Intelligent forest fire monitoring method and device, electronic equipment and medium

Abstract

The invention discloses an intelligent forest fire monitoring method and device, electronic equipment and a medium, and the monitoring method employs a multi-mode deep learning model based on a double-attention mechanism to enhance the capturing capability of early flame thermal radiation characteristics and smoke form characteristics, and greatly improves the recognition sensitivity. In combination with triple verification and a confidence coefficient decision-making mechanism, errors caused by environmental interference are effectively avoided, and the recognition reliability is comprehensively improved; meanwhile, based on intelligent gridding three-dimensional monitoring of terrain complexity and vegetation types, the blind area coverage rate is greatly reduced, normalized inspection and post-disaster quick response are performed through the unmanned aerial vehicle platform, the monitoring coverage range is enlarged, and the response efficiency is improved.

Description

Technical Field

[0001] This application belongs to the field of forest fire identification and early warning technology, specifically relating to an intelligent forest fire monitoring method, device, electronic equipment and medium. Background Technology

[0002] Against the backdrop of intensifying global climate change, forest fires have become a major threat to ecological security and human civilization. Therefore, there is an urgent need for advanced systems and technologies to monitor and identify forest fires. In recent years, with the gradual development of emerging technologies integrating machine learning and ecological behavior recognition, forest fire identification technology has also made continuous breakthroughs. Existing technologies mainly utilize multi-source monitoring methods such as satellite remote sensing, drone patrols, ground video surveillance, and thermal imaging, combined with artificial intelligence algorithms, to achieve early detection and warning of forest fire hazards.

[0003] However, existing forest fire identification technologies have the following main shortcomings in application and practical deployment:

[0004] (1) Insufficient intelligence and high false alarm rate: Traditional systems rely heavily on manual monitoring and analysis of video footage. Monitoring personnel need to process footage from hundreds of monitoring points simultaneously, which can easily lead to a high rate of missed alarms due to visual fatigue. Some devices that use pure video analysis lack intelligent filtering capabilities for interference such as clouds, fog, and reflections, resulting in a high false alarm rate.

[0005] (2) Hardware design defects and deployment limitations: Existing equipment generally suffers from problems such as bulky size and high power consumption, and the limited installation location can also lead to large monitoring blind spots.

[0006] (3) Insufficient fire simulation capability: Most systems only focus on fire identification and lack dynamic fire spread simulation that combines real-time meteorological and vegetation humidity data, making it difficult to provide accurate support for rescue route planning. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this application proposes an intelligent forest fire monitoring method, device, electronic equipment, and medium. It employs dual-optical-axis calibration and a multimodal deep learning model based on a dual-attention mechanism to enhance infrared band sensitivity and smoke morphology feature extraction capabilities, thereby improving the accuracy of early fire identification. Furthermore, it combines triple verification and confidence-weighted decision-making mechanisms to collaboratively determine fire conditions based on multi-source evidence, significantly reducing false alarms caused by environmental interference.

[0008] This application is achieved through the following technical solution:

[0009] A smart forest fire monitoring method includes:

[0010] Acquire multi-source data of the area to be monitored, including visible light images, infrared images, specific gas concentration data and their absorption spectrum data, environmental data, and meteorological data;

[0011] The monitored area is divided into several grids using an adaptive grid partitioning technique;

[0012] For each grid, the corresponding visible light image, infrared image, specific gas concentration data and its absorption spectrum data, and environmental data are input into a pre-trained multimodal deep learning model based on a dual attention mechanism to identify the fire situation and obtain the fire confidence and fire category.

[0013] For each grid, the corresponding historical environmental data, meteorological data, and fire confidence scores are input into a pre-trained LSTM model to predict future fire trends.

[0014] Fire warnings and responses are made based on fire identification results and fire trend prediction results. The fire identification results, fire trend prediction results and related responses are transmitted to the UAV platform for inspection path planning and emergency path planning.

[0015] In some implementations, the method of dividing the area to be monitored into several grids using adaptive grid partitioning technology includes:

[0016] Supported by a GIS geographic information system, the area to be monitored is divided into several basic grids of a preset size;

[0017] Taking into account factors such as terrain complexity, historical fire distribution, vegetation type, and flammability index, the basic grid is adaptively adjusted: for high-risk terrain areas with slopes greater than the threshold, the grid density is increased; in highly flammable vegetation areas with high resin content and strong flammability, the grid area is reduced; for areas with frequent historical fires, the grid density is increased; and in conjunction with real-time dynamic updates of satellite fire risk level and meteorological drought index, the grid density is increased when the temperature is higher than its threshold and the humidity is lower than its threshold.

[0018] In some implementations, the fire detection process includes:

[0019] The visible light image and the infrared image are spatially aligned using dual-optical-axis calibration technology;

[0020] The aligned visible light image and infrared image are respectively input into the CNN encoder of their corresponding branches for feature extraction to obtain visible light feature maps and infrared feature maps;

[0021] The visible light feature map is enhanced by a spatial attention branch, and the infrared feature map is enhanced by a channel attention branch.

[0022] The enhanced visible light feature map and infrared feature map are fused with the specific gas concentration data and its absorption spectrum data at the feature level and then input into the classifier for decision-making, outputting the fire confidence level and fire category.

[0023] In some implementations, the classifier integrates three verification mechanisms for decision-making: the first layer identifies abnormally high temperature areas based on the enhanced infrared feature map; the second layer verifies the visual and compositional characteristics of smoke based on the enhanced visible light feature map and gas spectral data; and the third layer analyzes the vegetation reflectance spectral characteristics through the rate of change of specific gas concentration and spectral characteristics, combined with the environmental data analysis, to comprehensively capture early combustion and hidden fire hazards.

[0024] A confidence-weighted fusion strategy is adopted. When any two or more verification results are triggered at the same time, the fire is finally confirmed and the fire confidence and fire category are output.

[0025] The fire confidence level is a value in the range of [0 to 1], representing the probability that a fire exists in the grid. The fire categories include open flame, smoldering, smoke, and no fire.

[0026] In some implementations, the fire early warning judgment and response based on fire identification results and fire trend prediction results includes:

[0027] Based on the fire identification results, it is determined whether a fire has occurred. Once a fire is determined to have occurred, a multi-level response mechanism is immediately triggered, including:

[0028] When multiple consecutive frames of image data within a single grid show suspected fire characteristics, a primary alarm is activated.

[0029] When multiple adjacent grids generate alarms simultaneously, drones are automatically dispatched to perform emergency tasks.

[0030] In some implementations, the fire identification results, fire trend prediction results, and related responses are transmitted to the drone platform for inspection path planning and emergency path planning.

[0031] In daily inspections, the drone swarm adopts a collaborative working mode of AI planning and grid-based flight. Its inspection path planning is based on an improved Spider-Bee algorithm, which uses risk-reward value as the core guiding factor. Through a risk gradient guidance strategy, the path planning is no longer limited to geographical space, but transforms into a response to dynamic risk fields. The AI ​​planning takes maximum risk coverage efficiency and minimum flight time as dual optimization objectives, integrates the fire identification results, fire trend prediction results, and real-time meteorological data to generate the optimal inspection sequence. The drones run lightweight AI algorithms through airborne edge computing devices to analyze the collected video streams in real time, realizing smoke recognition and abnormal situation detection.

[0032] Upon receiving an alarm signal, the nearest response principle is adopted to dispatch the drone closest to the suspected fire point to arrive at the target area within a preset time. After arriving at the target area, the drone automatically hovers and confirms the fire point on-site through infrared thermal imaging, and transmits the geographical coordinates of the fire point and the fire spread trend based on multi-frame image analysis to the command center.

[0033] In some implementations, the drone swarm adopts a fixed-wing and multi-rotor collaborative operation mode. The fixed-wing drones are responsible for large-area rapid scanning to obtain the overall situation of the fire, while the multi-rotor drones perform close-range observation and conduct detailed monitoring through hovering attitude.

[0034] The drones also combine GIS geographic information systems with infrared thermal imaging data, using 3D point cloud reconstruction technology to generate 3D models of the fire scene. Based on the principle of cellular automata and real-time meteorological data, they predict the fire spread path and dynamically adjust the drones' cruise paths and monitoring strategies.

[0035] Secondly, this application proposes an intelligent forest fire monitoring device, comprising:

[0036] The acquisition unit is configured to acquire multi-source data of the area to be monitored, including visible light images, infrared images, specific gas concentration data and their absorption spectrum data, environmental data and meteorological data;

[0037] The gridding processing unit is configured to divide the area to be monitored into several grids using adaptive grid partitioning technology.

[0038] The fire identification unit is configured to: for each grid, input the corresponding visible light image, infrared image, specific gas concentration data and its absorption spectrum data, and environmental data into a pre-trained multimodal deep learning model based on a dual attention mechanism to identify the fire and obtain the fire confidence and fire category.

[0039] The fire prediction unit is configured to: for each grid, input the corresponding historical environmental data, meteorological data and fire confidence into a pre-trained LSTM model to predict the future fire trend;

[0040] In addition, the emergency unit is configured to: determine and respond to fire warnings based on fire identification results and fire trend prediction results, and transmit the fire identification results, fire trend prediction results and related responses to the UAV platform for inspection path planning and emergency path planning.

[0041] Thirdly, this application proposes an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the above-described intelligent forest fire monitoring methods.

[0042] Fourthly, this application proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described embodiments of the intelligent forest fire monitoring method.

[0043] This application proposes an intelligent forest fire monitoring method. It employs a multimodal deep learning model based on a dual-attention mechanism to enhance the ability to capture early flame thermal radiation characteristics and smoke morphology characteristics, significantly improving recognition sensitivity. Combined with a triple verification and confidence decision mechanism, it effectively avoids errors caused by environmental interference, comprehensively improving recognition reliability. At the same time, based on intelligent gridded three-dimensional monitoring according to terrain complexity and vegetation type, it greatly reduces blind spot coverage. Through the use of UAV platforms for routine inspections and rapid post-disaster response, it increases the monitoring coverage and improves response efficiency.

[0044] Accordingly, the intelligent forest fire identification device, electronic device, and computer-readable storage medium proposed in this application also possess the aforementioned technical effects. Attached Figure Description

[0045] The accompanying drawings, which are included to provide a further understanding of the embodiments of this application and form part of this application, do not constitute a limitation on the embodiments of this application. In the drawings:

[0046] Figure 1 This is a schematic diagram of the monitoring method proposed in the embodiments of this application;

[0047] Figure 2 This is a schematic diagram of the fire identification process proposed in an embodiment of this application;

[0048] Figure 3 This is a schematic diagram of the drone workflow proposed in the embodiments of this application;

[0049] Figure 4This is a block diagram illustrating the principle of the monitoring device proposed in the embodiments of this application;

[0050] Figure 5 This is a schematic diagram of the electronic device proposed in the embodiments of this application;

[0051] Figure 6 This is a schematic diagram of a computer-readable storage medium proposed in an embodiment of this application.

[0052] Figure reference numerals and corresponding component names:

[0053] 400-Monitoring device, 401-Acquisition unit, 402-Grid processing unit, 403-Fire identification unit, 404-Fire prediction unit, 405-Emergency unit, 500-Electronic device, 510-Memory, 520-Processor, 511-Computer program A, 600-Computer-readable storage medium, 611-Computer program B. Detailed Implementation

[0054] In the following, the terms “comprising” or “may include” as used in the various embodiments of this application indicate the presence of a function, operation, or element of the invention and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in the various embodiments of this application, the terms “comprising,” “having,” and their cognates are intended only to indicate a specific feature, number, step, operation, element, component, or combination of the foregoing and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or adding one or more combinations of the foregoing.

[0055] In various embodiments of this application, the expression "or" or "at least one of A and / or B" includes any combination or all combinations of the words listed simultaneously. For example, the expression "A or B" or "at least one of A and / or B" may include A, may include B, or may include both A and B.

[0056] The terms used in the various embodiments of this application (such as "first," "second," etc.) may modify various constituent elements in the various embodiments, but do not limit the corresponding constituent elements. For example, the above terms do not limit the order and / or importance of the elements. The above terms are only used for the purpose of distinguishing one element from other elements. For example, a first user device and a second user device refer to different user devices, although both are user devices. For example, without departing from the scope of the various embodiments of this application, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element.

[0057] It should be noted that if a description is made of "connecting" one component to another, then the first component can be directly connected to the second component, and a third component can be "connected" between the first and second components. Conversely, when a component is "directly connected" to another component, it can be understood that there is no third component between the first and second components.

[0058] The terminology used in the various embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the various embodiments of this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. The terms (such as those defined in a generally used dictionary) are to be interpreted as having the same meaning as in the context of the relevant technical field and are not to be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the embodiments and accompanying drawings. The illustrative embodiments and descriptions of this application are only for explaining this application and are not intended to limit this application.

[0060] This application proposes an intelligent forest fire monitoring method. This method rapidly identifies forest fires and triggers multi-level early warnings and responses by analyzing multi-source data such as images, videos, thermal imaging, and environmental data in real time. Figure 1 As shown, the monitoring method includes the following steps:

[0061] Step 110: Acquire multi-source data of the area to be monitored; the multi-source data includes visible light images, infrared images, specific gas concentration data and their absorption spectrum data, environmental data, and meteorological data, etc.

[0062] Step 120: The area to be monitored is divided into several grids using adaptive grid division technology;

[0063] Step 130: For each grid, input the corresponding visible light image, infrared image, specific gas concentration data and its absorption spectrum data, and environmental data into a pre-trained multimodal deep learning model based on dual attention mechanism to identify fires and obtain fire confidence and fire category.

[0064] Step 140: For each grid, input the corresponding historical environmental data, meteorological data and fire confidence into the pre-trained LSTM model to predict the future fire trend.

[0065] Step 150: Based on the fire identification results and fire trend prediction results, make a fire early warning judgment and response, and transmit the fire identification results, fire trend prediction results and related responses to the UAV platform for inspection path planning and emergency path planning.

[0066] Furthermore, in step 110 of this application embodiment, basic data acquisition nodes are constructed by deploying visible light cameras, infrared thermal imagers, infrared spectral gas sensors, and temperature and humidity sensors in key forest areas. The visible light cameras acquire high-resolution color images and videos, and by analyzing the texture, color, motion trajectory, and diffusion pattern of the images or videos, suspected smoke areas are identified. The infrared thermal imagers acquire infrared radiation data, focusing particularly on the typical infrared characteristic band (750nm-950nm band) of initial combustion of wood materials, which can be converted into temperature distribution thermal imaging maps for the detection of abnormally high temperature areas. The infrared spectral gas sensors acquire concentration data and absorption spectrum data of specific gases, mainly monitoring the concentration and spectral characteristics of CO2, CO, and VOCs, for the spectral analysis of initial combustion gases in plants. The temperature and humidity sensors acquire environmental data such as air temperature and relative humidity values. Meteorological data can be directly obtained from the regional meteorological platform.

[0067] Furthermore, in step 120 of this embodiment, the adaptive mesh generation technique specifically includes:

[0068] Supported by a GIS geographic information system, a basic grid of 1-5 square kilometers is generated. Adaptive grid adjustments are made, taking into account factors such as terrain complexity, historical fire distribution, vegetation type, and flammability index. For high-risk terrain areas with slopes greater than 25°, where the terrain uplift effect leads to rapid fire spread and difficulty in firefighting, the grid density is increased by 30%. In highly flammable vegetation areas such as coniferous forests with high resin content and strong flammability, the grid area is reduced by 40%. Areas with a history of frequent fires are given even higher grid densities and are subject to focused monitoring. Simultaneously, dynamic updates are implemented based on satellite fire risk levels and meteorological drought indices. When the temperature exceeds 30°C and humidity falls below 30%, automatic grid refinement (i.e., increased grid density) is triggered. During drone inspections, dynamic grid subdivision can be initiated for suspected areas based on real-time data, refining the smallest analysis unit to the 10×10 pixel level.

[0069] Furthermore, in step 130 of this application embodiment, as follows: Figure 2 As shown, the fire identification process includes:

[0070] First, the visible light image and the infrared image are spatially aligned using dual optical axis calibration technology, with the error controlled within 2 pixels.

[0071] The aligned visible light image and infrared image are then input into their respective branches of the CNN encoder for feature extraction, yielding visible light feature maps and infrared feature maps, respectively. A dual attention mechanism is then used to enhance the visible light and infrared feature maps. The channel attention branch processes the infrared feature map, automatically learning and assigning higher weights to the 750–950 nm band to strengthen the model's sensitivity to early-stage flames. The spatial attention branch processes the visible light feature map, focusing on spatial regions such as the edges, textures, and irregular shapes of smoke diffusion.

[0072] The enhanced optical feature map, infrared feature map, specific gas concentration data and its absorption spectrum data, and environmental data are fused at the feature level and input into the classifier for decision-making, outputting fire confidence and fire category.

[0073] Furthermore, this classifier integrates three verification mechanisms to improve the reliability of the judgment: the first layer identifies abnormally high temperature areas based on enhanced infrared feature maps; the second layer verifies the visual and compositional characteristics of smoke based on enhanced visible light feature maps and gas spectral data; and the third layer analyzes the vegetation reflectance spectral characteristics through the rate of change of specific gas concentrations and spectral features, combined with environmental data analysis such as temperature and humidity (an environmental risk warning is triggered when the temperature is greater than 30℃ and the relative humidity is less than 30%), comprehensively capturing early combustion and hidden fire hazards. A confidence-based weighted fusion strategy is adopted, and a fire is finally confirmed when any two or more verification results are triggered simultaneously. The output includes fire confidence and fire category. The fire confidence is a value in the range [0-1], representing the probability of a fire in that grid; the higher the value, the greater the probability of a fire. The fire category is a classification label, including open flame, smoldering, smoke, and no fire.

[0074] Furthermore, in step 140 of this application embodiment, historical environmental data and meteorological data, as well as the historical fire confidence score output by a multimodal deep learning model based on a dual attention mechanism, are used as inputs to an LSTM model to predict the probability of a fire occurring in the grid within a certain period of time.

[0075] Furthermore, in step 150 of this application embodiment, intelligent drone airport equipment is deployed at high points in the forest for routine inspections and rapid post-disaster response. The airport integrates an automatic take-off and landing platform, a fast-charging battery device, and a meteorological monitoring module. The drone is equipped with a visible light / infrared / laser tri-light pod, a thermal imager, a gas sensor, and other equipment, supporting RTK centimeter-level positioning and 5G private network communication. In daily inspections, the drone adopts a collaborative working mode of AI planning + grid-based flight. Its inspection path planning is based on an improved spider-bee algorithm, which uses the risk-reward value (RVI) as the core guiding factor. Through a risk gradient guidance strategy, the path planning is no longer limited to geographical space but transforms into a response to a dynamic risk field. AI planning has the dual optimization objectives of maximizing risk coverage efficiency and minimizing flight time. It integrates the fire risk prediction value output by the LSTM model, the fire risk anomaly output by the multimodal deep learning model, and real-time meteorological data to generate the optimal inspection sequence. By adaptively densifying the grid and adjusting the flight path in high-risk areas, enhanced coverage of key areas is achieved, and high-frequency inspections are carried out no less than twice a day. The drone runs a lightweight AI algorithm through onboard edge computing equipment to analyze the collected video stream in real time, enabling smoke recognition and anomaly detection.

[0076] The system determines whether a fire has occurred based on fire confidence level and fire type. Once a fire is identified, a multi-level response mechanism is immediately triggered: when multiple consecutive frames of image data within a single grid show suspected fire characteristics, a primary alarm is activated; when multiple adjacent grids simultaneously generate linked alarms, drones are automatically dispatched to perform close-range reconnaissance missions. In emergency response mode, a dynamic task scheduling mechanism is activated. Upon receiving an alarm signal, the nearest-response principle is adopted, dispatching the drone closest to the suspected fire point (highest priority) to arrive at the target area within a preset time. Path planning employs a real-time replanning strategy, combined with a fast random tree algorithm, to achieve obstacle avoidance and shortest path generation. The drone swarm adopts a fixed-wing and multi-rotor collaborative operation mode: fixed-wing drones are responsible for large-area rapid scanning to obtain the overall fire situation; multi-rotor drones perform close-range observation, conducting detailed monitoring through hovering attitude.

[0077] Upon reaching the target area, the drone automatically hovers and conducts secondary confirmation via infrared thermal imaging. Within a preset time, it pushes the GPS coordinates of the fire point and a fire spread trend map based on multi-frame image analysis to the command center. Combining GIS geographic information system data and infrared thermal imaging data, a 3D fire model is generated using 3D point cloud reconstruction technology. Based on cellular automata principles and real-time meteorological data, the fire spread path is predicted, and the drone swarm's cruise path and monitoring strategy are dynamically adjusted, providing comprehensive data support for firefighting decisions and resource allocation. The drone workflow is as follows: Figure 3 As shown.

[0078] This application employs a multimodal deep learning model based on a dual-attention mechanism to enhance the capture of early flame thermal radiation characteristics and smoke morphology features, significantly improving recognition sensitivity. Combined with a triple verification and confidence-based decision mechanism, it effectively avoids errors caused by environmental interference, comprehensively improving recognition reliability. Simultaneously, intelligent gridded three-dimensional monitoring based on terrain complexity and vegetation type greatly reduces blind spot coverage. Routine inspections and rapid post-disaster response are conducted via UAV platforms, increasing monitoring coverage and improving response efficiency. Furthermore, intelligent path planning based on dynamic risk grids and improved optimization algorithms, combined with fixed-wing and multi-rotor collaborative reconnaissance modes, ensures rapid fire confirmation and synchronized perception of the entire situation. Through three-dimensional point cloud fire scene modeling and cellular automata spread prediction, it provides dynamic and accurate data support for command and decision-making, significantly improving the efficiency of firefighting resource allocation.

[0079] Based on the same technical concept described above, this application also proposes an intelligent forest fire monitoring device, such as... Figure 4 As shown, the monitoring device 400 includes:

[0080] The acquisition unit 401 is configured to acquire multi-source data of the area to be monitored; the multi-source data includes visible light images, infrared images, specific gas concentration data and their absorption spectrum data, environmental data, and meteorological data, etc. The specific data acquisition method is as described in step 110 above, and will not be repeated here.

[0081] The meshing processing unit 402 is configured to divide the area to be monitored into several meshes using adaptive meshing technology. The specific meshing process is as described in step 120 above, and will not be repeated here.

[0082] The fire identification unit 403 is configured to: for each grid, input the corresponding visible light image, infrared image, specific gas concentration data and its absorption spectrum data and environmental data into a pre-trained multimodal deep learning model based on a dual attention mechanism to identify the fire and obtain the fire confidence and fire category; the specific fire identification process is as described in step 130 above, and will not be repeated here.

[0083] The fire prediction unit 404 is configured to input the corresponding historical environmental data, meteorological data and fire confidence into the pre-trained LSTM model for each grid to predict the future fire trend; the specific fire trend prediction process is as described in step 140 above, and will not be repeated here.

[0084] Furthermore, emergency unit 405 is configured to: determine and respond to fire warnings based on fire identification results and fire trend prediction results, and transmit the fire identification results, fire trend prediction results, and related responses to the UAV platform for inspection path planning and emergency path planning. The specific warning determination and response, as well as the UAV inspection path planning and emergency path planning processes, are as described in step 150 above, and will not be repeated here.

[0085] Based on the same technical concept described above, this application also proposes an electronic device, such as... Figure 5 As shown, the electronic device 500 includes: a memory 510, a processor 520, and a computer program A511 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the computer program A511, it performs the following steps:

[0086] Acquire multi-source data for the area to be monitored; this multi-source data includes visible light images, infrared images, specific gas concentration data and their absorption spectrum data, environmental data, and meteorological data, etc.

[0087] The area to be monitored is divided into several grids using adaptive grid division technology;

[0088] For each grid, the corresponding visible light image, infrared image, specific gas concentration data and its absorption spectrum data, and environmental data are input into a pre-trained multimodal deep learning model based on a dual attention mechanism to identify fires and obtain fire confidence and fire category.

[0089] For each grid, the corresponding historical environmental data, meteorological data, and fire confidence scores are input into a pre-trained LSTM model to predict future fire trends.

[0090] Fire warnings and responses are made based on fire identification results and fire trend prediction results. The fire identification results, fire trend prediction results and related responses are transmitted to the UAV platform for inspection path planning and emergency path planning.

[0091] Optionally, when the processor 520 executes the computer program A511, it can implement any of the embodiments in the corresponding examples of the monitoring method described above.

[0092] It should be noted that the electronic device proposed in this application embodiment is a device used to implement the above-mentioned monitoring method. Therefore, based on the above-mentioned monitoring method proposed in this application embodiment, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this application embodiment. Therefore, the specific implementation method of the above-mentioned monitoring method will not be described in detail here. Any electronic device used by those skilled in the art to implement the above-mentioned monitoring method is within the scope of protection of this application.

[0093] Based on the same technical concept described above, embodiments of this application also propose a computer-readable storage medium, such as... Figure 6 As shown, the computer-readable storage medium 600 stores a computer program B611, which, when executed by a processor, performs the following steps:

[0094] Acquire multi-source data for the area to be monitored; this multi-source data includes visible light images, infrared images, specific gas concentration data and their absorption spectrum data, environmental data, and meteorological data, etc.

[0095] The area to be monitored is divided into several grids using adaptive grid division technology;

[0096] For each grid, the corresponding visible light image, infrared image, specific gas concentration data and its absorption spectrum data, and environmental data are input into a pre-trained multimodal deep learning model based on a dual attention mechanism to identify fires and obtain fire confidence and fire category.

[0097] For each grid, the corresponding historical environmental data, meteorological data, and fire confidence scores are input into a pre-trained LSTM model to predict future fire trends.

[0098] Fire warnings and responses are made based on fire identification results and fire trend prediction results. The fire identification results, fire trend prediction results and related responses are transmitted to the UAV platform for inspection path planning and emergency path planning.

[0099] Optionally, when the computer program B611 is executed by the processor, it can implement any of the embodiments corresponding to the above monitoring method.

[0100] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0101] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0102] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will 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 apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0103] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function 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.

[0104] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable 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.

[0105] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above description is only a specific embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent forest fire monitoring, characterized in that, include: Acquire multi-source data of the area to be monitored, including visible light images, infrared images, specific gas concentration data and their absorption spectrum data, environmental data, and meteorological data; The monitored area is divided into several grids using an adaptive grid partitioning technique; For each grid, the corresponding visible light image, infrared image, specific gas concentration data and its absorption spectrum data, and environmental data are input into a pre-trained multimodal deep learning model based on a dual attention mechanism to identify the fire situation and obtain the fire confidence and fire category. For each grid, the corresponding historical environmental data, meteorological data, and fire confidence scores are input into a pre-trained LSTM model to predict future fire trends. Fire warnings and responses are made based on fire identification results and fire trend prediction results. The fire identification results, fire trend prediction results and related responses are transmitted to the UAV platform for inspection path planning and emergency path planning.

2. The intelligent forest fire monitoring method according to claim 1, characterized in that, The method of dividing the area to be monitored into several grids using adaptive grid partitioning technology includes: Supported by a GIS geographic information system, the area to be monitored is divided into several basic grids of a preset size; Taking into account factors such as terrain complexity, historical fire distribution, vegetation type, and flammability index, the basic grid is adaptively adjusted: for high-risk terrain areas with slopes greater than the threshold, the grid density is increased; in highly flammable vegetation areas with high resin content and strong flammability, the grid area is reduced; for areas with frequent historical fires, the grid density is increased; and in conjunction with real-time dynamic updates of satellite fire risk level and meteorological drought index, the grid density is increased when the temperature is higher than its threshold and the humidity is lower than its threshold.

3. The intelligent forest fire monitoring method according to claim 1, characterized in that, The fire identification process includes: The visible light image and the infrared image are spatially aligned using dual-optical-axis calibration technology; The aligned visible light image and infrared image are respectively input into the CNN encoder of their corresponding branches for feature extraction to obtain visible light feature maps and infrared feature maps; The visible light feature map is enhanced by a spatial attention branch, and the infrared feature map is enhanced by a channel attention branch. The enhanced visible light feature map and infrared feature map are fused with the specific gas concentration data and its absorption spectrum data at the feature level and then input into the classifier for decision-making, outputting the fire confidence level and fire category.

4. The intelligent forest fire monitoring method according to claim 3, characterized in that, The classifier integrates three verification mechanisms for decision-making. The first layer identifies abnormally high temperature areas based on the enhanced infrared feature map. The second layer verifies the visual and compositional characteristics of smoke based on the enhanced visible light feature map and gas spectral data. The third layer analyzes the vegetation reflectance spectral characteristics through the rate of change of specific gas concentration and spectral characteristics, and combines the environmental data analysis to comprehensively capture early combustion and hidden fire hazards. A confidence-weighted fusion strategy is adopted. When any two or more verification results are triggered at the same time, the fire is finally confirmed and the fire confidence and fire category are output. The fire confidence level is a value in the range of [0 to 1], representing the probability that a fire exists in the grid. The fire categories include open flame, smoldering, smoke, and no fire.

5. A method for intelligent forest fire monitoring according to any one of claims 1-4, characterized in that, The aforementioned fire early warning judgment and response based on fire identification results and fire trend prediction results includes: Based on the fire identification results, it is determined whether a fire has occurred. Once a fire is determined to have occurred, a multi-level response mechanism is immediately triggered, including: When multiple consecutive frames of image data within a single grid show suspected fire characteristics, a primary alarm is activated. When multiple adjacent grids generate alarms simultaneously, drones are automatically dispatched to perform emergency tasks.

6. The intelligent forest fire monitoring method according to claim 5, characterized in that, The fire identification results, fire trend prediction results, and related responses are transmitted to the drone platform for inspection path planning and emergency path planning. In daily inspections, the drone swarm adopts a collaborative working mode of AI planning and grid-based flight. Its inspection path planning is based on an improved Spider-Bee algorithm, which uses risk-reward value as the core guiding factor. Through a risk gradient guidance strategy, the path planning is no longer limited to geographical space, but transforms into a response to dynamic risk fields. The AI ​​planning takes maximum risk coverage efficiency and minimum flight time as dual optimization objectives, integrates the fire identification results, fire trend prediction results, and real-time meteorological data to generate the optimal inspection sequence. The drones run lightweight AI algorithms through airborne edge computing devices to analyze the collected video streams in real time, realizing smoke recognition and abnormal situation detection. Upon receiving an alarm signal, the nearest response principle is adopted to dispatch the drone closest to the suspected fire point to arrive at the target area within a preset time. After arriving at the target area, the drone automatically hovers and confirms the fire point on-site through infrared thermal imaging, and transmits the geographical coordinates of the fire point and the fire spread trend based on multi-frame image analysis to the command center.

7. The intelligent forest fire monitoring method according to claim 6, characterized in that, The drone swarm adopts a coordinated operation mode of fixed-wing and multi-rotor drones. Fixed-wing drones are responsible for large-area rapid scanning to obtain the overall situation of the fire, while multi-rotor drones perform close-range observation and conduct detailed monitoring through hovering attitude. The drones also combine GIS geographic information systems with infrared thermal imaging data, using 3D point cloud reconstruction technology to generate 3D models of the fire scene. Based on the principle of cellular automata and real-time meteorological data, they predict the fire spread path and dynamically adjust the drones' cruise paths and monitoring strategies.

8. An intelligent forest fire monitoring device, characterized in that, include: The acquisition unit is configured to acquire multi-source data of the area to be monitored, including visible light images, infrared images, specific gas concentration data and their absorption spectrum data, environmental data and meteorological data; The gridding processing unit is configured to divide the area to be monitored into several grids using adaptive grid partitioning technology. The fire identification unit is configured to: for each grid, input the corresponding visible light image, infrared image, specific gas concentration data and its absorption spectrum data, and environmental data into a pre-trained multimodal deep learning model based on a dual attention mechanism to identify the fire and obtain the fire confidence and fire category. The fire prediction unit is configured to: for each grid, input the corresponding historical environmental data, meteorological data and fire confidence into a pre-trained LSTM model to predict the future fire trend; In addition, the emergency unit is configured to: determine and respond to fire warnings based on fire identification results and fire trend prediction results, and transmit the fire identification results, fire trend prediction results and related responses to the UAV platform for inspection path planning and emergency path planning.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the intelligent forest fire monitoring method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent forest fire monitoring method according to any one of claims 1-7.

Images