculation results (smoke concentration change sequence) of the smoke propagation equation. The visibility index is measured by a visibility meter on site. This represents the baseline visibility, a safe visibility threshold set according to safety regulations, such as the minimum visibility required for personnel evacuation. The concentration index of toxic gases is obtained from on-site gas composition detection data. This represents the baseline concentration, set based on the safety threshold concentration (such as IDLH value) of toxic gases (e.g., CO). The population density index is extracted from the future tourist location distribution sequence. The evacuation difficulty level is indicated by a pre-set value based on the scenic area map. , , , This represents the weighting coefficient of each indicator. It is a natural constant.

[0062] like Figure 2 As shown in the figure, this application provides a scenic area fire monitoring system based on Internet of Things big data, which is used to implement the scenic area fire monitoring method based on Internet of Things big data, including: a model building module, a heat map generation module, an instruction set determination module, a resource deployment module, and an iterative optimization module; The model building module is used to acquire real-time multi-source heterogeneous data within the scenic area and to build a dynamic digital twin model that is synchronized with the spatiotemporal state of the scenic area based on the real-time multi-source heterogeneous data. The heat map generation module is used to perform predictive inference based on the dynamic digital twin model and generate a heat map of the scenic area risk evolution over a future time period. The instruction set determination module is used to perform multi-strategy simulation scheduling evaluation based on the risk evolution heatmap to determine the optimal monitoring resource scheduling instruction set; The resource deployment module is used to execute the optimal scheduling instruction set to drive the deployment of scenic area monitoring resources, and simultaneously collect the on-site feedback data of the scenic area monitoring resources after the execution of the optimal scheduling instruction set. The iterative optimization module is used to compare the collected on-site feedback data of the scenic area with the risk evolution heat map of the scenic area, generate a prediction-on-site deviation value, and use the deviation value to iteratively optimize the tourist behavior rule base and risk propagation model.

[0063] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0064] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0065] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0066] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0067] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0068] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for fire monitoring in scenic areas based on Internet of Things big data, characterized in that, Includes the following steps: Acquire real-time multi-source heterogeneous data within the scenic area, and construct a dynamic digital twin model that is synchronized with the spatiotemporal state of the scenic area based on the real-time multi-source heterogeneous data; Based on the dynamic digital twin model, predictive simulations are performed to generate a heat map of the scenic area's risk evolution over future time periods. Based on the aforementioned risk evolution heatmap, perform multi-strategy simulation scheduling evaluation to determine the optimal set of monitoring resource scheduling instructions; The optimal scheduling instruction set is executed to drive the deployment of scenic area monitoring resources, and the on-site feedback data of the scenic area monitoring resources collected after the optimal scheduling instruction set is executed is collected simultaneously. The collected on-site feedback data of the scenic area is compared with the scenic area risk evolution heat map to generate a prediction-on-site deviation value, and the deviation value is used to iteratively optimize the tourist behavior rule base and risk propagation model.

2. The scenic area fire monitoring method based on IoT big data as described in claim 1, characterized in that, The steps for constructing a dynamic digital twin model synchronized with the spatiotemporal state of the scenic area based on the aforementioned real-time multi-source heterogeneous data include: Obtain multi-source numerical sequences within the scenic area and preprocess the multi-source numerical sequences; The preprocessed multi-source numerical sequence is input into 3D modeling software to generate a scenic area terrain mesh model, and a real-time data layer is superimposed to form a spatiotemporally synchronized representation, resulting in a dynamic digital twin model. The spatiotemporal synchronization representation is updated using a particle filter algorithm.

3. The scenic area fire monitoring method based on IoT big data as described in claim 1, characterized in that, The steps for performing predictive simulations to generate a heatmap of scenic area risk evolution over future time periods include: Extract the current spatiotemporal state data structure from the dynamic digital twin model; Load the pre-defined tourist behavior rules library and risk propagation model; By combining the current spatiotemporal state data structure with the tourist behavior rule base, the future tourist location distribution sequence is simulated and calculated. The future tourist location distribution sequence is input into the risk propagation model, the fire source diffusion equation is solved to generate the fire source propagation path sequence, and the smoke propagation equation is solved to generate the smoke concentration change sequence. By integrating the fire source propagation path sequence, smoke concentration change sequence, and future tourist location distribution sequence, a risk probability grid is calculated. The risk probability grid is extended over time to generate a risk evolution sequence covering future time periods; The risk evolution sequence is converted into a heatmap representation, where the color gradient of the heatmap corresponds to the risk value level.

4. The scenic area fire monitoring method based on IoT big data as described in claim 1, characterized in that, The steps for performing multi-strategy simulation scheduling evaluation to determine the optimal set of monitoring resource scheduling instructions include: Extract the future risk distribution matrix from the aforementioned risk evolution heatmap; Query the state vectors of fixed monitoring resources and mobile monitoring resources within the scenic area; Generate multiple sets of scheduling strategies, simulate the execution of each set of scheduling strategies, and calculate the coverage and response time metrics after scheduling. Compare the coverage metric and the response time metric, and select the scheduling strategy set with the largest sum of the metrics as the candidate instruction set; The candidate instruction set is subjected to constraint checks, and the candidate instruction set is adjusted until all constraints are satisfied, thereby generating the final optimal monitoring resource scheduling instruction set.

5. The scenic area fire monitoring method based on IoT big data as described in claim 1, characterized in that, The steps for synchronously collecting the on-site feedback data of the scenic area monitoring resources after executing the optimal scheduling instruction set include: Send reconfiguration commands to fixed monitoring resources and path planning commands to mobile monitoring resources; Monitor the status of the commands executed by the fixed and mobile monitoring resources; Collect and execute multi-source numerical sequences after executing the optimal scheduling instruction set; The multi-source numerical sequence after executing the optimal scheduling instruction set is preprocessed to obtain the preprocessed multi-source numerical sequence. The preprocessed multi-source numerical sequences are aggregated to form a scenic area on-site feedback dataset.

6. The scenic area fire monitoring method based on IoT big data as described in claim 1, characterized in that, The steps to generate the predicted-field deviation value include: Extract the scenic area on-site data structure from the scenic area on-site feedback dataset; Extract the predicted spatiotemporal state vector for the corresponding time point from the aforementioned scenic area risk evolution heat map; The temperature deviation component, humidity deviation component, smoke deviation component, gas deviation component, and tourist deviation component are used to generate the overall prediction-site deviation value.

7. The scenic area fire monitoring method based on IoT big data as described in claim 1, characterized in that, The steps for iteratively optimizing the tourist behavior rule base and risk propagation model using the deviation value include: Load the current tourist behavior rule base and the current risk propagation model; The prediction-reality deviation value is decomposed into a behavioral deviation component and a propagation deviation component; For the aforementioned behavioral deviation component, the gradient descent algorithm is applied to adjust the element values ​​of the tourist movement probability transition matrix and the element values ​​of the tourist aggregation threshold matrix. For the propagation deviation component, the least squares method is applied to update the coefficient values ​​of the fire source diffusion equation parameter set and the smoke propagation equation parameter set. Verify the adjusted and updated tourist behavior rule base and risk propagation model.

8. The scenic area fire monitoring method based on IoT big data as described in claim 3, characterized in that, The steps involved in simulating and calculating the future tourist location distribution sequence include: Obtain the current tourist distribution state tensor from the dynamic digital twin model; Load the spatiotemporal transition probability tensor from the tourist behavior rule base; Establish an improved Markov prediction equation: ； In the formula, This represents the tourist distribution vector at the current moment. Indicates the future The predicted tourist distribution vector at any given time. This represents the transition probability matrix that changes over time. This represents the field vector representing the tourist's movement velocity. Represents the diffusion coefficient matrix. Represents the tourist density gradient vector. This indicates normalization processing.

9. The scenic area fire monitoring method based on IoT big data as described in claim 3, characterized in that, The steps for calculating the risk probability grid include: The flame intensity and thermal radiation flux of each grid cell are extracted from the fire source propagation path sequence. The visibility impact index and toxic gas concentration index of each grid cell were extracted from the smoke concentration change sequence; Extract the population density index and evacuation difficulty index of each grid cell from the future tourist location distribution sequence; Establish a multi-factor risk fusion equation: ； In the formula, where This represents the final risk value of the grid cell. Indicates the flame intensity index. Indicates the distance to the fire source. Indicates the attenuation coefficient. Indicates smoke concentration index, Indicates visibility index, Indicates the baseline visibility. Indicators representing the concentration of toxic gases Indicates the baseline concentration. Indicators representing population density Indicates the evacuation difficulty coefficient. , , , This represents the weighting coefficient of each indicator. It is a natural constant.

10. A scenic area fire monitoring system based on Internet of Things (IoT) big data, used to implement the scenic area fire monitoring method based on IoT big data as described in any one of claims 1-9, characterized in that, include: Model building module, heatmap generation module, instruction set determination module, resource deployment module, iterative optimization module; The model building module is used to acquire real-time multi-source heterogeneous data within the scenic area and to build a dynamic digital twin model that is synchronized with the spatiotemporal state of the scenic area based on the real-time multi-source heterogeneous data. The heat map generation module is used to perform predictive inference based on the dynamic digital twin model and generate a heat map of the scenic area risk evolution over a future time period. The instruction set determination module is used to perform multi-strategy simulation scheduling evaluation based on the risk evolution heatmap to determine the optimal monitoring resource scheduling instruction set; The resource deployment module is used to execute the optimal scheduling instruction set to drive the deployment of scenic area monitoring resources, and simultaneously collect the on-site feedback data of the scenic area monitoring resources after the execution of the optimal scheduling instruction set. The iterative optimization module is used to compare the collected on-site feedback data of the scenic area with the risk evolution heat map of the scenic area, generate a prediction-on-site deviation value, and use the deviation value to iteratively optimize the tourist behavior rule base and risk propagation model.

Images