Mountain fire monitoring method and system based on laser radar and thermal infrared

Abstract

The invention provides a mountain fire monitoring method and system based on laser radar and thermal infrared. The method comprises the following steps: combining three-dimensional point cloud data with thermal radiation distribution data to generate a target fire diffusion path; an embedded real-time processing chip deployed in the detection unmanned aerial vehicle is utilized to perform real-time processing on the thermal radiation distribution data, a temperature abnormal area is identified, an interference area is removed, and a fire point position and a classification result are generated; and based on the target fire behavior diffusion path, the fire point position and the classification result, a dynamic scheduling instruction is sent to the fire extinguishing unmanned aerial vehicle cluster through a low-delay communication link so as to adjust the fire extinguishing agent throwing position, the fire extinguishing dosage and the coverage range of each fire extinguishing unmanned aerial vehicle in the fire extinguishing unmanned aerial vehicle cluster. According to the method, the three-dimensional point cloud and thermal radiation data of the forest fire area are collected in real time, the fire diffusion path is predicted in combination with an intelligent algorithm, and the fire extinguishing unmanned aerial vehicle cluster is dynamically scheduled, so that accurate positioning, interference filtering and optimal space distribution of fire extinguishing resources can be realized.

Description

Technical Field

[0001] This application relates to the technical field of wildfire monitoring, and particularly to a wildfire monitoring method and system based on lidar and thermal infrared. Background Art

[0002] Wildfires are one of the most serious natural disasters in forest ecosystems. Every year, a large area is damaged by wildfires globally, directly threatening biodiversity and the realization of ecological protection. Wildfire monitoring and suppression can reduce the harm caused by disasters. Therefore, how to combine modern equipment to achieve effective wildfire monitoring, timeliness of wildfire suppression, and rationality of suppression strategies is the current mainstream research direction.

[0003] Existing solutions use drones with fixed flight paths equipped with infrared sensors and visible light cameras to collect two-dimensional thermal imaging and image data of the fire scene through preset flight paths, perform rough fire point positioning in combination with an edge computing platform, and generate static fire suppression plans based on historical fire scene diffusion models. After the data is transmitted back, it relies on the central server of the ground station for fire trend prediction and task allocation.

[0004] However, due to relying on two-dimensional thermal imaging data and fixed flight path scanning, existing solutions are difficult to obtain the three-dimensional structure information of the vertical profile of the fire scene, resulting in insufficient prediction accuracy of the fire spread path; at the same time, the fusion dimension of infrared and visible light data is single, and the filtering ability for interference from smoke and high-temperature non-fire sources is limited. In addition, the centralized processing mode of the ground station introduces communication delays, making it difficult to meet the timeliness requirements of rapid response to dynamic fire scenes. Summary of the Invention

[0005] This application provides a wildfire monitoring method and system based on lidar and thermal infrared to solve the problems in the prior art, such as insufficient three-dimensional fire scene structure perception ability, single multi-source data fusion dimension, weak interference suppression ability, and response delay, which lead to poor wildfire monitoring effects and thus unsatisfactory suppression effects.

[0006] In the first aspect, this application provides a wildfire monitoring method based on lidar and thermal infrared, including:[[]]END]] By controlling the flight altitude and heading angle of the detection drone, the lidar and thermal infrared sensors deployed on the detection drone respectively collect three-dimensional point cloud data of the vertical profile of the wildfire area and thermal radiation distribution data of the wildfire area from multiple angles; Combining the three-dimensional point cloud data and the thermal radiation distribution data to generate a target fire spread path; Using the embedded real-time processing chip deployed in the detection drone to perform real-time processing on the thermal radiation distribution data, identify temperature abnormal areas and remove interference areas, and generate fire point positions and classification results; Based on the target fire spread path, the fire point location, and the classification result, send dynamic scheduling instructions to the fire extinguishing UAV cluster through a low-latency communication link to adjust the positions, fire extinguishing agent dosages, and coverage ranges of the fire extinguishing agents dropped by each fire extinguishing UAV in the fire extinguishing UAV cluster.

[0007] Optionally, the combining the three-dimensional point cloud data with the thermal radiation distribution data to generate a target fire spread path includes: Extract the elevation characteristics of the wildfire terrain from the three-dimensional point cloud data; Extract the temperature field gradient change characteristics from the thermal radiation distribution data; Based on the elevation characteristics of the wildfire terrain and the temperature field gradient change characteristics, combined with the historical meteorological data of the wildfire area, generate a target fire spread path.

[0008] Optionally, the elevation characteristics of the wildfire terrain include: elevation value sequence, slope change rate set, and slope direction angle set, the temperature field gradient change characteristics include: temperature extreme value region boundary, temperature gradient direction vector, and temperature gradient intensity distribution, and the historical meteorological data includes wind speed vector, humidity, and vegetation type; The generating a target fire spread path based on the elevation characteristics of the wildfire terrain and the temperature field gradient change characteristics, combined with the historical meteorological data of the wildfire area, includes: Perform differential calculation on the elevation value sequence to obtain the terrain elevation change rate distribution; Perform non-linear mapping on the slope change rate in the slope change rate set and the temperature gradient direction vector to generate a dynamic influence intensity parameter representing the influence of the terrain on the temperature field gradient change characteristics; Extract the topological structure characteristics from the temperature extreme value region boundary, and the topological structure characteristics include the number of connected domains and the boundary curvature radius; Perform azimuth angle partitioning on the slope direction angle set, calculate the covariance coefficient between the slope direction angle and the temperature gradient intensity distribution in each partition, and use the partition with the covariance coefficient greater than the preset coefficient threshold as the fire spread sensitive area; According to the terrain elevation change rate distribution, calculate the heat convection intensity of the fire front under different terrain conditions, calculate the spread rate of the fire front according to the vegetation type, humidity, and wind speed vector, and correct the direction of the fire front based on the dynamic influence intensity parameter to generate multiple candidate fire spread paths; According to the spatial coupling relationship between the distribution range of the fire spread sensitive area and the topological structure characteristics, perform probability simulation on each candidate fire spread path to screen out the target fire spread path that meets the preset spread conditions.

[0009] Optionally, non-linearly map the slope change rate and the temperature gradient direction vector in the set of slope change rates to generate a dynamic influence intensity parameter characterizing the gradient change characteristics of the terrain on the temperature field, including: Normalize all the slope change rates in the set of slope change rates to obtain a standardized slope change rate vector; Decompose the temperature gradient direction vector into a horizontal component and a vertical component, and calculate the first cosine similarity between the horizontal component and the standardized slope change rate vector and the second cosine similarity between the vertical component and the standardized slope change rate vector respectively; Exponentially weight the first cosine similarity and the second cosine similarity respectively to obtain a first weighted value and a second weighted value, and use the sum of the first weighted value and the second weighted value as the direction consistency parameter; Normalize the direction consistency parameter to obtain the dynamic influence intensity parameter.

[0010] Optionally, calculate the heat convection intensity of the fire front under different terrain conditions according to the terrain elevation change rate distribution, calculate the diffusion rate of the fire front according to the vegetation type, humidity and wind speed vector, and correct the direction of the fire front based on the dynamic influence intensity parameter to generate multiple candidate fire spread paths, including: Calculate the heat convection intensity correction coefficients at the first position and the second position in the volcanic area according to the formulas with different parameter values respectively; the first position is the position corresponding to the terrain elevation change rate greater than the preset change rate threshold in the terrain elevation change rate distribution, and the second position is the position corresponding to the terrain elevation change rate less than or equal to the preset change rate threshold in the terrain elevation change rate distribution; Take the product of the heat convection intensity correction coefficient, the intensity reference value and the correction coefficient corresponding to the vegetation type at each position as the heat convection intensity at each position; Calculate the diffusion rate according to the vegetation coefficient corresponding to the vegetation type, the rate reference value, humidity and wind speed vector by using a preset diffusion rate calculation formula; Calculate the extension length of the path according to the heat convection intensity and diffusion rate at each position; Determine the direction correction coefficient corresponding to the dynamic influence intensity parameter, and correct the direction of the fire front based on the direction correction coefficient to obtain the corrected direction; Initialize the path starting point as the current coordinate of the fire front, and generate multiple candidate fire spread paths based on the corrected direction and the extension length of the path, in combination with path screening conditions, where the path screening conditions include: the overlapping area between the path and the fire spread sensitive area is greater than a preset overlapping area threshold, the diffusion rate of the path is greater than a preset rate threshold, and the diffusion duration is greater than a preset duration threshold.

[0011] Optionally, the embedded real-time processing chip deployed in the detection UAV is used to perform real-time processing on the thermal radiation distribution data, identify the temperature anomaly area and remove the interference area, and generate the fire point position and classification result, including: Using the embedded real-time processing chip deployed in the detection UAV, perform median filtering on the thermal radiation distribution data to eliminate environmental noise interference and generate a denoised temperature field matrix; Process the denoised temperature field matrix to determine the target fire point area; Extract the temperature distribution feature vector from the target fire point area, and the temperature distribution feature vector includes the regional average temperature, the maximum temperature gradient, the standard deviation of the temperature distribution, and the ratio of the regional contour perimeter to the area; Use the K-means clustering algorithm to perform dynamic threshold division on the temperature distribution feature vector, and generate the fire point type label and confidence parameter according to the Euclidean distance matching between the clustering center and the historical fire point samples. Use the fire point type label as the classification result, and the fire point type label includes surface fire, smoldering fire and rekindled fire; Based on the spatial coordinates of the corresponding target fire point area in the three-dimensional point cloud data, calculate the three-dimensional geographical coordinates of the fire point center in the UAV navigation coordinate system, and use the three-dimensional geographical coordinates as the fire point position.

[0012] Optionally, the processing of the denoised temperature field matrix to determine the target fire point area includes: Perform binary segmentation on the denoised temperature field matrix based on a preset temperature threshold to extract candidate temperature anomaly areas, and use the region growing algorithm to merge adjacent candidate temperature anomaly areas to generate an initial fire point area set; Delete the fire point areas with an area smaller than the preset area threshold in the initial fire point area set to obtain an intermediate fire point area set; Screen out the fire point areas with a reflectivity lower than the preset reflectivity threshold from the intermediate fire point area set to obtain the target fire point area.

[0013] In a second aspect, the present application provides a wildfire monitoring system based on lidar and thermal infrared, including: An acquisition module, configured to control the flight altitude and heading angle of the detection UAV, so that the lidar and thermal infrared sensors deployed on the detection UAV respectively collect three-dimensional point cloud data of the vertical profile of the wildfire area and the thermal radiation distribution data of the wildfire area at multiple angles; A generation module, configured to combine the three-dimensional point cloud data and the thermal radiation distribution data to generate a target fire spread path; An identification module, configured to use an embedded real-time processing chip deployed in a detection UAV to perform real-time processing on the thermal radiation distribution data, identify temperature anomaly regions and remove interference regions, and generate fire point positions and classification results; An adjustment module, configured to send dynamic scheduling instructions to a fire extinguishing UAV cluster via a low-latency communication link based on the target fire spread path, the fire point positions, and the classification results, so as to adjust the positions, fire extinguishing agent dosages, and coverage ranges of the fire extinguishing agents dropped by the fire extinguishing UAVs in the fire extinguishing UAV cluster.

[0014] In a third aspect, the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a wildfire monitoring method based on lidar and thermal infrared as described in any one of the first aspects.

[0015] In a fourth aspect, the present application provides a computer storage medium storing a computer program, where when the computer program is executed by a computer, it implements a wildfire monitoring method based on lidar and thermal infrared as described in any one of the first aspects.

[0016] In an embodiment of the present application, a wildfire monitoring method based on lidar and thermal infrared is provided. The method includes: controlling the flight altitude and heading angle of a detection UAV to enable a lidar and a thermal infrared sensor deployed on the detection UAV to collect three-dimensional point cloud data of a vertical profile of a wildfire area and thermal radiation distribution data of the wildfire area from multiple angles respectively; combining the three-dimensional point cloud data and the thermal radiation distribution data to generate a target fire spread path; using an embedded real-time processing chip deployed in the detection UAV to perform real-time processing on the thermal radiation distribution data, identify temperature anomaly regions and remove interference regions, and generate fire point positions and classification results; and sending dynamic scheduling instructions to a fire extinguishing UAV cluster via a low-latency communication link based on the target fire spread path, the fire point positions, and the classification results, so as to adjust the positions, fire extinguishing agent dosages, and coverage ranges of the fire extinguishing agents dropped by the fire extinguishing UAVs in the fire extinguishing UAV cluster.

[0017] This application dynamically adjusts the flight parameters of the detection UAV, combines the multi-angle collaborative scanning of lidar and thermal infrared sensors, realizes the high-precision reconstruction of the three-dimensional point cloud of the vertical profile in the wildfire area and the omnidirectional coverage of the thermal radiation field, breaking through the limitation of the traditional plane data dimension; based on the fusion modeling of three-dimensional space and thermal field data, constructs the dynamic prediction ability of the wildfire spread path, enhances the credibility of wildfire deduction under complex terrain; uses airborne embedded real-time processing technology to perform online denoising and fire point classification on thermal radiation data, reduces the misjudgment rate of smoke and high-temperature non-fire sources, and improves the fire point positioning accuracy; relies on low-latency communication and dynamic scheduling algorithms to realize the real-time task replanning of the fire extinguishing UAV cluster, optimizes the spatial matching and dosage control accuracy of the fire extinguishing agent delivery, forms a closed-loop control link of "perception - decision - execution", and improves the response efficiency and resource utilization rate of wildfire fighting.

[0018] Furthermore, by extracting terrain elevation features from the three-dimensional point cloud data, analyzing the temperature field gradient features from the thermal radiation data, and combining historical meteorological data, a multi-dimensional wildfire spread prediction model is constructed. Specifically, it includes: generating dynamic influence intensity parameters based on the non-linear mapping of the terrain elevation change rate and the temperature gradient direction vector, screening the wildfire spread sensitive areas through the covariant coefficient of slope direction angle zoning, calculating the spread rate of the fire line front and the heat convection intensity by combining vegetation type, humidity and wind speed, and correcting the wildfire direction to generate candidate paths; finally, using the spatial coupling relationship between the topological structure features and the sensitive areas, screening out the target wildfire spread path through probability simulation. By finely integrating the three-dimensional structure of the terrain, the evolution of the temperature field gradient and the dynamic meteorological parameters, a wildfire spread model coupling multi-physical fields of terrain, thermodynamics and meteorology is established to improve the spatio-temporal accuracy of the fire line front prediction in complex environments; based on the non-linear mapping and covariant coefficient screening mechanism, enhances the quantitative characterization ability of the dynamic influence of the terrain on the wildfire direction, and reduces the path misjudgment caused by the misalignment of slope direction, slope and temperature gradient; through probability simulation and topological structure coupling analysis, realizes the physical constraint screening of candidate paths, avoids the overfitting risk of a single diffusion model, and finally generates a target wildfire spread path with high confidence, providing a reliable decision-making basis for the dynamic scheduling of fire extinguishing resources.

[0019] These aspects or other aspects of this application will be more clearly understood in the following description of the embodiments. Description of the Drawings

[0020] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0021] Figure 1Flowchart of a wildfire monitoring method based on lidar and thermal infrared provided by an embodiment of the present application; Figure 2 Structural schematic diagram of a wildfire monitoring system based on lidar and thermal infrared provided by an embodiment of the present application; Figure 3 Structural schematic diagram of a computing device provided by an embodiment of the present application. Detailed implementation manners

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application.

[0023] In some processes described in the specification, claims and above-mentioned drawings of the present application, a plurality of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 11, 12, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0025] To address the problems in the prior art, such as insufficient three-dimensional fire scene structure perception ability, single multi-source data fusion dimension, weak interference suppression ability, and response delay, which lead to poor wildfire monitoring effects and thus unsatisfactory fire extinguishing effects, the embodiments of this application provide a wildfire monitoring method based on lidar and thermal infrared. This method adopts the following concept: with multi-modal sensor collaborative acquisition and real-time decision-making as the core, design a dynamic flight control strategy for the detection UAV. By adaptively adjusting the flight altitude and heading angle, achieve multi-angle vertical scanning of the lidar and omnidirectional radiation coverage of the thermal infrared sensor, and construct a spatial coupling model of three-dimensional point cloud and thermal field data; develop an online cleaning algorithm for thermal radiation data based on an airborne embedded chip, combine temperature gradient threshold segmentation and morphological filtering to achieve interference suppression, synchronously fuse terrain elevation features and meteorological parameters to construct a fire spread dynamics equation; finally, relying on edge computing and low-latency communication technologies, establish a dynamic task allocation mechanism for the fire extinguishing UAV cluster, form a full-link closed-loop system from high-precision perception, multi-physical field modeling, real-time decision-making to precise execution, break through the data dimension limitations and response bottlenecks of traditional solutions, and thereby improve wildfire monitoring and fire extinguishing effects.

[0026] Figure 1 The flowchart of a wildfire monitoring method based on lidar and thermal infrared provided by the embodiments of this application is shown in Figure 1 As shown, this method includes: S11. By controlling the flight altitude and heading angle of the detection UAV, enable the lidar and thermal infrared sensor deployed on the detection UAV to collect three-dimensional point cloud data of the vertical profile of the wildfire area and the thermal radiation distribution data of the wildfire area from multiple angles respectively.

[0027] Among them, the thermal infrared sensor can refer to a device that detects the temperature distribution of the target area based on the principle of infrared thermal radiation, generates a temperature field image by receiving the radiation energy in a specific band, and is used to reflect the thermal distribution and gradient change characteristics of the wildfire area. The specific band can be mid-wave infrared, long-wave infrared, etc. The lidar can refer to a remote sensing device that obtains the target distance and shape by emitting laser pulses and measuring the time difference of the reflected signals. In the embodiments of this application, it can generate three-dimensional point cloud data based on the time-of-flight principle and is used to reconstruct the vertical profile structure of the terrain. The three-dimensional point cloud data refers to a spatial coordinate data set generated by lidar scanning, including elevation, slope, and aspect information of the target area, and is used for terrain modeling and fire spread analysis. The thermal radiation distribution data of the wildfire area refers to the two-dimensional matrix data of the temperature field collected by the thermal infrared sensor, which reflects the thermal radiation intensity and gradient direction of the fire scene and is used to identify temperature anomaly areas.

[0028] In the embodiment of the present application, first, the flight altitude and heading angle of the detection UAV are adjusted through a dynamic flight control algorithm to make it fly in a multi-angle hovering manner; second, the lidar is controlled to perform a vertical profile scan on the wildfire area, emit laser pulses and receive reflected signals to generate three-dimensional point cloud data including elevation, slope and aspect; at the same time, the thermal infrared sensor synchronously collects the thermal radiation energy of the wildfire area to generate temperature field distribution data; finally, the three-dimensional point cloud data and the thermal radiation distribution data are matched according to coordinates and timestamps through a spatio-temporal alignment algorithm to form a fusion data set.

[0029] S12. Combine the three-dimensional point cloud data and the thermal radiation distribution data to generate the target wildfire spread path.

[0030] Among them, the target wildfire spread path refers to the predicted path generated by fusing terrain, temperature field and meteorological data, and the high-confidence wildfire spread direction is selected through probability simulation.

[0031] In the embodiment of the present application, first, the terrain elevation features are extracted from the three-dimensional point cloud data, including the elevation value sequence, slope change rate and aspect angle set; second, gradient analysis is performed on the thermal radiation distribution data to extract the boundary of the temperature extreme region, gradient direction vector and intensity distribution; then, the terrain features and temperature field gradient are input into a non-linear mapping model to calculate the intensity parameter of the dynamic influence of the terrain on the temperature field; next, the wildfire spread sensitive areas are screened based on the covariance coefficient of aspect angle partitioning and temperature gradient intensity; finally, combined with the wind speed, humidity and vegetation type in the historical meteorological data, a wildfire spread dynamics model is constructed to generate multiple candidate paths and determine the target wildfire spread path through probability simulation.

[0032] S13. Use the embedded real-time processing chip deployed in the detection UAV to perform real-time processing on the thermal radiation distribution data, identify the temperature anomaly area and remove the interference area to generate the fire point location and classification result.

[0033] Among them, the embedded real-time processing chip refers to a high-performance computing module integrated in the detection UAV, which supports parallel processing of thermal radiation data for online denoising, fire point identification and classification. Removing the interference area means removing non-fire source high-temperature areas through morphological filtering and connected component analysis to improve the fire point detection accuracy. The fire point location and classification result refer to the geographical coordinates of the fire point and its type label obtained through temperature threshold segmentation and gradient analysis, including categories such as active fire points and ember areas.

[0034] In the embodiment of the present application, first, an embedded real-time processing chip is used to load a temperature gradient threshold segmentation algorithm to process the thermal radiation distribution data frame by frame, and identify abnormal areas where the temperature exceeds a preset threshold; secondly, a morphological filtering algorithm is used to perform dilation and erosion operations on the abnormal areas to eliminate isolated small-area interference areas; then, the fire point boundary is extracted based on connected component analysis, and the fire points are classified in combination with the temperature gradient direction; finally, a classification result including the geographical coordinates and type labels of the fire points is generated and spatially bound to the three-dimensional point cloud data.

[0035] S14. Based on the target fire spread path, the fire point location, and the classification result, send dynamic scheduling instructions to the fire extinguishing UAV cluster through a low-latency communication link to adjust the positions, fire extinguishing agent dosages, and coverage ranges of the fire extinguishing agents released by each fire extinguishing UAV in the fire extinguishing UAV cluster.

[0036] Among them, the low-latency communication link refers to a communication channel using high-speed wireless transmission technology, which is used to transmit fire situation data to the fire extinguishing UAV cluster in real time to ensure the timeliness of instruction issuance. The fire extinguishing UAV cluster refers to a collaborative operation system composed of multiple UAVs, equipped with a fire extinguishing agent bin and a dynamic scheduling module, and supports adjusting the delivery strategy according to instructions. The dynamic scheduling instruction refers to an optimized instruction generated based on the fire spread path and fire point classification, including the UAV delivery position, fire extinguishing agent dosage allocation, and coverage range parameters. The fire extinguishing agent dosage refers to the capacity of the fire extinguishing agent carried by a single UAV, and the priority is dynamically allocated according to the fire point type. The coverage range refers to the effective action area of the fire extinguishing agent spraying, and the coverage radius is controlled by adjusting the UAV flight height and nozzle angle.

[0037] In the embodiment of the present application, first, the target fire spread path, the fire point location, and the classification result are transmitted to the fire extinguishing UAV cluster through a low-latency communication link; secondly, based on the dynamic scheduling algorithm, calculate the spatial distance, fire extinguishing agent inventory, and coverage ability of each UAV to optimize the delivery priority; then, allocate the fire extinguishing agent dosage according to the fire point classification result and adjust the single UAV coverage radius; finally, control the fire extinguishing UAV cluster to release the fire extinguishing agent according to the instructions to form a spatially matched coverage area.

[0038] In the embodiment of the present application, first, the detection UAV adjusts the flight height and heading angle according to the fire field environment, the lidar generates three-dimensional point cloud data of the wildfire area, and the thermal infrared sensor synchronously collects thermal radiation distribution data. Secondly, fuse the terrain elevation characteristics and the temperature gradient direction, screen the fire spread sensitive areas, and generate the target fire spread path in combination with the wind speed and vegetation type. Then, the embedded chip processes the thermal radiation data in real time, identifies active fire points and removes interference areas, and generates a fire point classification result. Finally, issue instructions to the fire extinguishing UAV cluster through low-latency communication, dynamically allocate the fire extinguishing agent dosage, and adjust the coverage range to achieve precise fire extinguishing agent delivery.

[0039] By executing S11 - S14, in the embodiments of the present application, through the fusion of multi - angle three - dimensional point cloud and thermal radiation data, the vertical structure of the fire scene is accurately reconstructed and the dynamic changes of the thermal field are captured, improving the prediction accuracy of the fire spread path; the embedded real - time processing technology realizes rapid fire point identification and interference filtering, reducing the misjudgment rate; the dynamic scheduling algorithm optimizes the fire extinguishing agent delivery strategy, combined with low - latency communication, to form a closed - loop control from perception, decision - making to execution, greatly improving the efficiency of wildfire fighting and resource utilization.

[0040] In a possible embodiment, S12, combining the three - dimensional point cloud data with the thermal radiation distribution data to generate the target fire spread path, includes: Step 121, extracting the wildfire terrain elevation features from the three - dimensional point cloud data.

[0041] Among them, the wildfire terrain elevation features refer to the set of terrain parameters extracted from the three - dimensional point cloud data, including the elevation value sequence, the slope change rate set, and the slope direction angle set, which are used to quantify the dynamic impact of the terrain on the fire spread.

[0042] In the embodiments of the present application, first, the three - dimensional point cloud data is pre - processed by the point cloud density optimization algorithm to remove noise points and fill in the missing areas; secondly, the elevation interpolation technology is used to extract the elevation value sequence of the wildfire area to generate a continuous terrain elevation surface; then, based on the curvature calculation model, the slope change rate set is extracted from the elevation surface, and the terrain normal vector is projected onto the horizontal plane through the slope direction angle calculation formula to obtain the slope direction angle set; finally, the elevation value sequence, the slope change rate set, and the slope direction angle set are integrated to form a wildfire terrain elevation feature data set, improving the terrain modeling accuracy.

[0043] Step 122, extracting the temperature field gradient change features from the thermal radiation distribution data.

[0044] Among them, the temperature field gradient change features refer to the dynamic parameters of the thermal field analyzed based on the thermal radiation distribution data, including the temperature gradient direction vector, the gradient intensity distribution, and the boundary of the temperature extreme region, which are used to identify the fire spread trend.

[0045] In the embodiments of the present application, first, the thermal radiation distribution data is processed by Gaussian filtering to eliminate sensor noise; secondly, the gradient operator is used to calculate the gradient change of the temperature field in the horizontal direction to extract the temperature gradient direction vector and intensity distribution; then, the region growing algorithm is used to segment the temperature extreme region from the temperature field, and the edge detection technology is used to extract its boundary contour; finally, the time - series analysis is combined to quantify the temperature gradient change rate to generate a temperature field gradient change feature matrix, improving the sensitivity of temperature anomaly detection.

[0046] Step 123: Generate the target fire spread path based on the elevation characteristics of the wildfire terrain and the temperature field gradient change characteristics, in combination with the historical meteorological data of the wildfire area.

[0047] Among them, the historical meteorological data of the wildfire area refers to the set of historical environmental parameters of the fire area, including the wind speed vector, such as those affecting the fire spread direction and speed, humidity, and vegetation type, which are used to correct the dynamic boundary conditions of the fire spread model.

[0048] In the embodiment of the present application, first, divide the azimuth angle partition of the slope angle set in the elevation characteristics of the wildfire terrain, and calculate the covariance coefficient between the average slope angle and the temperature gradient intensity in each partition. The covariance coefficient is used as the fire spread sensitive area. Secondly, calculate the heat convection intensity at the fire front based on the distribution of the terrain elevation change rate, and combine the wind speed vector in the historical meteorological data multiplied by the combustibility coefficient of the vegetation type to correct the spread rate. Subsequently, generate multiple candidate fire spread paths through Monte Carlo simulation, and the probability weight of each path is determined by the spatial coupling degree of the sensitive area distribution and the topological structure characteristics. Finally, select the path with a probability exceeding 80% as the target fire spread path, reducing the prediction error.

[0049] In the embodiment of the present application, first, the three-dimensional point cloud data collected by the detection drone is interpolated to generate an elevation value sequence, and the slope change rate and slope angle are extracted to improve the terrain modeling accuracy. Secondly, the temperature gradient direction vector is extracted from the thermal infrared data through gradient analysis, and three temperature extreme value areas are segmented by combining the region growing algorithm. Subsequently, mark the two partitions with a covariance coefficient between the slope angle partition and the temperature gradient intensity greater than 0.7 as sensitive areas, and calculate the correction value of the fire line spread rate in combination with the historical wind speed of 8 m / s, humidity of 20%, and coniferous forest vegetation type. Finally, generate 5 candidate paths through Monte Carlo simulation, and screen the target path with a probability exceeding 80% to guide the precise delivery of fire extinguishing agents by the unmanned aerial vehicle cluster for fire fighting, shortening the rescue response time.

[0050] By executing Step 121 to Step 123, the embodiment of the present application constructs a multi-dimensional fire spread prediction model by integrating terrain elevation characteristics, temperature field gradient changes, and historical meteorological data, and conducts coupling analysis of the terrain and the thermal field, improving the path prediction accuracy; screening sensitive areas based on the covariance coefficient and Monte Carlo probability simulation, enhancing the robustness of the fire direction prediction in complex environments, and reducing the misjudgment rate; dynamically correcting the spread rate and calculating the heat convection intensity, improving the spatial matching of fire fighting resource scheduling.

[0051] In a possible embodiment, the elevation characteristics of the wildfire terrain include: an elevation value sequence, a set of slope change rates, and a set of aspect angles; the temperature field gradient change characteristics include: the boundary of the temperature extreme value region, the temperature gradient direction vector, and the temperature gradient intensity distribution; and the historical meteorological data includes the wind speed vector, humidity, and vegetation type. Step 123: Generate a target wildfire spread path based on the elevation characteristics of the wildfire terrain and the temperature field gradient change characteristics, in combination with the historical meteorological data of the wildfire area, including: Step a1: Perform a difference calculation on the elevation value sequence to obtain the terrain elevation change rate distribution.

[0052] Among them, the difference calculation operation refers to the process of subtracting adjacent data points in the terrain elevation sequence, which is used to extract the local change amount of the elevation value. The terrain elevation change rate distribution refers to a two-dimensional spatial matrix obtained through difference calculation, which reflects the increasing or decreasing trend of the terrain height per unit distance.

[0053] In the embodiment of the present application, first, a difference calculation is performed on the continuous terrain elevation value sequence, that is, the elevation change amount is obtained through the difference operation of adjacent elevation values to form the terrain elevation change rate distribution. This process can be calculated point by point using the first-order difference algorithm, and the elevation sequence is converted into a change rate matrix, which is used to characterize the local steepness and spatial continuity characteristics of the terrain undulation.

[0054] Step a2: Perform a non-linear mapping on the slope change rate in the set of slope change rates and the temperature gradient direction vector to generate a dynamic influence intensity parameter characterizing the dynamic influence of the terrain on the temperature field gradient change characteristics.

[0055] Among them, the slope change rate is a parameter obtained by calculating the directional derivative from the elevation change rate, which characterizes the severity of the slope change with the spatial position. The temperature gradient direction vector is a spatial vector field that describes the direction of the maximum temperature change in the temperature field, including magnitude and azimuth information. The non-linear mapping operation refers to the process of mapping two sets of heterogeneous data to a common feature space using a radial basis function or a neural network. The dynamic influence intensity parameter is a scalar field used to quantify the degree of the dynamic influence of the terrain on the temperature gradient, which is used to correct the wildfire spread direction.

[0056] In the embodiment of the present application, first, the set of slope change rates and the temperature gradient direction vector are input into the non-linear mapping model, and the two are correlated in the high-dimensional space using the kernel function, and the mapping parameters are optimized by the gradient descent method to generate the dynamic influence intensity parameter. After being normalized by the activation function, this parameter quantifies the non-linear enhancement or weakening effect of the terrain slope on the temperature field gradient change.

[0057] Step a3: Extract the topological structure features from the boundary of the temperature extreme value region, and the topological structure features include the number of connected domains and the boundary curvature radius.

[0058] Among them, the topological structure features include the set of the number of connected regions and the boundary curvature radius, reflecting the morphological complexity of the temperature extreme value regions.

[0059] In the embodiment of the present application, first, morphological closing operation is performed on the boundary of the temperature extreme value region to eliminate noise. Subsequently, the region growing algorithm is used to extract the connected regions, the number of connected regions is counted, and the curvature radius of each boundary point is calculated. The boundary curve is fitted by the least square method, and finally, a description matrix containing the topological distribution of the connected regions and the curvature change features is output.

[0060] Step a4: Perform azimuth angle partitioning on the set of slope aspect angles, calculate the covariance coefficient of the slope aspect angle and the temperature gradient intensity distribution within each partition, and use the partition with the covariance coefficient greater than the preset coefficient threshold as the fire spread sensitive region.

[0061] Among them, the azimuth angle partitioning is a spatial partitioning method that equally divides the 360-degree azimuth angle into several sectors, used to count the distribution law of the slope aspect angles. The covariance coefficient is an index characterizing the linear correlation degree of two sets of spatial data, calculated by the ratio of the covariance to the standard deviation. The covariance coefficient The calculation formula is: , where, is the set of slope aspect angles within the th azimuth angle partition, is the corresponding temperature gradient intensity distribution within the th azimuth angle partition, is the covariance of the slope aspect angle and the temperature gradient intensity , is the standard deviation of the slope aspect angles within the th partition, is the standard deviation of the temperature gradient intensity within the th partition.

[0062] In the embodiment of the present application, first, the set of slope aspect angles is divided into eight fan-shaped partitions according to the azimuth angle, the covariance of the slope aspect angle distribution and the temperature gradient intensity distribution within each partition is calculated, and then divided by the standard deviations of both to obtain the covariance coefficient. The partitions with strong positive correlation are selected through the preset coefficient threshold (exemplarily, it can be 0.8, 0.85, etc.), and after superimposing the geographical grid data, they are marked as the fire spread sensitive regions.

[0063] Step a5: According to the distribution of the terrain elevation change rate, calculate the heat convection intensity of the fire front under different terrain conditions, calculate the diffusion rate of the fire front according to the vegetation type, humidity, and wind speed vector, and correct the direction of the fire front based on the dynamic influence intensity parameter to generate multiple candidate fire spread paths.

[0064] Among them, the heat convection intensity is a heat transfer rate parameter calculated based on the air flow acceleration effect caused by terrain undulations.

[0065] In the embodiment of the present application, first, a heat convection intensity model is established based on the distribution of terrain elevation change rates, and the heat flow acceleration effect under different terrain conditions is calculated through a system of fluid mechanics equations. Secondly, the spatial interpolation results of the vegetation type combustion index, humidity attenuation factor, and wind speed vector are integrated to construct a diffusion rate field. Finally, the dynamic influence intensity parameter is used as the direction correction weight to iteratively generate multiple candidate fire spread paths.

[0066] Step a6: According to the spatial coupling relationship between the distribution range and topological structure characteristics of the fire spread sensitive area, perform probability simulation on each candidate fire spread path to screen out the target fire spread path that meets the preset spread conditions.

[0067] In the embodiment of the present application, first, a coupling degree evaluation function for the fire spread sensitive area and topological structure characteristics is constructed, and probability sampling is performed on the candidate paths through Monte Carlo simulation. Subsequently, the spatio-temporal overlap degree of each path with the high-curvature boundary and multi-connected domain area is calculated, and a Bayesian network is used to screen out the target fire spread path that meets the preset spread conditions.

[0068] In the embodiment of the present application, first, differential calculation is performed on the elevation data of a certain mountain area to generate a terrain elevation change rate distribution map. Subsequently, the slope change rate and the temperature gradient direction vector provided by the meteorological station are input into a pre-trained non-linear mapping model to output a dynamic influence intensity parameter field. Then, the boundary of the temperature extreme value area identified by the forest fire monitoring system is extracted, and its topological features are analyzed by calculating the number of connected domains and the radius of curvature. After azimuth angle partitioning, the covariance coefficient of the slope aspect angle and the temperature gradient in each sector area is calculated to delimit three fire spread sensitive areas. Based on the elevation change rate and meteorological data, the heat convection intensity and diffusion rate field are calculated, and five candidate paths are generated by combining dynamic parameter correction. Finally, the coupling probability of the candidate paths with the sensitive area and topological features is evaluated through Monte Carlo simulation, and two target paths that meet the spread conditions are selected for emergency evacuation planning.

[0069] By executing steps a1 to a6, the embodiment of the present application realizes high-precision prediction of the fire spread path through multi-source data fusion and multi-stage modeling. Combining the dynamic influence analysis of the thermodynamic field by terrain elevation changes, it effectively captures the evolution law of the fire line in a complex environment; through the coupling evaluation of topological features and sensitive areas, it improves the physical interpretability of path screening; finally, the constructed prediction model has both spatial resolution and computational efficiency, providing reliable decision-making support for disaster prevention and control.

[0070] In a possible embodiment, step a2 involves performing a non-linear mapping on the slope change rate in the slope change rate set and the temperature gradient direction vector to generate a dynamic influence intensity parameter characterizing the gradient change characteristics of the terrain on the temperature field, including: Step a21: Normalize all the slope change rates in the slope change rate set to obtain a normalized slope change rate vector.

[0071] The normalized slope change rate vector refers to the set of slope change rate data after normalization processing, which reflects the relative change degree of the terrain slope and is used to eliminate the dimension difference and enhance the data comparability.

[0072] In the embodiment of the present application, first, perform min-max normalization processing on all the values in the slope change rate set, subtract the minimum value from each slope change rate and then divide by the difference between the maximum value and the minimum value to obtain the normalized slope change rate vector; second, constrain the value range to between 0 and 1 through linear scaling to eliminate the dimension difference; finally, generate the normalized vector for subsequent direction consistency analysis to enhance the data comparability.

[0073] Step a22: Decompose the temperature gradient direction vector into a horizontal direction component and a vertical direction component, and calculate the first cosine similarity between the horizontal direction component and the normalized slope change rate vector and the second cosine similarity between the vertical direction component and the normalized slope change rate vector respectively.

[0074] The horizontal direction component refers to the projection component of the temperature gradient direction vector on the terrain plane and is used to analyze the correlation between heat transfer in the horizontal direction and the terrain slope. The vertical direction component refers to the component of the temperature gradient direction vector in the direction perpendicular to the terrain plane and reflects the heat vertical convection effect. The first cosine similarity refers to the cosine angle similarity measure between the normalized slope change rate vector and the horizontal direction component and is used to quantify the direction consistency between the two. The second cosine similarity refers to the cosine angle similarity measure between the normalized slope change rate vector and the vertical direction component and is used to quantify the correlation intensity between the vertical heat convection and the terrain.

[0075] In the embodiment of the present application, first, decompose the temperature gradient direction vector into a horizontal direction component and a vertical direction component. The horizontal component is calculated by projecting onto the terrain plane, and the vertical component is obtained by subtracting the magnitude of the horizontal component from the magnitude of the vector; second, calculate the first cosine similarity between the normalized slope change rate vector and the horizontal direction component, that is, the dot product of the two vectors divided by the product of their respective magnitudes; then, calculate the second cosine similarity between the normalized slope change rate vector and the vertical direction component; finally, quantify the spatial correlation between the terrain and the temperature gradient through the two-dimensional similarity.

[0076] Step a23: Exponentially weight the first cosine similarity and the second cosine similarity respectively to obtain a first weighted value and a second weighted value, and use the sum of the first weighted value and the second weighted value as the direction consistency parameter.

[0077] Among them, exponential weighting applies an exponential function weight to the cosine similarity, and amplifies the contribution of key components by adjusting the coefficient. The first weighted value refers to the result of exponentially weighting the first cosine similarity, reflecting the intensity of the synergistic effect between the horizontal terrain and the temperature gradient. The second weighted value refers to the result of exponentially weighting the second cosine similarity, reflecting the intensity of the coupling effect between the vertical terrain and the heat convection. The direction consistency parameter refers to the comprehensive parameter of the horizontal and vertical weighted values, characterizing the spatial synergy degree between the terrain slope and the temperature gradient direction.

[0078] In the embodiment of the present application, first, an exponential weight is applied to the first cosine similarity, and the weight coefficient is calculated based on the contribution ratio of the horizontal component to obtain the first weighted value; secondly, an independent exponential weight is applied to the second cosine similarity, and the weight coefficient is related to the heat convection effect of the vertical component to obtain the second weighted value; then, the first weighted value and the second weighted value are added together to generate the direction consistency parameter; finally, this parameter comprehensively reflects the spatial synergy degree between the terrain slope and the temperature gradient direction.

[0079] Step a24: Normalize the direction consistency parameter to obtain the dynamic influence intensity parameter.

[0080] Among them, in the embodiment of the present application, first, the direction consistency parameter is input into the S function for non-linear normalization processing to map the parameter range to between 0 and 1; secondly, the interference of extreme values is eliminated through dynamic threshold adjustment; finally, the dynamic influence intensity parameter is output, which is used to quantify the influence weight of the terrain on the fire spread direction.

[0081] In the embodiment of the present application, first, the slope change rate data collected by the drone is normalized to generate a standardized vector, and the temperature gradient direction is decomposed into horizontal and vertical components; secondly, the cosine similarity between the standardized vector and the horizontal component is calculated, and the first weighted value is generated by combining exponential weighting. At the same time, the similarity with the vertical component is calculated and weighted to obtain the second weighted value; then, the two are added together to generate the direction consistency parameter, which is normalized by the S function to the dynamic influence intensity parameter; finally, this parameter is input into the fire spread model to correct the prediction of the fire line direction and guide the optimization of the coverage range of the fire-fighting drone cluster.

[0082] By executing steps a21 to a24, the embodiment of the present application enhances the quantitative characterization capability of the coupling effect of terrain and thermal field through two-dimensional similarity analysis of the standardized slope change rate and the temperature gradient component; integrates the horizontal and vertical component contributions based on the exponentially weighted directional consistency parameter to improve the physical consistency of the fire spread direction prediction model; and reduces parameter sensitivity through dynamic normalization processing to ensure the robustness of the model in different fire environments.

[0083] In a possible embodiment, step a5, according to the distribution of terrain elevation change rate, calculate the thermal convection intensity of the fire front under different terrain conditions, calculate the diffusion rate of the fire front according to the vegetation type, humidity and wind speed vector, and correct the direction of the fire front based on the dynamic impact intensity parameter to generate multiple candidate fire diffusion paths, including: Step a51, respectively calculating the thermal convection intensity correction coefficient of the first position and the second position in the volcanic area according to the formula with different parameter values. The first position is the position corresponding to the terrain elevation change rate greater than the preset change rate threshold in the terrain elevation change rate distribution, and the second position is the position corresponding to the terrain elevation change rate less than or equal to the preset change rate threshold in the terrain elevation change rate distribution.

[0084] Among them, the thermal convection intensity correction coefficient refers to the adjustment parameter calculated in sections according to the difference in terrain elevation change rate, which is used to quantify the local enhancement or weakening effect of terrain undulation on thermal convection intensity.

[0085] In an embodiment of the present application, firstly, a first position and a second position are selected according to the distribution of terrain elevation change rate, wherein the first position corresponds to an area where the terrain elevation change rate is greater than a preset change rate threshold, and the second position corresponds to an area where the terrain elevation change rate is less than or equal to the threshold; secondly, piecewise functions are used to calculate the thermal convection intensity correction coefficients of the two types of areas, respectively, the correction coefficient formula for the first position is the terrain elevation change rate multiplied by the temperature gradient intensity, and the correction coefficient formula for the second position is the terrain elevation change rate divided by the temperature gradient intensity; finally, a set of correction coefficients for the two types of areas is generated to quantify the differentiated effects of terrain on thermal convection.

[0086] Step a52: The multiplication result of the thermal convection intensity correction coefficient of each position, the intensity reference value and the correction coefficient corresponding to the vegetation type is taken as the thermal convection intensity of each position.

[0087] The intensity reference value refers to a predefined thermal convection intensity reference, which is used to generate an actual intensity value in combination with a correction factor.

[0088] In the embodiments of the present application, first, obtain the heat convection intensity correction coefficients at each location, the predefined intensity reference value, and the correction coefficients corresponding to the vegetation types; secondly, multiply the three point by point according to the location to obtain the heat convection intensity at each location; finally, generate a continuous heat convection intensity distribution map through spatial interpolation to reflect the local enhancement or weakening effect of heat transfer at the fire front.

[0089] Step a53: Calculate the diffusion rate according to the vegetation coefficient corresponding to the vegetation type, the rate reference value, the humidity, and the wind speed vector by using a preset diffusion rate calculation formula.

[0090] Among them, the vegetation coefficient refers to the flammability quantification parameter corresponding to different vegetation types, which is obtained from experimental data or historical fire statistics. The rate reference value refers to the reference benchmark for the fire spread rate under the condition of no external interference, and is used to calculate the actual rate in combination with environmental parameters. The preset diffusion rate The calculation formula is: , where, is the rate reference value, is the vegetation coefficient, is the humidity, is the humidity attenuation coefficient, is the wind speed vector, is the wind speed enhancement coefficient, is the angle between the wind direction and the obstacle direction.

[0091] In the embodiments of the present application, first, find the corresponding vegetation coefficient according to the vegetation type, combine the rate reference value, the reciprocal of the humidity, and the magnitude of the wind speed vector, and input the preset diffusion rate calculation formula; secondly, the diffusion rate calculation formula is defined as the product of the vegetation coefficient, the rate reference value, and the product of the reciprocal of the humidity and the wind speed vector; finally, generate a spatial distribution of the diffusion rate to quantify the spread speed of the fire under different vegetation and meteorological conditions.

[0092] Step a54: Calculate the extension length of the path according to the heat convection intensity and the diffusion rate at each location.

[0093] Among them, the extension length of the path refers to the spatial distance that the fire spreads along a specific direction per unit time, which is jointly calculated by the heat convection intensity and the diffusion rate.

[0094] In the embodiments of the present application, first, superimpose the heat convection intensity and the diffusion rate at each location according to the location to calculate the path extension length; secondly, the extension length formula is the heat convection intensity multiplied by the diffusion rate and then divided by the reference time constant; finally, generate a path extension length matrix for evaluating the spatial expansion potential of the fire per unit time.

[0095] Step a55: Determine the direction correction coefficient corresponding to the dynamic influence intensity parameter, and correct the direction of the fire front based on the direction correction coefficient to obtain the corrected direction.

[0096] Among them, the corrected direction refers to the spread direction vector of the fire front after being adjusted by the dynamic influence intensity parameter, reflecting the coupling effect of the terrain and the thermal field.

[0097] In the embodiment of the present application, first, according to the mapping table of the dynamic influence intensity parameter and the preset direction correction coefficient, determine the direction correction coefficient of the current fire front; secondly, multiply the original direction vector of the fire front by the correction coefficient to obtain the corrected direction vector; finally, ensure the physical rationality of the direction correction through vector normalization processing.

[0098] Step a56: Initialize the path starting point as the current coordinate of the fire front, and generate multiple candidate fire spread paths based on the corrected direction and the extension length of the path, and the path screening conditions include: the overlapping area between the path and the fire spread sensitive area is greater than the preset overlapping area threshold, the spread rate of the path is greater than the preset rate threshold, and the spread duration is greater than the preset duration threshold.

[0099] Among them, initializing the path starting point means using the current coordinate of the fire front as the initial spatial position for generating candidate paths. The path screening conditions refer to the set of constraints that the candidate paths need to meet, including the threshold limits of the overlapping area with the sensitive area, the spread rate, and the spread duration. The overlapping area of the fire spread sensitive area refers to the spatial intersection area between the candidate path and the fire spread sensitive area, which is used to evaluate the rationality of the path. The preset rate threshold refers to the lowest allowable value of the fire spread rate, and the paths below this value will be excluded. The preset duration threshold refers to the lower limit of the fire spread duration, which is used to filter out short-term invalid paths.

[0100] In the embodiment of the present application, first, use the current coordinate of the fire front as the path starting point, and generate an initial path segment based on the corrected direction vector and the path extension length; secondly, traverse all paths that meet the screening conditions, and the screening conditions include that the overlapping area between the path and the fire spread sensitive area exceeds the preset threshold, the spread rate is greater than the preset threshold, and the spread duration is greater than the preset threshold; finally, output a set of multiple candidate fire spread paths for further screening by probability simulation.

[0101] In the embodiment of the present application, first, the first position and the second position are divided according to the terrain elevation change rate threshold, and the heat convection intensity correction coefficients are calculated respectively; secondly, the correction coefficient, the intensity reference value and the coniferous forest vegetation coefficient are multiplied to generate the heat convection intensity distribution; then, the diffusion rate is calculated based on the vegetation coefficient, the rate reference value, the humidity and the wind speed, and the heat convection intensity is superimposed to obtain the path extension length; then, the fire line direction is corrected according to the dynamic influence intensity parameter, and an initial path is generated with the fire line front coordinate as the starting point; finally, the paths with the overlapping area exceeding the threshold, the diffusion rate and the time length meeting the standards are screened, and multiple candidate fire spread paths are output to guide the precise interception of the UAV cluster.

[0102] By executing steps a51 to a56, the embodiment of the present application accurately quantifies the differential influence of terrain undulation on fire heat convection by calculating the heat convection intensity correction coefficient in segments; combines the diffusion rate calculation model of vegetation coefficient and meteorological parameters to enhance the environmental adaptability of fire spread speed prediction; based on the dynamic direction correction and multi-condition path screening mechanism, improves the physical rationality and spatial matching of candidate path generation, and provides a reliable basis for the dynamic scheduling of fire extinguishing resources.

[0103] In a possible embodiment, S13. Using the embedded real-time processing chip deployed in the detection UAV, the heat radiation distribution data is processed in real time to identify the temperature anomaly area and remove the interference area, and the fire point position and classification result are generated, including: Step 131. Using the embedded real-time processing chip deployed in the detection UAV, median filtering processing is performed on the heat radiation distribution data to eliminate environmental noise interference and generate a denoised temperature field matrix.

[0104] Among them, median filtering processing refers to an image denoising method that selects the median of the neighborhood pixels through a sliding window to replace the current pixel value, and is used to eliminate the salt-and-pepper noise in the heat radiation data. The denoised temperature field matrix is a two-dimensional temperature data matrix after median filtering processing, which retains the true temperature distribution in the fire point area and suppresses environmental interference.

[0105] In the embodiment of the present application, first, the embedded real-time processing chip deployed in the detection UAV is used to perform median filtering processing on the collected heat radiation distribution data, that is, each pixel point is traversed through a sliding window, and the radiation values of all pixels in its neighborhood are sorted by size and then the median is taken to replace the original pixel value to eliminate the interference of impulse noise and isolated abnormal points. After the processing is completed, a denoised temperature field matrix is generated, and this matrix retains the temperature distribution characteristics of the fire point area and suppresses the environmental noise.

[0106] Step 132. Process the denoised temperature field matrix to determine the target fire point area.

[0107] Among them, the target fire point area is a continuous high-temperature area extracted through threshold segmentation and morphological processing, representing the spatial range of potential fire point positions.

[0108] In the embodiment of the present application, first, an adaptive threshold segmentation is performed on the temperature field matrix after noise reduction. By calculating the mean and variance of the global temperature distribution, the segmentation threshold is dynamically adjusted to screen out continuous areas with temperature values higher than the threshold. Subsequently, morphological dilation and erosion operations are performed on the candidate areas to eliminate small holes, and the spatial range of the target fire point area is output after merging adjacent areas.

[0109] Step 133: Extract the temperature distribution feature vector from the target fire point area. The temperature distribution feature vector includes the regional average temperature, the maximum temperature gradient, the standard deviation of the temperature distribution, and the ratio of the regional contour perimeter to the area.

[0110] Among them, the temperature distribution feature vector is a four-dimensional vector containing the regional average temperature, the maximum temperature gradient, the standard deviation of the temperature distribution, and the contour perimeter-area ratio, which is used to quantify the thermodynamic and morphological characteristics of the fire point. The regional average temperature refers to the arithmetic mean of the temperature values of all pixels within the target fire point area, reflecting the overall heat radiation intensity. The maximum temperature gradient refers to the extreme value of the temperature change rate calculated through the gradient operator, characterizing the degree of temperature mutation at the fire point edge. The standard deviation of the temperature distribution is an index reflecting the degree of dispersion of the temperature values in the fire point area from the average temperature, reflecting the uniformity of the temperature distribution. The regional contour perimeter refers to the total length of the closed curve of the fire point area boundary, which is used to calculate the shape complexity in combination with the area.

[0111] In the embodiment of the present application, first, the temperature data within the target fire point area is statistically analyzed pixel by pixel, and the regional average temperature is calculated as an overall heat intensity index. Secondly, the Sobel operator is used to calculate the temperature gradient field, and the maximum gradient is extracted to characterize the degree of local temperature mutation. Subsequently, the standard deviation of the temperature values within the area is calculated to reflect the temperature distribution discreteness, and the regional contour is extracted through an edge detection algorithm. The contour perimeter is divided by the area to obtain the shape complexity ratio. Finally, the above four features are integrated into the temperature distribution feature vector.

[0112] Step 134: Use the K-means clustering algorithm to perform dynamic threshold division on the temperature distribution feature vector, and match according to the Euclidean distance between the cluster center and the historical fire point samples to generate a fire point type label and a confidence parameter. The fire point type label is used as the classification result, and the fire point type label includes surface fire, smoldering fire, and reignition fire.

[0113] Among them, the K-means clustering algorithm is an unsupervised classification algorithm based on distance measurement, which realizes the grouping of feature vectors by iteratively optimizing the cluster centers. The cluster center refers to the mean vector of the feature vectors of each category in the K-means algorithm, representing the typical features of that category. The Euclidean distance of the historical fire point samples refers to the straight-line distance between the current fire point feature vector and the historical samples in the feature space, which is used for category matching. The fire point type label refers to the fire point categories divided according to the matching results of clustering and historical data, including three combustion states: surface fire, smoldering fire, and rekindled fire. The confidence parameter is a probability index calculated based on the reciprocal of the Euclidean distance, reflecting the credibility of the classification result. A surface fire refers to a fire point type with visible flames and intense combustion, corresponding to features such as high temperature, high gradient, and high standard deviation. A smoldering fire refers to a fire point type with no open flames but continuous internal smoldering, showing features such as medium temperature and low gradient. A rekindled fire is a fire point type where a previously extinguished area reignites, with features such as small area and high contour complexity.

[0114] In the embodiment of the present application, first, the K-means clustering algorithm is used to perform unsupervised classification on the temperature distribution feature vectors. By iteratively updating the cluster centers until convergence, the feature vectors are divided into several categories. Secondly, calculate the Euclidean distance between each cluster center and the feature vectors of the historical fire point samples, select the category of the historical sample with the smallest distance as the type label of the current fire point, and generate a confidence parameter based on the reciprocal of the distance. Finally, output a classification result including three types of labels: surface fire, smoldering fire, and rekindled fire.

[0115] Step 135: Based on the spatial coordinates of the corresponding target fire point area in the three-dimensional point cloud data, calculate the three-dimensional geographical coordinates of the fire point center in the UAV navigation coordinate system, and use the three-dimensional geographical coordinates as the fire point location.

[0116] Among them, in the embodiment of the present application, first, based on the boundary coordinates of the target fire point area in the three-dimensional point cloud data, calculate the spatial coordinates of the fire point center through the centroid algorithm. Secondly, combine the geographical registration parameters of the UAV navigation coordinate system to convert the fire point center coordinates into three-dimensional geographical coordinates of longitude, latitude, and altitude. Finally, output the converted coordinates as the fire point location to the disaster emergency command system.

[0117] In the embodiment of the present application, first, the thermal radiation data of a certain forest area is collected by a thermal imaging sensor carried by a drone, and the median filtering process is performed by an embedded chip to generate a temperature field matrix after noise reduction. Subsequently, adaptive threshold segmentation and morphological optimization are used to extract three target fire point areas, and the temperature distribution feature vectors of each area are calculated respectively. Based on the K-means clustering algorithm, the feature vectors are divided into two categories. Combining the Euclidean distance matching of the historical fire point database, two of them are determined to be surface fires and one is a smoldering fire, and the confidence parameter is output. Finally, the fire point center coordinates are calculated according to the three-dimensional point cloud data, converted into longitude, latitude and altitude information, and transmitted to the fire command platform in real time for dynamic fire situation monitoring and fire fighting path planning.

[0118] By executing steps 131 to 135, the embodiment of the present application realizes the efficiency and accuracy of fire point detection and classification through multi-modal data fusion and hierarchical feature analysis. Combining real-time noise reduction processing and dynamic threshold segmentation effectively improves the positioning accuracy of target fire points; through multi-dimensional feature extraction and clustering matching, the physical interpretability of fire point type recognition is enhanced; the finally output three-dimensional geographical coordinates and type labels provide a reliable decision-making basis for fire situation judgment and resource scheduling.

[0119] In a possible embodiment, step 132, processing the temperature field matrix after noise reduction to determine the target fire point area, includes: Step b1, performing binary segmentation on the temperature field matrix after noise reduction based on a preset temperature threshold to extract candidate temperature anomaly areas, and using a region growing algorithm to merge adjacent candidate temperature anomaly areas to generate an initial fire point area set.

[0120] Among them, binary segmentation refers to the process of dividing the temperature field matrix into high-temperature anomaly points and background areas based on the temperature threshold, with high-temperature points assigned a value of 1 and the background assigned a value of 0. The candidate temperature anomaly area refers to the discrete high-temperature pixel set formed after binary segmentation, which represents the spatial distribution of potential fire point positions. The region growing algorithm refers to a spatial clustering algorithm that iteratively merges adjacent similar pixels and is used to generate a continuous closed area. The initial fire point area set refers to the set of fire point candidate areas merged by the region growing algorithm, which contains complete candidate targets without screening.

[0121] In the embodiment of the present application, first, the temperature field matrix after noise reduction is subjected to binary segmentation according to a preset temperature threshold, pixels higher than the threshold are marked as high-temperature anomaly points, and pixels lower than the threshold are classified into the background area, generating a candidate temperature anomaly area composed of discrete high-temperature pixels. Secondly, the region growing algorithm is used. Taking each high-temperature pixel as a seed point, based on the temperature similarity of adjacent pixels, the region is expanded, and adjacent candidate temperature anomaly areas with continuous space are merged, and finally an initial fire point area set is formed, which contains multiple independent and closed fire point candidate areas.

[0122] Step b2: Delete the fire point regions in the initial fire point region set whose area is smaller than the preset area threshold to obtain an intermediate fire point region set.

[0123] The intermediate fire point region set refers to the fire point region set after being filtered by the area threshold, which is the intermediate result after eliminating small-area noise.

[0124] In the embodiment of the present application, first, each fire point region in the initial fire point region set is traversed, the number of pixels it covers is calculated and converted into an actual area value. Secondly, the area of each region is compared with the preset area threshold, and the regions with an area smaller than the threshold are deleted to exclude the interference of isolated noise. Subsequently, the fire point regions that meet the area conditions are retained to generate an intermediate fire point region set, which serves as the basic data set for subsequent refined screening.

[0125] Step b3: Screen out the fire point regions with a reflectance lower than the preset reflectance threshold from the intermediate fire point region set to obtain the target fire point regions.

[0126] The target fire point regions refer to the final fire point regions screened in combination with the reflectance threshold, which are high-confidence results excluding non-fire point interference.

[0127] In the embodiment of the present application, first, the multi-spectral image data corresponding to the intermediate fire point region set is obtained, and the average reflectance of each fire point region is extracted. Secondly, the average reflectance is compared with the preset reflectance threshold, and the regions with a reflectance lower than the threshold are screened out to exclude the misjudgment interference of high-reflectance objects. Finally, the target fire point region set is output to ensure that the result only contains the real fire points that meet the combustion characteristics.

[0128] In the embodiment of the present application, first, the satellite thermal infrared image is binarized and segmented, and the pixels with a temperature higher than the preset threshold are extracted to form candidate temperature anomaly regions. The adjacent high-temperature pixels are merged through the region growing algorithm to generate a set containing six initial fire point regions. Subsequently, the area of each region is calculated and three regions smaller than the preset threshold are deleted, and three intermediate fire point regions are retained. Finally, in combination with the reflectance data of the visible light image, two regions with a reflectance lower than the threshold are screened out, excluding the interference of a high-reflectance water area, and two target fire point regions are output to the forest fire monitoring system for real-time fire situation dynamic tracking and emergency response.

[0129] By executing Step b1 to Step b3, the embodiment of the present application effectively improves the fire point detection accuracy through a multi-level screening mechanism. The binarization segmentation and the region growing algorithm ensure the integrity of the fire point region. The area threshold filtering eliminates noise interference, and the reflectance screening enhances the specificity of target identification, ultimately achieving the accurate positioning of real fire points and false alarm suppression in complex environments.

[0130] Figure 2Schematic diagram of a wildfire monitoring system based on lidar and thermal infrared provided by an embodiment of the present application, as Figure 2 shown. The system includes: An acquisition module 21, configured to control the flight altitude and heading angle of a detection unmanned aerial vehicle (UAV), so that a lidar and a thermal infrared sensor deployed on the detection UAV respectively collect three-dimensional point cloud data of a vertical profile of a wildfire area and thermal radiation distribution data of the wildfire area at multiple angles.

[0131] A generation module 22, configured to combine the three-dimensional point cloud data and the thermal radiation distribution data to generate a target wildfire spread path.

[0132] An identification module 23, configured to use an embedded real-time processing chip deployed in the detection UAV to perform real-time processing on the thermal radiation distribution data, identify temperature anomaly areas and remove interference areas, and generate fire point positions and classification results.

[0133] An adjustment module 24, configured to send dynamic scheduling instructions to a fire extinguishing UAV cluster through a low-latency communication link based on the target wildfire spread path, fire point positions, and classification results, so as to adjust the positions, fire extinguishing agent dosages, and coverage ranges of fire extinguishing agents released by each fire extinguishing UAV in the fire extinguishing UAV cluster.

[0134] Figure 2 The described wildfire monitoring system based on lidar and thermal infrared can execute Figure 1 the wildfire monitoring method based on lidar and thermal infrared described in the embodiment shown. The implementation principle and technical effects will not be elaborated again. For the wildfire monitoring system based on lidar and thermal infrared in the above embodiment, the specific manners in which each module and unit perform operations have been described in detail in the embodiment related to the method, and will not be elaborated here.

[0135] In a possible design, Figure 2 the wildfire monitoring system based on lidar and thermal infrared in the embodiment shown can be implemented as a computing device, as Figure 3 shown. The computing device can include a storage component 31 and a processing component 32.

[0136] The storage component 31 stores one or more computer instructions, where the one or more computer instructions are called and executed by the processing component 32.

[0137] The processing component 32 is configured to: by controlling the flight altitude and heading angle of the detection UAV, enable the lidar and thermal infrared sensor deployed on the detection UAV to collect three-dimensional point cloud data of the vertical profile of the wildfire area and thermal radiation distribution data of the wildfire area from multiple angles respectively; combine the three-dimensional point cloud data and the thermal radiation distribution data to generate a target wildfire spread path; use the embedded real-time processing chip deployed in the detection UAV to perform real-time processing on the thermal radiation distribution data, identify temperature anomaly areas and remove interference areas, and generate fire point positions and classification results; based on the target wildfire spread path, the fire point positions and the classification results, send dynamic scheduling instructions to the fire extinguishing UAV cluster through a low-latency communication link to adjust the positions, fire extinguishing agent dosages and coverage ranges of the fire extinguishing agents dropped by each fire extinguishing UAV in the fire extinguishing UAV cluster.

[0138] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0139] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disc.

[0140] Of course, the computing device necessarily may also include other components, such as input / output interfaces, display components, communication components, etc.

[0141] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above-mentioned peripheral interface module may be an output device, an input device, etc.

[0142] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0143] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above-mentioned processing component, storage component, etc. may be basic server resources leased or purchased from a cloud computing platform.

[0144] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above-mentioned Figure 1 A wildfire monitoring method based on lidar and thermal infrared shown in the embodiment.

[0145] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0146] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0147] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A wildfire monitoring method based on lidar and thermal infrared, characterized in that, Including: By controlling the flight altitude and heading angle of the detection UAV, the lidar and thermal infrared sensor deployed on the detection UAV respectively collect three-dimensional point cloud data of the vertical profile of the wildfire area and the thermal radiation distribution data of the wildfire area at multiple angles; Combining the three-dimensional point cloud data and the thermal radiation distribution data to generate a target wildfire spread path; Using the embedded real-time processing chip deployed in the detection UAV, the thermal radiation distribution data is processed in real time to identify the temperature anomaly area and remove the interference area, and the fire point position and classification result are generated; Based on the target wildfire spread path, the fire point position and the classification result, a dynamic scheduling instruction is sent to the fire extinguishing UAV cluster through a low-latency communication link to adjust the positions, fire extinguishing agent doses and coverage ranges of the fire extinguishing agents dropped by each fire extinguishing UAV in the fire extinguishing UAV cluster.

2. The method according to claim 1, characterized in that, The combining the three-dimensional point cloud data and the thermal radiation distribution data to generate a target wildfire spread path includes: Extracting the wildfire terrain elevation features from the three-dimensional point cloud data; Extracting the temperature field gradient change features from the thermal radiation distribution data; Based on the wildfire terrain elevation features and the temperature field gradient change features, combined with the historical meteorological data of the wildfire area, a target wildfire spread path is generated.

3. The method according to claim 2, wherein The wildfire terrain elevation features include: elevation value sequence, slope change rate set and slope direction angle set, the temperature field gradient change features include: temperature extreme value area boundary, temperature gradient direction vector and temperature gradient intensity distribution, and the historical meteorological data includes wind speed vector, humidity and vegetation type; The generating a target wildfire spread path based on the wildfire terrain elevation features and the temperature field gradient change features, combined with the historical meteorological data of the wildfire area includes: Performing differential calculation on the elevation value sequence to obtain the terrain elevation change rate distribution; Performing a non-linear mapping on the slope change rate in the slope change rate set and the temperature gradient direction vector to generate a dynamic influence intensity parameter characterizing the dynamic influence of the terrain on the temperature field gradient change features; Extracting the topological structure features from the temperature extreme value area boundary, and the topological structure features include the number of connected domains and the boundary curvature radius; Performing azimuth partitioning on the slope direction angle set, calculating the covariance coefficient of the slope direction angle and the temperature gradient intensity distribution in each partition, and taking the partition with the covariance coefficient greater than the preset coefficient threshold as the wildfire spread sensitive area; According to the terrain elevation change rate distribution, calculating the heat convection intensity of the fire front under different terrain conditions, calculating the spread rate of the fire front according to the vegetation type, humidity and wind speed vector, and correcting the direction of the fire front based on the dynamic influence intensity parameter to generate multiple candidate wildfire spread paths; According to the spatial coupling relationship between the distribution range of the wildfire spread sensitive area and the topological structure features, performing probability simulation on each candidate wildfire spread path to screen out the target wildfire spread path that meets the preset spread conditions.

4. The method according to claim 3, wherein The performing a non-linear mapping on the slope change rate in the slope change rate set and the temperature gradient direction vector to generate a dynamic influence intensity parameter characterizing the dynamic influence of the terrain on the temperature field gradient change features includes: Normalize all the slope change rates in the slope change rate set to obtain a standardized slope change rate vector; Decompose the temperature gradient direction vector into horizontal and vertical components, and calculate the first cosine similarity between the horizontal component and the standardized slope change rate vector, and the second cosine similarity between the vertical component and the standardized slope change rate vector respectively; Perform exponential weighting on the first cosine similarity and the second cosine similarity respectively to obtain a first weighted value and a second weighted value, and take the sum of the first weighted value and the second weighted value as the direction consistency parameter; Normalize the direction consistency parameter to obtain a dynamic influence intensity parameter.

5. The method according to claim 3, characterized in that, Calculating the heat convection intensity of the fire front under different terrain conditions according to the terrain elevation change rate distribution, calculating the diffusion rate of the fire front according to the vegetation type, humidity and wind speed vector, and correcting the direction of the fire front based on the dynamic influence intensity parameter to generate multiple candidate fire spread paths, including: Calculate the heat convection intensity correction coefficients at the first position and the second position in the volcanic area according to the formulas with different parameter values respectively; the first position is the position corresponding to the terrain elevation change rate greater than the preset change rate threshold in the terrain elevation change rate distribution, and the second position is the position corresponding to the terrain elevation change rate less than or equal to the preset change rate threshold in the terrain elevation change rate distribution; Take the product of the heat convection intensity correction coefficient, the intensity reference value and the correction coefficient corresponding to the vegetation type at each position as the heat convection intensity at each position; Calculate the diffusion rate according to the vegetation coefficient corresponding to the vegetation type, the rate reference value, the humidity and the wind speed vector by using the preset diffusion rate calculation formula; Calculate the extension length of the path according to the heat convection intensity and the diffusion rate at each position; Determine the direction correction coefficient corresponding to the dynamic influence intensity parameter, and correct the direction of the fire front based on the direction correction coefficient to obtain the corrected direction; Initialize the path starting point as the current coordinate of the fire front, and generate multiple candidate fire spread paths based on the corrected direction and the extension length of the path, in combination with the path screening conditions, where the path screening conditions include: the overlapping area of the path and the fire spread sensitive area is greater than the preset overlapping area threshold, the diffusion rate of the path is greater than the preset rate threshold, and the diffusion duration is greater than the preset duration threshold.

6. The method according to claim 1, characterized in that Using the embedded real-time processing chip deployed in the detection UAV, perform real-time processing on the thermal radiation distribution data, identify the temperature anomaly area and remove the interference area, and generate the fire point position and classification result, including: Use the embedded real-time processing chip deployed in the detection UAV to perform median filtering on the thermal radiation distribution data to eliminate environmental noise interference and generate a denoised temperature field matrix; Process the denoised temperature field matrix to determine the target fire point area; Extract the temperature distribution feature vector from the target fire point area, and the temperature distribution feature vector includes the regional average temperature, the maximum temperature gradient, the standard deviation of the temperature distribution, and the ratio of the regional contour perimeter to the area; The K-means clustering algorithm is used to dynamically divide the temperature distribution feature vectors, and based on the Euclidean distance matching between the cluster centers and historical fire point samples, fire point type labels and confidence parameters are generated. The fire point type labels are used as the classification results, and the fire point type labels include surface fires, smoldering fires, and rekindled fires. Based on the spatial coordinates of the corresponding target fire point area in the three-dimensional point cloud data, the three-dimensional geographical coordinates of the fire point center in the UAV navigation coordinate system are calculated, and the three-dimensional geographical coordinates are used as the fire point location.

7. The method according to claim 6, wherein The processing of the denoised temperature field matrix to determine the target fire point area includes: Based on a preset temperature threshold, the denoised temperature field matrix is binarized to extract candidate temperature anomaly areas, and the region growing algorithm is used to merge adjacent candidate temperature anomaly areas to generate an initial fire point area set. Fire point areas with an area smaller than a preset area threshold in the initial fire point area set are deleted to obtain an intermediate fire point area set. Fire point areas with a reflectivity lower than a preset reflectivity threshold are screened out from the intermediate fire point area set to obtain the target fire point area.

8. A wildfire monitoring system based on lidar and thermal infrared, characterized in that, It includes: A collection module for collecting three-dimensional point cloud data of the vertical profile of the wildfire area and the thermal radiation distribution data of the wildfire area from multiple angles by controlling the flight altitude and heading angle of the detection UAV, so that the lidar and thermal infrared sensor deployed on the detection UAV can collect the data respectively. A generation module for combining the three-dimensional point cloud data and the thermal radiation distribution data to generate a target fire spread path. An identification module for using the embedded real-time processing chip deployed in the detection UAV to perform real-time processing on the thermal radiation distribution data, identifying temperature anomaly areas and removing interference areas, and generating fire point locations and classification results. An adjustment module for sending dynamic scheduling instructions to the fire extinguishing UAV cluster through a low-latency communication link based on the target fire spread path, the fire point location, and the classification results, so as to adjust the positions, fire extinguishing agent dosages, and coverage ranges of the fire extinguishing agents dropped by each fire extinguishing UAV in the fire extinguishing UAV cluster.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a wildfire monitoring method based on lidar and thermal infrared as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by a computer, it implements a wildfire monitoring method based on lidar and thermal infrared as described in any one of claims 1 to 7.

Images