A fire hazard early warning processing method, device, medium and product

Abstract

The application provides a fire hazard early warning processing method, device, medium and product. The method comprises the following steps: acquiring real-time smoke concentration and environmental temperature of a target area; when any data exceeds a threshold value, extracting a change curve feature of the monitoring data, performing deviation analysis on a historical evolution curve feature, and generating a hazard confirmation result; after confirming the hazard, acquiring a hazard location, surrounding combustible material distribution data, and real-time location of a processing personnel within a dispatch range; inputting the above information into a hazard development prediction model, calculating a hazard spread risk level and a personnel estimated time consumption; if the risk level and the time consumption are less than preset threshold values, sending a site disposal instruction to the processing personnel; otherwise, sending an upgrade alarm instruction to a fire management terminal. The technical solution provided by the application can achieve objective confirmation of a hazard based on dynamic evolution characteristics, achieve high verification of an early warning signal and scientific hierarchical scheduling of on-site rescue forces through bidirectional deduction of fire spread and response cost.

Description

Technical Field

[0001] This application relates to the field of fire early warning technology, and in particular to a method, equipment, medium and product for early warning and processing of fire hazards. Background Technology

[0002] With the application of smart cities and the Industrial Internet of Things, the demand for fire monitoring in large commercial complexes and complex industrial environments is constantly increasing, which puts forward higher requirements for fire information collection and emergency response capabilities.

[0003] In existing technologies, fire early warning and emergency dispatch are typically based on detection methods using preset thresholds. When smoke concentration or temperature exceeds the threshold, an alarm is triggered, and back-end personnel verify and assign tasks. However, existing technologies have limited analytical capabilities for changes in fire characteristics under complex environments and lack sufficient information on the distribution and changes of dynamic combustibles indoors, making it difficult to support quantitative judgment of fire evolution trends and intelligent emergency decision-making. In the presence of interference sources or dynamic environmental changes, the accuracy of hazard verification and the scientific nature of rescue dispatch are compromised by high false alarm rates, delayed fire assessments, and insufficient safety support for dispatch routes, leading to secondary casualties. Summary of the Invention

[0004] In view of this, this application provides a method, equipment, medium, and product for handling fire hazard early warning, in order to solve the above problems.

[0005] Firstly, a method for early warning and handling of fire hazards is provided, the method comprising:

[0006] Acquire fire characteristic monitoring data for the target area, including real-time smoke concentration and real-time ambient temperature;

[0007] When the real-time smoke concentration exceeds the preset smoke alarm threshold, or the real-time ambient temperature exceeds the preset temperature alarm threshold, the current monitoring time is determined as the trigger time. The change curve features corresponding to the fire characteristic monitoring data within the preset time period including the trigger time are extracted, and the change curve features are compared with the pre-stored historical fire evolution curve features to generate the corresponding hazard confirmation results.

[0008] When the hazard confirmation result indicates that a hazard has occurred in the target area, the location of the hazard in the target area is obtained, and the distribution data of combustibles within the preset environmental feature retrieval range around the hazard location is extracted based on the hazard location, as well as the real-time location information of the personnel handling the hazard within the preset dispatch retrieval range.

[0009] The location of the hazard, the distribution data of combustibles, and the real-time location information of the personnel handling the situation are input into a pre-trained hazard development prediction model to calculate the risk level of the hazard's spread and the estimated time for the personnel handling the situation to reach the hazard location.

[0010] If the risk level of the spread is less than the preset risk threshold and the estimated time is less than the preset time threshold, then send an on-site handling instruction containing the location of the hazard to the mobile terminal carried by the handling personnel.

[0011] If the spread risk level is greater than or equal to the preset risk threshold or the estimated time is greater than or equal to the preset time threshold, an upgraded alarm command containing the location of the hazard will be sent to the preset fire management terminal.

[0012] The above technical solution uses a dual-threshold triggering mechanism and deviation analysis of historical curve characteristics to initially confirm the danger. Based on this, it integrates the dynamic distribution of combustibles around the danger location and the real-time location of personnel handling the situation. It uses a predictive model to simultaneously calculate the risk of fire spread and the time required for personnel to arrive. Finally, it implements a hierarchical dispatch strategy based on these two quantitative indicators, establishing a closed-loop processing mechanism from filtering out environmental false alarms and quantifying fire evolution to dynamic and accurate dispatching. This objectively avoids invalid alarms caused by single sensor alarms and the secondary risks of blindly assigning individual personnel to handle high-risk fire situations.

[0013] Optionally, a deviation analysis is performed between the changing curve characteristics and the pre-stored historical fire evolution curve characteristics to generate corresponding hazard confirmation results, specifically including:

[0014] Extract the temperature rise slope parameter and the smoke concentration peak feature parameter from the feature curve;

[0015] The absolute difference between the temperature rise slope parameter and the fire reference slope in the characteristics of historical fire evolution curve is calculated to obtain the first deviation value.

[0016] The absolute difference between the smoke concentration peak characteristic parameter and the fire reference peak characteristic in the historical fire evolution curve is calculated to obtain the second deviation value.

[0017] Based on the feature weight coefficients corresponding to the target region, a weighted summation is performed on the first deviation value and the second deviation value to obtain the weighted total deviation value;

[0018] When the weighted total deviation is less than the preset deviation threshold, a hazard is determined to have occurred in the target area, and the corresponding hazard confirmation result is output.

[0019] The above technical solution extracts feature parameters from two dimensions: the slope of temperature rise and the peak of smoke concentration. It then calculates the absolute difference and weighted summation with the historical real fire evolution characteristics. This objectively quantifies the degree of matching between the current environmental data change trend and the physical evolution law of the real fire. The algorithm logic directly eliminates irregular threshold fluctuations caused by non-fire interference sources such as cooking fumes, instantaneous dust, or water vapor. At the algorithm level, it achieves high signal-to-noise ratio verification of fire alarm signals.

[0020] Optionally, based on the location of the hazard, extract combustible material distribution data within a preset environmental feature retrieval range around the hazard location, specifically including:

[0021] Acquire static structural data of the building information model and real-time monitoring video stream data within the environmental feature retrieval range;

[0022] Image features are extracted from real-time monitoring video stream data using a pre-built semantic segmentation algorithm to determine the material classification labels of items in the video frame, and items whose material classification labels belong to the preset combustible material category are identified as target items.

[0023] Obtain the outline dimensions of the target item;

[0024] By overlaying the outline dimension parameters and material classification labels onto the static structural data of the building information model, combustible material distribution data is generated.

[0025] The above technical solution combines static structural data from building information modeling with real-time monitoring video streams, and uses semantic segmentation algorithms to extract material labels and outline dimensions of items within the video frame. This enables real-time capture and quantitative overlay of temporarily stacked flammable materials (such as illegally parked electric bicycles, temporarily piled cardboard boxes, etc.) in the target area. It overcomes the limitations of traditional fire protection systems that rely solely on static drawings, and provides highly accurate and real-time updated on-site environmental load data for subsequent fire spread prediction.

[0026] Optionally, obtain the real-time location information of personnel within a preset dispatch retrieval range from the location of the hazard, specifically including:

[0027] Acquire signal reception data from indoor positioning base stations deployed within the scheduling and retrieval range;

[0028] The signal reception data is parsed to extract the identity information of the mobile terminal corresponding to the signal reception data, as well as the arrival angle parameter and arrival time difference parameter of the mobile terminal.

[0029] The basic position coordinates of the mobile terminal are obtained by calculating the three-dimensional spatial coordinates based on the arrival angle parameter and the arrival time difference parameter.

[0030] The identity information is matched with the preset scheduling system data. When the scheduling system data indicates that the target person corresponding to the identity information is on duty, the target person is identified as the processing personnel, and the corresponding basic location coordinates are identified as the real-time location information of the processing personnel.

[0031] The above technical solution receives signals from indoor positioning base stations, uses the angle of arrival and time difference of arrival to calculate three-dimensional spatial coordinates to obtain high-precision positions, and forcibly combines the data from the scheduling system to verify the current duty status of the corresponding mobile terminal holder. This objectively reduces the probability that the system will mistakenly assign emergency response instructions to off-duty personnel, and improves the responsiveness of the dispatched objects and the accuracy of the basic position coordinates.

[0032] Optionally, the location of the hazard, the distribution data of combustible materials, and the real-time location information of the personnel handling the situation are input into a pre-trained hazard development prediction model to calculate the risk level of hazard spread and the estimated time for personnel to reach the hazard location, specifically including:

[0033] By aggregating node features of hazard location and combustible material distribution data through the graph neural network layer in the hazard development prediction model, a spatial spread feature vector is obtained.

[0034] By using the path deduction layer in the hazard development prediction model to perform pathfinding calculations on the hazard location and the real-time location information of the personnel handling the situation, a dynamic navigation feature vector is obtained.

[0035] The spatial spread feature vector and dynamic navigation feature vector are input into the fully connected layer of the hazard development prediction model for regression mapping, and the spread risk level and estimated time are obtained respectively.

[0036] The above technical solution decomposes the hazard development prediction model into a graph neural network layer, a path inference layer, and a fully connected layer, which independently handle the spatial spread and diffusion calculation of combustibles and the physical safety pathfinding calculation of personnel. Finally, regression mapping is performed to fuse the results, which objectively establishes a joint computing architecture that can simultaneously take into account the constraints of the physical evolution law of the fire scene and the physical displacement constraints of personnel. It directly outputs the risk level and time consumption parameters to support the underlying hierarchical scheduling decision-making.

[0037] Optionally, the spatial spread feature vector can be obtained by aggregating node features of the hazard location and combustible material distribution data through the graph neural network layer in the hazard development prediction model, specifically including:

[0038] Extract multiple independent rooms, spatial boundary coordinates of each independent room, and door and window passages connecting adjacent independent rooms from the physical spatial layout within the pre-acquired environmental feature retrieval range.

[0039] Define each independent room as a node, and define doors, windows and passageways as edges between corresponding nodes to construct a connected graph structure;

[0040] Based on spatial boundary coordinates, the data on the distribution of combustibles is analyzed to extract items located in the corresponding independent rooms, and the outline size parameters and material classification labels of the corresponding items are obtained.

[0041] Based on the contour size parameters and material classification labels, the flammable material density feature weights of the corresponding nodes are determined;

[0042] Obtain the preset ventilation volume parameters and real-time opening and closing status of the door and window passages, match the corresponding ventilation damping coefficient according to the real-time opening and closing status, and perform weighted calculation of the preset ventilation volume parameters and ventilation damping coefficients to obtain the edge feature weight of the corresponding edge.

[0043] Map the location of the hazard to a connected graph structure to determine the corresponding initial fire node;

[0044] Based on the flammable material density feature weight and edge feature weight, multi-level information transmission calculation is performed from the initial fire node to the associated neighboring nodes to obtain the spatial spread feature vector.

[0045] The above technical solution maps independent rooms in the physical space as nodes and doors, windows and passages as connected edges. It also converts the real-time extracted item attributes in the room into the flammable material density weight of the corresponding node and the real-time opening and closing status of doors and windows into the ventilation damping weight of the corresponding edge. This objectively restores the real physical laws that "the greater the fire environmental load, the more intense the combustion" and "the through wind accelerates the spread of fire when doors and windows are open". By using the multi-level information transmission of graph nodes, it outputs the spatial spread characteristics of fire that fit the actual building topology.

[0046] Optionally, the path deduction layer in the hazard development prediction model can be used to perform pathfinding calculations on the real-time location information of the hazard and the personnel handling the situation, resulting in a dynamic navigation feature vector, specifically including:

[0047] Extract the real-time location information of the personnel handling the situation as the starting point for navigation, and extract the location of the hazard as the ending point for navigation;

[0048] Based on the preset indoor navigation map, a navigation topology map containing the starting point and ending point of the navigation is extracted. The navigation topology map includes multiple navigation nodes and connected edges that connect adjacent navigation nodes.

[0049] Based on the correspondence of spatial coordinates, the spatial spread feature vector is mapped to the corresponding navigation node and connected edge in the navigation topology graph;

[0050] The fire spread hazard coefficient of the connected edges is determined based on the mapping results, and a passage resistance penalty coefficient is applied to the connected edges whose fire spread hazard coefficient is greater than the preset safety threshold.

[0051] Based on the navigation topology map with superimposed traffic resistance penalty coefficient, the minimum traffic cost path between the pathfinding start point and the pathfinding end point is calculated, and the path sequence features and traffic cost parameters of the minimum traffic cost path are extracted and fused to obtain a dynamic navigation feature vector.

[0052] The above technical solution maps the spatial spread hazard coefficient output from the front end to the indoor navigation topology map in reverse. It applies a passage resistance penalty coefficient to the connected edges that have or are about to have the risk of fire spread, and forces the pathfinding algorithm to avoid high-risk areas in order to calculate the path with the minimum passage cost. This objectively avoids the dangerous routes planned by the system that cross the fire spread area, and provides reliable safety redundancy for the navigation routes issued to the processing personnel.

[0053] In a second aspect, an electronic device is provided, including a processor, a memory, a user interface, and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any of the above.

[0054] Thirdly, a computer-readable storage medium is provided that stores instructions which, when executed, perform the method as described in any of the preceding descriptions.

[0055] Fourthly, a computer program product containing instructions is provided, which, when run on a server, causes the server to perform the method described in the first aspect and any possible implementation thereof.

[0056] Understandably, the electronic device provided in the second aspect, the computer-readable storage medium provided in the third aspect, and the computer program product provided in the fourth aspect are all used to execute the method provided in this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0057] In summary, implementing one or more technical solutions provided in this application has at least the following technical effects or advantages:

[0058] By integrating multi-source IoT sensor data, security video surveillance, building information modeling, and personnel scheduling and positioning data, this application constructs a multimodal fire early warning and dispatch data processing mechanism. This application improves the traditional single static threshold alarm mode into a dynamic dispatch model based on joint verification of multi-source heterogeneous data and spatial physical evolution deduction. This scheme can adaptively match the optimal response echelon and safe route based on the real-time calculated fire spread risk coefficient and dynamic navigation passage cost, thereby effectively improving the success rate of early fire intervention in complex building environments (such as large complexes and industrial plants) and reducing secondary safety risks during on-site dispatch. Attached Figure Description

[0059] Figure 1 This is an exemplary system architecture diagram of a fire hazard early warning and handling method disclosed in this application;

[0060] Figure 2 This is a flowchart illustrating a fire hazard early warning and handling method disclosed in this application;

[0061] Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in this application.

[0062] Explanation of reference numerals in the attached figures: 100, System architecture; 101, First terminal device; 102, Second terminal device; 103, Third terminal device; 104, Network; 105, Server; 301, Processor; 302, Communication bus; 303, User interface; 304, Network interface; 305, Memory. Detailed Implementation

[0063] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0064] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0065] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0066] Figure 1 An exemplary system architecture diagram is shown, illustrating an embodiment of a fire hazard early warning and processing method applicable to this application.

[0067] like Figure 1 As shown, the system architecture 100 may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0068] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as model training applications, video recognition applications, web browser applications, social platform software, etc.

[0069] Terminal devices 101, 102, and 103 can be either hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with displays, including but not limited to smartphones, tablets, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptops, and desktop computers, etc. When terminal devices 101, 102, and 103 are software, they can be installed in the aforementioned electronic devices. They can be implemented as multiple software programs or software modules (e.g., multiple software programs or software modules used to provide distributed services) or as a single software program or software module. No specific limitations are imposed here.

[0070] When terminals 101, 102, and 103 are hardware devices, video capture devices can also be installed on them. These video capture devices can be various devices capable of capturing video, such as cameras, sensors, etc. Users can use the video capture devices on terminals 101, 102, and 103 to capture video.

[0071] Server 105 can be a server that provides various services, such as a backend server for processing data displayed on terminal devices 101, 102, and 103. The backend server can analyze and process the received data and can feed back the processing results (such as recognition results) to the terminal devices.

[0072] It should be noted that a server can be either hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software programs or software modules (e.g., multiple software programs or software modules used to provide distributed services), or as a single software program or software module. No specific limitations are made here.

[0073] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included. In particular, if the target data does not need to be obtained remotely, the above system architecture may exclude the network and include only terminal devices or servers.

[0074] Figure 2 This is a flowchart illustrating a fire hazard early warning and handling method according to an embodiment of this application. This method can be implemented using a computer program or a microcontroller. The computer program can be integrated into an application or run as a standalone utility application. The specific steps of a fire hazard early warning and handling method are described in detail below.

[0075] S201: Acquire fire characteristic monitoring data of the target area, including real-time smoke concentration and real-time ambient temperature.

[0076] In this application embodiment, fire feature monitoring data refers to multi-dimensional environmental measurement indicators used to characterize whether there is a risk of fire occurrence and its physical evolution in a specific physical space. For example, it may be the density value of suspended particulate matter in the air, the temperature value of the degree of heat accumulation in the space, or the concentration of a specific flammable gas continuously collected by various front-end Internet of Things sensing devices.

[0077] Specifically, various types of IoT detection terminals are pre-deployed within a designated target area (representing a specific physical space requiring key fire safety monitoring, such as a specific floor of a building, an independent warehouse, a core computer room, or a complex commercial area). During routine monitoring, underlying environmental sensor signals from these detection terminals are continuously received or actively and periodically retrieved via communication network links. Furthermore, these sensor signals undergo protocol parsing, noise reduction, and timestamp alignment to extract key parameters: real-time smoke concentration reflecting the density of smoke particles in the air at the current moment, and real-time ambient temperature reflecting the thermodynamic state of the air at the current moment. Through the synchronous acquisition and digital encapsulation of these two core environmental parameters, fire characteristic monitoring data for the target area is ultimately obtained, providing the most basic on-site factual data support for subsequent hazard confirmation and dynamic dispatching.

[0078] S202: When the real-time smoke concentration exceeds the preset smoke alarm threshold or the real-time ambient temperature exceeds the preset temperature alarm threshold, the current monitoring time is determined as the trigger time. The change curve features corresponding to the fire characteristic monitoring data within the preset time period, including the trigger time, are extracted. The change curve features are then compared with the pre-stored historical fire evolution curve features to generate the corresponding hazard confirmation result.

[0079] For example, when the system detects that an environmental parameter exceeds the static alarm threshold, it does not immediately determine it as a real fire. Instead, it uses this single-point exceedance event as the wake-up node for the verification procedure. By extracting continuous time-series data before and after the trigger moment, the system elevates the originally isolated instantaneous data into a curve feature reflecting the dynamic evolution trend of the environmental parameter. This dynamic trend is then compared macroscopically with the actual fire standard development trajectory stored in the system. The aim is to identify from the timeline whether the current environmental anomaly is a short-term, irregular disturbance fluctuation (such as an instantaneous spike caused by local steam or dust) or a real fire spreading with a regular deterioration, thereby providing the system with a highly reliable hazard confirmation result based on dynamic trend evolution.

[0080] In one possible implementation, deviation analysis is performed between the changing curve features and pre-stored historical fire evolution curve features to generate corresponding hazard confirmation results. Specifically, this includes: extracting the temperature rise slope parameter and smoke concentration peak feature parameter from the changing curve features; calculating the absolute difference between the temperature rise slope parameter and the fire reference slope in the historical fire evolution curve features to obtain a first deviation value; calculating the absolute difference between the smoke concentration peak feature parameter and the fire reference peak feature in the historical fire evolution curve features to obtain a second deviation value; and performing a weighted summation calculation on the first and second deviation values ​​based on the feature weight coefficients corresponding to the target area to obtain a weighted total deviation value.

[0081] Specifically, the formula for calculating the weighted total deviation is as follows:

[0082]

[0083] in, Indicates the weighted deviation from the total value; This represents the currently extracted temperature rise slope parameter. Indicates the fire reference slope; This represents the peak characteristic parameters of the currently extracted smoke concentration. Indicates the characteristics of fire reference peaks; and Let represent the temperature feature weighting coefficient and the smoke feature weighting coefficient pre-configured by the system for the target area, respectively, and satisfy . By using the relative deviation ratio between the current parameters and the reference parameters for weighting, not only is the calculation bias caused by the different physical dimensions and numerical magnitudes of temperature and smoke concentration effectively eliminated, but it can also more accurately amplify small but critical environmental anomalies. When the total weighted deviation is less than a preset deviation threshold, a hazard is determined to have occurred in the target area, and the corresponding hazard confirmation result is output.

[0084] In this application embodiment, the historical fire evolution curve features refer to the standard data baseline or experience model that is collected in advance and stored in a structured manner, which can objectively reflect the dynamic evolution of physical environment parameters in the early stage of past real fires over time. For example, it can be the average slope value of the sharp rise in standard temperature statistically summarized from multiple confirmed real fire events in the past, as well as the peak shape parameters when the smoke concentration reaches the extreme value.

[0085] Specifically, the system first uses data fitting and extreme value detection algorithms to perform dimensionality reduction analysis on the previously acquired time series data, extracting the temperature rise slope parameter (used to represent the steepness or gentleness of temperature change over time) and the smoke concentration peak feature parameter (used to represent the extreme value and abrupt change pattern of smoke particle aggregation) from the change curve features. In the temperature dimension, the absolute difference between the extracted temperature rise slope parameter and the fire reference slope in the historical fire evolution curve features is calculated to quantify the distance between the current temperature growth rate and the actual fire baseline, obtaining the first deviation value; simultaneously, in the smoke dimension, the absolute difference between the smoke concentration peak feature parameter and the fire reference peak feature in the historical fire evolution curve features is calculated to obtain the second deviation value.

[0086] Furthermore, considering the varying tolerances of different spatial environments to environmental parameter fluctuations (e.g., kitchens are highly tolerant of smoke, while server rooms are sensitive to temperature), pre-configured feature weighting coefficients corresponding to the target area are retrieved. Based on these coefficients, a weighted summation of the first and second deviation values ​​is performed, resulting in a comprehensive quantitative weighted deviation value. This weighted deviation value is compared to a preset deviation threshold set by the system. When the weighted deviation value is less than the threshold, it indicates that the current environmental abrupt change trend closely matches the physical evolution of historical real fires. This successfully filters out accidental fluctuations caused by non-fire-related interference sources such as water vapor, dust, or brief open flames, thereby confirming a potential hazard in the target area and outputting the corresponding hazard confirmation result.

[0087] S203: When the hazard confirmation result indicates that a hazard has occurred in the target area, obtain the hazard location in the target area, extract combustible material distribution data within a preset environmental feature retrieval range around the hazard location based on the hazard location, and obtain the real-time location information of the personnel handling the hazard within a preset dispatch retrieval range from the hazard location.

[0088] For example, after the system outputs a confirmed fire alarm, it triggers a multi-dimensional environmental situation awareness mechanism. At this point, the system uses the determined location coordinates of the hazard as a reference center point, expands outwards to a preset range, and simultaneously triggers data acquisition commands in two dimensions: First, it extracts the dynamic combustible material distribution characteristics (i.e., environmental fire load) within the range, providing material parameter support for subsequent physical and thermodynamic spread simulations of the fire in three-dimensional space; second, it queries the real-time location coordinates of personnel with attendance permissions within the dispatch range. Through the simultaneous extraction of the aforementioned environmental material characteristics and personnel spatial status, the system establishes an initial simulation input matrix containing multiple underlying variables.

[0089] In one possible implementation, combustible material distribution data within a preset environmental feature retrieval range around the hazard location is extracted based on the hazard location. Specifically, this includes: acquiring static structural data of a building information model and real-time monitoring video stream data within the environmental feature retrieval range; extracting image features from the real-time monitoring video stream data using a pre-built semantic segmentation algorithm to determine the material classification labels of items within the video frame, and identifying items whose material classification labels belong to a preset combustible material category as target items; acquiring the outline dimension parameters of the target items; and overlaying the outline dimension parameters and material classification labels onto the static structural data of the building information model to generate combustible material distribution data.

[0090] In this embodiment of the application, combustible material distribution data refers to a comprehensive set of environmental features used to characterize the specific spatial location, material properties, and physical volume of flammable or combustible items in a specific three-dimensional physical space. For example, it can be a three-dimensional building topology digital base map superimposed with dynamic fire load information such as temporarily piled waste cardboard boxes in corridors and illegally parked lithium battery electric vehicles indoors.

[0091] Specifically, centered on the location of the hazard, an environmental feature retrieval range is defined according to a set spatial distance. Static structural data of the building information model (BIM) corresponding to this range is retrieved from the underlying database (this represents a pre-constructed basic spatial architecture model containing fixed building components such as walls, doors, windows, and corridors, along with their three-dimensional spatial coordinates). Simultaneously, real-time monitoring video stream data transmitted from on-site security equipment covering this range (this represents a sequence of dynamic visual images actually occurring at the current moment) is acquired. A pre-constructed semantic segmentation algorithm (a deep learning visual network model capable of pixel-level image classification) is used to extract image features frame-by-frame or by extracting individual frames from the captured real-time monitoring video stream data. This deeply analyzes the visual semantic information of each pixel block in the image, thereby accurately determining the material classification labels (e.g., wood, paper, plastic, or fabric) of items within the video frame. The extracted material classification labels are then matched and screened against a pre-configured list of flammable and explosive materials. Items with material classification labels belonging to a preset combustible material category are precisely filtered out and identified as target items. After locking onto the target object, target detection and spatial ranging technologies from computer vision are further utilized, combined with the internal and external calibration parameters of the camera, to calculate the object's length, width, height, and floor area in real three-dimensional space. This yields the object's outline dimensions (representing its physical volume, which is directly related to the subsequent calorific value and spread potential of the fire). Using a spatial coordinate mapping mechanism, the aforementioned outline dimensions and corresponding material classification labels are precisely superimposed onto the corresponding spatial nodes in the static structural data of the building information model, according to their actual relative positions. This transforms the static structural drawings, which originally only contained fixed building components, into a digital twin model that includes dynamic temporary fire load variables. Ultimately, this generates combustible material distribution data for subsequent accurate prediction of the spatial spread direction and speed of the fire.

[0092] In one possible implementation, image feature extraction is performed on real-time monitoring video stream data using a pre-built semantic segmentation algorithm to determine the material classification labels of objects within the video frame. Specifically, this includes: evaluating the visibility degradation index of the current frame image of the real-time monitoring video stream data and determining whether the visibility degradation index exceeds a preset smoke obscuration threshold; if the visibility degradation index exceeds the preset smoke obscuration threshold, a pre-deployed dark channel prior algorithm or a generative adversarial network-based image desmoke and defogging model is invoked to perform pixel-level transmittance mapping and scene reflectance restoration on the current frame image, resulting in a smoke-enhanced image; simultaneously, the thermal distribution matrix returned by an infrared thermal imaging sensor coaxially mounted with a visible light camera is retrieved, and the smoke-enhanced image and the thermal distribution matrix are fused across modalities to generate a multimodal fusion feature map; the multimodal fusion feature map is input into a pre-built semantic segmentation algorithm for pixel-level semantic classification and boundary segmentation to extract and determine the true material classification labels of objects obscured by smoke within the video frame.

[0093] In one possible implementation, obtaining the real-time location information of personnel within a preset dispatch retrieval range from the location of the hazard includes: acquiring signal reception data from indoor positioning base stations deployed within the dispatch retrieval range; parsing the signal reception data to extract the identity information of the mobile terminal corresponding to the signal reception data, as well as the arrival angle parameter and arrival time difference parameter of the mobile terminal; performing three-dimensional spatial coordinate calculation based on the arrival angle parameter and arrival time difference parameter to obtain the basic location coordinates of the mobile terminal; matching the identity information with preset scheduling system data, and if the scheduling system data indicates that the target personnel corresponding to the identity information are on duty, identifying the target personnel as the handling personnel and determining the corresponding basic location coordinates as the real-time location information of the handling personnel.

[0094] In the embodiments of this application, real-time location information refers to comprehensive positioning data used to characterize the precise three-dimensional physical space coordinates of a specific person at the current moment and their current duty availability status. For example, it may be the specific three-dimensional space coordinates of a security personnel in a certain park on a certain floor and in a certain defense zone, which is superimposed with the current on-duty status attribute.

[0095] Specifically, the system first continuously or periodically acquires signal reception data (the original message containing the underlying physical characteristics of the radio frequency signal captured by the base station) from indoor positioning base stations (representing fixed positioning anchor devices installed inside buildings that can receive and process wireless radio frequency signals) deployed within the scheduling and retrieval range through the underlying communication interface. The system then parses the signal reception data using a dedicated communication protocol stack, extracting and retrieving the unique hardware address or built-in account information of the mobile terminal corresponding to the signal reception data (representing portable wireless communication devices such as smartphones, explosion-proof walkie-talkies, or electronic name tags carried by security or fire personnel), as well as the arrival angle parameter (representing the specific spatial incident angle of the wireless radio frequency signal entering the base station antenna array) and arrival time difference parameter (representing the time difference between the arrival of the same radio frequency signal at different physical location base stations) presented in the physical layer ranging. Using a spatial geometric positioning algorithm, high-precision three-dimensional spatial coordinate calculation is performed based on the extracted arrival angle parameter and arrival time difference parameter to derive the absolute or relative coordinates of the device in the current building model, thus obtaining the basic position coordinates of the mobile terminal. To ensure that emergency dispatch instructions are issued to actually available personnel and to avoid rescue delays due to mis-dispatch, the previously parsed identity information is further linked and matched with the pre-set scheduling system data (a database of employee attendance time information pre-entered into the enterprise management backend, including employee shifts, leave, and on-duty status). If the scheduling system data clearly indicates that the target personnel corresponding to the identity information are currently on duty (i.e., excluding unavailable states such as off-duty rest, leave, or lack of handling authority), the target personnel are confirmed as qualified to receive and handle the alarm, and the corresponding basic location coordinates are finally determined as the real-time location information of the personnel, thus providing accurate and effective data support for subsequent rescue route planning and instruction issuance.

[0096] S204: Input the location of the hazard, the distribution data of combustibles, and the real-time location information of the personnel handling the hazard into a pre-trained hazard development prediction model to calculate the risk level of the hazard's spread and the estimated time for the personnel handling the hazard to reach the location of the hazard.

[0097] For example, after acquiring the aforementioned multidimensional on-site data, the system synchronously inputs this data into the input layer of the hazard development prediction model. Based on a digitized three-dimensional spatial topology, this model performs forward parallel computation in two dimensions: first, by combining spatial ventilation volume and flammable material density weights, it calculates the quantitative probability and affected area of ​​fire spreading across nodes in the spatial network; second, by combining physical spatial access barriers and high-risk fire zones, it calculates the minimum cost path and estimated time for personnel to reach the hazard location. Through these two joint computation links, the system directly outputs a quantified spread risk index and response time parameters, thereby providing objective data for subsequent automated distribution of disposal instructions.

[0098] In one possible implementation, the location of the hazard, the distribution data of combustible materials, and the real-time location information of the personnel handling the situation are input into a pre-trained hazard development prediction model to calculate the spread risk level of the hazard and the estimated time for the personnel handling the situation to reach the hazard location. Specifically, this includes: aggregating node features of the hazard location and combustible material distribution data through a graph neural network layer in the hazard development prediction model to obtain a spatial spread feature vector; performing pathfinding calculations on the hazard location and the real-time location information of the personnel handling the situation through a path deduction layer in the hazard development prediction model to obtain a dynamic navigation feature vector; and inputting the spatial spread feature vector and the dynamic navigation feature vector into a fully connected layer in the hazard development prediction model for regression mapping to obtain the spread risk level and the estimated time, respectively.

[0099] In this embodiment of the application, the hazard development prediction model refers to a deep learning composite network that is pre-trained based on a large amount of historical fire evolution data and building topology, and includes multiple specific functional network architectures to achieve multi-dimensional task joint evaluation. For example, it can be a joint architecture that superimposes spatial feature extraction and dynamic path planning dual processing branches, and is used to represent an intelligent evaluation engine that can simultaneously deduce the future physical spread intensity of the fire and the cost of emergency response based on the input static and dynamic environmental variables of the fire scene.

[0100] Specifically, the system first inputs the data on the location of the hazard, the distribution of combustibles, and the real-time location information of the personnel involved in the previous steps into a pre-trained hazard development prediction model for parallel feature extraction and calculation. In the internal forward propagation process of this model, on the one hand, the graph neural network layer in the hazard development prediction model (referring to a network layer that is good at processing non-Euclidean spatial topology data and learns through the relationship between nodes) performs node feature aggregation on the input hazard location and combustible material distribution data (used to represent the multi-level information transmission and feature weighting update of physical information such as the flammable material reserves and building spatial connectivity of the ignition point and its surroundings between the nodes and edges of the graph structure), thereby accurately extracting the spatial spread feature vector containing the physical diffusion potential and energy gradient of the fire; on the other hand, the path deduction layer in the hazard development prediction model (referring to a network architecture or algorithm layer that integrates spatial distance calculation and obstacle avoidance logic) performs pathfinding calculation on the hazard location (as the pathfinding endpoint) and the real-time location information of the personnel (as the pathfinding starting point). During the calculation process, the three-dimensional spatial distance of the indoor passage and the physical obstruction of the building are comprehensively considered to extract a dynamic navigation feature vector containing the movement trajectory route and the cost of passage resistance. After completing the independent feature extraction of the two branches, the spatial spread feature vector representing the evolution of the "fire" and the dynamic navigation feature vector representing the response capability of "humans" are spliced ​​and fused, and then input into the fully connected layer (a densely connected network structure that achieves feature space dimensionality reduction and transformation through the multiplication of multi-layer neuron weight matrices) in the hazard development prediction model for high-dimensional feature regression mapping (used to represent the transformation of abstract multi-dimensional feature vectors into concrete, continuous numerical values ​​or classification labels with actual physical meaning). Finally, the spread risk level of the hazard, which is used to quantify the probability of the fire getting out of control, and the estimated time for the personnel to reach the hazard location, which is used to quantify the time cost required for rescue personnel to reach the scene, are calculated simultaneously.

[0101] In one possible implementation, the graph neural network layer in the hazard development prediction model aggregates node features of the hazard location and combustible material distribution data to obtain a spatial spread feature vector. Specifically, this includes: extracting multiple independent rooms within the pre-acquired environmental feature retrieval range, the spatial boundary coordinates of each independent room, and the doors and windows connecting adjacent independent rooms; defining each independent room as a node and the doors and windows as edges between corresponding nodes to construct a connected graph structure; and parsing the combustible material distribution data based on the spatial boundary coordinates to extract items located in the corresponding independent rooms and obtain corresponding... The system calculates the outline dimensions and material classification labels of items; based on these parameters, it determines the flammable material density feature weights of corresponding nodes; it acquires the preset ventilation volume parameters and real-time opening and closing status of doors and windows, matches the corresponding ventilation damping coefficients according to the real-time opening and closing status, and performs a weighted calculation of the preset ventilation volume parameters and ventilation damping coefficients to obtain the edge feature weights of the corresponding edges; it maps the location of the hazard to the connected graph structure to determine the corresponding initial fire node; based on the flammable material density feature weights and edge feature weights, it performs multi-level information transmission calculations from the initial fire node to its associated neighboring nodes to obtain the spatial spread feature vector.

[0102] In the embodiments of this application, a connected graph structure refers to a mathematical network model used to digitally represent the spatial isolation and topological connectivity of a physical building. For example, it can be a graph data structure constructed by using each independent shop and storage room in a large complex building as graph nodes and the doors or ventilation ducts connecting each shop and the corridor as edges connecting each node, to simulate the physical spread path of fire and smoke.

[0103] Specifically, the system first extracts multiple independent rooms (representing enclosed or semi-enclosed spatial units with physical walls) from the pre-acquired environmental feature retrieval range, the spatial boundary coordinates of each independent room (representing the set of geometric polygon vertices defining the three-dimensional spatial range of the room), and the door and window passages connecting adjacent independent rooms (representing the physical medium passages through which fire and high-temperature smoke flow across space). Topological modeling is performed within the computational framework of a graph neural network, defining each independent room as a node and door and window passages as edges between corresponding nodes, thus constructing a connected graph structure. To assign realistic fire load attributes to the graph nodes, based on the spatial boundary coordinates, the system performs spatial inclusion analysis on the input combustible material distribution data, extracting items located within the corresponding independent rooms and obtaining the outline size parameters and material classification labels of the corresponding items. Furthermore, combining the physical volume reflected by the outline size parameters and the calorific value reflected by the material classification labels, the system quantifies and calculates the heat potential, thereby determining the flammable material density feature weight of the corresponding node (representing the degree of energy material accumulation that can fuel fire within a specific space). Simultaneously, to assess the fire's ability to penetrate across space, the inherent preset ventilation parameters of doors and windows (reflecting the ultimate airflow exchange capacity of the physical opening area of ​​doors and windows) and their real-time opening and closing states (such as fully open, closed, or partially closed) are obtained. Based on the real-time opening and closing states, corresponding ventilation damping coefficients (used to characterize the weakening effect of the current physical partition state on hot airflow and flame penetration) are matched. The preset ventilation parameters and ventilation damping coefficients are then weighted and calculated to obtain the edge feature weights of the corresponding edges (used to represent the smooth spread of fire and the probability of heat conduction between adjacent rooms). After assigning physical feature values ​​to the nodes and edges of the graph structure, the actual location of the hazard is mapped to the connected graph structure through coordinate system transformation, accurately locating the space where the fire source occurs and determining the corresponding initial fire node. In the model calculation process of graph neural network, based on the quantized flammable material density feature weights and edge feature weights, multi-level information transmission calculation is performed from the initial fire node to the associated neighboring nodes (referring to the deep learning graph operation process of simulating the energy of the fire field radiating across space along the connected channel and weighted fusion of the features of the neighboring nodes through graph convolution operators). After multiple rounds of iterative aggregation, high-dimensional features that comprehensively encompass the fire's range, spread direction, and the danger gradient of each node are extracted, and finally, the spatial spread feature vector is calculated and obtained.

[0104] Specifically, when performing multi-level information transfer calculations, the aggregation and update formula for node features can be expressed as:

[0105]

[0106] in, Represents a node In the The hidden layer state vector after the information is passed in order; Represents a node The state vector of itself in the previous hidden layer (when At that time, it is the initial flammable material density feature weight vector of that node. Represents nodes The set of connected neighboring nodes; Representing neighboring nodes The state vector of the previous hidden layer; Represents a node With nodes Edge feature weights of connected edges (used to control information from nodes) Passed to node (circulation rate); In a graph neural network, the first... The learnable parameter matrix of the layer (used to implement high-dimensional nonlinear mapping of the feature space). This represents a nonlinear activation function (e.g., ReLU). Through this formula, the model not only absorbs fire energy radiated from neighboring channels but also superimposes the combustion evolution state of the node space itself, and uses the parameter matrix... It achieves deep learning abstraction of high-dimensional features of fire scenes.

[0107] In one possible implementation, the path deduction layer in the hazard development prediction model performs pathfinding calculations on the hazard location and the real-time location information of the personnel handling the situation to obtain a dynamic navigation feature vector. Specifically, this includes: extracting the real-time location information of the personnel handling the situation as the pathfinding starting point and extracting the hazard location as the pathfinding ending point; based on a preset indoor navigation map, extracting a navigation topology map containing the pathfinding starting point and the pathfinding ending point, the navigation topology map including multiple navigation nodes and connected edges connecting adjacent navigation nodes; mapping the spatial spread feature vector to the corresponding navigation nodes and connected edges of the navigation topology map according to the correspondence of spatial coordinates; determining the fire spread hazard coefficient of the connected edges according to the mapping result, and applying a passage resistance penalty coefficient to the connected edges whose fire spread hazard coefficient is greater than a preset safety threshold; based on the navigation topology map after superimposing the passage resistance penalty coefficient, calculating the minimum passage cost path between the pathfinding starting point and the pathfinding ending point, and extracting the path sequence features and passage cost parameters of the minimum passage cost path, and fusing them to obtain the dynamic navigation feature vector.

[0108] In this embodiment, the dynamic navigation feature vector refers to a feature sequence of a safe passage route planned from the current physical location of the rescue personnel to the fire point, which takes into account the shortest spatial distance and the avoidance of secondary fires, under the constraints of the current moment and the future evolution trend of the fire. For example, it can be a multidimensional tensor with safety guidance attributes that superimposes parameters such as the congestion of the passage, the fire blocking status, and the passage time of each node.

[0109] Specifically, the system first extracts the real-time location information of the personnel obtained in the preceding positioning steps (reflecting the personnel's current absolute spatial coordinates) as the starting point for the algorithm's pathfinding, and simultaneously extracts the confirmed hazard locations as the pathfinding endpoints. It then retrieves the system database and, based on a pre-set indoor navigation map (a digital map file representing all accessible physical space elements within a building, such as corridors, stairwells, and elevators), extracts a local navigation topology map that fully covers and includes both the pathfinding starting and endpoints. This topology map includes multiple navigation nodes (representing key locations in accessible spaces such as corridor corners and room entrances) and connecting edges between adjacent navigation nodes (representing the actual walkable physical passages between two navigation nodes, with their inherent weights typically representing physical distance). To mitigate the risk of blindly traversing fire scenes using traditional static path planning, the system precisely maps the spatial spread feature vector (containing high-risk areas of fire spread and the probability of impact) output by the preceding graph neural network to the corresponding navigation nodes and connecting edges in the currently extracted navigation topology map, according to the corresponding mapping relationship of indoor three-dimensional spatial coordinates. By reading the physical evolution parameters mapped into the navigation network, the system dynamically calculates and determines the fire spread hazard coefficient of each connected edge based on the mapping results (used to quantify the probability that the passage will be affected by fire, blocked by smoke, or experience structural collapse within a specific time in the future). An early warning mechanism is established to apply a very large passage resistance penalty coefficient to connected edges with a fire spread hazard coefficient greater than a preset safety threshold (i.e., paths that are judged to be extremely high-risk or whose passage blockage rate exceeds the safety threshold and are therefore impassable). In graph calculation, this is reflected in infinitely amplifying the pathfinding cost weight of the connected edge, thereby achieving physical isolation at the algorithm level. Based on this navigation topology map, which incorporates fire environment resistance and a traffic resistance penalty coefficient, a heuristic search algorithm (such as A or Dijkstra's path optimization algorithm) is invoked to calculate the minimum traffic cost path connecting the pathfinding starting point to the pathfinding ending point, taking into account the shortest physical distance and completely avoiding high-risk blind spots. Feature analysis is then performed on this optimal path to extract the path sequence features (reflecting the spatial travel sequence between nodes) and traffic cost parameters (reflecting the expected time consumption and the secondary safety losses faced). These two high-dimensional features are then mathematically fused to finally generate and obtain a dynamic navigation feature vector.

[0110] In one possible implementation, before inputting the hazard location, combustible material distribution data, and real-time location information of the personnel into a pre-trained hazard development prediction model, the method further includes a step of jointly training the hazard development prediction model. Specifically, this includes: acquiring a training sample set, which includes a large number of historical fire simulation data records; each historical fire simulation data record contains: sample hazard location, sample combustible material distribution data, and sample personnel location as input features, as well as corresponding manually labeled real spread risk level labels and real personnel arrival time labels; constructing an initial hazard development prediction model, and inputting the input features into the initial training model. In the initial hazard development prediction model, forward propagation calculations are performed through its internal graph neural network layer and path inference layer to output the predicted spread risk level and the predicted arrival time of personnel. The prediction errors are calculated separately: the cross-entropy loss function is used to calculate the first loss value between the predicted spread risk level and the actual spread risk level label; the mean squared error loss function or the smoothed L1 loss function is used to calculate the second loss value between the predicted arrival time of personnel and the actual arrival time label; based on the initial configuration and the task weight coefficients that support dynamic adaptive updates, the first loss value and the second loss value are dynamically weighted and summed to obtain the multi-task joint loss value.

[0111] Specifically, the formula for calculating the total joint loss of multiple tasks can be expressed as:

[0112]

[0113] in, This represents the total combined loss across multiple tasks. This is the first loss value (cross-entropy loss) used to calculate the risk level classification. This is the second loss value (mean squared error loss) used to calculate the expected time-consuming regression. and These represent the actual risk level label and the predicted risk level, respectively. and These represent the actual arrival time and the predicted arrival time, respectively. and The task weights are dynamically adjusted. A backpropagation algorithm and a gradient descent optimizer (such as the Adam optimizer) are used to minimize the total joint loss of multiple tasks. The network weight parameters in the graph neural network layers, path deduction layers, and fully connected layers are iteratively updated until the total joint loss of multiple tasks converges to a preset range, thus completing the training of the hazard development prediction model.

[0114] S205: If the spread risk level is less than the preset risk threshold and the estimated time is less than the preset time threshold, then send an on-site handling instruction containing the location of the hazard to the mobile terminal carried by the handling personnel.

[0115] In this embodiment of the application, the on-site handling instruction refers to an electronic dispatch task automatically generated by the fire decision-making backend and issued to the front-line security or emergency personnel, requiring them to use the surrounding fire extinguishing equipment for initial firefighting and on-site confirmation, provided that the current fire is determined to be in an early controllable state and personnel can arrive safely and in a timely manner. For example, it can be a structured alarm work order message that includes the indoor three-dimensional coordinates of the fire point, the best safe passage route map, and a suggestion to carry a portable dry powder fire extinguisher to extinguish the fire.

[0116] Specifically, after obtaining the prediction results output by the preceding model, the system first extracts the spread risk level (used to represent the quantitative probability of the current fire being out of control and spreading to the outside area, calculated by combining the fire load and the building space topology), and compares it numerically with the preset risk threshold set in the system (a critical parameter boundary used to distinguish between the initial stage of the fire that can be extinguished manually by a single soldier and the deflagration stage that requires the intervention of a professional fire brigade). At the same time, it extracts the estimated time (used to represent the theoretical time cost required to reach the fire point by walking or taking a specific means of transportation based on the current physical coordinates of the personnel and after fully considering the resistance of the fire passage), and compares it synchronously with the preset time threshold (a maximum allowable response time limit set to ensure that the fire is controlled in the initial stage). If, after this dual-logic judgment, it is confirmed that the current spread risk level is strictly less than the preset risk threshold (i.e., the current fire intensity is low, surrounding combustibles have not been ignited on a large scale, and there is no risk of explosion or collapse), and at the same time, the estimated time is strictly less than the preset time threshold (i.e., the matched rescue force can safely and quickly reach the target defense zone before the fire is truly out of control), then the current emergency situation is determined to meet the high safety conditions for early intervention and rapid containment by frontline security personnel. The location of the emergency (a specific physical spatial coordinate node representing the actual fire) verified in the pre-process is extracted. This location information, along with emergency response specifications, optimal navigation maps, and other data messages, is structured and packaged, and then pushed to the screen of the mobile terminal (a portable device such as an explosion-proof smartphone, electronic patrol device, or smart bracelet with wireless signal reception, work order display, and navigation interaction functions) carried by the target personnel via a wireless communication network channel. This ultimately completes the sending of on-site handling instructions containing the location of the emergency to the mobile terminal carried by the personnel, thereby achieving very early and precise intervention in the fire while ensuring personnel safety.

[0117] S206: If the spread risk level is greater than or equal to the preset risk threshold or the estimated time is greater than or equal to the preset time threshold, an upgraded alarm command containing the location of the hazard is sent to the preset fire management terminal.

[0118] In this embodiment of the application, the escalation alarm command refers to a high-level alarm signal or dispatch message triggered when it is determined that the current fire scene situation has exceeded the early safe firefighting capabilities of individual soldiers or front-line security personnel, or that the best initial extinguishing opportunity has been missed. This is intended to call for intervention from higher-level professional fire and rescue forces. For example, it may be an emergency reinforcement request work order that is directly pushed to the city's 119 fire command center platform or the central control center of a large park, and is displayed in a large-screen pop-up window with the highest level of audible and visual warnings.

[0119] Specifically, after obtaining the risk and time prediction results output by the preceding model, the system performs an "OR" judgment for safety circuit breaking through multi-dimensional evaluation indicators. It extracts the calculated spread risk level and determines whether it is greater than or equal to a preset risk threshold; simultaneously, it extracts the calculated estimated time and determines whether it is greater than or equal to a preset time threshold. If any of the above high-risk conditions are met (i.e., the current fire intensity is extremely high and faces a large-scale risk of uncontrolled spread, or although the fire is in its early stages, physical barriers prevent rescue forces from arriving within the golden time frame), the routine process of issuing manual firefighting instructions to individual personnel is immediately terminated to prevent secondary casualties. At this point, the system directly retrieves the confirmed location of the hazard from the pre-process, encapsulates the high-risk location data along with underlying situational information such as the predicted direction of fire spread, and pushes it to the pre-set fire management terminal (which refers to the integrated command and control console screen and associated communication platform deployed in high-level fire control rooms, disaster prevention center monitoring rooms, or professional fire brigades) through a dedicated secure communication link, bypassing the conventional dispatch level. This completes the sending of an upgraded alarm command containing the location of the hazard to the pre-set fire management terminal, thereby triggering the highest level of emergency fire plan as quickly as possible and ensuring that professional cluster forces can intervene in the event of a serious hazard at the first moment.

[0120] This embodiment also discloses an electronic device, as shown in the reference. Figure 3 The electronic device may include: at least one processor 301, at least one communication bus 302, user interface 303, network interface 304, and at least one memory 305.

[0121] The communication bus 302 is used to enable communication between these components.

[0122] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0123] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0124] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 305, and by calling data stored in memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0125] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. Figure 3 As shown, the memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a fire hazard early warning and handling method.

[0126] exist Figure 3In the electronic device shown, the user interface 303 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 301 can be used to call an application program stored in the memory 305 for a fire hazard early warning processing method. When executed by one or more processors 301, the electronic device executes one or more methods as described in the above embodiments.

[0127] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0128] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0129] In some embodiments of this application, a computer-readable storage medium is provided, including instructions that, when executed on the electronic device, cause the electronic device to perform a fire hazard early warning processing method according to an embodiment of this application.

[0130] In some embodiments of this application, a computer program product is also provided, which, when run on an electronic device, causes the electronic device to execute a fire hazard early warning processing method according to an embodiment of this application.

[0131] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some service interfaces; indirect couplings or communication connections between apparatuses or units may be electrical or other forms.

[0132] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0133] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0134] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory 305 and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory 305 includes various media capable of storing program code, such as a USB flash drive, external hard drive, magnetic disk, or optical disk.

[0135] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the disclosure in this specification. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are considered exemplary only, and the scope of this application is defined by the claims.

Claims

1. A method for early warning and handling of fire hazards, characterized in that, The method includes: Acquire fire characteristic monitoring data of the target area, including real-time smoke concentration and real-time ambient temperature; When the real-time smoke concentration exceeds the preset smoke alarm threshold, or the real-time ambient temperature exceeds the preset temperature alarm threshold, the current monitoring time is determined as the trigger time. The change curve features corresponding to the fire feature monitoring data within the preset time period including the trigger time are extracted, and the change curve features are compared with the pre-stored historical fire evolution curve features to generate the corresponding hazard confirmation result. When the hazard confirmation result indicates that a hazard has occurred in the target area, the location of the hazard in the target area is obtained, and based on the location of the hazard, the distribution data of combustibles within a preset environmental feature retrieval range around the location of the hazard is extracted, as well as the real-time location information of the personnel handling the hazard within a preset dispatch retrieval range from the location of the hazard is obtained. The location of the hazard, the distribution data of the combustibles, and the real-time location information of the personnel handling the situation are input into a pre-trained hazard development prediction model to calculate the spread risk level of the hazard and the estimated time for the personnel handling the situation to reach the location of the hazard. If the spread risk level is less than a preset risk threshold and the estimated time is less than a preset time threshold, then send an on-site handling instruction containing the location of the hazard to the mobile terminal carried by the handling personnel. If the spread risk level is greater than or equal to a preset risk threshold or the estimated time is greater than or equal to a preset time threshold, an upgraded alarm command containing the location of the hazard is sent to a preset fire management terminal.

2. The method according to claim 1, characterized in that, The step of performing a deviation analysis between the changing curve features and the pre-stored historical fire evolution curve features to generate a corresponding hazard confirmation result specifically includes: Extract the temperature rise slope parameter and the smoke concentration peak feature parameter from the characteristics of the change curve; The absolute difference between the temperature rise slope parameter and the fire reference slope in the historical fire evolution curve characteristics is calculated to obtain the first deviation value. The absolute difference between the smoke concentration peak feature parameter and the fire reference peak feature in the historical fire evolution curve feature is calculated to obtain the second deviation value. Based on the feature weight coefficients corresponding to the target region, a weighted summation is performed on the first deviation value and the second deviation value to obtain the weighted total deviation value; When the weighted total deviation is less than the preset deviation threshold, it is determined that a danger has occurred in the target area, and the corresponding danger confirmation result is output.

3. The method according to claim 1, characterized in that, The step of extracting combustible material distribution data within a preset environmental feature retrieval range around the location of the hazard specifically includes: Obtain static structural data of the building information model and real-time monitoring video stream data within the environmental feature retrieval range; The real-time monitoring video stream data is processed by a pre-built semantic segmentation algorithm to extract image features, determine the material classification labels of items in the video frame, and identify items whose material classification labels belong to a preset combustible material category as target items. Obtain the outline dimension parameters of the target item; The outline dimension parameters and the material classification labels are superimposed on the static structural data of the building information model to generate the combustible material distribution data.

4. The method according to claim 1, characterized in that, The step of obtaining the real-time location information of the personnel handling the situation within a preset dispatch retrieval range from the location of the hazard specifically includes: Obtain signal reception data from indoor positioning base stations deployed within the scheduling retrieval range; The signal reception data is parsed to extract the identity information of the mobile terminal corresponding to the signal reception data, as well as the arrival angle parameter and arrival time difference parameter of the mobile terminal. The basic position coordinates of the mobile terminal are obtained by performing three-dimensional spatial coordinate calculation based on the arrival angle parameter and the arrival time difference parameter. The identity information is matched with preset scheduling system data. When the scheduling system data indicates that the target person corresponding to the identity information is on duty, the target person is identified as the processing personnel, and the corresponding basic location coordinates are identified as the real-time location information of the processing personnel.

5. The method according to claim 1, characterized in that, The step of inputting the location of the hazard, the distribution data of the combustible material, and the real-time location information of the personnel involved into a pre-trained hazard development prediction model to calculate the spread risk level of the hazard and the estimated time required for the personnel involved to reach the location of the hazard specifically includes: The graph neural network layer in the hazard development prediction model aggregates node features of the hazard location and combustible material distribution data to obtain a spatial spread feature vector. The path deduction layer in the hazard development prediction model performs pathfinding calculations on the location of the hazard and the real-time location information of the personnel handling the situation, thereby obtaining a dynamic navigation feature vector. The spatial spread feature vector and the dynamic navigation feature vector are input into the fully connected layer of the hazard development prediction model for regression mapping to obtain the spread risk level and the estimated time, respectively.

6. The method according to claim 5, characterized in that, The step of aggregating node features of the hazard location and combustible material distribution data through the graph neural network layer in the hazard development prediction model to obtain a spatial spread feature vector specifically includes: Extract multiple independent rooms in the physical spatial layout within the pre-acquired environmental feature retrieval range, the spatial boundary coordinates of each independent room, and the door and window passages connecting adjacent independent rooms; Each of the independent rooms is defined as a node, and the doors, windows and passageways are defined as edges between the corresponding nodes, thus constructing a connected graph structure; Based on the spatial boundary coordinates, the combustible material distribution data is analyzed to extract items located in the corresponding independent rooms, and the outline size parameters and material classification labels of the corresponding items are obtained. Based on the contour size parameters and the material classification labels, the flammable material density feature weights of the corresponding nodes are determined. Obtain the preset ventilation volume parameters and real-time opening and closing status of the door and window passage, match the corresponding ventilation damping coefficient according to the real-time opening and closing status, and perform weighted calculation on the preset ventilation volume parameters and the ventilation damping coefficient to obtain the edge feature weight of the corresponding edge. Map the location of the hazard to the connected graph structure to determine the corresponding initial fire node; Based on the flammable material density feature weight and the edge feature weight, multi-level information transmission calculations are performed from the initial fire node to the associated neighboring nodes to obtain the spatial spread feature vector.

7. The method according to claim 5, characterized in that, The step of performing pathfinding calculations on the location of the hazard and the real-time location information of the personnel handling the situation through the path deduction layer in the hazard development prediction model to obtain a dynamic navigation feature vector specifically includes: Extract the real-time location information of the personnel handling the incident as the starting point for navigation, and extract the location of the hazard as the ending point for navigation; Based on a preset indoor navigation map, a navigation topology map containing the navigation start point and the navigation end point is extracted. The navigation topology map includes multiple navigation nodes and connected edges connecting adjacent navigation nodes. According to the correspondence of spatial coordinates, the spatial spread feature vector is mapped to the corresponding navigation node and connected edge of the navigation topology graph; The fire spread hazard coefficient of the connected edge is determined based on the mapping result, and a passage resistance penalty coefficient is applied to the connected edge whose fire spread hazard coefficient is greater than a preset safety threshold. Based on the navigation topology map after superimposing the traffic resistance penalty coefficient, the minimum traffic cost path between the pathfinding start point and the pathfinding end point is calculated, and the path sequence features and traffic cost parameters of the minimum traffic cost path are extracted and fused to obtain the dynamic navigation feature vector.

8. An electronic device, characterized in that, Including processor and memory; The memory is used to store computer program code, the computer program code including computer instructions, and the processor invokes the computer instructions to cause the electronic device to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium storing computer instructions, characterized in that, When the computer instructions are executed on the electronic device, the electronic device causes the electronic device to perform the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on an electronic device, it causes the electronic device to perform the method as described in any one of claims 1-7.

Images