Fire extinguishing method and system based on visual positioning

Abstract

The application discloses a fire extinguishing method and system based on visual positioning, and relates to the technical field of fire extinguishing. The method comprises the following steps: when a fire occurs in a chemical enterprise, a cluster of unmanned aerial vehicles is used to collect visual information from multiple angles and in all directions, an image recognition model on board is used for real-time analysis, a fire analysis result is generated and uploaded to a cloud server; in the cloud server, the fire analysis result is subjected to visual positioning to generate a fire visual positioning result; based on the fire visual positioning result, a multi-target enzyme action optimization algorithm is used to iteratively optimize fire extinguishing decisions to generate an optimal fire extinguishing strategy; and according to the fire visual positioning result and the optimal fire extinguishing strategy, a fire water cannon is used to execute fire extinguishing. The application solves the problems of weak fire sensing, easy secondary disaster caused by decision making and low striking precision in the prior art.

Description

Technical Field

[0001] This invention relates to the field of fire extinguishing technology, and in particular to a fire extinguishing method and system based on visual positioning. Background Technology

[0002] Chemical plants typically store and handle large quantities of flammable, explosive, toxic, and hazardous chemicals. Once a fire breaks out, the flames spread rapidly and can easily trigger a chain of explosions and leaks of highly toxic substances.

[0003] Traditional chemical firefighting methods have the following limitations: 1) Weak perception capability: Fire scenes are often accompanied by thick smoke and high temperatures, making it difficult for firefighters to get close. Single-point monitoring cameras have blind spots and cannot obtain the precise three-dimensional location of the fire scene and the overall situation of the fire.

[0004] 2) High degree of blind decision-making: The selection of extinguishing agents (such as water, foam, dry powder, etc.) is highly dependent on the commander's experience. If the selected extinguishing agent is incompatible with the hazardous chemicals stored nearby (such as substances that burn upon contact with water), it can easily trigger greater secondary disasters.

[0005] 3) Low accuracy: Traditional fire monitors often rely on manual adjustment of elevation angle and pressure, making it difficult to accurately and continuously cover and strike the core area of ​​the fire source in complex environments.

[0006] Therefore, there is an urgent need for an automated fire-fighting technology solution that can accurately locate fire scenes visually and make intelligent decisions and precise strikes based on the characteristics of chemical plant areas. Summary of the Invention

[0007] This invention provides a visual positioning-based fire extinguishing method and system. This invention solves the problems of weak detection in chemical fires, decision-making that easily leads to secondary disasters, and low accuracy in fire suppression in existing technologies.

[0008] In a first aspect, embodiments of the present invention provide a fire extinguishing method based on visual positioning, the method comprising: When a fire occurs in a chemical plant, a cluster of drones is used to collect multi-angle, all-round visual data of the fire scene. The onboard image recognition model is used for real-time analysis to generate fire scene analysis results, which are then uploaded to a cloud server. On the cloud server, the fire scene analysis results are visually located to generate fire scene visual location results; Based on the visual location results of the fire scene, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize the fire extinguishing decision and generate the optimal fire extinguishing strategy. Based on the visual location results of the fire scene and the optimal fire extinguishing strategy, fire monitors were used to carry out fire extinguishing.

[0009] The technical solution provided in this application has at least the following beneficial effects: By collaboratively acquiring dual-light video streams through a drone swarm, and utilizing the MT-CNN algorithm to segment fire sources, assess intensity, and identify hazardous chemical equipment, combined with BIM models and bundle adjustment, a precise conversion from two-dimensional images to a three-dimensional world coordinate system was achieved, solving the problem of fire source location in dense smoke environments. An innovative multi-objective enzyme-based optimization algorithm was introduced, incorporating "incompatibility between the extinguishing agent and surrounding high-risk threatened equipment" as a highly weighted penalty term into the fitness function, in addition to fire extinguishing efficiency and cost, thus fundamentally preventing explosions caused by incorrect extinguishing agent selection. Secondary disasters caused by explosions or toxic gas leaks; the multi-objective enzyme action optimization algorithm integrates the gray wolf cooperative idea and the transformation mechanism, ensuring population diversity through chaotic initialization, triggering transformation through potential field difference, and combining penetration component, vertical component and encirclement traction correction term, which greatly improves the algorithm's global search capability and convergence speed in complex multi-objective space; based on the absolute three-dimensional coordinates of the fire source and the optimal strategy parameters, the elevation angle, rotation angle and outlet pressure of the fire monitor are automatically calculated, realizing the seamless connection of "cloud decision-making-end execution", which greatly improves the fire extinguishing efficiency.

[0010] In one alternative implementation, when a fire occurs at a chemical plant, a swarm of drones is used to conduct multi-angle, all-around visual data acquisition of the fire scene. Onboard image recognition models are then used for real-time analysis to generate fire scene analysis results, which are uploaded to a cloud server, including: When a fire occurs in a chemical plant, a drone swarm consisting of several drones is deployed at the fire site, and a consensus-based swarm algorithm is used to allocate tasks to the drone swarm and generate data collection tasks for each drone. In each of the aforementioned UAVs, based on the pre-stored BIM terrain data of the chemical plant area, complementary observation trajectory planning is performed according to the data acquisition task to generate several collision-free and complementary flight trajectories. Using drones, dual-light video streams are synchronously acquired along the corresponding flight path, and a timestamp synchronization mechanism is introduced to align all image frames acquired by the drones at each moment on the time axis to obtain an aligned dual-light video stream. The aligned dual-light video stream includes visible light images and thermal infrared images of consecutive frames. The aligned dual-light video streams are input into the corresponding UAV's onboard image recognition model for real-time analysis, generating corresponding fire scene analysis results, which are then uploaded to the cloud server.

[0011] In one optional implementation, the fire analysis results include fire source area masking, fire intensity index, and distribution of surrounding hazardous chemical equipment; The airborne image recognition model is constructed based on the MT-CNN algorithm and includes an input layer, a bimodal feature encoding backbone module, a fire source semantic segmentation branch, a fire intensity regression branch, and a surrounding equipment detection branch connected to the bimodal feature encoding backbone module, as well as an output layer.

[0012] In one optional implementation, the formula for obtaining the fire source region mask is:

[0013] In the formula, The first in the fire scene analysis results j Drones in l A mask for the fire source area at any given moment; This represents the fused dual-modal features. The convolutional layer operation function for splitting branches; For activation functions; l For time indication; j For drone indication; The formula for obtaining the fire intensity index is:

[0014] In the formula, The first in the fire scene analysis results j Drones in l The fire intensity index at any given moment; The flame area; Masking the fire source area Total pixel area; For pixels p In thermal infrared images grayscale values ​​in; p Pixel indicator; This is a fire source area; This represents the maximum range of the thermal infrared sensor. This is the weighting coefficient for the fire intensity index; The distribution of surrounding hazardous chemical equipment includes a bounding box set and category labels.

[0015] In one alternative implementation, the fire scene analysis results are visually located on a cloud server to generate visual location results for the fire scene, including: On the cloud server, the fire scene analysis results uploaded by each drone are received, and a weighted fusion algorithm based on confidence is used to generate a global situation. The global situation includes a global fire source mask, a global fire intensity, and a detection set of surrounding equipment. The global fire source mask in the global situation is combined with the drone pose data of the drone swarm, and the bundle adjustment method is introduced to optimize the point cloud coordinates to construct a three-dimensional point cloud model of the fire scene. The three-dimensional point cloud model of the fire site was registered with the pre-stored BIM model of the chemical enterprise by ICP, and the rigid body transformation matrix was solved. Based on the rigid body transformation matrix, the three-dimensional fire source point cloud in the three-dimensional point cloud model of the fire scene is transformed to the world coordinate system to obtain the transformed fire source point cloud; Cluster analysis was performed on the transformed fire source point cloud to calculate the geometric center of the main cluster and obtain the absolute coordinates of the fire source. Calculate the minimum bounding box of the transformed fire source point cloud, and calculate the volume of the minimum bounding box to obtain the fire source volume; Based on the distribution of surrounding hazardous chemical equipment and the absolute coordinates of the fire source, spatial distance is determined, hazardous chemical equipment within the radiant heat range of the fire is identified, and a list of high-risk threatened equipment is generated. By combining the absolute coordinates of the fire source, the volume of the fire source, and the list of high-risk threatened equipment, the visual location results of the fire scene are obtained.

[0016] In one optional implementation, spatial distance is determined based on the distribution of surrounding hazardous chemical equipment and the absolute coordinates of the fire source. Hazardous chemical equipment within the radiant heat range of the fire is identified, and a list of highly threatened equipment is generated, including: Extract the center pixel coordinates of each device from the bounding box set of the surrounding hazardous chemical equipment distribution, and use the back projection formula of the camera projection model, combined with the UAV pose and ground height estimation, to calculate the approximate world coordinates of the device. Iterate through all identified devices and calculate the spatial Euclidean distance between their approximate world coordinates and the absolute coordinates of the fire source; If the spatial Euclidean distance is less than the safe distance threshold, the device is identified as a high-risk threatened device, its attribute information is retrieved from the BIM database, and it is added to the list of high-risk threatened devices.

[0017] In one alternative implementation, based on the visual location results of the fire scene, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize fire extinguishing decisions and generate the optimal fire extinguishing strategy, including: Based on the visual positioning results of the fire scene, a fitness function of the multi-objective enzyme action optimization algorithm is set, and the fire extinguishing decision is encoded as the position vector of the enzyme molecule in the multi-objective enzyme action optimization algorithm. The fire extinguishing decision includes the type of extinguishing agent, the mixing ratio, and the pressure of the fire monitor. Based on the fitness function, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize fire extinguishing decisions and obtain the Pareto optimal solution set. The Pareto optimal solution set is weighted and filtered to obtain the optimal solution. The position vector of the optimal solution is decoded to obtain the optimal fire extinguishing strategy, which includes the optimal extinguishing agent type, the optimal mixing ratio, and the optimal fire monitor pressure.

[0018] In one alternative implementation, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize fire extinguishing decisions based on a fitness function, yielding a Pareto optimal solution set, including: The chaotic sequence is generated using Logistic mapping and then mapped to the parameter space of the enzyme molecule in the multi-objective enzyme action optimization algorithm to obtain the initial population. Using the fitness function, the fitness value of each initial enzyme molecule in the initial population is calculated, and the historical best position of each initial enzyme molecule is obtained based on the fitness value. Based on the cooperative concept of gray wolves, the optimal, second-best, and third-best enzyme molecules are selected from the initial population according to the non-dominant ranking and crowding distance. Based on the historical best position and the best, second-best, and third-best enzyme molecules, calculate the initial basic movement step size of the enzyme molecule and the potential field difference between the pre-updated position and the current position. If the difference is greater than the potential field surge threshold, trigger the allosteric transformation and proceed to the next step. The basic moving step size is decomposed into penetration and vertical components, and combined with the potential field difference, the deformation contraction coefficient and expansion coefficient are dynamically calculated. The initial population is updated based on the penetration component, vertical component, allosteric contraction coefficient, and expansion coefficient to obtain the updated population. Using a fitness function, the fitness value of each updated enzyme molecule in the updated population is calculated. Based on the fitness value, non-dominated sorting and crowding calculation are performed, dominated inferior solutions are eliminated, and the frontier solution set is output as the Pareto optimal solution set.

[0019] In one alternative implementation, based on the visual location results of the fire scene and the optimal fire suppression strategy, fire monitors are used to perform fire suppression, including: Based on the absolute coordinates of the fire source and the installation location of the fire monitor in the visual positioning results of the fire scene, calculate the horizontal distance and relative height difference for fire extinguishing; Calculate the initial velocity of the water jet from the fire monitor based on the optimal fire monitor pressure and the optimal extinguishing agent type in the optimal fire extinguishing strategy. Based on the horizontal distance, relative height difference, and initial velocity of the water jet, solve for the pitch angle and horizontal rotation angle; Based on the pitch angle and horizontal rotation angle, control the spray position of the fire monitor and control the valves in the fire monitor pipeline to switch to the storage tank pipeline corresponding to the optimal extinguishing agent type; Adjust the opening of the proportional mixing valve in the fire monitor pipeline to set the mixing ratio to the optimal mixing ratio; Start the variable frequency pump of the fire monitor, gradually stabilize the output pressure to the optimal fire monitor pressure, open the outlet valve of the fire monitor, so that the extinguishing agent covers the fire source area according to the predetermined trajectory, and carry out fire extinguishing.

[0020] Secondly, embodiments of the present invention provide a vision-based fire extinguishing system for implementing a fire extinguishing method. The system includes: The fire scene analysis unit is used to collect multi-angle, all-round visual data of the fire scene when a fire occurs in a chemical enterprise using a drone swarm. The unit uses an onboard image recognition model to perform real-time analysis, generate fire scene analysis results, and upload them to a cloud server. The visual positioning unit is used to perform visual positioning of the fire scene analysis results on the cloud server and generate fire scene visual positioning results. The fire extinguishing decision unit is used to iteratively optimize the fire extinguishing decision based on the visual positioning results of the fire scene using a multi-objective enzyme-based optimization algorithm to generate the optimal fire extinguishing strategy. The fire extinguishing execution unit is used to execute fire extinguishing using fire monitors based on the visual positioning results of the fire scene and the optimal fire extinguishing strategy.

[0021] A third aspect of this invention provides an electronic device, which includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, such that the at least one processor can perform the method proposed in the first aspect of the present invention.

[0022] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in the first aspect of the present invention. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart of the steps of a fire extinguishing method based on visual positioning provided in an embodiment of the present invention; Figure 3 This is a functional unit diagram of a vision-based fire extinguishing system provided in an embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0025] The present invention will be further described below with reference to the accompanying drawings.

[0026] Reference Figure 1 , Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention.

[0027] like Figure 1 As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0028] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0029] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and an electronic program for a vision-based fire extinguishing system.

[0030] exist Figure 1In the electronic device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device. The electronic device calls the electronic program of the visual positioning-based fire extinguishing system stored in the memory 1005 through the processor 1001 and executes the visual positioning-based fire extinguishing method provided in the embodiment of the present invention.

[0031] Reference Figure 2 The present invention provides a visual positioning-based fire extinguishing method, the method comprising: S201: When a fire occurs in a chemical plant, a cluster of drones is used to conduct multi-angle, all-round visual acquisition of the fire scene, and the onboard image recognition model is used for real-time analysis to generate fire scene analysis results, which are then uploaded to the cloud server. S202: On the cloud server, the fire scene analysis results are visually located to generate fire scene visual location results; S203: Based on the visual location results of the fire scene, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize the fire extinguishing decision and generate the optimal fire extinguishing strategy. S204: Based on the visual location results of the fire scene and the optimal fire extinguishing strategy, use fire monitors to carry out fire extinguishing.

[0032] The technical solution provided in this application has at least the following beneficial effects: By collaboratively acquiring dual-light video streams through a swarm of drones, and utilizing a Multi-Task Convolutional Neural Network (MT-CNN) algorithm, fire source segmentation, intensity assessment, and hazardous chemical equipment identification are achieved. This is combined with Building Information Modeling (BIM). Modeling (BIM) and bundle adjustment achieve precise conversion from 2D images to a 3D world coordinate system, solving the problem of fire source location in dense smoke environments. An innovative multi-objective enzyme-based optimization algorithm is introduced, which, in addition to fire extinguishing efficiency and cost, incorporates "incompatibility between the extinguishing agent and surrounding high-risk threatened equipment" as a highly weighted penalty term into the fitness function, fundamentally preventing secondary disasters such as explosions or toxic gas leaks caused by incorrect extinguishing agent selection. The multi-objective enzyme-based optimization algorithm integrates the gray wolf cooperative thinking and the transformation mechanism, ensuring population diversity through chaotic initialization, triggering transformation through potential field differences, and combining penetration components, vertical components, and encirclement traction correction terms to significantly improve the algorithm's global search capability and convergence speed in complex multi-objective spaces. Based on the absolute 3D coordinates of the fire source and optimal strategy parameters, the algorithm automatically calculates the elevation angle, rotation angle, and outlet pressure of the fire monitor, achieving seamless integration of "cloud-based decision-making and end-side execution," greatly improving fire extinguishing efficiency.

[0033] In one alternative implementation, when a fire occurs at a chemical plant, a swarm of drones is used to conduct multi-angle, all-around visual data acquisition of the fire scene. Onboard image recognition models are then used for real-time analysis to generate fire scene analysis results, which are uploaded to a cloud server, including: S2011: When a fire occurs in a chemical plant, a drone swarm consisting of several drones is deployed at the fire site, and a consensus-based swarm algorithm is used to allocate tasks to the drone swarm and generate data collection tasks for each drone. In this embodiment, when the fire alarm system is triggered or the fire is manually confirmed, the cloud server automatically wakes up the firefighting drone nest and deploys a cluster of several drones. A consensus-based clustering algorithm (such as the Consensus-Based Bundle Algorithm (CBBA)) is used to divide the three-dimensional space of the fire scene into multiple observation sub-regions. Conflict-free data acquisition tasks are generated by combining the battery level and location of each drone. S2012: For each UAV, based on the pre-stored BIM terrain data of the chemical plant area, complementary observation trajectory planning is performed according to the data acquisition task to generate several collision-free and complementary flight trajectories. In this embodiment, at the single-machine level, based on the pre-stored BIM terrain data of the chemical plant area (including tank height, pipeline direction, and plant outline), an improved fast expanding random tree algorithm is used to plan complementary observation trajectories. When generating path nodes, the algorithm introduces a "view complementarity" evaluation function to ensure that the angle between the observation lines of adjacent UAVs is greater than a preset threshold (such as 30 degrees), thereby generating several flight trajectories that will not collide with the tall equipment in the plant area and can cover the fire source from multiple sides. S2013: Using a drone, synchronously acquire dual-light video streams along the corresponding flight path, and introduce a timestamp synchronization mechanism to align all image frames acquired by the drone at each moment on the time axis to obtain an aligned dual-light video stream, which includes visible light images and thermal infrared images of consecutive frames. In this embodiment, the high-frequency pulse per second (PPS) signal provided by the UAV's flight control system is used as the reference clock to stamp each frame of visible light image and thermal infrared image with an absolute timestamp; for micro-hour differences, optical flow method is used to perform sub-pixel level image alignment, and finally a dual-light video stream with strict time axis alignment is obtained. S2014: The aligned dual-light video stream is input into the corresponding UAV's onboard image recognition model for real-time analysis, generating the corresponding fire scene analysis results, and uploading them to the cloud server.

[0034] In one optional implementation, the fire analysis results include fire source area masking, fire intensity index, and distribution of surrounding hazardous chemical equipment; The airborne image recognition model is constructed based on the MT-CNN algorithm and includes an input layer, a dual-modal feature encoding backbone module, a fire source semantic segmentation branch, a fire intensity regression branch, and a surrounding equipment detection branch connected to the dual-modal feature encoding backbone module, as well as an output layer. The input layer is used to receive the aligned dual-light video stream; The dual-modal feature coding backbone module is equipped with a dual-stream input architecture and a feature fusion unit. The visible light stream in the dual-stream input architecture is used to extract visible light features from the visible light image in the aligned dual-light video stream; The thermal infrared stream in the dual-stream input architecture is used to extract thermal infrared features from the thermal infrared image in the aligned dual-light video stream; The feature fusion unit is used to weighted fuse visible light features and thermal infrared features using a feature pyramid fusion strategy to obtain fused dual-modal features. The fire source semantic segmentation branch is equipped with a decoder structure and a segmentation prediction head; The decoder structure is used to gradually restore the fused bimodal features to the original image resolution using transposed convolution, and introduces skip connections to fuse the shallow high-resolution features with the deep semantic features to obtain semantic features, so as to ensure the fineness of the fire source edge. The segmentation prediction head is used to compress the feature channels to 1 using 1×1 convolution, and outputs a probability map through the Sigmoid activation function based on semantic features to obtain the fire source region mask. The fire intensity regression branch includes a global feature aggregation layer and a regression prediction head. The global feature aggregation layer is used to compress the fused bimodal features into a global feature vector using global average pooling, capturing the overall statistical characteristics of the fire scene. An attention weighting mechanism is set to enhance the features of the fire source area identified by the segmentation branch, resulting in enhanced features of the fire source area and suppressing background noise interference. The regression prediction head is used to nonlinearly combine the enhanced features and semantic features to obtain the nonlinearly combined features. Based on the nonlinearly combined features, the flame area and infrared integral term are calculated to obtain the normalized fire intensity index. The peripheral device detection branch is equipped with a region suggestion network, a region of interest (ROI) alignment layer, and a multi-task prediction head; Region Proposal Network (RPN) is used to generate candidate regions that may contain targets based on fused bimodal features. The RPN leverages the saliency of thermal infrared features to quickly locate metal devices with high heat capacity or specific shapes. The ROI Align layer is used to extract a fixed-size feature map for each candidate region, thus solving the deformation problem caused by changes in the drone's viewpoint. The multi-task prediction head is used to output the class probability distribution based on the feature map using the classification sub-head and the bounding box regression sub-head to output the bounding box coordinate offset. This is used to refine the equipment position and obtain the bounding box set of surrounding hazardous chemical equipment and its class labels.

[0035] In one optional implementation, the formula for obtaining the fire source region mask is:

[0036] In the formula, The first in the fire scene analysis results j Drones in l The fire source area mask at any given time, where a value of 1 represents the fire source and a value of 0 represents the background; This refers to the weighted fusion of visible light features from visible light images extracted using an airborne image recognition model and thermal infrared features from thermal infrared images to obtain the fused dual-modal features; This is the convolutional layer operation function for segmenting branches, used to map high-dimensional feature maps into single-channel feature maps with the same size as the original image; The activation function maps the output value to the interval [0,1], representing the probability that each pixel belongs to the fire source; l For time indication; j For drone indication; The formula for obtaining the fire intensity index is:

[0037] In the formula, The first in the fire scene analysis results j Drones in l The fire intensity index at any given time, ranging from 0 to 1. A value closer to 1 indicates a more intense fire. The flame area is represented by the fire source area mask. The total number of pixels with a value of 1, which is the projected area of ​​the flame on the image; Masking the fire source area The total pixel area is used for normalization to eliminate the influence of image resolution differences; For pixels p In thermal infrared images The grayscale value in the image is positively correlated with the actual temperature. p Pixel indicator; This is a fire source area; This is the maximum range of the thermal infrared sensor, used to normalize the temperature integral term; The weighting coefficient for the fire intensity index satisfies... For example, in a dense smoke environment, it can be increased It focuses on temperature information; in open flame scenarios, it can increase To emphasize the combustion area; The distribution of surrounding hazardous chemical equipment includes a bounding box set. and category labels ,in, For the first A bounding box, For the first The category labels corresponding to each bounding box; This represents the total number of bounding boxes.

[0038] In one alternative implementation, the fire scene analysis results are visually located on a cloud server to generate visual location results for the fire scene, including: S2021: On the cloud server, the fire scene analysis results uploaded by each drone are received. A weighted fusion algorithm based on confidence level is used to generate a global situational awareness. The global situational awareness includes a global fire source mask, a global fire intensity, and a set of surrounding equipment detection data. The formula is as follows:

[0039] In the formula, This system serves as a global fire source masking system, a global fire intensity monitoring system, and a detection set for surrounding equipment. Union operator; N The total number of drones; For the first j Confidence weights for drones; S2022: This method combines the global fire source mask from the overall situational awareness with the drone pose data from the drone swarm, and introduces bundle adjustment to optimize the point cloud coordinates, constructing a 3D point cloud model of the fire scene. The formula is as follows:

[0040] In the formula, For the first j The rotation matrix and translation vector of the UAV in the world coordinate system; Fire source area The Middle k The three-dimensional spatial coordinates of the fire source in the world coordinate system; This is to identify the rotation matrix, translation vector, and three-dimensional spatial coordinates of the fire source point for the UAV in the world coordinate system. For the first j On the image plane of the drone, corresponding to the first k The actual observed pixel coordinates of the fire source; For camera projection functions; The square of the reprojection error; k This is an indicator of the fire source. S2023: Perform iterative closest point (ICP) registration between the 3D point cloud model of the fire scene and the pre-stored BIM model of the chemical enterprise, and solve for the rigid body transformation matrix. The formula is as follows:

[0041] In the formula, To assign an objective function to the ICP registration; The rigid body transformation matrix; Let be the rotation matrix and translation vector of the rigid body transformation matrix; 3D point cloud model of the fire scene The Middle Three-dimensional point; For BIM model The Middle Three-dimensional point; For three-dimensional point indication; S2024: Based on the rigid body transformation matrix, the 3D fire source point cloud in the 3D point cloud model of the fire scene is transformed to the world coordinate system to obtain the transformed fire source point cloud. The formula is:

[0042] In the formula, This is the converted fire source point cloud; Fire source area The Middle The three-dimensional spatial coordinates of the three-dimensional fire source point in the world coordinate system; A three-dimensional point cloud model of the fire scene; S2025: Perform cluster analysis on the transformed fire source point cloud, calculate the geometric center of the main cluster, and obtain the absolute coordinates of the fire source. The formula is:

[0043] In the formula, The absolute coordinates of the fire source; The total number of three-dimensional fire source point clouds; The three-dimensional absolute coordinates of the fire source center in the world coordinate system; S2026: Calculate the minimum bounding box of the transformed fire source point cloud, and calculate the volume of the minimum bounding box to obtain the fire source volume; S2027: Based on the distribution of surrounding hazardous chemical equipment and the absolute coordinates of the fire source, determine the spatial distance, identify hazardous chemical equipment within the range of the fire's radiant heat, and generate a list of high-risk threatened equipment. S2028: By combining the absolute coordinates of the fire source, the volume of the fire source, and the list of high-risk threatened equipment, the visual location results of the fire scene are obtained.

[0044] In one optional implementation, spatial distance is determined based on the distribution of surrounding hazardous chemical equipment and the absolute coordinates of the fire source. Hazardous chemical equipment within the radiant heat range of the fire is identified, and a list of highly threatened equipment is generated, including: S20271: Extract the center pixel coordinates of each device from the bounding box set of the surrounding hazardous chemical equipment distribution, and use the back projection formula of the camera projection model, combined with the UAV pose and ground height estimation, to calculate the approximate world coordinates of the device. S20272: Traverse all identified devices and calculate the spatial Euclidean distance between their approximate world coordinates and the absolute coordinates of the fire source. The formula is:

[0045] In the formula, For the first Spatial Euclidean distance of the equipment; For the first bounding box in the set Approximate world coordinates of the device; The absolute coordinates of the fire source; For equipment indication; S20273: If the spatial Euclidean distance is less than the safe distance threshold, the device is identified as a high-risk threatened device, its attribute information is retrieved from the BIM database, and it is added to the list of high-risk threatened devices.

[0046] In one alternative implementation, based on the visual location results of the fire scene, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize fire extinguishing decisions and generate the optimal fire extinguishing strategy, including: S2031: Based on the visual positioning results of the fire scene, a fitness function for a multi-objective enzyme action optimization algorithm is set, and the fire extinguishing decision is encoded as the position vector of the enzyme molecule in the multi-objective enzyme action optimization algorithm. The fire extinguishing decision includes the extinguishing agent type, mixing ratio, and fire monitor pressure, as shown in the formula:

[0047] In the formula, enzyme molecules X fitness value; X For fire suppression decision-making, the corresponding enzyme molecule position vector in the multi-objective enzyme action optimization algorithm. ; This includes the type of extinguishing agent, the mixing ratio, and the pressure of the fire monitor. enzyme molecules X The fire extinguishing efficiency value; enzyme molecules X The cost value; enzyme molecules X The secondary environmental hazard value;

[0048] In the formula, Fire extinguishing agent type and mixing ratio The overall efficiency coefficient is as follows; The flow rate of the fire monitor under the pressure of the fire monitor; The volume of the fire source in the visual location results of the fire scene; This refers to the overall fire intensity in the fire scene analysis results. This is the fire intensity coefficient;

[0049] In the formula, Fire extinguishing agent type The unit cost of the extinguishing agent; The duration of the fire monitor's spray;

[0050] In the formula, To determine the incompatible decision function, query the "Chemical Emergency Response Database" to determine the first... High-risk threatened equipment With extinguishing agent type Compatibility; This is an indicator function that converts the compatibility determination result into a numerical value. If the incompatibility determination is True (there is a risk), then... =1, if the judgment is False (safe), then =0; List of high-risk, threatened devices; List of high-risk, threatened devices The first in High-risk, threatened equipment; For a high penalty coefficient, a very large positive number (such as 10) 6 Or 10 9 ); S2032: Based on the fitness function, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize fire extinguishing decisions and obtain the Pareto optimal solution set. S2033: A weighted selection process is performed on the Pareto optimal solution set to obtain the optimal solution. The location vector of the optimal solution is then decoded to obtain the optimal fire extinguishing strategy. This optimal fire extinguishing strategy includes the optimal extinguishing agent type, the optimal mixing ratio, and the optimal fire monitor pressure, as shown in the following formula:

[0051] In the formula, The optimal firefighting strategy; The optimal fire extinguishing agent type, optimal mixing ratio, and optimal fire monitor pressure are determined in the optimal fire extinguishing strategy. The weighted selection coefficients are used for the selection process. For the first i enzyme molecules The fire extinguishing effectiveness, cost, and secondary environmental hazard values; For the first i Enzyme molecules; i This is the indicator quantity for enzyme molecules.

[0052] In one alternative implementation, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize fire extinguishing decisions based on a fitness function, yielding a Pareto optimal solution set, including: S20321: Use Logistic mapping to generate chaotic sequences, and map the chaotic sequences to the parameter space of enzyme molecules in the multi-objective enzyme action optimization algorithm to obtain the initial population; The formula is:

[0053] In the formula, For the first n+ 1. n There are several chaotic variables whose values ​​range from [0, 1]. The stability coefficient is typically 4. This sequence is ergodic and random, ensuring that the initial population is uniformly distributed in the solution space, avoiding getting trapped in local optima, which is superior to traditional random initialization. n For chaotic variable indicators;

[0054] In the formula, For the initial population, the first i An initial enzyme molecule; For the first i One chaotic variable; These are the upper and lower bounds of the parameter space; i Indicator amount for enzyme molecules; t This is an indicator of the number of iterations. S20322: Using the fitness function, calculate the fitness value of each initial enzyme molecule in the initial population, and obtain the historical best position of each initial enzyme molecule based on the fitness value. S20323: Based on the gray wolf cooperative concept, the optimal, second-optimal, and third-optimal enzyme molecules are selected from the initial population according to the non-dominant ranking level and crowding distance. The formula is as follows:

[0055] In the formula, For the first t The iteration of the ... i A new enzyme molecule The gray wolf hierarchical selection function, in the first iteration, For the first i An initial enzyme molecule; for Non-dominated ranking order; for Crowding distance, The distance at maximum congestion; For the first t The optimal enzyme molecule, the second-best enzyme molecule, and the third-best enzyme molecule in the next iteration; S20324: Based on the historical best position and the best, second-best, and third-best enzyme molecules, calculate the initial basic movement step size of the enzyme molecule and the potential field difference between the pre-updated position and the current position. If the difference is greater than the potential field surge threshold, trigger the allosteric change and proceed to the next step. In this embodiment, based on the historical best position and the best, second-best, and third-best enzyme molecules, the initial base movement step size of the enzyme molecule is calculated using the following formula:

[0056] In the formula, For the first t+ The base movement step size for one iteration; For the first t The iteration of the ... i A new enzyme molecule, in the first iteration, For the first i An initial enzyme molecule; For the first t The iteration of the ... i The best historical position; t This represents the current iteration number; The self-cognitive acceleration factor is a positive constant that regulates the step size by which an enzyme molecule moves toward its optimal position in its historical sequence. The larger the enzyme molecule is, the more likely it is to linger near a local optimum (i.e., a previously found good position); The social cognitive acceleration coefficient is a positive constant that regulates the step size at which enzyme molecules move towards the optimal, suboptimal, and third-optimal enzyme molecules. The larger the value, the faster the algorithm converges, but too large a value can lead to premature convergence. As a random perturbation vector, it is a randomly generated random number (or random vector) in the interval [0,1], which breaks the absolutely fixed and rigid trajectory and gives the enzyme molecule the characteristics of "Brownian motion" in the optimization process, enabling it to have the ability to explore randomly and jump out of local optimal solutions. For the first t The gray wolf collaboration guides the direction in the next iteration;

[0057] In the formula, For the first t The first, second, and third potential movement vectors of the next iteration; For the first t The next iteration The distance vector between the optimal enzyme molecule, the second-best enzyme molecule, and the third-best enzyme molecule; This represents the vector of the first, second, and third control coefficients; These are the first, second, and third oscillation coefficients; A random number between [0, 1]; For control coefficients and oscillation coefficients; For the first t The convergence factor of the next iteration; t This represents the current iteration number;

[0058] In the formula, These are the maximum and minimum values ​​of the convergence factor; This is the threshold for the number of iterations; Calculate the initial pre-update position of the enzyme molecule, and then calculate the potential difference between the pre-update position and the current position, using the following formula:

[0059] In the formula, For the first t+ The first iteration i The pre-update position of each updated enzyme molecule; For the first t The potential field difference in the next iteration; The secondary environmental hazard values ​​are for the pre-updated location and the current location; It is the minimum value; If the potential difference is greater than the potential surge threshold, then the structure is triggered and the step size decomposition step is entered. S20325: Decompose the basic moving step size into penetration and vertical components, and combine the potential field difference to dynamically calculate the deformation contraction coefficient and expansion coefficient.

[0060]

[0061] In the formula, This represents the penetration component along the negative direction of the potential field gradient; The vertical component is perpendicular to the gradient direction; The gradient direction for the secondary environmental hazard values ​​at the pre-updated location; Based on the moving step size The angle between the gradient and the negative gradient direction; The normal vector is the direction of the gradient.

[0062] In the formula, These are the coefficients of thermal expansion and contraction. It is the expansion elasticity coefficient; It is an exponential function; S20326: The initial population is updated based on the penetration component, vertical component, morphological contraction coefficient, and expansion coefficient to obtain the updated population. In this embodiment, a gray wolf encirclement and traction correction term is introduced to calculate the final step size after the modification, as shown in the formula:

[0063] In the formula, Correction item for the gray wolf encirclement and traction; For the first t The potential field difference in the next iteration; This represents the final step size after the transformation. This is the weighting coefficient for the encirclement and traction. Based on the final step size after the transformation, the initial population is updated to obtain the updated population, using the following formula:

[0064] In the formula, For the first t +1 iterations of the first iteration i A newer enzyme molecule; S20327: Using the fitness function, calculate the fitness value of each updated enzyme molecule in the updated population, and perform non-dominated sorting and crowding calculation based on the fitness value. Eliminate dominated inferior solutions and output the frontier solution set as the Pareto optimal solution set.

[0065] In one alternative implementation, based on the visual location results of the fire scene and the optimal fire suppression strategy, fire monitors are used to perform fire suppression, including: S2041: Based on the absolute coordinates of the fire source and the installation location of the fire monitor in the visual positioning results of the fire scene, calculate the horizontal distance and relative height difference for fire extinguishing. The formula is:

[0066] In the formula, Horizontal distance; This refers to the relative height difference; absolute coordinates of the fire source Coordinates on the X, Y, Z axes; The coordinates of the fire monitor's installation location on the X, Y, and Z axes; S2042: Based on the optimal fire monitor pressure and optimal extinguishing agent type in the optimal fire extinguishing strategy, calculate the initial velocity of the fire monitor's water jet. The formula is:

[0067] In the formula, The initial velocity of the water jet from the fire monitor; The optimal extinguishing agent type in the optimal fire extinguishing strategy. The density of the extinguishing agent; The optimal fire monitor pressure in the optimal fire extinguishing strategy; For flow coefficient; S2043: Based on the horizontal distance, relative height difference, and initial velocity of the water jet, calculate the pitch angle and horizontal rotation angle using the following formulas:

[0068] In the formula, The elevation angle of the fire monitor; It is the gravitational acceleration constant; It is the arctangent function; It is the horizontal rotation angle; The horizontal rotation angle in world coordinates; This is the initial yaw angle; S2044: Based on the pitch and yaw angles, control the spray position of the fire monitor and control the valves in the fire monitor pipeline to switch to the optimal extinguishing agent type. Corresponding storage tank piping; S2045: Adjust the opening of the proportional mixing valve in the fire monitor pipeline to set the mixing ratio to the optimal mixing ratio. ; S2046: Start the variable frequency pump for the fire monitor, and gradually stabilize the output pressure to the optimal fire monitor pressure. Open the outlet valve of the fire monitor to allow the extinguishing agent to cover the fire source area according to the predetermined trajectory and carry out fire extinguishing.

[0069] This invention also provides a vision-based fire extinguishing system 300, referring to... Figure 3 The system may include the following units: The fire analysis unit 301 is used to collect multi-angle, all-round visual data of the fire scene when a fire occurs in a chemical enterprise using a cluster of drones, perform real-time analysis using an onboard image recognition model, generate fire analysis results, and upload them to a cloud server. The visual positioning unit 302 is used to perform visual positioning on the fire scene analysis results on the cloud server and generate fire scene visual positioning results. Firefighting decision unit 303 is used to iteratively optimize firefighting decisions based on the visual positioning results of the fire scene using a multi-objective enzyme-based optimization algorithm to generate the optimal firefighting strategy. The fire extinguishing execution unit 304 is used to execute fire extinguishing using fire monitors based on the visual positioning results of the fire scene and the optimal fire extinguishing strategy.

[0070] Based on the same inventive concept, another embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the vision-based fire extinguishing method of the present invention.

[0071] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EI) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. The communication interface is used for communication between the aforementioned terminal and other devices. The memory can include Random Access Memory (RAM), or non-volatile memory, such as at least one disk storage device. Optionally, the memory can also be at least one storage device located remotely from the aforementioned processor.

[0072] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0073] Furthermore, to achieve the above objectives, embodiments of the present invention also propose a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the vision-based fire extinguishing method of the present invention.

[0074] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable hardware devices (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0075] The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (apparatus), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0076] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0078] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. "And / or" indicates that either one or both can be chosen. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the element.

[0079] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A fire extinguishing method based on visual positioning, characterized in that, The method includes: When a fire occurs in a chemical plant, a cluster of drones is used to collect multi-angle, all-round visual data of the fire scene. The onboard image recognition model is used for real-time analysis to generate fire scene analysis results, which are then uploaded to a cloud server. On the cloud server, the fire scene analysis results are visually located to generate fire scene visual location results; Based on the visual location results of the fire scene, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize the fire extinguishing decision and generate the optimal fire extinguishing strategy. Based on the visual location results of the fire scene and the optimal fire extinguishing strategy, fire monitors were used to carry out fire extinguishing.

2. The fire extinguishing method based on visual positioning according to claim 1, characterized in that, When a fire occurs at a chemical plant, a swarm of drones is used to conduct multi-angle, all-around visual data acquisition of the fire scene. Onboard image recognition models are used for real-time analysis, generating fire scene analysis results which are then uploaded to a cloud server, including: When a fire occurs in a chemical plant, a drone swarm consisting of several drones is deployed at the fire site, and a consensus-based swarm algorithm is used to allocate tasks to the drone swarm and generate data collection tasks for each drone. In each of the aforementioned UAVs, based on the pre-stored BIM terrain data of the chemical plant area, complementary observation trajectory planning is performed according to the data acquisition task to generate several collision-free and complementary flight trajectories. Using drones, dual-light video streams are synchronously acquired along the corresponding flight path, and a timestamp synchronization mechanism is introduced to align all image frames acquired by the drones at each moment on the time axis to obtain an aligned dual-light video stream. The aligned dual-light video stream includes visible light images and thermal infrared images of consecutive frames. The aligned dual-light video streams are input into the corresponding UAV's onboard image recognition model for real-time analysis, generating corresponding fire scene analysis results, which are then uploaded to the cloud server.

3. The fire extinguishing method based on visual positioning according to claim 2, characterized in that, The fire analysis results include fire source area masking, fire intensity index, and distribution of surrounding hazardous chemical equipment. The airborne image recognition model is constructed based on the MT-CNN algorithm and includes an input layer, a bimodal feature encoding backbone module, a fire source semantic segmentation branch, a fire intensity regression branch, and a surrounding equipment detection branch connected to the bimodal feature encoding backbone module, as well as an output layer.

4. The fire extinguishing method based on visual positioning according to claim 3, characterized in that, The formula for obtaining the fire source area mask is: In the formula, The first in the fire scene analysis results j Drones in l A mask for the fire source area at any given moment; This represents the fused dual-modal features. The convolutional layer operation function for splitting branches; For activation functions; l For time indication; j For drone indication; The formula for obtaining the fire intensity index is: In the formula, The first in the fire scene analysis results j Drones in l The fire intensity index at any given moment; The flame area; Masking the fire source area Total pixel area; For pixels p In thermal infrared images grayscale values ​​in; p For pixel point indication; This is a fire source area; This represents the maximum range of the thermal infrared sensor. This is the weighting coefficient for the fire intensity index; The distribution of surrounding hazardous chemical equipment includes a bounding box set and category labels.

5. The fire extinguishing method based on visual positioning according to claim 4, characterized in that, On the cloud server, the fire scene analysis results are visually located to generate visual location results for the fire scene, including: On the cloud server, the fire scene analysis results uploaded by each drone are received, and a weighted fusion algorithm based on confidence is used to generate a global situation. The global situation includes a global fire source mask, a global fire intensity, and a detection set of surrounding equipment. The global fire source mask in the global situation is combined with the drone pose data of the drone swarm, and the bundle adjustment method is introduced to optimize the point cloud coordinates to construct a three-dimensional point cloud model of the fire scene. The three-dimensional point cloud model of the fire site was registered with the pre-stored BIM model of the chemical enterprise by ICP, and the rigid body transformation matrix was solved. Based on the rigid body transformation matrix, the three-dimensional fire source point cloud in the three-dimensional point cloud model of the fire scene is transformed to the world coordinate system to obtain the transformed fire source point cloud; Cluster analysis was performed on the transformed fire source point cloud to calculate the geometric center of the main cluster and obtain the absolute coordinates of the fire source. Calculate the minimum bounding box of the transformed fire source point cloud, and calculate the volume of the minimum bounding box to obtain the fire source volume; Based on the distribution of surrounding hazardous chemical equipment and the absolute coordinates of the fire source, spatial distance is determined, hazardous chemical equipment within the radiant heat range of the fire is identified, and a list of high-risk threatened equipment is generated. By combining the absolute coordinates of the fire source, the volume of the fire source, and the list of high-risk threatened equipment, the visual location results of the fire scene are obtained.

6. The fire extinguishing method based on visual positioning according to claim 5, characterized in that, Based on the distribution of surrounding hazardous chemical equipment and the absolute coordinates of the fire source, spatial distance is determined to identify hazardous chemical equipment within the radiant heat range of the fire, generating a list of high-risk threatened equipment, including: Extract the center pixel coordinates of each device from the bounding box set of the surrounding hazardous chemical equipment distribution, and use the back projection formula of the camera projection model, combined with the UAV pose and ground height estimation, to calculate the approximate world coordinates of the device. Iterate through all identified devices and calculate the spatial Euclidean distance between their approximate world coordinates and the absolute coordinates of the fire source; If the spatial Euclidean distance is less than the safe distance threshold, the device is identified as a high-risk threatened device, its attribute information is retrieved from the BIM database, and it is added to the list of high-risk threatened devices.

7. The fire extinguishing method based on visual positioning according to claim 6, characterized in that, Based on the visual location results of the fire scene, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize fire extinguishing decisions and generate the optimal fire extinguishing strategy, including: Based on the visual positioning results of the fire scene, a fitness function of the multi-objective enzyme action optimization algorithm is set, and the fire extinguishing decision is encoded as the position vector of the enzyme molecule in the multi-objective enzyme action optimization algorithm. The fire extinguishing decision includes the type of extinguishing agent, the mixing ratio, and the pressure of the fire monitor. Based on the fitness function, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize fire extinguishing decisions and obtain the Pareto optimal solution set. The Pareto optimal solution set is weighted and filtered to obtain the optimal solution. The position vector of the optimal solution is decoded to obtain the optimal fire extinguishing strategy, which includes the optimal extinguishing agent type, the optimal mixing ratio, and the optimal fire monitor pressure.

8. The fire extinguishing method based on visual positioning according to claim 7, characterized in that, Based on the fitness function, a multi-objective enzyme-based optimization algorithm is used to iteratively optimize fire extinguishing decisions, obtaining a Pareto optimal solution set, including: The chaotic sequence is generated using Logistic mapping and then mapped to the parameter space of the enzyme molecule in the multi-objective enzyme action optimization algorithm to obtain the initial population. Using the fitness function, the fitness value of each initial enzyme molecule in the initial population is calculated, and the historical best position of each initial enzyme molecule is obtained based on the fitness value. Based on the cooperative concept of gray wolves, the optimal, second-best, and third-best enzyme molecules are selected from the initial population according to the non-dominant ranking and crowding distance. Based on the historical best position and the best, second-best, and third-best enzyme molecules, calculate the initial basic movement step size of the enzyme molecule and the potential field difference between the pre-updated position and the current position. If the difference is greater than the potential field surge threshold, trigger the allosteric transformation and proceed to the next step. The basic moving step size is decomposed into penetration and vertical components, and combined with the potential field difference, the deformation contraction coefficient and expansion coefficient are dynamically calculated. The initial population is updated based on the penetration component, vertical component, allosteric contraction coefficient, and expansion coefficient to obtain the updated population. Using a fitness function, the fitness value of each updated enzyme molecule in the updated population is calculated. Based on the fitness value, non-dominated sorting and crowding calculation are performed, dominated inferior solutions are eliminated, and the frontier solution set is output as the Pareto optimal solution set.

9. The fire extinguishing method based on visual positioning according to claim 8, characterized in that, Based on the visual location results of the fire scene and the optimal fire-fighting strategy, fire monitors are used to carry out fire-fighting operations, including: Based on the absolute coordinates of the fire source and the installation location of the fire monitor in the visual positioning results of the fire scene, calculate the horizontal distance and relative height difference for fire extinguishing; Calculate the initial velocity of the water jet from the fire monitor based on the optimal fire monitor pressure and the optimal extinguishing agent type in the optimal fire extinguishing strategy. Based on the horizontal distance, relative height difference, and initial velocity of the water jet, solve for the pitch angle and horizontal rotation angle; Based on the pitch angle and horizontal rotation angle, control the spray position of the fire monitor and control the valves in the fire monitor pipeline to switch to the storage tank pipeline corresponding to the optimal extinguishing agent type; Adjust the opening of the proportional mixing valve in the fire monitor pipeline to set the mixing ratio to the optimal mixing ratio; Start the variable frequency pump of the fire monitor, gradually stabilize the output pressure to the optimal fire monitor pressure, open the outlet valve of the fire monitor, so that the extinguishing agent covers the fire source area according to the predetermined trajectory, and carry out fire extinguishing.

10. A vision-based fire extinguishing system for implementing the fire extinguishing method as described in any one of claims 1-9, characterized in that, The system includes: The fire scene analysis unit is used to collect multi-angle, all-round visual data of the fire scene when a fire occurs in a chemical enterprise using a drone swarm. The unit uses an onboard image recognition model to perform real-time analysis, generate fire scene analysis results, and upload them to a cloud server. The visual positioning unit is used to perform visual positioning of the fire scene analysis results on the cloud server and generate fire scene visual positioning results. The fire extinguishing decision unit is used to iteratively optimize the fire extinguishing decision based on the visual positioning results of the fire scene using a multi-objective enzyme-based optimization algorithm to generate the optimal fire extinguishing strategy. The fire extinguishing execution unit is used to execute fire extinguishing using fire monitors based on the visual positioning results of the fire scene and the optimal fire extinguishing strategy.

Images