Firefighting and Rescue Intelligent Management Method and System Based on Multimodal AI Large Model

Abstract

The present invention relates to the technical field of fire management, and in particular to a fire rescue intelligent management method and system based on a multi-modal AI large model. The method includes the following steps: obtaining real-time environmental monitoring data of the fire scene; parsing environmental parameters of the real-time environmental monitoring data of the fire scene and predicting the gas change trend to generate the change trend of on-site gas parameters; obtaining the thermal imaging image of the fire scene; performing visual recognition of temperature distribution and calculating the temperature change rate of consecutive frames based on the thermal imaging image of the fire scene, so as to generate the temperature change rate of the fire source area; predicting the multi-time fire potential trend according to the temperature change rate of the fire source area and the change trend of on-site gas parameters, so as to generate a multi-time fire potential trend prediction map; identifying the points in the area to be rescued based on the thermal imaging image of the fire scene. The present invention realizes the real-time and accurate allocation of on-site fire rescue resources, and improves the rescue management efficiency and safety.

Description

Technical Field

[0001] The present invention relates to the technical field of fire management, and particularly to a fire rescue intelligent management method and system based on a multi-modal AI large model. Background Art

[0002] As an important means to ensure the safety of people's lives and property, fire rescue work has always faced a complex and high-risk working environment. With the acceleration of urbanization and the frequent occurrence of natural disasters, fire rescue tasks have become increasingly heavy and difficult. In traditional fire rescue management, it relies on manual command, equipment dispatching, and on-site judgment. Although these means were able to meet certain needs in the past, there is still a large room for improvement in terms of efficiency, accuracy, and emergency response speed.

[0003] Especially in sudden disasters such as fires, rescue tasks often require rapid response and real-time adjustment of rescue strategies. Traditional methods cannot accurately integrate and process various complex information on the site, such as the development of the fire, building structure, location of rescue personnel, meteorological data, etc. This makes rescue decisions often lag, affecting the efficiency and safety of fire rescue. Therefore, there is an urgent need for a new intelligent method to achieve comprehensive and real-time analysis and management of the fire rescue process, so as to improve rescue efficiency, optimize resource allocation, and ensure personnel safety. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes a fire rescue intelligent management method and system based on a multi-modal AI large model to solve at least one of the above technical problems.

[0005] To achieve the above object, the present invention provides a fire rescue intelligent management method based on a multi-modal AI large model, including the following steps:

[0006] Step S1: Obtain real-time environmental monitoring data of the fire scene; analyze environmental parameters of the real-time environmental monitoring data of the fire scene, and predict the gas change trend to generate the change trend of on-site gas parameters;

[0007] Step S2: Obtain a thermal imaging image of the fire scene; perform visual recognition of temperature distribution and calculate the temperature change rate of consecutive frames based on the thermal imaging image of the fire scene, so as to generate the temperature change rate of the fire source area;

[0008] Step S3: Perform multi-time fire potential trend prediction based on the temperature change rate of the fire source area and the change trend of on-site gas parameters, so as to generate a multi-time fire potential trend prediction map;

[0009] Step S4: Identify the points in the area to be rescued based on the thermal imaging image of the fire scene; perform optimal rescue route analysis for each area of the points in the area to be rescued based on the multi-time fire trend prediction map, and generate multiple area rescue tasks;

[0010] Step S5: Identify the fire-fighting resources at the fire scene; perform dynamic resource allocation for multiple area rescue tasks according to the fire-fighting resources at the fire scene, and conduct real-time rescue monitoring to obtain real-time rescue monitoring data;

[0011] Step S6: Extract instant feedback information and optimize intelligent resource allocation based on the real-time rescue monitoring data, so as to construct an intelligent fire-fighting rescue optimization strategy.

[0012] Through the acquisition of real-time environmental monitoring data, the information in the fire-fighting rescue process of the present invention is more timely and accurate. By analyzing environmental parameters, the change trend of harmful gases at the fire scene can be identified, the gas concentration and diffusion trend can be predicted in time, providing a crucial early warning for the safety of rescue personnel. Using the prediction of gas change trend provides a scientific basis for fire-fighting command decision-making, avoiding rescue personnel from entering high-risk areas and ensuring the efficient execution of rescue tasks. The thermal imaging image provides real-time temperature data of the fire scene, which can help rescue personnel accurately understand the location of the fire source and its temperature change. By calculating the temperature change rate of consecutive frames, the change speed of the fire can be captured, the spread trend of the fire can be predicted, providing real-time dynamic guidance for on-site rescue. Monitoring the temperature change rate helps to identify the hot spots of the fire and guide the reasonable allocation of fire-fighting resources. Combining the temperature change rate in the fire source area and the gas change trend for fire prediction can provide a more comprehensive and accurate fire development trend. Through multi-time prediction, the fire-fighting command center can clearly understand the future development direction of the fire, formulate countermeasures in advance, and reduce the uncertainty and risk in rescue. The fire trend prediction map provides a priority ranking of rescue tasks for firefighters, ensuring that resources are preferentially allocated to the places where they are most needed. Thermal imaging image analysis can accurately identify the areas to be rescued at the fire scene, especially those areas with high temperature and serious fire near the fire source. Optimal rescue route analysis can prevent rescue personnel from entering dangerous areas, ensuring that they can complete the task through the safest and fastest route, reducing the risk and time consumption in the rescue process. The optimization of the rescue route greatly improves the rescue efficiency, especially at a fire scene that is complex and full of uncertain factors. By identifying the available fire-fighting resources at the scene (such as fire extinguishers, fire trucks, fire-fighters, etc.), resources can be allocated in real time and accurately, ensuring that every rescue task can be supported. Dynamic resource allocation adjusts according to the changes at the fire scene, effectively avoiding resource waste and unnecessary delays. Real-time rescue monitoring can feedback the progress of the rescue, ensure the safety of fire-fighters, and make timely adjustments to the on-site situation, making the entire rescue process more flexible and efficient.

[0013] In this specification, a fire rescue intelligent management system based on a multimodal AI large model is provided, which is used to execute the fire rescue intelligent management method based on the multimodal AI large model as described above, and includes:

[0014] A gas trend prediction module, which is used to obtain real-time environmental monitoring data of the fire scene; analyze environmental parameters of the real-time environmental monitoring data of the fire scene, and predict the gas change trend to generate the change trend of on-site gas parameters.

[0015] A temperature distribution recognition module, which is used to obtain a thermal imaging image of the fire scene; perform visual recognition of the temperature distribution and calculate the temperature change rate of consecutive frames based on the thermal imaging image of the fire scene, so as to generate the temperature change rate of the fire source area.

[0016] A fire trend prediction module, which is used to perform multi-time fire trend prediction according to the temperature change rate of the fire source area and the change trend of on-site gas parameters, so as to generate a multi-time fire trend prediction map.

[0017] A rescue route analysis module, which is used to identify the points in the area to be rescued based on the thermal imaging image of the fire scene; perform optimal rescue route analysis for each area of the points in the area to be rescued based on the multi-time fire trend prediction map, and generate multiple area rescue tasks.

[0018] A dynamic resource allocation module, which is used to identify the fire fighting resources at the fire scene; perform dynamic resource allocation for multiple area rescue tasks according to the fire fighting resources at the fire scene, and perform real-time rescue monitoring to obtain real-time rescue monitoring data.

[0019] A resource allocation optimization module, which is used to extract instant feedback information and optimize intelligent resource allocation based on the real-time rescue monitoring data, so as to construct an intelligent fire rescue optimization strategy.

[0020] Through the collection of real-time environmental monitoring data, this invention ensures that the command center and rescue personnel can obtain the changes in on-site gas composition and concentration in the first place, providing data support for safety decision-making. Through the prediction of gas change trends, it can timely warn of the accumulation of harmful gases, such as carbon monoxide, nitrogen oxides, etc., so as to avoid high-risk areas in advance and protect the safety of rescue personnel. Gas trend prediction helps commanders more scientifically evaluate the scope of fire spread, the intensity of the fire source, and the changes in dangerous areas, thus formulating more effective rescue and evacuation strategies. Through temperature distribution identification, it can confirm the heat source area of the fire in real time, helping rescue personnel quickly lock in the fire source and the main direction of fire development. By calculating the rate of change of temperature in consecutive frames, it evaluates the speed of fire spread, helps predict the development trend of the fire, and thus makes timely responses when the fire spreads rapidly. Monitoring the rate of temperature change in the fire source area enables on-site command to flexibly adjust resources and rescue strategies to cope with the changes in the fire. Combining the fire source temperature and gas changes can provide a more comprehensive fire prediction, not only paying attention to the temperature of the fire source itself, but also taking into account gas diffusion and the speed of fire spread. By predicting the multi-timepoint trend of the fire, it can understand in advance the possible development of the fire at different time points, providing more forward-looking information for fire commanders and supporting rescue decision-making. The fire trend chart helps the command headquarters clarify the development direction of the fire, make preparations for the allocation of fire resources and the deployment of rescue personnel in advance, and ensure the timeliness and effectiveness of the response. Through thermal imaging images, it can clarify which areas have higher temperatures and which areas have trapped people, thus quickly identifying the areas that need to be rescued first. Based on the fire prediction, the rescue route analysis can calculate the shortest and safest path, avoiding rescue personnel from entering overly dangerous or rapidly spreading fire areas, thus improving rescue efficiency and safety. In a complex fire scene, it can generate rescue tasks for multiple areas and ensure the coordination of each task, ensuring the orderly progress of rescue operations. According to the changes in the fire scene, it dynamically allocates various types of fire resources (such as fire extinguishers, fire trucks, personnel, etc.) to ensure the most effective use of resources. As the situation at the fire scene changes, it can flexibly adjust the allocation of resources to ensure that at different stages, the rescue force can cover the areas most in need. Real-time rescue monitoring data can provide key information such as the usage of fire resources and the changes in the fire scene, helping commanders make quick responses. Using AI analysis algorithms, it optimizes the current resource allocation situation based on real-time data and adjusts the resource configuration to adapt to the latest dynamics of the fire scene. Through intelligent decision-making, it can avoid resource waste or unreasonable allocation caused by human judgment errors, ensuring that the use of each resource is as efficient as possible. As the fire scene information is continuously updated, the resource allocation optimization module can continuously adjust and optimize the strategy according to the actual situation to ensure the efficiency and flexibility of the rescue process. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1Schematic diagram of the step flow of a fire rescue intelligent management method based on a multi-modal AI large model according to the present invention;

[0022] Figure 2 Schematic diagram of the detailed implementation steps of step S1;

[0023] Figure 3 Schematic diagram of the detailed implementation steps of step S2;

[0024] Figure 4 Schematic diagram of the detailed implementation steps of step S3. Specific implementation manners

[0025] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0026] The embodiments of the present application provide a fire rescue intelligent management method and system based on a multi-modal AI large model. The execution subjects of the fire rescue intelligent management method and system based on the multi-modal AI large model include but are not limited to: mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc. that can be regarded as general computing nodes of the present application. The data processing platform includes but is not limited to at least one of: an audio and image management system, an information management system, and a cloud data management system.

[0027] Please refer to Figures 1 to 4 , the present invention provides a fire rescue intelligent management method based on a multi-modal AI large model. The fire rescue intelligent management method based on the multi-modal AI large model includes the following steps:

[0028] Step S1: Obtain real-time environmental monitoring data of the fire scene; analyze the environmental parameters of the real-time environmental monitoring data of the fire scene, and predict the gas change trend to generate the change trend of the on-site gas parameters;

[0029] Step S2: Obtain the thermal imaging image of the fire scene; perform visual recognition of the temperature distribution and calculate the temperature change rate of consecutive frames based on the thermal imaging image of the fire scene, so as to generate the temperature change rate of the fire source area;

[0030] Step S3: Perform multi-time fire potential trend prediction according to the temperature change rate of the fire source area and the change trend of the on-site gas parameters, so as to generate a multi-time fire potential trend prediction map;

[0031] Step S4: Identify the points in the area to be rescued based on the thermal imaging image of the fire scene; perform optimal rescue route analysis for each area of the points in the area to be rescued based on the multi-time fire potential trend prediction map, and generate multiple area rescue tasks;

[0032] Step S5: Identify the fire-fighting resources at the fire scene; dynamically allocate resources for multiple regional rescue tasks based on the fire-fighting resources at the fire scene, and conduct real-time rescue monitoring to obtain real-time rescue monitoring data;

[0033] Step S6: Extract instant feedback information and optimize intelligent resource allocation based on the real-time rescue monitoring data, so as to construct an intelligent fire-fighting rescue optimization strategy.

[0034] Through the acquisition of real-time environmental monitoring data, the information in the fire-fighting rescue process of the present invention is more timely and accurate. By analyzing the environmental parameters, the changing trend of harmful gases at the fire scene can be identified, the gas concentration and diffusion trend can be predicted in a timely manner, and crucial early warnings can be provided for the safety of rescue personnel. Using the prediction of the gas changing trend provides a scientific basis for fire-fighting command decision-making, avoids rescue personnel from entering high-risk areas, and ensures the efficient execution of rescue tasks. The thermal imaging image provides real-time temperature data of the fire scene, which can help rescue personnel accurately understand the location of the fire source and its temperature change. By calculating the temperature change rate of consecutive frames, the changing speed of the fire can be captured, the spreading trend of the fire can be predicted, and real-time dynamic guidance can be provided for on-site rescue. The monitoring of the temperature change rate helps to identify the hot spots of the fire and guide the reasonable allocation of fire-fighting resources. Combining the temperature change rate in the fire source area and the gas changing trend for fire prediction can provide a more comprehensive and accurate fire development trend. Through multi-time-point prediction, the fire-fighting command center can clearly understand the future development direction of the fire, formulate countermeasures in advance, and reduce the uncertainty and risks in rescue. The fire trend prediction map provides a priority ranking for the rescue tasks of firefighters, ensuring that resources are preferentially allocated to the places where they are most needed. The thermal imaging image analysis can accurately identify the areas to be rescued at the fire scene, especially those areas with high temperature and serious fire near the fire source. The optimal rescue route analysis can prevent rescue personnel from entering dangerous areas, ensure that they can complete the task through the safest and fastest route, and reduce the risks and time consumption in the rescue process. The optimization of the rescue route greatly improves the rescue efficiency, especially at the fire scene with complex and uncertain factors. By identifying the available fire-fighting resources at the scene (such as fire extinguishers, fire trucks, fire-fighters, etc.), resources can be allocated in real time and accurately, ensuring that every rescue task can be supported. The dynamic resource allocation is adjusted according to the changes at the fire scene, effectively avoiding resource waste and unnecessary delays. The real-time rescue monitoring can feedback the progress of the rescue, ensure the safety of fire-fighters, and make timely adjustments to the on-site situation, making the entire rescue process more flexible and efficient.

[0035] In the embodiment of the present invention, refer to Figure 1 , which is a schematic diagram of the step flow of a fire-fighting rescue intelligent management method based on a multi-modal AI large model of the present invention. In this example, the steps of the fire-fighting rescue intelligent management method based on the multi-modal AI large model include:

[0036] Step S1: Obtain real-time environmental monitoring data of the fire scene; analyze the environmental parameters of the real-time environmental monitoring data of the fire scene, and predict the gas change trend to generate the change trend of on-site gas parameters;

[0037] In this embodiment, suitable environmental monitoring devices are selected, such as multi-gas detectors, temperature and humidity sensors, and smoke sensors. These devices can monitor the environmental parameters at the fire scene in real time, including temperature, humidity, oxygen concentration, carbon monoxide concentration, and volatile organic compounds (VOCs), etc. Ensure the calibration and maintenance of the monitoring devices to guarantee the accuracy of the data. For example, the calibration period of the gas sensor should be set to once a month to ensure that the sensor meets the standards before use. Deploy the monitoring devices at key positions at the fire scene, such as near the fire source, the smoke diffusion area, and the area where trapped people may be. Ensure that the devices can cover the entire scene and can transmit data in real time. Set the data acquisition frequency, for example, collect data once every 1 second to ensure a quick response to the changes in the fire development. The data is transmitted to the data processing system in real time through a wireless network or a wired connection. Use a data management system to record and store the real-time monitoring data. The data management system should have data backup and recovery functions to ensure the security and integrity of the data. Regularly check the integrity of the data records to ensure that there is no data loss or abnormal situation, and record the working status and data stream of each sensor. Preprocess the collected real-time monitoring data, including denoising, missing value filling, and outlier detection. The missing values can be filled using simple mean substitution or median substitution methods to ensure the continuity of the data. Ensure that the timestamps of the data are consistent for subsequent analysis. For example, convert all data to the UTC time format uniformly to avoid data confusion caused by time zone problems. Extract key environmental parameters from the preprocessed data, such as temperature, humidity, oxygen concentration, and harmful gas concentration, etc. Use statistical analysis methods to calculate the mean, standard deviation, and change range of each parameter to understand the basic characteristics of the on-site environment. Set the analysis period, for example, conduct a summary analysis of the environmental parameters every 5 minutes, and record the change trends and fluctuations of the environmental parameters. Conduct trend analysis on the extracted environmental parameters, especially the gas concentration data. Time series analysis methods can be used to identify the trends of gas concentration changes over time. Apply moving average or exponential smoothing methods to smooth the data to reduce the impact of short-term fluctuations on trend identification. Set the smoothing window size, for example, select a 5-minute window to better observe the change trends of gas concentration. Select a suitable prediction model, such as the autoregressive integrated moving average (ARIMA) model or the long short-term memory (LSTM) network. These models can effectively capture the change patterns of gas concentration and conduct future trend predictions. Set the parameters of the model, such as the autoregressive order and the moving average order in the ARIMA model, and usually select the optimal parameters based on historical data. Use historical monitoring data to train the selected prediction model. Ensure that the time span of the training data is long enough to capture the change trends of gas concentration. Conduct cross-validation on the model to ensure its prediction ability on different datasets. 70% of the data can be used for training and 30% for validation to evaluate the performance of the model.Use the trained prediction model to predict the future gas concentration changes. Set the prediction time range, for example, predict the gas concentration changes within the next 30 minutes. Record the prediction results, including the predicted gas concentration values and their confidence intervals, for subsequent decision-making reference. Generate a prediction report that details the trend of gas concentration changes and their potential impact on the safety of the fire scene.

[0038] Step S2: Obtain the thermal imaging image of the fire scene; perform visual recognition of the temperature distribution and calculate the temperature change rate of consecutive frames based on the thermal imaging image of the fire scene, so as to generate the temperature change rate of the fire source area;

[0039] In this embodiment, a high-quality thermal imaging camera is selected to ensure that it has sufficient resolution and sensitivity. Generally, a thermal imaging device with a resolution of 640x480 pixels is selected to capture clear temperature distribution images at the fire scene. The device is calibrated to ensure the accuracy of its temperature measurement before use. The operating temperature range of the device is set, for example, from -20°C to +1200°C, to meet the requirements of different fire scenarios. A thermal imaging camera is set up at the fire scene to ensure that it can cover the fire source and the surrounding area. According to the actual situation of the fire scene, a suitable shooting angle and height are selected to obtain the best viewing angle. The image acquisition frequency is set, for example, one frame per second, to capture the dynamic changes in the fire source area. The images are transmitted in real time to the data processing center via wireless network or wired connection. A data management system is established to store and manage the acquired thermal imaging images. To ensure the security and integrity of the data, a regular backup mechanism is set. Metadata such as the timestamp, shooting location, and environmental conditions of each frame of the image are recorded for reference during subsequent analysis. The acquired thermal imaging images are preprocessed, including denoising, enhancing contrast, and color mapping. Gaussian filtering can be used to remove random noise in the image and improve the visualization effect of the temperature distribution. Image processing parameters, such as the standard deviation of the filter and the enhancement degree, are set to ensure that the preprocessed image is clearly distinguishable. Computer vision techniques are applied for visual recognition of the temperature distribution. Deep learning models (such as convolutional neural networks) can be used to classify and label the temperature distribution. When training the model, a thermal imaging dataset containing different fire sources and temperature distributions is used. The recognition threshold is set, for example, areas with a temperature higher than 60°C are marked as the fire source area for subsequent analysis. The identified fire source areas and their corresponding temperature values are organized into a report, which details the temperature distribution and the location of the fire source. A temperature distribution map is generated to visually display the temperature characteristics of the fire source area. The key parameters and results during the recognition process are recorded to provide a basis for subsequent calculation of the temperature change rate. The temperature data of the fire source area is extracted from consecutive thermal imaging images. For each frame of the image, the temperature value of the fire source area is recorded, and time series data is generated. The time interval for data extraction is set, for example, the temperature data is extracted once per second, to analyze the temperature change of the fire source. Based on the extracted temperature data, the temperature change rate of the fire source area is calculated. The simple difference method can be used to calculate the temperature change between adjacent frames. For example, the temperature change rate = (temperature of the current frame - temperature of the previous frame) / time interval. The threshold for the change rate is set, for example, a temperature change rate exceeding 5°C / second is regarded as a high-risk fire source for key monitoring. The calculated temperature change rate is organized into a report, which details the temperature change of the fire source area, identifies possible high-temperature areas and their change trends. A temperature change rate map is generated to visually display the temperature changes at different time points, helping decision-makers quickly understand the dynamics of the fire source and providing a basis for subsequent rescue and control measures.

[0040] Step S3: Perform multi - time ignition potential trend prediction based on the temperature change rate in the fire source area and the change trend of on - site gas parameters, so as to generate a multi - time ignition potential trend prediction map;

[0041] In this embodiment, the temperature change rate data and the change trend data of on-site gas parameters in the fire source area are collected. These data should include the temperature changes, oxygen concentration, carbon monoxide concentration, and concentration change records of other relevant gases at multiple time points. Ensure that the timestamps of the data are consistent for subsequent analysis. For example, unify the temperature change rate and gas parameter data into the UTC time format to facilitate the merging of different data sources. Clean the collected temperature change rate and gas parameter data to remove outliers and missing values. The mean filling method can be used to handle the missing values to ensure the integrity of the data. Use the standardization method to process the data and convert the data with different dimensions into the same scale. For example, standardize the gas concentration and temperature data to the range of 0 to 1 to facilitate the training of subsequent models. Before data analysis, generate preliminary visualization charts of the temperature change rate and gas parameters. These charts can include the temperature change trend chart and the gas concentration change curve to visually display the basic characteristics and change trends of the data. Set the visualization parameters, such as a time window of 5 minutes, to better observe the changes within a short period. Select a suitable fire trend prediction model, such as a multiple linear regression model, an ARIMA model, or a long short-term memory (LSTM) model. According to the nature and requirements of the data, select a model that can capture the characteristics of time series. Set the input parameters of the model, including the temperature change rate, gas concentration, and other relevant factors, to ensure that the model can comprehensively reflect the influencing factors of fire changes. Use historical data to train the selected prediction model. Ensure that the training data has a sufficient time span to capture the laws of fire changes. For example, use the temperature change rate and gas parameter data of the past 24 hours for training. Conduct cross-validation on the model to ensure its prediction ability on different data sets. 70% of the data can be selected for training, and 30% of the data can be used for validation to evaluate the accuracy and robustness of the model. Use the trained model to predict the fire trend at multiple future time points. Set the prediction time range, such as predicting the fire changes every 5 minutes within the next 1 hour. Record the prediction results at each time point, including the fire intensity, possible fire spread direction, and relevant gas concentration. These results will be used for subsequent decision support. Organize the results of multi-time fire trend prediction into a report, detailing the fire intensity and gas concentration predictions at each time point. Ensure the integrity and clarity of the information for decision-makers to reference. Generate a prediction data table listing the fire intensity, temperature, and main gas concentration at each time point to help decision-makers quickly obtain key information. Generate a multi-time fire trend prediction chart based on the prediction results to visually display the future fire change trend. These charts can be in the form of line charts or heat maps to clearly show the changes in fire intensity and gas concentration. Mark the key time points and change amplitudes on the charts to facilitate decision-makers to quickly understand the dynamic characteristics of fire development.

[0042] Step S4: Identify the points in the area to be rescued based on the thermal imaging image of the fire scene; analyze the optimal rescue routes for each area to be rescued one by one based on the multi-time fire trend prediction map, and generate multiple area rescue tasks;

[0043] In this embodiment, the thermal imaging image of the fire scene is preprocessed, including denoising and image enhancement. Gaussian filtering or median filtering can be used to remove background noise, and at the same time, the contrast of the image is improved through histogram equalization to make the area to be rescued more obvious. Set processing parameters, such as the standard deviation of the filter and the enhancement intensity, to ensure that the image quality is high enough for subsequent area recognition. Apply image recognition algorithms, such as deep learning-based object detection techniques (such as YOLO or Faster R-CNN), to analyze the preprocessed thermal imaging image to identify the areas to be rescued. These areas usually show higher temperatures and may have trapped people or fire sources. Set the detection threshold, such as areas with temperatures higher than 70°C are marked as areas to be rescued that need attention, to ensure the accuracy and reliability of the recognition process. Extract the fire trend change information of each area to be rescued from the multi-time fire trend prediction map, including fire intensity, possible diffusion direction, and gas concentration change. These data will be used to evaluate the risk levels of each area. Ensure that the time stamps of the fire data are consistent with the recognition time of the areas to be rescued for effective comparison and analysis. Select a suitable optimal rescue route analysis algorithm, such as the A* search algorithm or Dijkstra algorithm, to plan the route based on the risk level, distance, and available resources of each area to be rescued. Set analysis parameters, including the weight of each path, considering factors such as fire intensity, passage obstacles, and resource availability, to ensure that the safest and most effective rescue route is calculated. For each area to be rescued, use the selected algorithm to perform optimal rescue route analysis. Record the starting point, ending point, passed nodes, and estimated travel time of each path. Set the tolerance of path analysis, such as allowing a path error within 1 meter, to ensure that the identified path is feasible and safe. Ensure that the path of each area can be quickly adjusted to handle emergencies. Organize the optimal rescue route and relevant information of each area into a report, detailing the rescue tasks of each area to be rescued, including required resources, estimated time, and potential risks, etc. Generate a rescue task list to ensure the clarity and readability of the information and provide decision support for on-site commanders. Ensure that the rescue priority of each area is included in the task list to allocate resources reasonably.

[0044] Step S5: Identify the fire-fighting resources at the fire scene; perform dynamic resource allocation for multiple area rescue tasks according to the fire-fighting resources at the fire scene, and conduct real-time rescue monitoring to obtain real-time rescue monitoring data;

[0045] In this embodiment, a list of fire-fighting resources is formulated, including all available personnel, equipment and supplies. Resource classification should include firefighters, fire trucks, fire extinguishers, rescue equipment and logistical support, etc. Set resource identification criteria. For example, each firefighter should have the corresponding qualification certificates and experience, and the equipment should meet national safety standards. Ensure the accuracy and traceability of resource information. Evaluate the status and availability of fire-fighting resources through on-site inspections and real-time monitoring systems (such as drones and sensors). Record the location, quantity and working status of each resource. Set evaluation parameters, such as the health status of firefighters and the functional integrity of equipment, to ensure that resources can be effectively used in emergencies. Establish a fire-fighting resource database and enter the identified resource information. The database should include basic information, status, historical usage records, etc. of the resources to ensure systematic management of the information. Regularly update the resource database to ensure the timeliness and accuracy of the data. For example, update the resource status information every hour to keep abreast of the on-site situation at any time. Set priorities for rescue tasks in each area according to the urgency and complexity of the rescue tasks. A risk assessment model can be used to classify the tasks into high, medium and low priorities. Set evaluation parameters, such as the number of trapped people, fire intensity and resource requirements, to ensure the scientific nature of the priority evaluation. Establish a dynamic resource allocation model and select a suitable optimization algorithm (such as linear programming, dynamic programming or genetic algorithm) for resource allocation. The model should consider factors such as resource availability, task priority and response time. Set the input parameters of the model, including the quantity of resources, task requirements and priorities, etc., to ensure that the model can generate an optimal resource allocation plan. Run the resource allocation model and make dynamic adjustments according to real-time data and task requirements. Prioritize the allocation of resources to high-priority tasks to ensure the rational use of resources. Record the results of each resource allocation, including the quantity, area and time of allocation, etc., to ensure the transparency and traceability of the allocation process. Establish a real-time rescue monitoring system, integrate on-site monitoring equipment (such as cameras, sensors and drones), and obtain real-time data. This data includes the action status of firefighters, fire changes and on-site personnel conditions, etc. Determine the monitoring frequency. For example, update the data every 1 minute to ensure timely response to the on-site situation. Collect and analyze the obtained real-time rescue monitoring data to evaluate the rescue effect and resource usage. Use data analysis tools to generate real-time monitoring reports, including rescue progress, resource usage and fire changes. Set analysis parameters, such as the action efficiency of rescue personnel and the utilization rate of resources, etc., to evaluate the dynamic changes on-site. Adjust rescue strategies and resource allocation in a timely manner according to real-time monitoring data to ensure the flexibility and effectiveness of rescue operations. Record the key parameters and decision-making basis during the adjustment process for subsequent evaluation and improvement. Generate real-time feedback reports to provide the latest rescue status information to the command center and on-site commanders to ensure the scientific nature and timeliness of decision-making.

[0046] Step S6: Extract instant feedback information and optimize intelligent resource allocation based on real-time rescue monitoring data, so as to construct an intelligent fire rescue optimization strategy.

[0047] In this embodiment, real-time rescue data is collected from the on-site monitoring system, including the location information of firefighters, equipment status, fire situation changes, gas concentration, etc. These data are transmitted to the central command system in real time through sensors and monitoring devices. Integrate data from different sources, such as image information obtained through IoT devices and drones, to ensure the comprehensiveness and accuracy of the data. Set the data collection frequency, for example, update it every 5 minutes to ensure the timeliness of information. Use data processing algorithms to analyze the collected real-time monitoring data and extract key information, such as the rate of change of the fire situation, the number of trapped people, the usage of resources, and the working status of rescue personnel. Set analysis parameters. When the rate of change of the fire situation exceeds a specific threshold (for example, 5°C / minute), trigger an alarm mechanism to remind the commander to respond quickly. Identify potential risk points and bottlenecks through data mining techniques. Organize the extracted immediate feedback information into a report, which details the real-time situation at the scene, including fire situation changes, rescue progress, and resource usage. The report should include charts and visual information to help the commander quickly understand the current situation. Set the report update frequency, for example, generate an updated report every 30 minutes to ensure that the command center can obtain real-time dynamic information at the scene. Based on the real-time monitoring data, build an intelligent resource allocation optimization model, select suitable algorithms, such as linear programming, genetic algorithms, or deep learning methods, to achieve dynamic resource allocation. Set the input parameters of the model, including task priorities, available resource quantities, fire situation change rates, and rescue objectives, etc., to ensure that the model can comprehensively reflect the on-site situation. Use historical rescue data to train the resource allocation model to ensure that the model can make effective predictions in different situations. For example, use the data from the past few fire rescues to train the model to identify optimal resource allocation strategies. Conduct cross-validation on the model to ensure its prediction ability on different data sets. Set training parameters, such as the training cycle and learning rate, to ensure that the model can achieve the best performance. Run the optimization model and dynamically adjust resource allocation according to real-time data and task requirements. Prioritize the allocation of resources to high-priority task areas to ensure the reasonable use of resources. Record the results of each allocation, including the quantity, area of resource allocation, and estimated rescue time, to ensure the transparency and traceability of the allocation process. Based on the real-time feedback information and resource allocation results, formulate an intelligent fire rescue optimization strategy. The strategy should cover multiple aspects such as resource configuration, rescue steps, and emergency response mechanisms. Set strategy evaluation criteria, such as rescue efficiency, resource utilization rate, and the safety of trapped people, etc., and evaluate the effectiveness of the strategy through quantitative indicators. During the implementation process, monitor the rescue effect in real time and adjust the optimization strategy in a timely manner according to the changes in the on-site situation. For example, if the fire spread speed accelerates, immediately increase the resource input in high-risk areas. Set monitoring parameters, such as the resource usage in each area and the fire situation change rate, to ensure that the rescue strategy can flexibly respond to emergencies.

[0048] In this embodiment, refer toFigure 2 , which is a schematic diagram of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0049] Step S11: Obtain real-time environmental monitoring data of the fire scene based on sensors;

[0050] Step S12: Analyze the environmental parameters of the real-time environmental monitoring data of the fire scene to generate real-time parameter characteristics of the fire scene;

[0051] Step S13: Analyze the gas component information according to the real-time parameter characteristics of the fire scene and extract the environmental gas component information;

[0052] Step S14: Analyze the change of gas components in the environmental gas component information to obtain the change characteristics of gas components;

[0053] Step S15: Calculate the multi-dimensional gas concentration according to the real-time parameter characteristics of the fire scene to generate multiple gas concentration time series curves;

[0054] Step S16: Predict the gas change trend based on the gas component change characteristics and multiple gas concentration time series curves to generate the change trend of on-site gas parameters.

[0055] In this embodiment, at the fire scene, the arrangement and selection of sensors are carried out first. Select sensors suitable for fire monitoring, such as temperature sensors, smoke sensors, gas sensors (such as CO, CO 2 , O 2 , NO 2etc.) to ensure comprehensive monitoring of environmental changes. Determine the installation locations of sensors, giving priority to areas where fires are likely to occur, such as flammable material storage areas and ventilation openings. Ensure that the sensors can quickly capture environmental changes during a fire. Build a real-time data acquisition system and connect the data of each sensor to a central data processing unit. Select high-frequency data acquisition devices and set the sampling frequency to 1 Hz to ensure that changes in environmental parameters can be captured in real time. Ensure the stability of data transmission and use wireless transmission technologies (such as Zigbee, LoRa) to ensure effective connection at the fire scene. Start the data acquisition system and monitor the environmental parameters at the fire scene in real time. Record the output data of each sensor, including temperature, humidity, gas concentration, etc., to ensure the integrity and accuracy of the data. Set up a data storage mechanism and store the real-time monitoring data on a server or in the cloud for subsequent analysis and processing. Ensure the security and traceability of the data. Preprocess the obtained real-time environmental monitoring data, including denoising, smoothing, and outlier detection. Use the moving average method to smooth the data and reduce the fluctuations caused by environmental interference. Set thresholds to identify and remove abnormal data that does not conform to reality. For example, if the reading of the temperature sensor exceeds the set maximum value, it needs to be marked as an outlier. Extract key features from the processed environmental monitoring data, including statistical features such as mean, maximum, minimum, and standard deviation. These features will be used to describe the environmental conditions at the fire scene. Combine time series analysis to extract the change trends and periodic features of the data, and record the changes in environmental parameters at different time periods. Organize the extracted environmental parameter features into a report, detailing the real-time environmental status at the fire scene, including changes in parameters such as temperature, humidity, and gas concentration. Generate visual charts, such as line charts and bar charts, to intuitively display the change trends of parameter features and provide support for subsequent analysis. Extract the relevant data of the gas sensor from the real-time environmental monitoring data, including carbon dioxide (CO 2 ) and carbon monoxide (CO), oxygen (O 2)and the concentration values of other harmful gases. Set the time range for data extraction, for example, extract gas composition data once per minute to form time series data of gas composition. Select a suitable gas composition analysis method, such as principal component analysis (PCA) or clustering analysis, to identify the relationships between different gas components. This method can help extract representative gas composition characteristics. Set analysis parameters, for example, select the first two principal components to explain the main reasons for gas composition changes. Organize the results of gas composition analysis into a report, recording the concentration changes of each gas component and its corresponding environmental parameter characteristics. Generate visual charts to show the trends of different gas components over time, providing support for subsequent gas composition change analysis. Conduct change analysis on the extracted gas composition information. Calculate the change rate of each gas component and analyze its increasing or decreasing trend over time. Set the calculation formula for change characteristics, for example, change rate = (current concentration - previous concentration) / previous concentration, to quantify the change characteristics of gas components. Use time series analysis methods to conduct trend analysis on gas composition changes. Use the autoregressive moving average (ARMA) model to evaluate the change pattern of gas components over time. Record change characteristics, including peak values, valley values, and their corresponding time points, to identify critical moments of gas composition changes. Organize the gas composition change characteristics into a report, describing in detail the change laws of different gas components and their potential impacts on the fire scene environment. Generate visual charts to show the trends of gas composition change characteristics, providing data support for subsequent gas concentration calculations. According to real-time monitoring data, select a suitable gas concentration calculation method, such as using the ideal gas state equation or a linear regression model based on the output of gas sensors. Set the calculation formula to ensure that it can accurately reflect the concentration changes of each gas. Implement multi-dimensional gas concentration calculations, independently calculating for different gas components. Record the change of the concentration of each gas over time to form multiple gas concentration time series curves. Set a time window, for example, record the concentration value once per minute to ensure the timeliness and accuracy of the data. Generate time series curves for the calculated gas concentration data to visually show the trends of different gas concentrations over time. Use visualization software to generate charts to ensure the charts are clear and easy to read. Record the key characteristics of the time series curves, such as the maximum concentration, minimum concentration, and their corresponding time points, providing a reference for subsequent prediction of gas change trends. Select a suitable gas change trend prediction model, such as a time series prediction model (such as ARIMA) or a machine learning model (such as support vector machine, neural network). Set the model parameters to ensure that it can accurately capture the trend of gas concentration changes. Use historical gas concentration data to train the selected prediction model. Divide the data set into a training set and a validation set to ensure the accuracy and generalization ability of the model after training. Record the prediction results of the model, including the predicted gas concentration values and their change trends, and compare them with the actual monitoring data to evaluate the accuracy of the model.Organize the prediction results into a report, describing in detail the gas change trends and their potential impacts on the fire scene environment. Record the key parameters of the prediction, such as the maximum and minimum gas concentrations and their change rates. Generate visual charts to show the predicted trends of gas changes, providing support for on-site safety monitoring and decision-making.

[0056] In this embodiment, refer to Figure 3 , which is a schematic diagram of the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0057] Step S21: Obtain the thermal imaging image of the fire scene based on the drone;

[0058] Step S22: Remap the pixel gray values of the thermal imaging image of the fire scene to construct a brightness-optimized thermal imaging image;

[0059] Step S23: Perform visual recognition of the temperature distribution on the brightness-optimized thermal imaging image to generate the on-site temperature distribution field;

[0060] Step S24: Accurately identify the hot spots in the on-site temperature distribution field and mark the on-site fire source points;

[0061] Step S25: Calculate the spatial positioning of the fire source points on-site to generate the spatial coordinates of each fire source;

[0062] Step S26: Calculate the temperature change rate of consecutive frames based on the spatial coordinates of each fire source to generate the temperature change rate in the fire source area.

[0063] In this embodiment, a suitable drone is selected to ensure its ability to carry a thermal imaging camera. The drone should have good flight stability and a long endurance time to enable effective thermal imaging monitoring at the fire scene. A high-resolution thermal imaging camera is configured to ensure that it can capture subtle temperature changes. The sensitivity of the camera should reach 0.1°C to accurately record the thermal distribution at the fire scene. A flight plan for the drone is formulated, including flight altitude, speed, and path. Generally, a cruising altitude of 30 to 50 meters is selected to obtain a wide coverage area and detailed thermal imaging images. At the fire scene, the drone is launched for flight to ensure the stability of the camera during flight. The flight status of the drone is monitored in real time to ensure that it can complete the thermal imaging task safely and effectively. During flight, thermal imaging images are collected in real time. The drone should take thermal images regularly, setting to save one image every 5 seconds to ensure capturing the dynamic changes at the fire scene. The collected thermal imaging images are stored in the built-in memory or external storage device of the drone for subsequent processing and analysis. First, preprocess the collected thermal imaging images, including denoising and enhancing contrast. Use a filtering algorithm (such as Gaussian filtering) to remove the noise in the image to improve the image quality. Ensure the clarity and details of the image, especially in the heat source area at the fire scene, for subsequent analysis. Perform pixel gray value remapping to linearly or non-linearly adjust the gray values in the thermal imaging image to optimize the brightness and contrast of the image. Set the target gray range, for example, map the gray values from 0 - 255 to 100 - 255 to highlight the heat source area. Apply color mapping technology to convert the optimized gray values into a pseudo-color image to improve the visual effect and readability. This helps to quickly identify the heat source area. Generate a thermal imaging image with optimized brightness and save it in a suitable image format (such as TIFF or PNG) to ensure that the image quality is not lost. Record the processing parameters and conversion methods of the image for subsequent analysis and verification of the image processing effect. Extract temperature data from the thermal imaging image with optimized brightness. According to the calibration data of the thermal imaging device, convert the optimized pixel values into actual temperature values to ensure the accuracy of the temperature corresponding to each pixel. Set the temperature range, usually including the minimum and maximum temperatures at the fire scene, for subsequent temperature distribution analysis. Based on the extracted temperature data, construct a temperature distribution field. Use a grid method to divide the thermal imaging image into multiple grids, and each grid corresponds to a temperature value. Generate a temperature distribution map, using isotherms or heat maps to display the temperature changes in different regions, visually presenting the temperature distribution characteristics at the fire scene. Verify the generated temperature distribution field by comparing it with the temperature values measured on-site to ensure the accuracy and reliability of the temperature distribution field. Record the verification results and adjust the extraction and conversion methods of the temperature data as needed to improve the accuracy of the temperature distribution field.Select a suitable hotspot recognition algorithm, such as a threshold-based method or a clustering analysis method (e.g., K-means clustering), to accurately identify the hotspot areas in the temperature distribution field. Set the recognition parameters, such as the temperature threshold for hotspots. Usually, select the areas with a temperature 10°C higher than the surrounding ambient temperature as hotspots. Apply the selected hotspot recognition algorithm to analyze the temperature distribution field, identify the hotspot areas and mark them. Mark the identified hotspot areas on the image with different colors or symbols. Record the temperature value of each hotspot and its position in the image for subsequent fire source point positioning and analysis. Verify the marked hotspot areas to ensure they match the actual fire source positions. Verification can be carried out by comparing with on-site measured data to ensure the accuracy of hotspot recognition. According to the verification results, adjust the parameters and algorithms of hotspot recognition to improve the recognition effect and accuracy. Select a suitable fire source spatial positioning method, such as a geometric positioning method or a three-dimensional reconstruction technology based on multi-angle thermal imaging data. Ensure that the selected method can accurately calculate the spatial coordinates of the fire source. Determine the required input data, including the image coordinates of hotspots and the flight altitude of the drone, etc. According to the selected method, perform spatial positioning calculations on the marked fire source points. Convert the two-dimensional coordinates in the thermal imaging image into three-dimensional spatial coordinates, considering the flight parameters and viewing angle of the drone. Record the spatial coordinates of each fire source point, including its X, Y, and Z coordinate values, to ensure the accuracy of the coordinates. Organize the calculated fire source spatial coordinates to generate a positioning report of the fire source points, detailing the spatial positions of each fire source and their corresponding temperature values. Record the parameters and methods of the spatial positioning calculation for subsequent analysis and verification. Collect multiple frames of thermal imaging images and extract the temperature values of each fire source point at different time frames. Ensure that the collected images cover the change process of the fire source points. Set the time interval for data collection, such as collecting once every 5 seconds, to ensure that the temperature changes in the fire source area can be captured. Analyze the temperature data of each fire source point and calculate its temperature change rate. Change rate = ΔT / Δt, where ΔT is the temperature change and Δt is the time interval. Record the temperature change rate of each fire source point to ensure the accuracy and consistency of the calculation. Organize the calculated temperature change rates to generate a temperature change rate report for the fire source area, detailing the temperature change conditions of each fire source point and their potential impact on the fire development. Record the parameters and methods of the temperature change rate calculation to provide a basis for subsequent fire monitoring and emergency response.

[0064] In this embodiment, the specific steps of step S22 are as follows:

[0065] Detect the image noise points in the thermal imaging image of the fire scene and mark the abnormal image noise points;

[0066] Perform adaptive noise filtering on the abnormal image noise points to generate a filtered and noise-reduced thermal imaging image;

[0067] Calculate the gray histogram distribution of the filtered and noise-reduced thermal imaging image to obtain the gray histogram distribution value;

[0068] Perform non-linear gray range stretching on the gray histogram distribution value to obtain the gray stretching parameter;

[0069] Based on the gray stretching parameter, remap the gray value of each pixel of the filtered and noise-reduced thermal imaging image to construct a brightness-optimized thermal imaging image.

[0070] In this embodiment, a suitable noise detection algorithm is selected, such as median filtering or local variance analysis. These algorithms can effectively identify and mark abnormal noise points in the thermal imaging image. Detection parameters are set, such as the noise threshold. Usually, the threshold is set to twice the standard deviation of the temperature values in the image to determine which pixel values are considered abnormal. The selected noise detection algorithm is applied to the thermal imaging image. First, the statistical features of the neighborhood around each pixel are calculated, such as the mean and standard deviation. The pixels that exceed the set threshold are marked as abnormal noise points. Different colors or symbols can be used to mark them on the image for subsequent processing and analysis. The number and location of the detected abnormal noise points are recorded, and a noise detection report is generated. This will provide a basis for subsequent filtering and noise reduction processing. Preliminary analysis is carried out to evaluate the impact of the noise on the overall image quality and ensure that important heat source information is not missed during the processing. A suitable adaptive filtering algorithm is selected, such as adaptive median filtering or Wiener filtering. These algorithms can dynamically adjust the filtering parameters according to the local image characteristics, thereby effectively removing noise. The window size of the filter is determined. Usually, a neighborhood of 3x3 or 5x5 is selected to balance the computational efficiency and filtering effect. The adaptive filter is applied to the thermal imaging image, traversing each pixel and calculating the statistical features of the surrounding neighborhood. The filtering parameters are dynamically adjusted according to these features to remove abnormal noise points. The results of each filtering operation are recorded to ensure that the filtering process can effectively reduce noise while retaining the heat source information. The generated filtered and noise-reduced thermal imaging image is verified by comparing it with the original image to ensure that the noise is effectively removed and the heat source features are retained. According to the verification results, the filtering parameters are adjusted to improve the filtering effect. The optimized parameters are recorded for subsequent analysis and application. The gray value statistics of the filtered and noise-reduced thermal imaging image are carried out. Each pixel in the image is traversed, and its gray value is counted into the corresponding histogram bin. Usually, the gray value range is set to 0-255 to cover all possible pixel values. 256 bins are selected to ensure that each gray value has a corresponding statistic. The occurrence frequency of each gray value is calculated, and a gray histogram is generated. The frequency of each gray value is recorded for subsequent analysis and processing. The characteristic parameters of the histogram are calculated, such as the mean, variance, and kurtosis, to describe the overall gray distribution characteristics of the image. A visualization chart of the histogram is generated to intuitively display the distribution of gray values. Ensure that the chart is clear and easy to read, and the frequencies of each gray value are marked. The calculation results and visualization chart of the histogram are recorded for subsequent analysis and optimization. A non-linear gray range stretching method is selected, such as gamma correction or logarithmic transformation. This method can effectively enhance the contrast of the image and highlight important temperature features. The stretching parameters are set. For example, the gamma value is usually set between 0.5 and 2 to find the optimal contrast enhancement effect. According to the selected non-linear stretching method, the gray histogram distribution values are processed. The formula is applied to convert the gray values, the stretching results of each gray value are recorded, a new gray histogram is generated, and a comparative analysis is carried out with the original histogram.Verify the grayscale histogram after stretching to ensure that it can effectively enhance the contrast of the image. Observe whether the distribution of the histogram is more concentrated and whether it can better reflect the temperature characteristics in the image. Select a suitable grayscale value remapping method and perform pixel-by-pixel mapping according to the grayscale stretching parameters obtained in the previous step. Determine the mapping formula, such as linear mapping or non-linear mapping based on gamma correction, to ensure that the image brightness is optimized. Traverse each pixel in the filtered and noise-reduced thermal imaging image and perform grayscale value remapping according to the grayscale stretching parameters. Record the new grayscale value of each pixel during the mapping process. Generate a thermal imaging image with optimized brightness and save it in a high-quality image format to ensure that the image information is not lost. Verify the generated thermal imaging image with optimized brightness to ensure that it is superior to the original image in terms of contrast and details. Evaluation can be carried out through visual inspection and computer analysis methods. Generate a visualization report of the optimized image, display the comparison before and after optimization, and record the key parameters and results during the optimization process for subsequent application and analysis.

[0071] In this embodiment, refer to Figure 4 , which is a schematic diagram of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:

[0072] Step S31: Evolve the fire spread path of the on-site fire source point to obtain the fire spread path;

[0073] Step S32: Fit the fire spread distribution according to the temperature change rate in the fire source area and the fire spread path, and construct a fire spread distribution change diagram;

[0074] Step S33: Mine the fire situation evolution of the fire spread distribution change diagram to obtain the fire situation evolution law;

[0075] Step S34: Based on a preset multi-modal AI large model, simulate the fire spread evolution of the fire situation evolution law and the change trend of on-site gas parameters to generate fire spread evolution simulation data;

[0076] Step S35: Predict the multi-time fire trend of the fire spread evolution simulation data to generate a multi-time fire trend prediction diagram.

[0077] In this embodiment, a suitable fire spread model is selected, such as a heat diffusion model or a fluid dynamics model. These models can simulate the diffusion behavior of flames under different environmental conditions, considering factors such as wind speed, humidity, and temperature. Determine the input parameters of the model, such as the initial temperature of the ignition source, the temperature of the surrounding environment, wind speed, and wind direction. Usually, the wind speed is set at 5 meters per hour and the humidity is 30%. Based on the selected diffusion model, perform simulation calculations on the ignition source. Use numerical methods (such as the finite difference method or the finite element method) to solve the equations and obtain the fire spread paths at different time points. Record the fire spread situation at each time step (e.g., every minute) to generate a fire spread path diagram, showing the change of the fire over time. Analyze the calculated fire spread paths to identify the main direction, speed, and potentially affected areas of the fire spread. Ensure that the path diagram can clearly reflect the dynamic characteristics of the fire spread. Generate a visualization chart to show the fire spread path and its changes, providing an intuitive basis for subsequent fitting of the fire spread distribution. Collect the temperature change rate data of the on-site fire source area, including the temperature changes at different time points. These data can be obtained through a thermal imaging monitoring system to ensure their accuracy. Set the data extraction time interval, such as recording the temperature change rate every 5 minutes, to capture the changes during the fire spread process. Use statistical analysis methods or machine learning techniques (such as regression analysis or clustering analysis) to fit the fire spread distribution. Combine the temperature change rate with the fire spread path to establish a mathematical model of the fire spread. Set the fitting parameters, such as the influence weight of the temperature change rate, to ensure that the model can reasonably reflect the characteristics of the fire spread. Generate a fire spread distribution change diagram according to the fitting results, intuitively showing the distribution of the fire at different time points. The diagram should mark the areas with large temperature changes and the main direction of the fire spread. Record the key parameters and results during the fitting process to provide a basis for subsequent mining of the fire situation evolution. Select a suitable fire situation evolution analysis method, such as time series analysis or state transition model. These methods can help identify the key change points and evolution laws during the fire spread process. Set the analysis parameters, such as the time window (e.g., 5 minutes), to analyze the evolution trend and change characteristics of the fire. Conduct an in-depth analysis of the fire spread distribution change diagram to identify the patterns and laws during the fire spread process. Use graph algorithms to analyze the paths, speeds, and influencing factors of the fire propagation. Record the key evolution laws found, including the acceleration period, deceleration period, and possible turning points of the fire spread, ensuring the integrity and accuracy of the information. According to the analysis results, generate a fire situation evolution law report, detailing the changes and potential influencing factors during the fire spread process. Record the key parameters of the evolution laws to provide a reference basis for subsequent simulations and predictions. Select a suitable multi-modal AI large model, such as a combination of a deep learning-based image recognition model and a gas parameter analysis model, to comprehensively simulate the fire spread.Configure the input parameters of the model, including the evolution law of the fire situation, on-site gas parameters, and historical data, to ensure that the model can comprehensively reflect the dynamic changes at the fire scene. Input the collected evolution law of the fire situation and gas parameters into the AI model for simulating the spread and evolution of the fire. Set the simulation time range, for example, simulate the fire situation changes within the next 1 hour. Record the key data during the simulation process, including the fire spread speed, temperature changes, and gas concentration changes, etc., to ensure the accuracy and real-time nature of the data. Analyze the generated simulation data of the fire spread and evolution to ensure its rationality and effectiveness. Compare the simulation results with the actual monitoring data to evaluate the prediction ability of the model. Select a suitable trend prediction model, such as the ARIMA model or the long short-term memory (LSTM) network. These models can perform trend prediction based on historical data and consider the characteristics of time series. Set the input parameters of the model, including the previous simulation data, gas concentration changes, and fire spread path data, for accurate trend prediction. Use the selected trend prediction model to analyze the simulation data of the fire spread and evolution to generate the fire trend prediction results at multiple future time points. Set the prediction time range, such as predicting the fire dynamics within the next 30 minutes to 1 hour. Record the prediction results at each time point, including the fire spread speed, temperature changes, and potential impact areas, to provide data support for decision-making. Organize the prediction results into a multi-time fire trend prediction chart to visually display the change trend of the fire and the expected impact areas. Ensure that the chart is clear and easy to read, and mark the key data and time points.

[0078] In this embodiment, step S4 includes the following steps:

[0079] Step S41: Identify the points in the area to be rescued based on the thermal imaging image of the fire scene;

[0080] Step S42: Conduct target detection on the trapped persons at the points in the area to be rescued to obtain the persons to be rescued in each area;

[0081] Step S43: Evaluate the rescue priority of each area based on the persons to be rescued in each area, so as to obtain the rescue priority of each area;

[0082] Step S44: Identify the available rescue paths based on the thermal imaging image of the fire scene and extract multiple available rescue routes;

[0083] Step S45: Analyze the optimal rescue route for each area for the multiple available rescue routes according to the multi-time fire trend prediction chart, so as to obtain the optimal rescue route for each area;

[0084] Step S46: Generate multiple area rescue tasks according to the rescue priority of each area and the optimal rescue route of each area.

[0085] In this embodiment, the thermal imaging images of the fire scene are preprocessed, including denoising and enhancing contrast, to ensure that the image quality is high enough for subsequent analysis. Median filtering or Gaussian filtering algorithms can be used to remove noise. Set denoising parameters, such as the filter window size of 3x3, to balance processing speed and image quality. Select a suitable image recognition algorithm, such as a deep learning-based object detection algorithm (e.g., YOLO or Faster R-CNN), to identify the areas to be rescued. When training the model, use a thermal imaging dataset containing different fire scenarios to improve the recognition accuracy. Set threshold parameters, such as the confidence threshold for object detection to be set above 0.5, to ensure that only reliable points in the areas to be rescued are identified. Apply the trained model to infer the processed thermal imaging images to identify the areas to be rescued. Record the position coordinates of each area and its corresponding temperature value to assist subsequent analysis. Generate a regional recognition report, listing in detail the coordinates, features, and relevant environmental information of each area to be rescued to ensure the scientific and systematic nature of subsequent rescue work. Select a suitable object detection model specifically for identifying trapped persons. Convolutional neural networks (CNNs) or improved YOLO models can be used to ensure that the model can effectively identify trapped persons in a complex fire environment. Set the training dataset, including thermal imaging images with and without trapped persons, to improve the robustness and accuracy of the model. Perform object detection on the identified areas to be rescued and extract information about trapped persons in each area. Record the coordinates, size, and temperature value in the thermal imaging image of each object. Set the confidence threshold for object detection, such as set above 0.6, to ensure that only reliable trapped persons are identified. Organize the detected information about trapped persons into a report, listing the number, location, status, and temperature information of the persons to be rescued in each area. Ensure the integrity and accuracy of the information. Record the parameters and results during the detection process to provide reliable data support for subsequent rescue priority assessment. Determine the criteria for regional rescue priority assessment, including the number of trapped persons, status (such as conscious or not), distance from the fire source, and the degree of danger of the surrounding environment, etc. Set evaluation parameters, such as giving a higher weight (1 to 5 points) to the status of trapped persons, while giving a relatively lower weight (1 to 3 points) to the degree of environmental danger. Calculate the priority for each area to be rescued according to the set evaluation criteria. Considering various factors comprehensively, calculate the overall rescue priority score for each area. Record the priority assessment results for each area to ensure the accuracy and consistency of the data for subsequent rescue plan formulation. Organize the priority assessment results into a report, listing in detail the rescue priority of each area, the status of the persons to be rescued, and relevant environmental information. Ensure the clarity and readability of the information for subsequent decision-making. Generate a priority chart to visually display the rescue priorities of each area, providing a basis for subsequent rescue path identification and task allocation.Select a suitable path recognition algorithm, such as the A* algorithm or Dijkstra algorithm, which can calculate the best rescue route based on environmental data. Determine the algorithm input parameters, including the map of the fire scene, the positions of obstacles (such as walls, fire sources, etc.), and the walkable areas. According to the environmental information identified in the thermal imaging image, use the selected algorithm to identify available rescue paths. Record the starting point, ending point, and passing nodes of each path. Set the tolerance of path recognition, for example, allowing a path error within 1 meter to ensure that the recognized path is feasible and safe. Organize the identified multiple available rescue routes into a report, detailing the characteristics, lengths, and potential risks of each path. Generate a visualization chart to show the positions and risk assessments of each path in the fire scene. Use the multi-time fire trend prediction chart to analyze the impact of fire spread on each available rescue route. Identify the speed, direction, and possible dangerous areas of fire spread. Set the analysis time window, for example, evaluate the impact of the fire on the rescue route every 5 minutes to ensure the timeliness of the data. Conduct an optimal rescue route analysis for each available rescue route in each area. Considering the fire spread, the positions of trapped people, and the rescue priorities comprehensively, calculate the safety and effectiveness of each path. Record the evaluation results of each path to ensure the scientific nature and accuracy of the analysis process. Organize the analysis results of the optimal rescue route into a report, listing in detail the optimal paths, potential risks, and the status of trapped people in each area. Ensure the integrity and clarity of the information. Generate an optimal path map to visually show the optimal rescue routes in each area, providing a basis for subsequent rescue task allocation. etc. Set the allocation strategy to ensure that high-priority areas can be rescued in a timely manner. Set the task parameters, for example, each rescue team is responsible for 1-2 areas, and consider the personnel configuration and professional skills of the rescue team. Generate specific rescue tasks according to the rescue priorities and optimal rescue routes in each area. Ensure that each task includes information such as the target area, task description, and execution time. Record the details of each rescue task to ensure the accuracy and consistency of the information, providing support for subsequent implementation. Organize the generated multiple area rescue tasks into a task list, detailing the objectives, priorities, executors, and route information of each task. Ensure the clarity and readability of the task list. Generate a rescue task report to show the allocation of each task and the potential risk assessment, providing decision-making support for on-site command.

[0086] In this embodiment, step S5 includes the following steps:

[0087] Step S51: Identify the fire-fighting resources at the fire scene; conduct a personnel configuration and equipment status analysis of the fire-fighting resources at the fire scene to obtain the on-site fire-fighting rescue resource evaluation value;

[0088] Step S52: Dynamically allocate resources for multiple area rescue tasks according to the on-site fire-fighting rescue resource evaluation value, thereby obtaining dynamic rescue resource allocation data;

[0089] Step S53: Make a rescue decision based on the dynamic rescue resource allocation data, and conduct real-time rescue monitoring to obtain real-time rescue monitoring data.

[0090] In this embodiment, available fire-fighting resources are identified through on-site surveys and monitoring systems (such as drones and thermal imaging devices). This includes firefighters, fire trucks, fire extinguishers, rescue equipment, etc. Ensure that each resource is recorded in detail, including quantity, location, and status. Develop resource identification criteria, classify the resources, and set the identification conditions for each type of resource. For example, firefighters should possess corresponding qualification certificates, and the equipment status should meet safety standards. Analyze the on-site fire-fighting personnel configuration, including the number of people in each team, professional skills, and task allocation. Ensure the accuracy of the information by retrieving on-site monitoring data and personnel registration information. Set evaluation parameters, such as personnel skill levels, work experience, and health status, to ensure that each team can effectively execute tasks. Conduct a detailed analysis of the status of on-site equipment, and check the availability and functional status of fire extinguishers, fire trucks, and other rescue equipment. Record the working conditions and maintenance records of each piece of equipment to ensure its effective use in case of an emergency. Combine the personnel configuration and equipment status to calculate the evaluation value of on-site fire-fighting and rescue resources. The evaluation value can adopt the weighted scoring method to quantitatively evaluate factors such as personnel skills, equipment status, and quantity, and generate a comprehensive score. Based on the evaluation value of on-site fire-fighting and rescue resources, establish a dynamic resource allocation model. Select appropriate optimization algorithms, such as linear programming or genetic algorithms, to achieve the optimal allocation of resources. Determine the input parameters of the model, including the rescue task priorities in each area, the types and quantities of required resources, and the evaluation values of available resources. Run the resource allocation model, calculate the amount of fire-fighting resources required in each area, and make dynamic adjustments according to the evaluation values. For example, for areas with high priorities, allocate more firefighters and equipment first. Record the resource allocation situation in each area to ensure the transparency and accuracy of the allocation process. Dynamically adjust the evaluation values and resource allocation to cope with changes in on-site situations. Organize the dynamic rescue resource allocation data into a report, detailing the resource allocation situation in each area, including personnel, equipment, and their status. Ensure the integrity and traceability of the information. Generate resource allocation charts to visually display the resource configuration in each area, providing data support for subsequent rescue decisions. Based on the dynamic rescue resource allocation data, make rescue decisions. Considering the rescue priorities and resource allocation situations in each area, determine the rescue strategy for each area. Set decision-making parameters, such as rescue objectives, time limits, and resource utilization efficiency, to ensure the scientificity and rationality of the decisions. Establish a real-time rescue monitoring system, integrating on-site monitoring devices (such as cameras, drones, and sensors) to obtain real-time data. This data includes the action status of firefighters, fire spread changes, and on-site personnel conditions. Set the monitoring frequency, for example, update the data every 1 minute, to ensure a timely response to on-site situations. Analyze the obtained real-time rescue monitoring data, evaluate the rescue effect and resource utilization situation. Use data analysis tools to generate real-time monitoring reports, including rescue progress and resource utilization. According to the analysis results, promptly adjust the rescue decisions and resource allocation to ensure the flexibility and effectiveness of the rescue operation.Record the key parameters and decision-making basis during the adjustment process for subsequent evaluation and improvement.

[0091] In this embodiment, step S6 includes the following steps:

[0092] Step S61: Extract instant feedback information based on real-time rescue monitoring data to obtain real-time rescue feedback information;

[0093] Step S62: Identify emergency situations for the real-time rescue feedback information to generate emergency situation data;

[0094] Step S63: Locate the emergency area based on the emergency situation data to obtain emergency situation area location information;

[0095] Step S64: Optimize intelligent resource allocation based on the emergency situation area location information, thereby constructing an intelligent fire rescue optimization strategy.

[0096] In this embodiment, a on-site monitoring system (such as cameras, sensors, drones, etc.) is used to collect real-time rescue monitoring data. These data include various information such as the activity status of firefighters, fire situation changes, and the conditions of trapped people. Set the data collection frequency, for example, collect once every 1 minute, to ensure the timeliness and accuracy of information. This is crucial for emergency response because the situation at the fire scene changes rapidly. Conduct preliminary processing on the collected real-time monitoring data, including denoising and data cleaning. Use data cleaning techniques to remove incomplete or abnormal data points to improve data quality. Adopt data mining techniques to extract key information from the processed data and generate real-time rescue feedback information. This information may include the work efficiency of firefighters, the fire spread speed, and the number of trapped people. Organize the extracted real-time rescue feedback information into a report, detailing the current rescue progress and on-site conditions. The report should include the fire situation, personnel status, and resource usage in each area. Establish criteria for emergency situation identification to identify which situations require special attention. The criteria can include rapid fire spread, an increase in the number of trapped people, resource shortages, etc. Set identification parameters, such as the fire spread rate exceeding a certain threshold (e.g., more than 5 meters per minute), or the number of trapped people exceeding a set value (such as 10 people), to trigger emergency situation identification. Use data analysis techniques to identify whether there is an emergency situation based on the real-time rescue feedback information. Machine learning methods, such as decision trees or support vector machines, can be used to train a model to identify emergency situations. Record the results of each identification to ensure the accuracy and integrity of the data. Generate emergency situation data, including the type of identified emergency situation, occurrence time, and affected area. Organize the emergency situation data into a report, detailing the identified emergency situation and its potential impact on the rescue work. Ensure that the information is clear and easy to read so that relevant personnel can quickly understand the situation. Select a suitable area positioning algorithm, such as clustering analysis or spatial analysis, to determine the areas where emergency situations occur concentratedly. These algorithms can help identify the main affected areas of the fire and trapped people. Set analysis parameters, such as the clustering radius and the minimum number of samples, to accurately identify the emergency areas. Conduct area positioning analysis on the site based on the emergency situation data to identify the most severely affected areas. The K-means clustering algorithm can be used to divide the monitoring data into multiple clusters to identify high-risk areas. Record the location information, impact degree, and relevant data of each emergency area to assist in subsequent resource allocation and rescue decision-making. Such as linear programming or genetic algorithms, to ensure that the resource allocation can be dynamically adjusted according to the real-time situation. Determine the input parameters of the model, including the resource requirements, available resource quantities, and priorities of each emergency area. Run the resource allocation optimization model and allocate resources according to the emergency situation area positioning information. Prioritize the allocation of resources to high-priority emergency areas to ensure the reasonable use of resources. Record the results of each allocation to ensure the transparency and accuracy of the information. Dynamically adjust the resource allocation to cope with changes in the on-site situation.

[0097] In this embodiment, a fire rescue intelligent management system based on a multimodal AI large model is provided, which is used to execute the fire rescue intelligent management method based on the multimodal AI large model as described above, and includes:

[0098] A gas trend prediction module, which is used to obtain real-time environmental monitoring data at the fire scene; analyze environmental parameters of the real-time environmental monitoring data at the fire scene, and predict the gas change trend to generate the on-site gas parameter change trend;

[0099] A temperature distribution recognition module, which is used to obtain a thermal imaging image of the fire scene; perform visual recognition of the temperature distribution and calculate the temperature change rate of consecutive frames based on the thermal imaging image of the fire scene, so as to generate the temperature change rate of the fire source area;

[0100] A fire trend prediction module, which is used to perform multi-time fire trend prediction according to the temperature change rate of the fire source area and the on-site gas parameter change trend, so as to generate a multi-time fire trend prediction map;

[0101] A rescue route analysis module, which is used to identify the points in the area to be rescued based on the thermal imaging image of the fire scene; perform optimal rescue route analysis for each area of the points in the area to be rescued based on the multi-time fire trend prediction map, and generate multiple area rescue tasks;

[0102] A dynamic resource allocation module, which is used to identify the fire fighting resources at the fire scene; perform dynamic resource allocation for multiple area rescue tasks according to the fire fighting resources at the fire scene, and perform real-time rescue monitoring to obtain real-time rescue monitoring data;

[0103] A resource allocation optimization module, which is used to extract instant feedback information and optimize intelligent resource allocation based on the real-time rescue monitoring data, so as to construct an intelligent fire rescue optimization strategy.

[0104] Through the collection of real-time environmental monitoring data, the present invention ensures that the command center and rescue personnel can obtain the changes in on-site gas composition and concentration in the first place, providing data support for safety decision-making. Through the prediction of gas change trends, it is possible to timely warn of the accumulation of harmful gases, such as carbon monoxide, nitrogen oxides, etc., so as to avoid high-risk areas in advance and protect the safety of rescue personnel. The gas trend prediction helps the commander more scientifically evaluate the scope of fire spread, the intensity of the fire source and the changes in dangerous areas, so as to formulate more effective rescue and evacuation strategies. Through temperature distribution identification, the heat source area of the fire can be confirmed in real time, helping rescue personnel quickly lock in the fire source and the main direction of the fire development. By calculating the rate of change of temperature in consecutive frames, the speed of fire spread is evaluated, helping to predict the development trend of the fire, so as to make timely responses when the fire spreads rapidly. Monitoring the rate of temperature change in the fire source area enables on-site command to flexibly adjust resources and rescue strategies to cope with the changes in the fire. Combining the fire source temperature and gas changes can provide a more comprehensive fire prediction, not only paying attention to the temperature of the fire source itself, but also taking into account gas diffusion and the speed of fire spread. By predicting the multi-timepoint trend of the fire, it is possible to understand in advance the possible development of the fire at different time points, providing more forward-looking information for fire commanders and supporting rescue decision-making. The fire trend chart helps the command headquarters clarify the development direction of the fire, make preparations in advance for the allocation of fire resources and the deployment of rescue personnel, ensuring the timeliness and effectiveness of the response. Through thermal imaging images, it is possible to clarify which areas have higher temperatures and which areas have trapped people, so as to quickly identify the areas that need to be rescued first. Based on the fire prediction, the rescue route analysis can calculate the shortest and safest path, avoiding rescue personnel from entering overly dangerous or rapidly spreading fire areas, thus improving rescue efficiency and safety. In a complex fire scene, rescue tasks for multiple areas can be generated and coordinated to ensure the orderly progress of rescue operations. According to the changes in the fire scene, various types of fire resources (such as fire extinguishers, fire trucks, personnel, etc.) are dynamically allocated to ensure the most effective use of resources. As the situation at the fire scene changes, the allocation of resources can be flexibly adjusted to ensure that at different stages, the rescue force can cover the areas most in need. Real-time rescue monitoring data can provide key information such as the usage of fire resources and the changes in the fire scene, helping the commander to make a quick response. Using AI analysis algorithms, the current resource allocation situation is optimized based on real-time data, and the resource configuration is adjusted to adapt to the latest dynamics of the fire scene. Through intelligent decision-making, it is possible to avoid waste of resources or unreasonable allocation caused by human judgment errors, ensuring that the use of each resource is as efficient as possible. As the fire scene information is continuously updated, the resource allocation optimization module can continuously adjust and optimize the strategy according to the actual situation, ensuring the efficiency and flexibility of the rescue process.

[0105] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Accordingly, all changes that fall within the meaning and scope of the equivalent elements of the application documents are intended to be embraced within the present invention.

[0106] As described above, these are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A fire rescue intelligent management method based on a multimodal AI large model, characterized in that: The following steps are involved: Step S1: Acquire real-time environmental monitoring data of the fire scene; Analyze environmental parameters of real-time environmental monitoring data at the fire scene and predict gas change trends to generate on-site gas parameter change trends; Step S2: Acquire a thermal imaging image of the fire scene; perform visual recognition of temperature distribution and calculation of temperature change rate of consecutive frames based on the thermal imaging image of the fire scene, thereby generating a temperature change rate of the fire source area; Step S3: Predict the fire intensity trend at multiple points in time according to the temperature change rate of the fire source area and the change trend of the on-site gas parameters, thereby generating a multi-point fire intensity trend prediction graph; Step S4: Identify the area to be rescued based on the thermal imaging image of the fire scene; Based on the multi-time point fire trend prediction map, the optimal rescue route is analyzed for each area to be rescued, and multiple regional rescue tasks are generated; Step S5: Identify firefighting resources at the fire scene; Dynamically allocate resources to multiple regional rescue tasks based on fire scene firefighting resources, conduct real-time rescue monitoring, and obtain real-time rescue monitoring data; Step S6: Extracting instant feedback information and optimizing intelligent resource allocation based on real-time rescue monitoring data, thereby building an intelligent fire rescue optimization strategy; Among them, the specific steps of step S3 are: Step S31: Evolving the fire spread path of the on-site fire source point to obtain the fire spread path; Step S32: fitting the fire spread distribution according to the temperature change rate of the fire source area and the fire spread path, and constructing a fire spread distribution change diagram; Step S33: performing fire situation evolution mining on the fire spread distribution change graph to obtain the fire situation evolution law; Step S34: Simulate the evolution of fire spread based on the fire situation evolution law and the change trend of on-site gas parameters based on the preset multimodal AI large model to generate fire spread evolution simulation data; Step S35: performing multi-time point fire trend prediction on the fire spread evolution simulation data, thereby generating a multi-time point fire trend prediction graph; The specific steps of step S4 are: Step S41: Identify the area to be rescued based on the thermal imaging image of the fire scene; Step S42: Detect trapped persons at the points in the area to be rescued, and obtain persons to be rescued in each area; Step S43: performing a regional rescue priority assessment based on the persons to be rescued in each area, thereby obtaining a rescue priority for each area; Step S44: identifying available rescue paths based on the fire scene thermal imaging image, and extracting multiple available rescue routes; Step S45: Analyze the optimal rescue route for each area of ​​multiple available rescue routes according to the multi-time point fire trend prediction map, so as to obtain the optimal rescue route for each area; Step S46: Generate multiple regional rescue tasks according to the rescue priority of each area and the optimal rescue route of each area.

2. The fire rescue intelligent management method based on multimodal AI big model according to claim 1 is characterized in that: The specific steps of step S1 are: Step S11: acquiring real-time environmental monitoring data of the fire scene based on sensors; Step S12: performing environmental parameter analysis on the real-time environmental monitoring data of the fire scene to generate real-time parameter features of the fire scene; Step S13: Analyze gas composition information according to the real-time parameter characteristics of the fire scene to extract environmental gas composition information; Step S14: Analyze the change of the ambient gas composition information to obtain the characteristics of the change of the gas composition; Step S15: performing multi-dimensional gas concentration calculation according to the real-time parameter characteristics of the fire scene to generate multiple gas concentration time series curves; Step S16: Predicting the gas change trend based on the gas composition change characteristics and multiple gas concentration time series curves to generate the on-site gas parameter change trend.

3. The fire rescue intelligent management method based on multimodal AI big model according to claim 1 is characterized in that: The specific steps of step S2 are: Step S21: Acquire a thermal imaging image of the fire scene based on the drone; Step S22: remapping the pixel grayscale values ​​of the fire scene thermal imaging image to construct a brightness optimized thermal imaging image; Step S23: performing visual recognition of temperature distribution on the brightness optimized thermal imaging image, thereby generating an on-site temperature distribution field; Step S24: accurately identify hot spots in the on-site temperature distribution field and mark the on-site fire source points; Step S25: performing fire source spatial positioning calculation on the on-site fire source point to generate each fire source spatial coordinate; Step S26: Calculate the temperature change rate of consecutive frames according to each fire source spatial coordinate, so as to generate the temperature change rate of the fire source area.

4. The fire rescue intelligent management method based on multimodal AI big model according to claim 3 is characterized in that: The specific steps of step S22 are: Perform image noise detection on thermal imaging images of fire scenes and mark abnormal image noise points; Adaptively filter abnormal noise points in the image to generate a filtered and denoised thermal imaging image; Calculate the grayscale histogram distribution of the filtered and denoised thermal imaging image to obtain the grayscale histogram distribution value; Performing nonlinear grayscale range stretching on the grayscale histogram distribution value to obtain grayscale stretching parameters; The grayscale value of the filtered and denoised thermal imaging image is remapped pixel by pixel based on the grayscale stretching parameter to construct a brightness optimized thermal imaging image.

5. The fire rescue intelligent management method based on multimodal AI big model according to claim 1 is characterized in that: The specific steps of step S5 are: Step S51: Identify firefighting resources at the fire scene; perform personnel allocation and equipment status analysis on the firefighting resources at the fire scene to obtain an assessment value of the firefighting and rescue resources at the scene; Step S52: dynamically allocating resources to multiple regional rescue tasks according to the on-site fire rescue resource evaluation value, thereby obtaining dynamic rescue resource allocation data; Step S53: Make rescue decisions based on the dynamic rescue resource allocation data, perform real-time rescue monitoring, and obtain real-time rescue monitoring data.

6. The fire rescue intelligent management method based on multimodal AI big model according to claim 1 is characterized in that: The specific steps of step S6 are: Step S61: extracting instant feedback information based on real-time rescue monitoring data to obtain real-time rescue feedback information; Step S62: identifying emergency situations based on the real-time rescue feedback information and generating emergency situation data; Step S63: positioning the emergency area based on the emergency situation data to obtain emergency situation area positioning information; Step S64: Perform intelligent resource allocation optimization based on the emergency situation area positioning information, thereby constructing an intelligent fire rescue optimization strategy.

7. A fire rescue intelligent management system based on a multimodal AI large model, characterized in that: The method for executing the fire rescue intelligent management method based on the multimodal AI big model as claimed in claim 1 comprises: The gas trend prediction module is used to obtain real-time environmental monitoring data at the fire scene; analyze environmental parameters of the real-time environmental monitoring data at the fire scene, and predict gas change trends to generate on-site gas parameter change trends; The temperature distribution recognition module is used to obtain thermal imaging images of the fire scene; perform visual recognition of temperature distribution and calculation of the temperature change rate of consecutive frames based on the thermal imaging images of the fire scene, thereby generating the temperature change rate of the fire source area; The fire trend prediction module is used to predict the fire trend at multiple points in time according to the temperature change rate of the fire source area and the change trend of the on-site gas parameters, thereby generating a multi-point fire trend prediction graph; The rescue route analysis module is used to identify the points of the area to be rescued based on the thermal imaging images of the fire scene; analyze the optimal rescue route for each area to be rescued based on the multi-time point fire trend prediction map, and generate multiple regional rescue tasks; Dynamic resource allocation module, used to identify fire-fighting resources at the fire scene; dynamically allocate resources to multiple regional rescue tasks based on the fire-fighting resources at the fire scene, and conduct real-time rescue monitoring to obtain real-time rescue monitoring data; The resource allocation optimization module is used to extract instant feedback information and optimize intelligent resource allocation based on real-time rescue monitoring data, thereby building an intelligent fire rescue optimization strategy.

Images