Fire alarm method combined with multi-camera collaboration

Abstract

The present invention discloses a fire alarm method combined with multi-camera collaboration, which relates to the technical field of fire alarm. The method obtains a camera network in a target area, constructs a multi-agent agent for fire alarm management, and is embedded in each network node in the camera network. The method determines the alarm condition based on the fire characteristics, constructs a lightweight fire alarm module and is built into the alarm response agent. The multi-agent agent assists the multi-agent agent, controls the camera network to monitor the target area, locates the target point with fire characteristics and performs fire tracking alarm management. The method is used to solve the technical problems of insufficient supervision decision-making ability and monitoring coordination ability in the prior art, which lead to insufficient accuracy and timeliness of disaster alarms and inability to adapt to the fire supervision needs in complex environments. The method significantly improves the accuracy and response efficiency of fire monitoring through coordinated monitoring and dynamic fire tracking.

Description

Technical Field

[0001] The present invention relates to the technical field of fire alarm technology, and in particular to a fire alarm method combined with multi-camera collaboration. Background Art

[0002] With technological advancements, multi-camera collaborative monitoring is gaining popularity and being applied to disaster monitoring. However, existing multi-camera collaborative technology suffers from shortcomings in data processing and analysis, making it difficult to quickly and accurately extract effective fire characteristics from massive amounts of video data, hindering timely and accurate alarm responses. Furthermore, the coordination between different cameras is not intelligent enough, and monitoring strategies cannot be dynamically adjusted according to changes in the fire situation.

[0003] Secondly, the system lacks dynamic fire tracking capabilities, enabling only flame recognition in static images and unable to track fire evolution in real time, especially in the early stages of a fire's spread. Furthermore, effective collaborative monitoring capabilities have been lacking. This results in insufficient global fire situational awareness, particularly in large areas or complex environments, limiting fire monitoring coverage.

[0004] Therefore, existing technologies in fire monitoring still have technical problems such as insufficient supervision decision-making capabilities and monitoring coordination capabilities, resulting in insufficient accuracy and timeliness of disaster alarms, and are unable to adapt to the fire supervision needs in complex environments. Summary of the Invention

[0005] This application provides a fire alarm method combined with multi-camera collaboration, which is used to solve the technical problems of insufficient supervision decision-making capabilities and monitoring coordination capabilities in the existing technology, resulting in insufficient accuracy and timeliness of disaster alarms and inability to adapt to the fire supervision needs in complex environments.

[0006] In view of the above problems, the present application provides a fire alarm method that combines multi-camera collaboration.

[0007] The present application provides a fire alarm method combined with multi-camera collaboration, the method comprising: obtaining a camera network in a target area, wherein the camera network is composed of distributed infrared and visual cameras; constructing a multi-agent agent for fire alarm management, and embedding it in each network node in the camera network, wherein the multi-agent agent at least includes field of view control - fire deduction - communication interaction - alarm response; determining alarm conditions based on fire characteristics, and constructing a lightweight fire alarm module, wherein the fire alarm module is built into an alarm response agent; assisting the multi-agent agent, controlling the camera network to monitor the target area, locating target points with fire characteristics, executing fire tracking and fire alarm management based on the fire alarm module, which includes single network node alarms and global alarms.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] The fire alarm method combined with multi-camera collaboration provided in the embodiment of the present application obtains the camera network of the target area, constructs a multi-agent agent for fire alarm management, and embeds it in each network node in the camera network. The multi-agent agent at least includes field of view control-fire deduction-communication interaction-alarm response; determines the alarm conditions based on the fire characteristics, constructs a lightweight fire alarm module, and the fire alarm module is built into the alarm response agent; assists the multi-agent agent to control the camera network to monitor the target area, locates the target point with fire characteristics, performs fire tracking and fire alarm management based on the fire alarm module, and is used to solve the technical problems of insufficient supervision decision-making ability and monitoring coordination ability in the prior art, resulting in insufficient accuracy and timeliness of disaster alarms, and the inability to adapt to the fire supervision needs of complex environments. Collaborative monitoring and dynamic fire tracking significantly improve the accuracy and response efficiency of fire monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flow chart of a fire alarm method combining multi-camera collaboration is provided for this application;

[0011] Figure 2 This application provides a schematic diagram of the fire tracking and fire alarm management process in a fire alarm method combined with multi-camera collaboration. DETAILED DESCRIPTION

[0012] This application provides a fire alarm method combined with multi-camera collaboration, obtains the camera network of the target area, constructs a multi-agent agent for fire alarm management, and embeds it in each network node in the camera network, determines the alarm conditions based on fire characteristics, constructs a lightweight fire alarm module and builds it into the alarm response agent, assists the multi-agent agent, controls the camera network to monitor the target area, locates the target point with fire characteristics and performs fire tracking alarm management, which is used to solve the technical problems of insufficient supervision decision-making ability and monitoring coordination ability in the existing technology, resulting in insufficient accuracy and timeliness of disaster alarms and the inability to adapt to the fire supervision needs in complex environments.

[0013] Example: Figure 1 As shown, the present application provides a fire alarm method combined with multi-camera collaboration, the method comprising:

[0014] S1: Obtain a camera network of the target area, wherein the camera network is composed of distributed infrared and visual cameras.

[0015] In the fire alarm method of the present invention, the camera network is composed of a distributed combination of infrared and visual cameras. Specifically, the target area refers to a specific area where fire monitoring and alarms are required, typically including building interiors or exteriors, industrial parks, forests, urban areas, etc. The camera network deployed within this area uses the coordinated operation of infrared and visual cameras to achieve multi-dimensional monitoring of fire detection.

[0016] Infrared cameras are used to detect heat fluctuations and can effectively identify fire sources at night or in low-light conditions. Infrared cameras use the distinct temperature characteristics of infrared radiation to identify fires. This is especially true in the early stages of a fire, when temperature fluctuations are significant. Infrared cameras can sensitively capture the heat radiation from the fire source and determine the fire's location. For example, by identifying abnormal temperature rises in high-temperature areas, they can provide early warning signs of a fire.

[0017] Vision cameras monitor target areas using visible light images. Combined with image feature recognition, they can identify obvious fire characteristics such as flames and smoke. Vision cameras are not only suitable for daytime monitoring but can also accurately capture and locate fire sources at greater distances, thanks to their resolution, viewing angle, and focusing capabilities. For example, they can scan and capture the dynamic changes of flames over a large area, providing intuitive data support for fire alarms.

[0018] Therefore, the camera network in the target area combines and distributes infrared and visual cameras, forming a multi-layered fire monitoring system. This system complements infrared and visual cameras, ensuring effective fire detection both at night and in low-light conditions, as well as during the day and in brightly lit environments. This combination enables the camera network to more comprehensively and accurately monitor potential fire risks within the target area, providing sufficient monitoring data to support subsequent fire alarms and responses.

[0019] S2: Construct a multi-agent agent for fire alarm management and embed it in each network node in the camera network. The multi-agent agent at least includes view control, fire situation simulation, communication interaction, and alarm response.

[0020] In the fire alarm method of the present invention, the multi-agent proxy refers to an agent component configured on each camera network node in the fire alarm system. Each agent component has autonomous decision-making and task allocation capabilities, and can intelligently analyze and process data collected by the camera, thereby effectively managing and responding to fire alarm events. The multi-agent proxy can work independently or collaboratively to complete different monitoring, alarm, and response tasks. Each network node in the camera network is configured with a multi-agent proxy.

[0021] In this step, the multi-agent agent includes at least four core modules: a visual control agent, a fire scenario simulation agent, a communication interaction agent, and an alarm response agent. Each module has specific functions and works together to ensure the efficient operation of the entire fire monitoring and alarm system.

[0022] The View Control Agent is responsible for controlling and adjusting the camera's field of view. This module allows the agent to dynamically adjust camera parameters such as rotation, tilt, and focus based on fire monitoring needs, maximizing coverage and improving fire detection accuracy. For example, if a camera detects abnormal temperature or smoke changes, the View Control module can automatically adjust the camera's focal length or viewing angle to focus on the suspected fire source, providing clearer images and data support.

[0023] The Fire Prediction Agent is used to predict fire trends based on real-time fire data. This module analyzes fire characteristics, such as the heat of the fire source, the spread of smoke, and the changes in flames, and uses data analysis and machine learning techniques to infer the likely development of a fire. This prediction process helps the system determine the speed and scope of a fire's spread, as well as its potential evolutionary path, supporting alarm decisions. For example, if the location of a fire source changes, the prediction module can predict the direction of the fire's spread and alert monitoring nodes in other areas in advance.

[0024] The communication and interaction agent is responsible for enabling data sharing and collaboration between camera network nodes. Through this module, agents can exchange fire data and alarm information, forming a global collaborative monitoring system. The communication and interaction module ensures smooth information flow between different camera nodes, preventing the impact of single node failures or false alarms, and improving the robustness and reliability of the entire system. For example, when a camera detects a fire and triggers an alarm, the communication and interaction module promptly transmits this information to other adjacent nodes, initiating a joint monitoring and response mechanism within the area.

[0025] The alarm response agent is the final execution module of the multi-agent system. Upon confirming a fire, it is responsible for issuing the appropriate alarm signal according to the specified alarm conditions. Depending on the severity and risk level of the fire, appropriate alarm methods can be selected, such as local alarm, remote alarm, and automatic linkage. For example, when a monitoring node determines that a fire has reached a high-risk threshold, the alarm response module will immediately trigger an alarm in the area or notify the fire management terminal through the communication interaction module for subsequent processing.

[0026] Multi-agent systems enable efficient and flexible management of fire alarm tasks, enabling rapid perception, accurate assessment, and timely response to fire risks within a target area. Each agent is embedded in a node in the camera network, and through the functional division and coordination of various modules, they together form an intelligent and efficient fire alarm management system.

[0027] S3: Determine alarm conditions based on fire characteristics and construct a lightweight fire alarm module, wherein the fire alarm module is built into the alarm response agent.

[0028] In the fire alarm method of the present invention, alarm conditions are determined based on fire characteristic data collected during fire monitoring, combined with specific judgment rules to generate alarm conditions suitable for fire alarm management. The fire alarm module, based on these alarm condition settings, is responsible for processing the fire data and triggering an alarm response.

[0029] First, fire characteristics refer to various characteristic data that can indicate the occurrence of a fire. Common ones include color, temperature, and appearance. Color refers to the color change of the fire source within the visible light range, temperature refers to the heat generated by the fire source, and appearance refers to the shape of the flame or the distribution of smoke. By collecting and analyzing these fire characteristics in real time, it is possible to accurately determine the occurrence and development of a fire.

[0030] Based on the fire characteristics, the alarm conditions are determined. Specifically, for example, when the fire source temperature exceeds a set threshold, the flame color reaches a specific chromaticity range, or the flame's appearance characteristics meet a preset standard, the system will determine that a fire has occurred and entered a high-risk state, triggering an alarm, and using the threshold condition as the alarm condition.

[0031] Preferably, the alarm conditions include a first alarm condition for a high-risk alert and a second alarm condition based on risk control and early warning. The setting of these alarm conditions, namely the first and second alarm conditions, allows for flexible responses to different stages of a fire's development, avoiding false alarms and missed alarms. The first alarm condition has a higher risk level than the second alarm condition and is used to address more urgent fire incidents, ensuring a rapid system response when a fire occurs.

[0032] The lightweight fire alarm module is the core of fire alarm management. It primarily uses a data-driven approach to conduct static and dynamic decision-making training based on collected fire data. Specifically, the alarm module first performs preliminary training by retrieving historical disaster alarm records in a data-driven manner to generate the module's initial decision rules. Based on this, static decision distillation training optimizes the alarm module based on the initial alarm condition, enabling it to respond and make decisions quickly in the early stages of a fire.

[0033] At the same time, dynamic decision distillation training is performed. This training is based on the dynamic changes in the fire situation, ensuring that the alarm module can flexibly respond to evolving fire characteristics. For example, the dynamic state of a fire source expanding or flame formation changing. Ultimately, this dual training of static and dynamic decision making ensures that the alarm system can make accurate and rapid alarm decisions in a variety of complex fire situations.

[0034] Ultimately, the constructed fire alarm module was embedded into the alarm response agent, becoming the core component of its response tasks. The module's lightweight design means it has low computational and storage requirements, enabling efficient operation on resource-constrained devices while ensuring the timeliness and accuracy of fire alarms. Based on the fire alarm module's judgment, the alarm response agent will execute appropriate alarm responses, including local alerting and remote notification.

[0035] Furthermore, to determine the alarm conditions based on the fire characteristics, step S3 of this application includes:

[0036] Determine fire characteristics, wherein the fire characteristics include at least chromaticity, temperature, and shape; set a first threshold group based on the fire characteristics as a first alarm condition for fire alarm; set a second threshold group based on the fire characteristics as a second alarm condition for fire warning, wherein the risk level of the first threshold group is higher than that of the second threshold group; and use the first alarm condition and the second alarm condition as the alarm conditions.

[0037] In the fire alarm method of the present invention, the fire characteristics include at least chromaticity, temperature and shape. These three characteristics are the basic basis for judging the occurrence of fire and can accurately reflect the different manifestations of fire.

[0038] Chromaticity refers to the color variation of a fire source within the visible light range. The color of the flame and smoke can indicate the temperature of the fire source and its severity. Color can also be used to distinguish between the background and the fire area. Temperature is the heat released by the fire source, and a rise in temperature is a clear sign of a fire. Appearance refers to the shape of the flame or the distribution of the smoke. During a fire, the shapes of flames and smoke often exhibit certain regularities. For example, rapidly spreading flames often exhibit irregular shapes, while smoke can exhibit various morphological characteristics, such as thick smoke or curved smoke flows. By monitoring these morphological characteristics, the type and development trend of a fire can be further confirmed.

[0039] Next, a first threshold group based on the fire characteristics is set as the first alarm condition for fire alerts. The first threshold group is determined based on the typical values ​​of the various parameters in the fire characteristics. For example, when the flame chromaticity reaches a certain range, the temperature rises to a certain critical value, or the flame shape exhibits certain typical characteristics, the system will determine that a fire has occurred and is in an emergency state. The values ​​corresponding to these fire characteristics will be used as the first alarm condition. This alarm condition has a higher risk level than the second alarm condition, ensuring a timely response to fire incidents and ensuring the safety of people and property.

[0040] The second threshold group is used to determine early warning signals for fire conditions and is typically set for low-risk fire events. For example, a slight increase in the temperature of the fire source or a preliminary change in the color of the flames, but not yet reaching a high-risk level, may trigger a warning mechanism based on the second alarm condition rather than an immediate alarm, thereby providing more time for emergency response. The second alarm condition has a lower risk level than the first alarm condition and is intended to prevent the fire from escalating while providing an early warning.

[0041] Using the first and second alarm conditions as the alarm conditions, when a fire occurs, the system automatically determines whether to trigger the high-risk first alarm condition or the low-risk second alarm condition based on real-time fire characteristics, taking into account factors such as color, temperature, and appearance. This hierarchical alarm mechanism allows the system to flexibly respond to fire incidents of varying severity, avoid false alarms and missed alarms, and ensure accurate and timely fire management.

[0042] Furthermore, to construct a lightweight fire alarm module, step S3 of this application includes:

[0043] By retrieving disaster alarm records, an initialization alarm module is constructed in a data-driven training manner; static decision distillation training is performed on the initialization alarm module based on the alarm conditions to determine the first alarm branch; dynamic decision distillation training is performed on the initialization alarm module based on the trend change of the fire characteristics to determine the second alarm branch; the first alarm branch and the second alarm branch are run in parallel to determine the fire alarm module.

[0044] In the fire alarm method of the present invention, an initialization alarm module is constructed using a data-driven training approach. Specifically, disaster alarm records refer to disaster alarm data that has occurred in the target area. By collecting and processing historical fire data, key fire characteristics and alarm patterns are extracted, such as temperature changes, flame color changes, smoke distribution, and other information, along with alarm response information. Sample training is then performed until convergence to determine the initialization alarm module.

[0045] Due to the different alarm types and diversified scenarios, in order to improve the pertinence and efficiency of alarm decisions, the initialization alarm module is distilled and trained from the static, that is, the judgment and decision-making at a unit moment, and the dynamic, that is, the judgment and decision-making based on feature trends.

[0046] Specifically, static decision distillation training is performed on the initialization alarm module, that is, only the logical rules related to static decisions in the initialization alarm module are filtered and retained. Based on the disaster alarm record, static samples are screened, and combined with the static samples, rule logic filtering is performed on the initialization alarm module, that is, the rules related to the static sample training decision process are retained to determine the first alarm branch.

[0047] Similarly, the initialization alarm module is trained for dynamic decision distillation, that is, only the logical rules related to dynamic decision-making in the initialization decision module are filtered and retained. Based on the disaster alarm records, dynamic samples are screened, that is, alarm records based on the trend of fire conditions. The trend of fire conditions refers to the various dynamic changes that occur during the development of the fire source, such as temperature rise, flame spread, and smoke changes. The initialization alarm module is subjected to rule logic filtering, that is, the rules related to the dynamic sample training decision-making process are retained to determine the first alarm branch.

[0048] By monitoring these dynamic changes in real time, the alarm module can adjust its decision-making rules to respond to evolving fire conditions. For example, if the fire source temperature gradually rises but has not yet reached the emergency alarm threshold, the system may activate the secondary alarm branch to issue a warning instead of directly issuing an alarm. The dynamic decision-making mechanism of the secondary alarm branch can gradually adjust the alarm strategy based on the changing trends of the fire in real time, preventing premature or delayed responses.

[0049] The first and second alarm branches operate in parallel to collaboratively provide alarm decisions. Through these steps, a fire alarm module integrating static and dynamic decision-making is constructed. This module not only meets conventional fire alarm needs but also flexibly responds to complex situations during fire evolution, improving the intelligence and accuracy of the fire alarm system.

[0050] S4: Assisting multi-agent agents, controlling the camera network to monitor the target area, locating target points with fire characteristics, performing fire tracking and fire alarm management based on the fire alarm module, including single network node alarm and global alarm.

[0051] In the fire alarm method of the present invention, each network node in the camera network, through the collaborative work of multiple agents, can efficiently monitor the target area in real time and detect fire characteristics. This multi-agent collaboration enables the entire camera network to cover a wider area and provide the ability to quickly respond and intelligently handle fire incidents.

[0052] First, through intelligent agent collaboration, each node in the camera network is managed to achieve dynamic adjustment and coordinated monitoring. For example, when a node detects a suspected fire, the agents at other nodes can adjust their viewing angles based on this information, expanding the monitoring range and ensuring comprehensive tracking of the fire source.

[0053] Multi-agent collaboration means that the field of view control agent can determine whether to adjust the field of view, focus, or expand the field of view based on the fire situation in the current monitored area. The fire situation prediction agent makes predictions based on the monitoring information and then collaborates with the alarm response agent to initiate an alarm. Furthermore, if the prediction decision involves the field of view of other network nodes, the communication interaction agent is used to exchange information. This multi-agent collaboration ensures the autonomous operation of each node in the camera network. Furthermore, the communication interaction agent ensures data exchange and collaborative management across the camera network.

[0054] Controlling a network of cameras to monitor a target area can pinpoint fire signatures, effectively determining the specific location of the fire source. This fire signature detection relies on information captured by the cameras, such as color, temperature, and shape. Target location is based on infrared or visual imaging data from the camera network, using the regional spatial coordinate system as a reference to determine the target point's coordinate position. Once the target point is located, the system further tracks the fire source, leveraging collaboration between cameras to track the spread of the fire.

[0055] Fire tracking involves confirming the fire's source location (the target point) and then tracking the fire's progression in real time through a network of cameras. Network nodes automatically adjust their monitoring angles and focus based on the fire's location and evolving trends. For example, if a fire source changes location or spreads, other nodes in the network quickly respond by adjusting their angles to track its movement. Fire tracking effectively provides dynamic information about a fire's progression, ensuring the system is constantly responsive to changes.

[0056] Next, fire alarm management is performed based on the fire alarm module. Alarm management is initiated based on fire event monitoring and tracking. Alarm management determines the alarm method and triggering timing based on the alarm conditions and risk level set by the fire alarm module. This includes single-node alarms and global alarm modes.

[0057] In the case of a single network node alarm, the system will only send an alarm signal to the camera node that detects the fire source. For example, when a camera node confirms the fire source and meets the alarm conditions, the node will directly trigger a local alarm, reporting through the node's alarm or to the fire management system.

[0058] Global alarms, on the other hand, are triggered when a fire spreads or when multiple camera nodes are working together. In this mode, all participating monitoring nodes share data through communication modules to identify the distribution of fire sources throughout the entire area. Fire data from multiple nodes is combined to generate a global fire alarm message, which is then sent to the fire management terminal, initiating a wide-scale emergency response.

[0059] Through these steps, multi-agent collaboration and camera network coordination enable precise location, continuous tracking, and effective alarming of fire incidents. Through precise target location and multi-mode alarm response, the system ensures a quick and accurate response to fires, preventing their spread and protecting people and property.

[0060] Further, such as Figure 2 As shown, executing fire tracking and fire alarm management based on the fire alarm module, step S4 of this application includes:

[0061] A target point with fire characteristics is located, and a first node of the camera network is determined, wherein the field of view of the first node is the target point; for the first node, a second node is calibrated by executing multi-agent collaboration, and the first node and the second node are coordinated to perform alarm management under fire tracking, wherein the second node is at least one and is determined based on the field of view of the fire evolution of the target point.

[0062] In the fire alarm method of the present invention, each camera node in the camera network is equipped with a certain field of view, or the monitoring range it can cover. When a camera node captures fire characteristics (such as rising temperature, spreading smoke, or changing flame color), it first analyzes and locates the precise location of the fire source. This location becomes the target point.

[0063] Next, the monitoring node at the fire source is identified as the first node, and real-time monitoring and fire analysis are performed through this node's field of view. This node will be responsible for initially tracking the fire source and ensuring continuous monitoring of the target point.

[0064] After the first node is determined, a multi-agent collaborative mechanism selects and calibrates additional camera nodes related to the fire source. These nodes are referred to as second nodes. The selection of second nodes depends on the location of the target point monitored by the first node and the fire's development trend. For example, if the first node detects a gradual increase in the temperature of the fire source, fire simulation will determine whether additional cameras in adjacent areas or viewpoints need to be activated for auxiliary monitoring. These newly added nodes become second nodes. The second node calibration process involves selecting a camera node with the best field of view for tracking the fire source based on the first node's monitoring range.

[0065] Similarly, as the fire monitoring spreads, collaborative nodes are added outward based on the second node.

[0066] By coordinating the first and second nodes, as well as subsequently added network nodes, alarm management under fire tracking is performed. Based on the communication interaction agent of each network node, interactive coordination between nodes is achieved to ensure comprehensive monitoring of the fire source. During the fire tracking process, the first node continuously monitors the core area of ​​the fire source, while the second node expands the monitoring range and provides additional information support. If the fire source changes within the field of view of the first node, the second node will dynamically adjust its monitoring perspective, update the location of the fire source in real time, and share monitoring data with the first node. Through this collaboration, the system can ensure that all changes in the fire source are within the monitoring range, thereby providing accurate fire evolution information.

[0067] Among them, the second node is at least one, which is determined based on the fire evolution field of the target point. This step means that the number and position of the second nodes are not only based on the initial position of the fire source, but also need to consider the evolution field of the fire. The fire evolution field means that as the fire spreads, the range, temperature, smoke and other characteristics of the fire source may change. Therefore, the selection and adjustment of the second node must be flexible to respond to these changes. For example, if the fire source begins to spread to a new area, the system will dynamically select more camera nodes as the second nodes according to the evolution path of the fire source to provide more accurate monitoring. Each newly added second node is optimized based on the fire evolution of the target point, thereby ensuring that the fire tracking system can adapt to the different development stages of the fire in real time.

[0068] In summary, the ability to accurately locate the fire source, dynamically select and coordinate multiple nodes for fire tracking ensures efficient and comprehensive monitoring and management of the entire fire process. This not only improves the timeliness of fire response, but also enhances the fire alarm system's adaptability in complex environments through coordinated multi-node cooperation.

[0069] Furthermore, after determining the first node of the camera network, step S4 of the present application includes:

[0070] Identify the fire characteristic value of the target point, trigger the first alarm branch in the alarm response agent to perform a static judgment based on the alarm condition, and determine the first alarm information; determine the fire characteristic trend based on the continuous monitoring of the first node, and perform a dynamic judgment in combination with the second alarm branch in the alarm response agent to determine the second alarm information; and control the first node to execute an alarm response based on the first alarm information and the second alarm information.

[0071] In the fire alarm method of the present invention, monitoring nodes (such as the first and second nodes) monitor target points in real time, collecting fire characteristic data in the area, primarily including temperature, color, smoke density, and shape. The specific vector values ​​of these characteristics are used as fire characteristic values, reflecting the basic state of the fire source. Combined with the first alarm branch, a threshold crossing determination is performed based on alarm conditions (such as temperature exceeding a set threshold or flame color change), executing a judgment process based on static alarm conditions. For example, if a node detects that the fire source temperature exceeds the set first alarm threshold, it determines the first alarm message based on this condition, triggering an emergency alarm response.

[0072] Next, based on continuous node monitoring, the fire characteristic trends are determined. This involves continuously monitoring the fire source to observe the changing trends of fire characteristics. Fire characteristic trends include changes in the rate of temperature rise, flame spread, and smoke concentration, all of which reflect the evolution of the fire. By monitoring these dynamic changes, the second alarm branch is activated for dynamic judgment. Unlike static judgment, dynamic judgment considers the evolution of the fire source over time. For example, if the temperature of the fire source continues to rise over a certain period of time at a high rate, and the smoke concentration gradually increases, a second alarm message is generated, indicating that a more serious fire may be imminent.

[0073] Finally, based on the first alarm information and the second alarm information, the first node is controlled to execute an alarm response. This step determines whether to issue an alarm signal based on the judgment results of the first alarm branch and the second alarm branch, and controls the node's alarm response according to the alarm information. Specifically, after receiving the first alarm information and the second alarm information, the system controls the first node to initiate the corresponding alarm response, taking into account the severity and evolution trend of the fire. If the first alarm information indicates that the fire source is already in an emergency state, and the second alarm information indicates that the fire source is rapidly expanding or intensifying, the first node will immediately execute an alarm response, issue an alarm to the fire management terminal, and trigger relevant emergency measures. In addition, if the judgment results of the two are inconsistent, the system will decide whether to execute the alarm response based on the set priority (usually the first alarm information has a higher priority) and adjust the urgency of the response measures.

[0074] Through the above steps, the multi-level alarm mechanism based on static and dynamic judgment can make flexible and accurate responses at different stages of fire development, avoiding premature or late alarms, thereby improving the efficiency and accuracy of fire prevention and control.

[0075] Furthermore, including global alarm, step S4 of this application includes:

[0076] According to the communication interaction agent, fire data interaction is performed based on the camera network to determine N fire fragments; the N fire fragments are traversed, and splicing based on relative viewing positions is performed to determine the entire fire distribution area; global alarm information based on the entire fire distribution area is generated, and the entire fire distribution area is displayed on the fire management terminal, and an alarm response based on the global alarm information is performed.

[0077] In the fire alarm method of the present invention, a communication and interaction agent enables data sharing and information exchange between camera nodes. When multiple camera nodes simultaneously monitor a fire, they aggregate their respective monitoring data (such as fire source location, fire characteristics, temperature, smoke concentration, etc.) through the communication and interaction agent. The data fragments captured by each camera node are called fire fragments. These fire fragments contain the local fire information observed by a single camera node. When multiple fragments are combined, they form complete fire monitoring data. Therefore, the role of the communication and interaction agent is to facilitate data exchange between different camera nodes and ensure comprehensive collection of fire information.

[0078] Next, the N fire fragments are traversed, where N is the number of cameras with fire in the camera field of view, and splicing based on the relative field of view position is performed to determine the entire fire distribution area. Specifically, the fire fragments obtained from each camera node are processed. Since the fields of view of each camera node are different, a single node can only monitor the local area of ​​the fire source. Therefore, it is necessary to splice together multiple local fire fragments by analyzing the relative field of view positions of the fire fragments. Preferably, according to the regional spatial coordinate system constructed above in this application, the N fire fragments obtained are subjected to a coordinate system distribution conversion to form a fire monitoring area covering the entire fire area, which serves as the entire fire distribution area.

[0079] For example, if a camera node monitors one part of a fire source, while an adjacent camera node monitors another part, the system will stitch together the monitoring data based on the relative positions of these nodes to create a complete fire distribution map. This stitching process not only relies on the camera node's field of view but also takes into account the dynamic changes of the fire source in time and space, ensuring that the stitched data accurately reflects the entire fire source.

[0080] Next, global alarm information is generated from the spliced ​​global fire distribution data. This information encompasses fire-related data captured by all camera nodes. This information includes not only the specific location and changing trends of the fire source, but also the risk level and potential expansion range. The overall fire risk is assessed, illustratively by weighting the fire feature values ​​of each disaster fragment and summing them as global disaster data. This information is then combined with the first warning branch to determine if a threshold has been crossed, generating the global alarm information. Optimally, this information is weighted based on location importance and the acceleration of feature growth.

[0081] The global alarm information and the global fire distribution are displayed on the fire management terminal for operators to review. By displaying the global fire distribution, managers can clearly see the dynamic progress of the entire fire incident and its impact on different areas, thereby better formulating emergency response plans.

[0082] Finally, based on this global alarm information, a disaster alert is issued, initiating the corresponding alarm response mechanism. If the global alarm information indicates that the fire risk has reached a certain threshold, the system will issue an alarm through the fire management terminal or directly through a connected alarm. Alarm information can include the specific location of the fire, its expansion trend, and the predicted spread path. The alarm response is not limited to triggering audible and visual alarms but can also incorporate multiple measures such as the activation of firefighting equipment and evacuation guidance. Through this comprehensive response mechanism, the system ensures that all relevant personnel and equipment can respond quickly when a fire occurs, reducing the damage caused by the fire.

[0083] In summary, by generating comprehensive fire distribution maps and alarm information, comprehensive fire monitoring and emergency response are achieved. This alarm management method based on data interaction and information splicing not only improves the accuracy and timeliness of fire alarms, but also enhances the overall effectiveness of multi-camera collaboration, providing more efficient technical support for fire prevention and control.

[0084] Furthermore, the application steps also include:

[0085] Establish a connection between the camera network and the alarm, wherein the alarm is distributed and installed in the target area; set an alarm mode, and perform alarm management on the alarm and the fire management terminal according to the alarm mode.

[0086] In the fire alarm method of the present invention, a stable communication connection is established between each camera in the camera network and the nearest alarm installed therein according to the proximity principle. Exemplarily, the connection can be achieved through wireless communication, local area network connection or other compatible communication protocols. The alarms are usually distributed, covering various locations in the target area. Based on the fire information received, the alarms can perform alarm functions such as sound and light alarms and vibration prompts to promptly notify the surrounding personnel of the occurrence of the fire. The distributed assembly of the alarms can ensure that when a fire occurs, the alarm signal can be quickly and evenly transmitted to the entire target area, thereby providing a comprehensive alarm response.

[0087] The alarm mode refers to the response method triggered in different fire situations, and may include various forms such as on-site alarm, remote alarm, and graded alarm.

[0088] Based on the specific requirements of the alarm model, the alarm information from each alarm device and fire management terminal is centrally managed and controlled. For example, if the fire management terminal receives a high-risk alarm, the system will initiate emergency response measures, including issuing evacuation instructions to relevant personnel or activating the automatic fire extinguishing system through the fire management terminal. Simultaneously, the alarm device will also perform corresponding on-site alarm actions, such as flashing warning lights and emitting warning sounds, to ensure that on-site personnel receive the fire alarm and respond as quickly as possible. The system's alarm management function coordinates the interaction between cameras, alarm devices, and fire management terminals, thereby achieving comprehensive, multi-level fire early warning and response.

[0089] In summary, this ensures that when a fire occurs, relevant equipment can quickly respond based on the specific fire situation, greatly improving the accuracy and timeliness of fire response. The automated collaboration and flexible adjustments in this process provide efficient and reliable technical support for fire prevention and control.

[0090] Furthermore, the steps of this application also include: setting a first alarm mode and a second alarm mode, wherein the first alarm mode is an on-site alarm based on the alarm, and the second alarm mode is an online alarm based on the fire management terminal; according to the alarm scenario, determining the first sub-mode and the second sub-mode, wherein the first sub-mode is a fire alarm based on the first alarm condition, and the second sub-mode is a fire warning based on the second alarm condition.

[0091] The fire alarm method of this invention provides two different alarm modes, each designed to respond to different scenarios after a fire occurs. The first alarm mode is on-site alarm, which uses alarms (such as audible and visual alarms, vibration alarms, etc.) installed in the target area to immediately issue an on-site alarm upon fire. The purpose of on-site alarm is to notify on-site personnel as quickly as possible, enabling them to take emergency measures such as evacuation or fire extinguishing.

[0092] The second alarm mode is online alarming, which involves sending alarm information via the fire management terminal to a remote monitoring center or emergency response department. In this mode, alarm information is delivered to management personnel or relevant departments via data packets, text messages, phone calls, or push notifications, enabling timely decision-making and initiation of emergency response measures. This mode's advantage lies in its ability to transcend physical limitations, enabling remote processing and coordination of fire information. It is particularly suitable for fire emergency management requiring cross-regional coordination.

[0093] Next, based on the alarm scenario, the system determines the first and second sub-modes. The first sub-mode is a fire alarm based on the first alarm condition, and the second sub-mode is a fire warning based on the second alarm condition. In this step, the system further refines the alarm mode based on the specific fire situation and the alarm scenario.

[0094] Specifically, the first and second alarm modes can be further subdivided into the first and second sub-modes, respectively. The first sub-mode triggers an on-site alarm when fire characteristics meet certain conditions (such as temperature or smoke concentration reaching a specific threshold) in the first or second alarm modes. At this point, the alarm will directly issue an emergency fire alarm based on the determination of these characteristic values. For example, if the temperature is too high or the smoke concentration is abnormal, the on-site sound and light alarm will quickly activate to alert personnel to the fire.

[0095] The second sub-mode triggers a fire warning based on the secondary alarm condition. At this point, the fire management terminal analyzes data to determine whether the fire is in the early warning stage, such as if the fire source is small or spreading slowly, posing no immediate threat. The warning information may include preliminary fire data, projected area of ​​spread, and fire characteristics requiring attention. The alarm response in this mode is relatively mild, intended to provide managers with a preliminary assessment of fire development and allow for proactive prevention and preparation.

[0096] In summary, this system can flexibly respond to the alarm needs of different fire scenarios. By setting different alarm modes and sub-modes, the alarm is more targeted and flexible, providing a more accurate and timely fire response. This hierarchical alarm response method improves the comprehensiveness and efficiency of fire prevention and control.

[0097] Furthermore, after executing the fire alarm management, the application steps also include:

[0098] According to the camera network, the alarm response information is tracked and monitored; the alarm response information is identified and the risk control status is determined. If the risk control is favorable, the alarm execution is terminated; if the risk control is unfavorable or there is no response, an external alarm is executed.

[0099] In the fire alarm method of the present invention, the camera network continuously monitors and tracks the alarm response information after a fire occurs. When a fire occurs, each camera node collects real-time data within its field of view as monitoring alarm response information, that is, the on-site reaction after the alarm.

[0100] Next, the alarm response information is identified and the risk control status is determined. Specifically, by performing data analysis based on the collected monitoring alarm response information, the risk status of the fire, that is, the state quantity of the disaster characteristics, is identified. Risk control status determination refers to determining the change in the degree of risk based on the change in fire information. For example, an increase in temperature value increases the degree of risk. Comprehensively analyze the risk level and development trend of the current fire situation, and determine whether it meets the conditions for continuing to alarm. Specifically, the risk of the fire is assessed in real time based on multiple parameters such as the speed of the fire source expansion, the severity of the fire, the smoke concentration, and temperature changes. If the fire has been effectively controlled, or the expansion of the fire source has been curbed, the system will determine that the risk control is favorable, that is, the fire control situation is good and no further alarm response is required.

[0101] If the system determines that risk control is favorable, the alarm execution is terminated. In this case, the system decides to terminate the originally scheduled alarm execution based on the risk control judgment. For example, if the system detects that the fire has been extinguished, or the fire source is no longer spreading, and the fire risk level has decreased, the current alarm has been processed and the instruction to stop further alarms will not be triggered other alarm devices or notify management personnel. At this time, the system will feedback the current fire control status through the fire management terminal, avoiding unnecessary waste of resources and personnel panic.

[0102] However, if the system determines that risk control is poor or unresponsive, meaning the fire risk is not effectively controlled or the source of the fire has not been promptly addressed, the system will decide to initiate an external alert. At this point, an alert is issued to relevant external entities (such as the fire department, emergency management center, etc.) to ensure that professional rescue forces can respond promptly and take emergency measures. For example, external alerts may include automatically dialing emergency numbers, sending text messages, or pushing emergency fire information via online platforms. This measure is intended to address situations where a fire spreads or becomes uncontrollable, ensuring that relevant personnel and rescue departments are immediately aware of the actual fire situation and can take effective measures.

[0103] Through these steps, the system can dynamically adjust alarm response strategies and flexibly implement alarm management based on the real-time development of the fire situation. When the fire is well under control, the system can promptly terminate the alarm to avoid unnecessary interference. If the fire spreads or is not effectively controlled, the system can promptly initiate an external alarm to ensure that external rescue forces can quickly intervene, thus ensuring the effectiveness and safety of fire prevention and control.

[0104] The fire alarm method combined with multi-camera collaboration provided in this application has the following technical effects:

[0105] 1. Comprehensive monitoring of target areas is achieved through a distributed network of infrared and visual cameras. Real-time visual and thermal data from fires are collected, forming a foundational data network for fire monitoring. This enhances the comprehensiveness and flexibility of fire monitoring, enabling effective fire detection in diverse environments.

[0106] 2. Multiple agents are embedded in camera network nodes, each with independent functional modules (such as field of view control, fire simulation, communication interaction, and alarm response). These agents work collaboratively to drive and control each network node, jointly completing target area monitoring, fire simulation, and alarm response when a fire occurs. This achieves efficient and automated fire detection and alarming. Monitoring strategies can be flexibly adjusted based on specific fire scenarios, avoiding delays and errors caused by manual operation. Fire data from the camera network is exchanged to form a complete fire distribution map, which in turn generates global alarm information.

[0107] 3. By determining fire characteristics and setting multiple alarm thresholds, a two-level alarm system is formed, corresponding to fire warning and fire early warning respectively. The intensity and type of alarms can be automatically adjusted according to the different stages of the fire, avoiding excessive alarms or missed alarms, and improving alarm accuracy and response efficiency.

[0108] 4. Build a lightweight fire alarm module, using distilled training based on static and dynamic decisions to create an alarm system with automatic learning capabilities. This optimizes alarm response logic based on historical fire data and fire trends, making the system more flexible and adaptable in practical applications. This allows for more accurate handling of complex fire scenarios and enhanced early warning capabilities. Furthermore, risk control assessments and external alarm mechanisms ensure efficient emergency response.

[0109] In summary, a flexible, efficient and accurate fire monitoring and response system is formed.

[0110] Through the above detailed description of the fire alarm method combined with multi-camera collaboration in this specification, those skilled in the art can clearly understand the fire alarm method combined with multi-camera collaboration in this embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0111] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fire alarm method combining multi-camera collaboration is characterized in that: The method comprises: Obtaining a camera network of the target area, wherein the camera network is composed of distributed infrared and visual cameras; Construct a multi-agent agent for fire alarm management and embed it in each network node in the camera network. The multi-agent agent includes at least view control, fire situation simulation, communication interaction, and alarm response. Determine alarm conditions based on fire characteristics and construct a lightweight fire alarm module, wherein the fire alarm module is built into the alarm response agent; Auxiliary multi-agent agent controls the camera network to monitor the target area, locates the target point with fire characteristics, performs fire tracking and fire alarm management based on the fire alarm module, including single network node alarm and global alarm; The execution of fire tracking and fire alarm management based on the fire alarm module includes: Locating a target point where fire characteristics are present, and determining a first node of the camera network, wherein the field of view of the first node is the target point; For the first node, a second node is calibrated by executing multi-agent collaboration, and the first node and the second node are coordinated to perform alarm management under fire tracking, wherein the second node is at least one and is determined based on the fire evolution view of the target point; After determining the first node of the camera network, the method includes: Identify the fire characteristic value of the target point, trigger the first alarm branch in the alarm response agent to perform static judgment based on the alarm condition, and determine the first alarm information; Determine the fire characteristic trend based on the continuous monitoring of the first node, perform dynamic judgment in conjunction with the second alarm branch in the alarm response agent, and determine the second alarm information; Controlling the first node to execute an alarm response according to the first alarm information and the second alarm information; This includes global alarms, including: According to the communication interaction agent, fire situation data interaction based on the camera network is performed to determine N fire situation fragments; Traversing the N fire fragments, performing splicing based on relative view positions, and determining the entire fire distribution area; Generate global alarm information based on the global fire distribution, and execute the display of the global fire distribution and the alarm response based on the global alarm information on the fire management terminal.

2. The fire alarm method combined with multi-camera collaboration as claimed in claim 1, characterized in that: Determine alarm conditions based on fire characteristics, including: Determining fire characteristics, wherein the fire characteristics include at least color, temperature, and shape; Setting a first threshold value group based on the fire characteristics as a first alarm condition for a fire alarm; Setting a second threshold value group based on the fire characteristics as a second alarm condition for fire early warning, wherein the risk level of the first threshold value group is higher than that of the second threshold value group; The first alarm condition and the second alarm condition are used as the alarm condition.

3. The fire alarm method combined with multi-camera collaboration as claimed in claim 2, characterized in that: Build a lightweight fire alarm module, including: By retrieving disaster alarm records and using data-driven training, an initialization alarm module is constructed; Performing static decision distillation training on the initialization alarm module based on the alarm condition to determine a first alarm branch; Based on the trend change of the fire characteristics, dynamic decision distillation training is performed on the initialization alarm module to determine the second alarm branch; The first alarm branch and the second alarm branch are operated in parallel to determine the fire alarm module.

4. The fire alarm method combined with multi-camera collaboration as claimed in claim 1, characterized in that: The method further comprises: Establishing a connection between the camera network and alarms, wherein the alarms are distributed and installed in the target area; Set the alarm mode and perform alarm management for the alarm and fire management terminal according to the alarm mode.

5. The fire alarm method combined with multi-camera collaboration as claimed in claim 4, characterized in that: Setting a first alarm mode and a second alarm mode, wherein the first alarm mode is an on-site alarm based on an alarm device, and the second alarm mode is an online alarm based on the fire management terminal; According to the alarm scenario, a first sub-mode and a second sub-mode are determined, wherein the first sub-mode is a fire alarm based on a first alarm condition, and the second sub-mode is a fire warning based on a second alarm condition.

6. The fire alarm method combined with multi-camera collaboration as claimed in claim 1, characterized in that: After implementing fire alarm management, including: Tracking and monitoring alarm response information based on the camera network; Identify the alarm response information and determine the risk control status. If the risk control is favorable, terminate the alarm execution; if the risk control is unfavorable or there is no response, execute an external alarm.

Images