A park fire linkage early warning method of multi-modal risk identification

Abstract

The application relates to a park fire linkage early warning method for multi-modal risk identification, which can be used in the technical field of fire early warning. The method comprises the following steps: acquiring multi-modal monitoring data of a park; the multi-modal monitoring data comprises temperature data, image data and smoke concentration data; performing data preprocessing and weighted fusion processing on the multi-modal monitoring data to obtain a risk feature matrix of the park; inputting the risk feature matrix into a pre-trained risk prediction model to perform risk prediction processing and obtain risk prediction information of the park; determining a fire risk level of the park according to the risk prediction information and a preset safety threshold benchmark library; and performing fire linkage early warning processing on the park according to the fire risk level. The method can improve the accuracy of park fire early warning.

Description

Technical Field

[0001] This application relates to the field of fire early warning technology, and in particular to a multimodal risk identification method for coordinated fire early warning in industrial parks. Background Technology

[0002] With the continuous expansion of the scale of modern industrial park construction and the increasing complexity of its functions, fire safety management within the park has become particularly important.

[0003] Currently, the monitoring and early warning of fire risks in industrial parks mainly rely on single-type sensors or manual periodic inspections. However, when faced with complex environmental changes and emergencies within the park, traditional methods are unable to accurately reflect the actual safety hazards on site, resulting in low accuracy of fire early warnings in industrial parks. Summary of the Invention

[0004] Therefore, it is necessary to provide a multimodal risk identification method, system, computer equipment, computer-readable storage medium, and computer program product for park fire linkage early warning that can improve the accuracy of park fire early warning, in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a multimodal risk identification method for coordinated early warning of fire prevention in industrial parks. The method includes:

[0006] Acquire multimodal monitoring data of the park; the multimodal monitoring data includes temperature data, image data, and smoke concentration data;

[0007] The multimodal monitoring data is preprocessed and weighted fusion processed to obtain the risk feature matrix of the park.

[0008] The risk feature matrix is ​​input into a pre-trained risk prediction model for risk prediction processing to obtain the risk prediction information of the park.

[0009] Based on the risk prediction information and the preset safety threshold benchmark library, the fire risk level of the park is determined;

[0010] Based on the fire risk level, fire alarm and early warning procedures will be implemented for the park.

[0011] In one embodiment, the step of performing data preprocessing and weighted fusion processing on the multimodal monitoring data to obtain the risk feature matrix of the park includes:

[0012] The multimodal monitoring data is preprocessed to obtain preprocessed monitoring data;

[0013] Based on the scene feature information corresponding to the park, the preprocessed monitoring data is weighted and fused to obtain the risk feature matrix.

[0014] In one embodiment, the step of preprocessing the multimodal monitoring data to obtain preprocessed monitoring data includes:

[0015] The temperature data is filtered and denoised to obtain the target temperature data;

[0016] The image data is subjected to image enhancement processing to obtain the target image data;

[0017] The smoke concentration data is processed to remove outliers, and the target smoke concentration data is obtained.

[0018] Based on the target temperature data, the target image data, and the target smoke concentration data, the preprocessed monitoring data is obtained.

[0019] In one embodiment, the step of performing weighted fusion processing on the preprocessed monitoring data based on the scene feature information corresponding to the park to obtain the risk feature matrix includes:

[0020] Based on the scene feature information, determine the fusion weights corresponding to each data modality in the preprocessed monitoring data;

[0021] Based on the fusion weights, the preprocessed monitoring data is weighted and fused to obtain the risk feature matrix.

[0022] In one embodiment, determining the fire risk level of the park based on the risk prediction information and a preset safety threshold benchmark library includes:

[0023] The risk prediction information is compared with the various safety thresholds in the safety threshold benchmark library to obtain the comparison results;

[0024] Based on the comparison results, determine the number and degree of abnormal modes in the multimodal monitoring data;

[0025] The fire risk level of the park is determined based on the number of abnormal modes and the degree of abnormality.

[0026] In one embodiment, the fire risk level includes a first risk level, a second risk level, or a third risk level;

[0027] The fire alarm linkage and early warning processing for the park based on the fire risk level includes:

[0028] When the fire risk level is the first risk level, the park will be subject to the first level of fire alarm linkage; the first level of fire alarm linkage includes activating the audible and visual alarm and triggering video verification of the abnormal area;

[0029] When the fire risk level is the second risk level, the park shall be subject to the second level of fire alarm response; the second level of fire alarm response includes activating the area smoke exhaust equipment and emergency lighting equipment on the basis of the first level of fire alarm response.

[0030] When the fire risk level is the third risk level, the park shall be subject to the third-level fire linkage early warning process; the third-level fire linkage early warning process includes, on the basis of the second-level fire linkage early warning process, activating the sprinkler linkage equipment and unlocking the evacuation access control.

[0031] Secondly, this application also provides a multimodal risk identification and fire-fighting linkage early warning system for industrial parks. The system includes:

[0032] The data acquisition module is used to acquire multimodal monitoring data of the park; the multimodal monitoring data includes temperature data, image data, and smoke concentration data.

[0033] The data fusion module is used to perform data preprocessing and weighted fusion processing on the multimodal monitoring data to obtain the risk feature matrix of the park;

[0034] The risk prediction module is used to input the risk feature matrix into a pre-trained risk prediction model for risk prediction processing to obtain the risk prediction information of the park.

[0035] The risk level determination module is used to determine the fire risk level of the park based on the risk prediction information and a preset safety threshold benchmark library.

[0036] The early warning processing module is used to perform fire-related early warning processing for the park based on the fire risk level.

[0037] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0038] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0039] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the above aspects.

[0040] The aforementioned multimodal risk identification method, system, computer equipment, computer-readable storage medium, and computer program product for park fire linkage early warning, by acquiring multimodal monitoring data including temperature data, image data, and smoke concentration data, facilitates a comprehensive perception of the park's fire status from multiple dimensions, avoiding the limitations of a single data source. By performing data preprocessing and weighted fusion processing on the multimodal monitoring data to obtain a risk feature matrix, environmental interference is reduced and key features are highlighted. Inputting the risk feature matrix into a pre-trained risk prediction model for processing, and combining it with a preset safety threshold benchmark library to determine the fire risk level, facilitates accurate assessment of fire risks. Fire linkage early warning processing based on the fire risk level allows for the implementation of matched response measures for different levels of risk, thereby improving the accuracy of park fire early warning. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating a multimodal risk identification method for fire-fighting linkage early warning in a park, as shown in one embodiment.

[0043] Figure 2 This is a flowchart illustrating the steps of the weighted fusion process in one embodiment;

[0044] Figure 3 This is a flowchart illustrating a multimodal risk identification-based fire-fighting linkage early warning method for industrial parks, as shown in another embodiment.

[0045] Figure 4 This is a structural block diagram of a park fire alarm linkage system for multimodal risk identification in one embodiment;

[0046] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0049] In one exemplary embodiment, such as Figure 1 As shown, a multimodal risk identification-based fire alarm linkage method for industrial parks is provided. This embodiment illustrates the application of this method to a terminal; however, it is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc.; the server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. In this embodiment, the method includes the following steps:

[0050] Step S101: Acquire multimodal monitoring data of the park; the multimodal monitoring data includes temperature data, image data, and smoke concentration data;

[0051] Step S102: Perform data preprocessing and weighted fusion processing on the multimodal monitoring data to obtain the risk feature matrix of the park;

[0052] Step S103: Input the risk feature matrix into the pre-trained risk prediction model for risk prediction processing to obtain the risk prediction information of the park;

[0053] Step S104: Determine the fire risk level of the park based on the risk prediction information and the preset safety threshold benchmark library;

[0054] Step S105: Based on the fire risk level, conduct fire alarm and early warning procedures for the park.

[0055] Among them, multimodal monitoring data can be a collection of various types of data collected by infrared monitoring equipment, visual monitoring equipment, and smoke detection equipment in the park area to reflect the fire safety status.

[0056] Among them, the temperature data can be the ambient temperature of the target area in the park and the temperature parameters of the object surface, which are collected in real time by infrared monitoring equipment.

[0057] The image data can be visible light images of the target area collected by visual monitoring equipment, including images used to extract flame or smoke features.

[0058] Among them, the smoke concentration data can be a value reflecting the smoke content in the air, which is collected in real time by smoke sensing devices.

[0059] Data preprocessing can be a process of denoising, calibrating, and standardizing the format of the acquired multimodal monitoring data in order to remove abnormal fluctuation values ​​and improve data quality.

[0060] Weighted fusion processing can be a process of assigning weights to the reliability of each data point in multimodal monitoring data under different scenarios and integrating the multimodal monitoring data into a unified data structure.

[0061] The risk feature matrix can be a matrix structure that integrates multi-dimensional features such as temperature data, image data, and smoke concentration data, generated after weighted fusion processing.

[0062] Among them, the risk prediction model can be constructed using deep learning networks (such as a model combining convolutional neural networks and long short-term memory networks) to perform in-depth feature analysis on the input risk feature matrix in order to determine the risk situation.

[0063] Among them, the risk prediction information can be the output of the risk prediction model, reflecting the preliminary prediction results of the current fire safety status of the park.

[0064] The safety threshold benchmark library can be a built-in reference database containing the normal range or critical value of various fire safety indicators.

[0065] Among them, the fire risk level can be a level of risk severity classified by comparing risk prediction information with a safety threshold benchmark library, such as different levels such as no risk, early warning risk, and emergency risk.

[0066] Among them, fire alarm linkage processing can be the process of triggering corresponding alarm prompts and controlling relevant fire equipment to perform corresponding actions based on the determined fire risk level.

[0067] Optionally, the terminal acquires multimodal monitoring data of the park in real time through monitoring equipment deployed in various areas of the park. The multimodal monitoring data includes temperature data, image data, and smoke concentration data. To improve data quality and unify data format, the multimodal monitoring data undergoes data preprocessing and weighted fusion processing to obtain a risk feature matrix of the park that integrates multidimensional features. The risk feature matrix is ​​then input into a pre-trained risk prediction model for risk prediction processing to obtain risk prediction information for the park. Based on the risk prediction information and a preset safety threshold benchmark library, a comparative analysis is performed to determine the fire risk level of the park. Based on the determined fire risk level, the park is given a corresponding level of fire linkage early warning processing to achieve automated response to fire safety in the park.

[0068] The aforementioned multimodal risk identification-based fire-fighting linkage early warning method for industrial parks utilizes multimodal monitoring data, including temperature, image, and smoke concentration data, to comprehensively perceive the fire status of the park from multiple dimensions, avoiding the limitations of a single data source. Preprocessing and weighted fusion of the multimodal monitoring data to obtain a risk feature matrix helps reduce environmental interference and highlight key features. Inputting the risk feature matrix into a pre-trained risk prediction model and combining it with a preset safety threshold benchmark library to determine the fire risk level facilitates accurate assessment of fire risks. Fire-fighting linkage early warning processing based on the fire risk level allows for tailored response measures to different levels of risk, thereby improving the accuracy of fire-fighting early warnings for industrial parks.

[0069] In one exemplary embodiment, reference is made to Figure 2 The multimodal monitoring data undergoes preprocessing and weighted fusion to obtain the risk characteristic matrix of the park, including:

[0070] Step S201: Perform data preprocessing on the multimodal monitoring data to obtain preprocessed monitoring data;

[0071] Step S202: Based on the scene feature information corresponding to the park, the preprocessed monitoring data is weighted and fused to obtain a risk feature matrix.

[0072] The preprocessed monitoring data can be a dataset that has undergone noise reduction, calibration, and format standardization to remove abnormal fluctuation values.

[0073] Among them, scene feature information can be parameter information describing the environmental characteristics of different areas of the park, such as high temperature environment, enclosed space and other characteristics.

[0074] Optionally, the terminal performs data preprocessing such as noise reduction, calibration and format standardization on the acquired multimodal monitoring data to obtain preprocessed monitoring data; obtains scene feature information corresponding to the park, evaluates the reliability of various data under different scenarios based on scene feature information, and performs weighted fusion processing on the preprocessed monitoring data to obtain a risk feature matrix integrating multidimensional features.

[0075] The technical solution provided in this embodiment is beneficial for rationally allocating data weights based on the specific characteristics of the scenario, thereby improving the accuracy of the risk feature matrix.

[0076] In an exemplary embodiment, data preprocessing is performed on the multimodal monitoring data to obtain preprocessed monitoring data, including: filtering and denoising the temperature data to obtain target temperature data; performing image enhancement processing on the image data to obtain target image data; performing outlier removal processing on the smoke concentration data to obtain target smoke concentration data; and obtaining preprocessed monitoring data based on the target temperature data, target image data, and target smoke concentration data.

[0077] Among them, the filtering and noise reduction process can be a process of removing noise interference from temperature data using algorithms such as median filtering.

[0078] The target temperature data can be a smoother and more accurate temperature parameter obtained after filtering and noise reduction.

[0079] Image enhancement processing can be a process of optimizing visual image quality and improving image clarity and contrast.

[0080] The target image data can be an image that contains clear flame or smoke features after image enhancement processing.

[0081] Outlier removal can be a process of identifying and removing abnormal fluctuations in smoke concentration data caused by environmental interference.

[0082] The target smoke concentration data can be a value that reflects the true smoke concentration after outlier removal.

[0083] Optionally, the terminal performs filtering and noise reduction processing on the collected temperature data to remove data noise and obtain target temperature data; performs image enhancement processing on the image data to optimize visual image quality and obtain target image data; performs outlier removal processing on the smoke concentration data to remove abnormal fluctuation values ​​and obtain target smoke concentration data; and performs format standardization and summarization on the target temperature data, target image data, and target smoke concentration data to obtain preprocessed monitoring data.

[0084] The technical solution provided in this embodiment is beneficial for adopting matching optimization methods for different types of data, thereby improving the accuracy and effectiveness of the preprocessed monitoring data.

[0085] In an exemplary embodiment, the preprocessed monitoring data is weighted and fused according to the scene feature information corresponding to the park to obtain a risk feature matrix, including: determining the fusion weight corresponding to each data modality in the preprocessed monitoring data according to the scene feature information; and performing weighted fusion processing on the preprocessed monitoring data according to the fusion weight to obtain a risk feature matrix.

[0086] Among them, data modes can be different types of data contained in the preprocessed monitoring data, such as temperature modes, image modes, and smoke concentration modes.

[0087] The fusion weight can be a reliability proportion value assigned to each data modality based on scene feature information.

[0088] Optionally, the terminal assesses the reliability of various data under different scenarios based on the scenario feature information corresponding to the park, determines the fusion weight corresponding to each data modality in the preprocessed monitoring data, and performs weighted fusion processing on various parameters and feature vectors in the preprocessed monitoring data according to the determined fusion weight, integrating multi-source data into a unified risk feature matrix.

[0089] The technical solution provided in this embodiment helps to highlight key data features and weaken interfering factors, thereby improving the ability of the risk feature matrix to represent real risks.

[0090] In an exemplary embodiment, the fire risk level of the park is determined based on risk prediction information and a preset safety threshold benchmark library, including: comparing the risk prediction information with various safety thresholds in the safety threshold benchmark library to obtain comparison results; determining the number and degree of abnormal modes in the multimodal monitoring data based on the comparison results; and determining the fire risk level of the park based on the number and degree of abnormal modes.

[0091] The comparison results can reflect whether the various indicators in the risk prediction information exceed the corresponding safety thresholds in the safety threshold benchmark library.

[0092] The number of anomalous modalities can be the number of data modalities that exceed the safety threshold in the comparison results, such as single-modal anomalousness or multi-modal collaborative anomalousness.

[0093] The degree of abnormality can be the extent or severity of exceeding the safety threshold, such as local high temperature not reaching the critical value or seriously exceeding the standard.

[0094] Optionally, the terminal compares the risk prediction information output by the risk prediction model with each safety threshold in the preset safety threshold benchmark library to obtain the comparison results; based on the comparison results, it counts the number of abnormal modes in the multimodal monitoring data that exceed the safety thresholds and assesses the degree of abnormality of each abnormal data; based on the number and degree of abnormality, it determines the fire risk level of the park.

[0095] The technical solution provided in this embodiment is conducive to multi-dimensional quantitative assessment of the current risk status, thereby improving the objectivity and accuracy of the determined fire risk level.

[0096] In an exemplary embodiment, the fire risk level includes a first risk level, a second risk level, or a third risk level. Based on the fire risk level, fire alarm linkage processing is performed on the park, including: If the fire risk level is first risk level, a first-level fire alarm linkage is performed; the first-level fire alarm linkage is performed by activating audible and visual alarms and triggering video verification of abnormal areas; if the fire risk level is second risk level, a second-level fire alarm linkage is performed; the second-level fire alarm linkage is performed by activating area smoke extraction equipment and emergency lighting equipment, based on the first-level fire alarm linkage processing; if the fire risk level is third risk level, a third-level fire alarm linkage is performed; the third-level fire alarm linkage is performed by activating sprinkler linkage equipment and unlocking evacuation access control, based on the second-level fire alarm linkage processing.

[0097] The first risk level can be the level of the warning risk, such as the case of abnormal single-modal data that has not reached the critical value.

[0098] The second risk level can be a level corresponding to an emergency risk, such as an anomaly in multimodal data collaboration.

[0099] The third risk level can be a level corresponding to severe emergency risk, such as a situation where multimodal data is severely abnormal and clear fire characteristics are captured.

[0100] Among them, audible and visual alarms can emit sound and flashing light signals to alert personnel to abnormal equipment operation.

[0101] Among them, video verification can be an operation that controls visual monitoring equipment to automatically focus on abnormal areas and upload real-time images for confirmation.

[0102] Among them, the area smoke exhaust equipment can be a ventilation device such as a fan used to exhaust smoke from abnormal areas.

[0103] Among them, the sprinkler linkage equipment can be a fire protection system used for automatic sprinkler fire extinguishing.

[0104] Evacuation access control can be a channel door lock device used to control the entry and exit of personnel, which facilitates the rapid evacuation of personnel after unlocking.

[0105] Optionally, the terminal determines the fire risk level. If the fire risk level is Level 1, it initiates Level 1 fire alarm response for the park, activating audible and visual alarms to alert of the anomaly and triggering video verification of the abnormal area to upload real-time footage. If the fire risk level is Level 2, it initiates Level 2 fire alarm response for the park, activating area smoke extraction equipment and turning on emergency lighting equipment in addition to the Level 1 response. If the fire risk level is Level 3, it initiates Level 3 fire alarm response for the park, activating sprinkler systems in and around the abnormal area to extinguish the fire and unlocking evacuation access control to guide personnel evacuation.

[0106] The technical solution provided in this embodiment, by implementing corresponding levels of fire linkage early warning processing for different fire risk levels, is conducive to establishing a graded response mechanism, thereby improving the pertinence and efficiency of emergency response.

[0107] The following example illustrates the multimodal risk identification and park fire linkage early warning method provided in this application.

[0108] In park management, fire safety is a core element in ensuring the safety of people's lives and property. Existing park fire early warning systems often employ single monitoring methods, such as relying solely on smoke detectors or a single type of camera, which has several limitations: First, single monitoring is prone to false alarms and missed alarms. For example, visual monitoring is easily affected by light interference, leading to misjudgments of smoke, and smoke detectors have delayed responses under ventilated conditions. Second, risk identification is limited to a single dimension, failing to comprehensively consider multiple fire risk indicators such as abnormal temperature, flame characteristics, and smoke concentration, resulting in incomplete risk prediction. Third, the linkage response mechanism is rigid, lacking intelligent linkage between early warning and fire-fighting equipment (sprinklers, smoke exhaust, access control), requiring manual intervention and delaying emergency response time. Fourth, the ability to identify early potential risks (such as localized high temperatures and trace amounts of smoke) is insufficient, making it difficult to achieve "early detection and early response."

[0109] To meet the fire early warning needs of modern industrial parks with high population density and complex building structures, it is crucial to develop a multimodal risk identification-based fire-fighting linkage early warning method. This method integrates data from infrared monitoring, visual surveillance, and smoke alarms to achieve multi-dimensional intelligent risk identification, enabling rapid response from fire-fighting equipment, improving early warning accuracy and emergency response efficiency, and reducing losses caused by fire accidents. Therefore, this embodiment is of great significance for improving the intelligence level and safety of industrial park fire management.

[0110] This embodiment relates to the field of fire early warning technology in industrial parks, specifically a multimodal risk identification-based fire-fighting linkage early warning method for industrial parks. The purpose of this embodiment is to provide a multimodal risk identification-based fire-fighting linkage early warning method for industrial parks. By integrating multimodal information from infrared monitoring, visual surveillance, and smoke detector data, it achieves accurate identification of fire risks and intelligent linkage early warning, solving the problems of high false alarm rates and delayed linkage in existing systems based on single monitoring.

[0111] 1. The specific technical solution is as follows:

[0112] 1.1. System Composition:

[0113] It includes a multimodal data acquisition module, a data fusion processing module, a risk identification module, a linkage early warning module, a human-computer interaction module, and fire equipment execution units. The multimodal data acquisition module includes infrared monitoring cameras, visual monitoring cameras, and smoke alarm fire control panels, installed in key areas of the park (such as warehouses, corridors, and computer rooms); the data fusion processing module integrates and optimizes multi-source monitoring data; the risk identification module determines the fire risk level based on the fused data; the linkage early warning module triggers tiered early warnings and fire equipment linkage; the human-computer interaction module provides user operation and information feedback; and the fire equipment execution units include existing fire equipment in the park such as sprinkler systems, smoke exhaust fans, emergency lighting, and evacuation access control systems.

[0114] 1.2. Multimodal data acquisition:

[0115] To achieve comprehensive and accurate risk monitoring, this embodiment uses a multimodal data acquisition module to simultaneously acquire three types of core fire protection data, covering three major risk dimensions: temperature, visual, and smoke.

[0116] Infrared monitoring camera: Real-time acquisition of infrared thermal imaging data of target areas in the park, extraction of environmental and object surface temperature parameters, sampling frequency set to 5Hz, temperature measurement range -20℃~300℃, accuracy ±0.5℃, focusing on monitoring abnormal conditions such as equipment overheating and local high temperature;

[0117] Visual surveillance camera: Collects visible light image data of the target area at a sampling frequency of 10fps, supports image extraction of flame features (color, shape, flashing frequency) and smoke features (grayscale, texture, diffusion speed), and is adaptable to different lighting environments such as day and night;

[0118] Smoke alarm fire control panel: It connects to the distributed smoke sensors in the park, collects smoke concentration values ​​(unit: ppm), sensor working status and alarm signals in real time, and the data transmission frequency is 1Hz to ensure timely detection of smoke anomalies.

[0119] The collected data is transmitted to the data fusion processing module via wired Ethernet or 5G wireless communication links, and the transmission process uses encryption protocols to ensure data security.

[0120] 1.3. Data Fusion and Intelligent Risk Identification:

[0121] Traditional single-data identification is susceptible to interference. This embodiment uses a multimodal data fusion algorithm and a machine learning model to achieve accurate risk identification.

[0122] Data preprocessing: The data fusion processing module performs noise reduction, calibration and format standardization on the collected temperature data, image data and smoke concentration data. For example, median filtering is used to remove noise from infrared data, image enhancement technology is used to optimize visual image quality, and abnormal fluctuation values ​​in smoke sensing data are removed.

[0123] Multimodal fusion: A weighted fusion algorithm is used to assign weights to each data point based on the reliability of different scenarios (e.g., the weight of infrared data is increased in high-temperature environments, and the weight of smoke detection data is increased in enclosed spaces), integrating temperature parameters, image feature vectors, and smoke concentration values ​​into a unified risk feature matrix.

[0124] Risk identification model: Employing a deep learning model combining CNN (Convolutional Neural Network) and LSTM (Long Short-Term Memory Network), trained on massive datasets (including normal scenario data, initial fire data, and data from disruptive scenarios), the model assesses risk levels. The model output is categorized into three risk levels: no risk (all data meet safety thresholds), early warning risk (abnormal single-modal data, such as localized high temperatures not reaching critical values), and emergency risk (multi-modal data exhibiting coordinated anomalies, such as high temperature + excessive smoke concentration + visual images capturing suspected flames).

[0125] Threshold benchmark library: Built-in safety threshold benchmark library based on fire protection standards and park scene characteristics, which can be customized according to park type.

[0126] 1.4. Tiered and Linked Early Warning Mechanism:

[0127] This embodiment establishes a closed-loop linkage mechanism of "early warning-response-feedback", triggering corresponding actions according to the risk level to ensure the timeliness and pertinence of the response:

[0128] Risk warning (Level 1): The linkage warning module activates the audible and visual alarm (the park broadcasts a message saying "An anomaly has occurred in a certain area, please have staff check it," and the area indicator light flashes), and the human-computer interaction module pushes warning information (including the location and type of abnormal data); at the same time, video verification is activated, the visual monitoring camera automatically focuses on the abnormal area, and uploads real-time images to the management platform;

[0129] Emergency Risk (Level 2): ​​In addition to executing Level 1 warning operations, the linkage fire-fighting equipment execution unit initiates preliminary handling: starts the area smoke exhaust fan, and automatically turns on the emergency lighting; sends an emergency warning to the park fire management center and on-duty personnel through the human-machine interaction module;

[0130] Severe Emergency Risk (Level 3): Based on the Level 2 response, trigger comprehensive fire-fighting coordination: activate the sprinkler system in and around the abnormal area, automatically unlock evacuation access control, and guide personnel to evacuate; coordinate with the park's broadcast system to continuously play evacuation instructions.

[0131] 1.5. Human-Computer Interaction and Remote Control:

[0132] To enhance management flexibility, the system is equipped with comprehensive human-computer interaction functions:

[0133] Information Display: The human-computer interaction module displays real-time monitoring data, risk levels, and equipment operating status to users through the management platform interface and mobile APP, and supports simultaneous viewing of infrared thermal images, visual monitoring screens, and smoke concentration curves;

[0134] Command reception: Users can send manual control commands through the management platform, such as confirming warnings, turning off alarms, starting / stopping fire-fighting equipment, and adjusting monitoring thresholds;

[0135] Remote review: Users can remotely view real-time images of abnormal areas via a mobile app, manually determine the type of risk, and avoid accidental linkage.

[0136] Record Tracking: Automatically stores all early warning information, linkage operation records, and data change curves, and supports historical data query and export, providing a basis for fire hazard investigation.

[0137] 1.6. Data Processing Center:

[0138] As the core of the system, the data processing center uses high-performance servers with multi-threaded parallel processing capabilities, supporting real-time analysis and storage of massive amounts of data. The data processing center has a built-in redundant backup mechanism to ensure that critical data is not lost in extreme situations such as network outages or power failures, and also supports offline data caching with automatic synchronization after network recovery.

[0139] 1.7. Scalability and Customizability:

[0140] The system design supports modular expansion, allowing for the addition of infrared / visual cameras and smoke sensors based on the size of the park, or the integration of other fire protection equipment (such as gas detectors and heat detectors); it also supports the customization of risk identification model parameters and linkage logic according to the park scenario (industrial, commercial, residential), adapting to the fire management needs of different parks.

[0141] 2. Workflow:

[0142] 2.1. Multimodal data acquisition:

[0143] The multimodal data acquisition module (infrared monitoring camera, visual monitoring camera, smoke alarm fire control panel) collects temperature data, image data, and smoke concentration data in various areas of the park in real time and transmits them to the data processing center through the communication link.

[0144] 2.2. Data Fusion and Preprocessing:

[0145] The data processing center performs noise reduction, calibration, and format conversion on the collected multi-source data, and integrates them into a unified risk feature matrix using a weighted fusion algorithm to remove interfering data and improve data quality.

[0146] 2.3. Risk Identification and Level Assessment:

[0147] The deep learning model analyzes the fused feature matrix and, combined with a safety threshold benchmark library, outputs the risk level judgment results: no risk, early warning risk, and emergency risk.

[0148] 2.4. Hierarchical and coordinated response:

[0149] The linkage early warning module triggers corresponding early warning operations based on the risk level, links the fire-fighting equipment execution unit to initiate response measures, and pushes information to the user through the human-computer interaction module.

[0150] 2.5. Human-Computer Interaction and Feedback:

[0151] Users can view risk information and execute manual control commands through the management platform or mobile APP. The system records the operation process and the handling results, forming a closed-loop management.

[0152] 2.6. Post-launch tracking and optimization:

[0153] The system automatically stores all data and operation records, supporting historical traceability; by continuously collecting actual operation data, it iteratively optimizes the risk identification model to improve identification accuracy.

[0154] 3. Technical features:

[0155] 3.1. Multimodal fusion for accurate recognition:

[0156] By integrating three core data sources—infrared, visual, and smoke detection—it overcomes the limitations of single monitoring methods. Combined with a deep learning model, it effectively reduces false alarm and false negative rates, achieving a risk identification accuracy of ≥98%.

[0157] 3.2. Hierarchical linkage for efficient response:

[0158] Establish a three-tiered risk response mechanism, matching corresponding early warning and response measures according to the severity of the risk, avoiding over-response or under-response, and ensuring that the emergency risk response time is ≤3 seconds.

[0159] 3.3. Human-machine collaboration, flexible operation:

[0160] It supports a combination of automatic linkage and manual control, allowing users to remotely review and intervene in the handling process, balancing intelligence and humanization, and reducing management difficulty.

[0161] 3.4. Scene adaptability and strong scalability:

[0162] It has a built-in customizable threshold benchmark library and linkage logic to adapt to different types of parks; it supports modular expansion, allowing the addition of monitoring equipment or connection to other fire protection systems to meet the needs of park expansion or functional upgrades.

[0163] 3.5. Closed-loop management, convenient traceability:

[0164] It achieves a closed-loop process of "data collection-identification-early warning-handling-feedback-optimization", with all data and operations being traceable, providing data support for the optimization of fire management.

[0165] Advantages of this embodiment:

[0166] Improve the accuracy of risk identification: By using multimodal data fusion and deep learning models, it effectively filters out false alarms caused by environmental interference (such as fog and dust), accurately captures the initial characteristics of fires, and solves the problem of identification bias caused by single monitoring methods.

[0167] Shorten emergency response time: The tiered linkage mechanism enables "early detection and early response", and fire-fighting equipment is automatically activated under emergency risks without human intervention, improving efficiency by more than 80% compared to traditional manual alarm handling.

[0168] Reduce management costs: Reduce the frequency and scope of manual inspections, and reduce manpower input through intelligent monitoring and linkage; accurate early warning avoids resource waste caused by ineffective handling (such as property damage caused by accidental activation of sprinklers).

[0169] Enhance the fire safety level of the park: achieve all-time and all-round monitoring of fire risks, cover early potential hazards, and coordinate multi-dimensional responses to minimize the risk of fire spread and protect the safety of people's lives and property.

[0170] High adaptability and flexibility: It supports customized configurations for different types of parks, has strong scalability, meets the needs of park function upgrades or scale expansion in the later stage, and has a wide range of applications.

[0171] In this embodiment:

[0172] A multimodal risk identification-based fire alarm and early warning system for industrial parks includes:

[0173] The multimodal data acquisition module includes an infrared monitoring camera, a visual monitoring camera, and a smoke alarm fire control panel, used to collect temperature data, image data, and smoke concentration data in the park.

[0174] The data fusion processing module is used to perform noise reduction, calibration, format conversion, and weighted fusion processing on multi-source acquired data.

[0175] The risk identification module has a built-in deep learning model and a safety threshold benchmark library for determining the fire risk level;

[0176] The linkage early warning module is used to trigger graded early warnings and link them with fire-fighting equipment based on the risk level;

[0177] The human-computer interaction module is used to display monitoring data, receive user commands, and provide feedback on processing results;

[0178] Firefighting equipment execution units, including sprinkler systems, smoke exhaust fans, emergency lighting, evacuation access control, etc., are used to perform firefighting operations;

[0179] The data processing center is used to receive, store, and analyze data and instructions from various modules, providing computing power support.

[0180] The multimodal data acquisition module includes: an infrared monitoring camera with a sampling frequency of ≥5Hz, a temperature measurement range of -20℃ to 300℃, and an accuracy of ±0.5℃, used to collect temperature data of the target area; a visual monitoring camera with a sampling frequency of ≥10fps, supporting the extraction of flame and smoke image features and adapting to different lighting environments; and a smoke alarm fire control panel with a data transmission frequency of ≥1Hz, used to collect smoke concentration values ​​and sensor operating status.

[0181] The data fusion processing module employs a weighted fusion algorithm to allocate weights to each modality of data based on the characteristics of the park scene, integrating temperature parameters, image feature vectors, and smoke concentration values ​​into a unified risk feature matrix.

[0182] The risk identification module employs a deep learning model combining CNN and LSTM, which uses sample training to determine risk levels and outputs three levels of results: no risk, warning risk, and emergency risk. It also includes a built-in customizable safety threshold benchmark library to adapt to different park types.

[0183] The linkage early warning module establishes a three-level linkage mechanism: early warning risks trigger audible and visual alarms and video verification; emergency risks trigger preliminary equipment handling and emergency notification; and severe emergency risks trigger comprehensive fire linkage and external alarms.

[0184] The human-computer interaction module supports operation on both the management platform and the mobile APP, and provides functions such as data display, command reception, remote review, and record traceability.

[0185] The system is scalable and customizable, supporting the addition of new monitoring equipment and integration with other fire protection systems. The model parameters and linkage logic can be adjusted according to the park scenario.

[0186] A multimodal risk identification method for fire-fighting linkage early warning in industrial parks includes the following steps:

[0187] (1) Real-time acquisition of park temperature data, image data, and smoke concentration data through a multimodal data acquisition module;

[0188] (2) The data fusion processing module preprocesses and weights the collected data to generate a risk feature matrix;

[0189] (3) The risk identification module uses a deep learning model and a safety threshold benchmark library to determine the risk level;

[0190] (4) The linkage early warning module triggers graded early warning and linkage operation with fire-fighting equipment according to the risk level;

[0191] (5) The human-computer interaction module displays information to the user, receives instructions, and provides feedback on the processing results;

[0192] (6) The system records data and operations, and iteratively optimizes the recognition model.

[0193] refer to Figure 3 It includes a modal data acquisition module (including infrared monitoring cameras, visual monitoring cameras, and smoke alarm fire control panel), a data fusion processing module, a risk identification and alarm level module, a human-computer interaction module, a linkage early warning module, a fire equipment execution unit, a data processing center, early warning information push, equipment control commands, and data storage. It also includes steps such as data cleaning / feature extraction and risk analysis / level determination.

[0194] Implementation method of this embodiment:

[0195] The following steps are included, but are not limited to, those taken as an example, in an industrial park:

[0196] Equipment installation and deployment: Install infrared monitoring cameras (temperature measurement range -20℃~300℃) and visual monitoring cameras (supporting 1080P high-definition shooting and image feature extraction) in key areas such as workshops, warehouses, power distribution rooms, and corridors in the park; connect the existing smoke alarm fire control panel in the park to the system to ensure that the smoke sensors cover all areas of the park in a distributed manner; establish communication connection between all devices and the data processing center server via Ethernet.

[0197] System initialization configuration: Import the industrial park safety threshold benchmark library into the data processing center, set the smoke concentration safety threshold to ≤300ppm, the electrical equipment surface temperature safety threshold to ≤85℃, and the workshop ambient temperature safety threshold to ≤40℃; divide the park into 12 monitoring areas according to the park layout, and assign a unique identifier and map location to each area; train the deep learning model, import 100,000 sets of sample data (including 60,000 sets of normal scenario data, 30,000 sets of initial fire data, and 10,000 sets of interference scenario data), and deploy the model to the data processing center after training is completed.

[0198] Multimodal data acquisition: After the system is started, the infrared monitoring camera acquires temperature data at a frequency of 5Hz, the visual monitoring camera acquires image data at a frequency of 10fps, and the smoke alarm fire control panel acquires smoke concentration data at a frequency of 1Hz. All data is transmitted to the data processing center in real time.

[0199] Data fusion and risk identification: The data processing center preprocesses the collected data (median filtering for noise reduction and image enhancement), integrates it into a risk feature matrix through a weighted fusion algorithm, and inputs it into a deep learning model; the model analyzes the feature matrix, compares it with a safety threshold benchmark library, and outputs the risk level.

[0200] Tiered and coordinated response:

[0201] If a risk warning is detected (e.g., the local temperature in the power distribution room is 70℃, which is below the 85℃ threshold, and other modal data are normal): the power distribution room's audible and visual alarm is activated, the management platform pushes the warning information, the visual monitoring camera focuses on the power distribution room, and uploads real-time images;

[0202] If an emergency risk is identified (e.g., smoke concentration in the workshop is 350ppm + infrared monitoring temperature is 90℃ + visual image captures suspected smoke): activate the workshop's audible and visual alarm, shut off the workshop's power valve, start the smoke exhaust fan, push an emergency warning to the on-duty personnel's APP, and simultaneously upload real-time images;

[0203] If the risk is determined to be a serious emergency (e.g., warehouse smoke concentration of 500ppm + infrared temperature of 150℃ + visual detection of flame characteristics): Based on the emergency risk response, activate the sprinkler system in the warehouse and surrounding area, unlock the evacuation access control, broadcast evacuation instructions over the park's public address system, and send alarm information to the fire and rescue department.

[0204] Human-computer interaction and feedback: On-duty personnel can view early warning information and real-time images through the management platform to confirm the risk type; if it is a false alarm, send a "close alarm" command; if it is a real risk, track the handling progress until the risk is eliminated; the system automatically records the data and operations of the entire process and generates a handling report.

[0205] Model optimization: The system regularly collects actual operating data to iteratively train the deep learning model and continuously optimize the recognition accuracy.

[0206] Through the above steps, this embodiment achieves multimodal and accurate identification and hierarchical linkage early warning of fire risks in the park, improves the level of intelligent fire management and emergency response efficiency, and ensures fire safety in the park.

[0207] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0208] Based on the same inventive concept, this application also provides a multimodal risk identification-based park fire linkage early warning system for implementing the multimodal risk identification-based park fire linkage early warning method described above. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the multimodal risk identification-based park fire linkage early warning system provided below can be found in the limitations of the multimodal risk identification-based park fire linkage early warning method described above, and will not be repeated here.

[0209] In one exemplary embodiment, such as Figure 4 As shown, a multimodal risk identification-based fire alarm system for industrial parks is provided. This multimodal risk identification-based fire alarm system 400 may include:

[0210] The data acquisition module 401 is used to acquire multimodal monitoring data of the park; the multimodal monitoring data includes temperature data, image data and smoke concentration data;

[0211] Data fusion module 402 is used to perform data preprocessing and weighted fusion processing on multimodal monitoring data to obtain the risk feature matrix of the park;

[0212] The risk prediction module 403 is used to input the risk feature matrix into a pre-trained risk prediction model for risk prediction processing to obtain the risk prediction information of the park.

[0213] The risk level determination module 404 is used to determine the fire risk level of the park based on risk prediction information and a preset safety threshold benchmark library.

[0214] The early warning processing module 405 is used to conduct fire-related early warning processing for the park based on the fire risk level.

[0215] In an exemplary embodiment, the data fusion module 402 is further configured to perform data preprocessing on the multimodal monitoring data to obtain preprocessed monitoring data; and to perform weighted fusion processing on the preprocessed monitoring data according to the scene feature information corresponding to the park to obtain a risk feature matrix.

[0216] In an exemplary embodiment, the data fusion module 402 is further configured to perform filtering and noise reduction processing on the temperature data to obtain target temperature data; perform image enhancement processing on the image data to obtain target image data; perform outlier removal processing on the smoke concentration data to obtain target smoke concentration data; and obtain preprocessed monitoring data based on the target temperature data, target image data, and target smoke concentration data.

[0217] In an exemplary embodiment, the data fusion module 402 is further configured to determine the fusion weight corresponding to each data modality in the preprocessed monitoring data based on the scene feature information; and to perform weighted fusion processing on the preprocessed monitoring data based on the fusion weight to obtain a risk feature matrix.

[0218] In an exemplary embodiment, the level determination module 404 is further configured to compare the risk prediction information with various safety thresholds in the safety threshold benchmark library to obtain the comparison result; determine the number and degree of abnormal modes in the multimodal monitoring data based on the comparison result; and determine the fire risk level of the park based on the number and degree of abnormal modes.

[0219] In an exemplary embodiment, the fire risk level includes a first risk level, a second risk level, or a third risk level; the early warning processing module 405 is further configured to perform a first-level fire linkage early warning processing for the park when the fire risk level is the first risk level; the first-level fire linkage early warning processing includes activating audible and visual alarms and triggering video verification of abnormal areas; when the fire risk level is the second risk level, the park is configured to perform a second-level fire linkage early warning processing; the second-level fire linkage early warning processing includes activating area smoke extraction equipment and emergency lighting equipment based on the first-level fire linkage early warning processing; when the fire risk level is the third risk level, the park is configured to perform a third-level fire linkage early warning processing; the third-level fire linkage early warning processing includes activating sprinkler linkage equipment and unlocking evacuation access control based on the second-level fire linkage early warning processing.

[0220] The modules in the aforementioned multimodal risk identification-based park fire alarm system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0221] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a multimodal risk identification and park fire alarm linkage method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0222] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0223] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0224] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above-described method embodiments.

[0225] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0226] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0227] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0228] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A multimodal risk identification method for fire-fighting linkage early warning in industrial parks, characterized in that, The method includes: Acquire multimodal monitoring data of the park; the multimodal monitoring data includes temperature data, image data, and smoke concentration data; The multimodal monitoring data is preprocessed and weighted fusion processed to obtain the risk feature matrix of the park. The risk feature matrix is ​​input into a pre-trained risk prediction model for risk prediction processing to obtain the risk prediction information of the park. Based on the risk prediction information and the preset safety threshold benchmark library, the fire risk level of the park is determined; Based on the fire risk level, fire alarm and early warning procedures will be implemented for the park.

2. The method according to claim 1, characterized in that, The process of preprocessing and weighted fusion of the multimodal monitoring data to obtain the risk feature matrix of the park includes: The multimodal monitoring data is preprocessed to obtain preprocessed monitoring data; Based on the scene feature information corresponding to the park, the preprocessed monitoring data is weighted and fused to obtain the risk feature matrix.

3. The method according to claim 2, characterized in that, The process of preprocessing the multimodal monitoring data to obtain preprocessed monitoring data includes: The temperature data is filtered and denoised to obtain the target temperature data; The image data is subjected to image enhancement processing to obtain the target image data; The smoke concentration data is processed to remove outliers, and the target smoke concentration data is obtained. Based on the target temperature data, the target image data, and the target smoke concentration data, the preprocessed monitoring data is obtained.

4. The method according to claim 2, characterized in that, The step of performing weighted fusion processing on the preprocessed monitoring data based on the scene feature information corresponding to the park to obtain the risk feature matrix includes: Based on the scene feature information, determine the fusion weights corresponding to each data modality in the preprocessed monitoring data; Based on the fusion weights, the preprocessed monitoring data is weighted and fused to obtain the risk feature matrix.

5. The method according to claim 1, characterized in that, The step of determining the fire risk level of the park based on the risk prediction information and a preset safety threshold benchmark library includes: The risk prediction information is compared with the various safety thresholds in the safety threshold benchmark library to obtain the comparison results; Based on the comparison results, determine the number and degree of abnormal modes in the multimodal monitoring data; The fire risk level of the park is determined based on the number of abnormal modes and the degree of abnormality.

6. The method according to any one of claims 1 to 5, characterized in that, The fire risk levels include the first risk level, the second risk level, or the third risk level; The fire alarm linkage and early warning processing for the park based on the fire risk level includes: When the fire risk level is the first risk level, the park will be subject to the first level of fire alarm linkage; the first level of fire alarm linkage includes activating the audible and visual alarm and triggering video verification of the abnormal area; When the fire risk level is the second risk level, the park shall be subject to the second level of fire alarm response; the second level of fire alarm response includes activating the area smoke exhaust equipment and emergency lighting equipment on the basis of the first level of fire alarm response. When the fire risk level is the third risk level, the park shall be subject to the third-level fire linkage early warning process; the third-level fire linkage early warning process includes, on the basis of the second-level fire linkage early warning process, activating the sprinkler linkage equipment and unlocking the evacuation access control.

7. A multimodal risk identification-based fire alarm system for industrial parks, characterized in that, The system includes: The data acquisition module is used to acquire multimodal monitoring data of the park; the multimodal monitoring data includes temperature data, image data, and smoke concentration data. The data fusion module is used to perform data preprocessing and weighted fusion processing on the multimodal monitoring data to obtain the risk feature matrix of the park; The risk prediction module is used to input the risk feature matrix into a pre-trained risk prediction model for risk prediction processing to obtain the risk prediction information of the park. The risk level determination module is used to determine the fire risk level of the park based on the risk prediction information and a preset safety threshold benchmark library. The early warning processing module is used to perform fire-related early warning processing for the park based on the fire risk level.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images