Unmanned aerial vehicle fire monitor fire source positioning system based on infrared thermal imaging

Abstract

The invention provides an unmanned aerial vehicle fire monitor fire source positioning system based on infrared thermal imaging, and the system comprises a fire scene situation collection module which obtains thermal image data and high-precision point cloud data; the fire diffusion model building module is used for generating a radiation intensity distribution diagram; the thermal image processing module is used for generating a structured fire source risk feature image; the LSTM time sequence model processing module is used for generating a smooth diffusion thermodynamic diagram to predict the spreading direction and speed; the combat map generation module is used for realizing data visualization; and the unmanned aerial vehicle flight path planning module realizes comprehensive coverage of a fire monitor striking path. Aiming at the problem that the prior art depends on static fire source data and is difficult to capture time-space dynamic evolution of a fire behavior, the method comprises the following steps: constructing a multi-dimensional fire behavior diffusion model, and fusing fire scene air flow vector projection analysis, thermal radiation background concentration dynamic superposition and LSTM time sequence modeling modes; accurate prediction of the diffusion path of the fire from the low-risk area to the high-risk area is realized, and instant support is provided for fire extinguishing action.

Description

Technical Field

[0001] The present invention relates to the technical field of fire extinguishing, and specifically to an unmanned aerial vehicle (UAV) fire fighting gun fire source positioning system based on infrared thermal imaging. Background Art

[0002] Traditional fire fighting gun fire source positioning systems mainly rely on static fire source positions and single temperature data. This limitation makes it difficult to cope with the spatio-temporal dynamic changes of the fire.

[0003] In complex environments, such as high-density urban building clusters, chemical plants, and the interface between forests and cities, the spread of fire is affected by multiple factors, including wind speed, terrain, distribution of flammable materials, building density, and the characteristics of chemical substances, etc.

[0004] Specifically, in high-density urban building clusters, fires spread rapidly upwards due to the chimney effect of high-rise buildings, and traditional systems cannot capture this dynamic change in real time, resulting in the failure to deploy fire extinguishing resources in a timely manner. In chemical plant fires, the key problems that traditional systems have difficulty in identifying are hidden fire sources and the risk of re-ignition in chemical storage tank areas, which lead to the out-of-control of the fire and even the occurrence of secondary disasters. In addition, fires in the interface between forests and cities are greatly affected by wind speed and vegetation distribution. Traditional systems cannot accurately predict the spread direction and speed of the fire, thus delaying the fire extinguishing opportunity.

[0005] In response to this problem, the prior art CN118552873A proposes an urban fire early warning analysis method, which mainly includes: real-time collecting video images of the fire scene through a camera, preprocessing the images, using an improved YOLOv5s network model for smoke detection, and determining whether the number of consecutive smoke detections exceeds a threshold to trigger a fire early warning.

[0006] However, it does not consider another main problem of traditional systems: the limitation of a single data source. Relying solely on temperature data or fire source positions cannot comprehensively reflect the complexity of the fire scene, especially under changing environmental conditions. This increases the uncertainty and risk of fire fighting operations. Specifically, the method proposed in the prior art CN118552873A only relies on video images collected by cameras, lacks the integration of multi-source data such as infrared thermal imaging of the fire scene and LiDAR point cloud data, and cannot comprehensively capture the "dynamic development" of the fire scene. In addition, this method is mainly based on single analysis, lacks the ability to dynamically predict the spread of the fire, and is difficult to adapt to complex environments (such as high-density building clusters, chemical plants, and the interface between forests and cities), resulting in problems such as fire source positioning deviation, misjudgment of the spread path, and lag in resource scheduling in practical applications.

[0007] In addition, the prior art CN118230533B proposes an intelligent analysis system based on fire information. Through a fire data acquisition module, a data analysis and processing module, and a fire level assessment module, it can monitor the fire information in the fire area in real time, analyze the fire scale and spread risk, evaluate the fire danger level, and formulate rescue measures according to the evaluation results.

[0008] Similarly, the prior art CN118552873A mainly relies on explicit indicators such as temperature, smoke concentration, and fire source area obtained at fixed / static detection points, and has obvious defects in dynamically predicting the spread of the fire. Specifically, the system mainly relies on the real-time acquisition and analysis of static data (such as temperature, smoke concentration, wind speed, and flammable substance content), lacks the ability to dynamically model the spread of the fire, does not consider the changing trend of the fire over time for simulating and predicting the spread path, and cannot effectively predict the non-linear spread behavior of the fire in complex environments. As a result, when dealing with large-scale and complex fire scenarios, it cannot provide forward-looking predictions of the fire spread, thus affecting the timeliness and accuracy of rescue measures.

[0009] In summary, the limitations of traditional fire-fighting gun fire source positioning systems in complex environments are mainly reflected in the following aspects: unable to capture the dynamic changes of the fire in real time, difficult to identify hidden fire sources and re-ignition risks, lack of prediction ability integrating multiple factors, and lag in resource scheduling. These problems are particularly prominent in modern fire prevention and control, and urgently need to be solved by introducing high-precision real-time data fusion, dynamic prediction models, and intelligent fire extinguishing strategies. Summary of the Invention

[0010] Aiming at the deficiencies of the prior art, the purpose of the present invention is to provide an unmanned aerial vehicle (UAV) fire-fighting gun fire source positioning system based on infrared thermal imaging to solve the problems raised in the above background technology.

[0011] To achieve the above purpose, the present invention is implemented through the following technical solutions: An unmanned aerial vehicle fire-fighting gun fire source positioning system based on infrared thermal imaging, including: A fire scene situation acquisition module, which acquires high-precision point cloud data representing the thermal image data of the target fire scene and the spatial information of the location; A fire spread model construction module, based on the point cloud data, quantifies the spatial relationship between the air flow direction in the fire scene and the data acquisition points, and screens out the upstream acquisition points consistent with the fire source spread trend; based on the thermal image data, considering the attenuation effect of the fire source radiation intensity at the upstream acquisition points with the change of their distances, calculates the thermal radiation background concentration of each upstream acquisition point, optimizes and dynamically solves the fire source radiation intensity through gradient descent, fuses the preset parameters of multiple diffusion model libraries, constructs a radiation intensity matrix, generates a radiation intensity distribution map with geographical coordinates, and forms a processable fire source image by characterizing the fire spread path of the target fire scene through the change of radiation intensity. A thermal image processing module performs morphological processing on the fire source image, conducts convolution operations through a structural kernel that simulates the fluctuation characteristics of the fire source, and extracts key fire source feature data including the fire source edge and diffusion direction in the fire source image through radiation energy gradient analysis, quantifies the fire spread trend. Finally, a pixel point is selected from the key fire source feature data as the analysis starting point, calculates the radiation energy correlation degree between each pixel point and the starting point, conducts a risk level division of the fire spread, and generates a structured fire source risk feature image through binary mapping according to the risk level to clarify the fire front position and risk distribution; An LSTM time series model processing module updates the state using a gating mechanism, calculates the probability of cross-level fire spread based on the fire source risk feature image, and generates a smooth heat diffusion map to predict the spread direction and speed; A combat map generation module overlays the heat diffusion map on the actual geographical coordinates to achieve data visualization; A drone flight path planning module combines the fire area and the diffusion probability, sets an objective function, and achieves full coverage of the fire fighting cannon strike path.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Aiming at the problems of the prior art relying on static fire source data and being difficult to capture the spatio-temporal dynamic evolution of the fire situation, the present invention realizes the accurate prediction of the fire spread path from the low-risk area to the high-risk area by constructing a multi-dimensional fire spread model, integrating the projection analysis of the fire field air flow vector, the dynamic superposition of the thermal radiation background concentration, and the LSTM time series modeling method. Specifically, the present invention constructs a fire field fluid perception unit, calculates the spatial projection factor between the fire field air flow vector and the point cloud acquisition point in real time, dynamically screens the upstream fire source influence area in combination with the thermal radiation attenuation law, quantifies the multi-dimensional radiation contribution of the fire source to the spatial grid. Compared with the traditional fixed threshold model, the present invention realizes the dynamic superposition calculation of the fire situation background concentration, solves the misjudgment problem caused by the sudden change of the fire field air flow, and provides real-time data support for the prediction of the fire spread path; 2. Considering the risk problem that the thermal imaging of the existing system is vulnerable to smoke interference, the present invention designs a sine-Gaussian hybrid structure kernel through a morphological processing unit, simulates the fluctuation characteristics of the fire source to strengthen the local dynamic characteristics, combines dynamic threshold binary segmentation of the active area, and further calculates the change rate of multi-directional radiation energy (longitude, latitude, main / secondary diffusion direction), extracts the fire front edge gradient feature, and divides the fire field into high, medium, and low risk three-level areas through the weight distribution of the radiation energy correlation degree, making the system have the advantage of high accuracy in identifying hidden fire sources; 3. Through real-time data processing and LSTM time series modeling, the present invention uses an infrared thermal imager and a LiDAR device carried by a drone to collect the thermal radiation distribution and three-dimensional spatial data of the fire scene in real time, captures the dynamic change law of the fire trend through the LSTM time series model, and combines interpolation smoothing to process the diffusion probability distribution, enabling the system to quickly update the fire spread thermal diagram and provide immediate support for fire extinguishing operations; 4. By collecting thermal imaging and LiDAR point cloud data at a high frequency of 30Hz by the drone, dynamically annotating the fire source level, spread vector arrows, and safety channel streamer guidance, and combining RTK-GPS centimeter-level positioning and particle system simulation of the fire cannon trajectory, the present invention generates an operation map containing spatial coordinates, timestamps, and temperature gradients, greatly reducing the fire extinguishing response time. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The disclosure of the present invention will be described with reference to the accompanying drawings. It should be understood that the drawings are only for illustrative purposes and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components. Among them: Figure 1 is a schematic structural framework diagram of a drone fire cannon fire source positioning system proposed in an embodiment of the present invention; Figure 2 is a visual schematic diagram of a radiation intensity matrix established in an embodiment of the present invention to quantify the spatial influence of a fire source in a fire scene on each point cloud data acquisition point in the target fire scene; Figure 3 is a schematic diagram of a comprehensive radiation intensity distribution generated in an embodiment of the present invention to analyze the column gradient change of the fire source radiation intensity in the image of the fire source active area; Figure 4 is based on Figure 2 the radiation influence value shown in the proposed radiation intensity matrix Figure 1 ; Figure 5 is based on Figure 2 the radiation influence value shown in the proposed radiation intensity matrix Figure 2 ; Figure 6 is a schematic diagram of using a path planning algorithm to ensure the safe and efficient flight and strike path of a drone proposed in an embodiment of the present invention; Figure 7 is a schematic diagram of morphological processing of a fire source image proposed in an embodiment of the present invention; Figure 8 is a schematic diagram showing the fire spread direction and potential risk areas intuitively based on the operation map generation module proposed in an embodiment of the present invention Figure 1 ; Figure 9 This is proposed in an embodiment of the present invention. Based on the combat map generation module, it intuitively shows the schematic of the fire spread direction and potential risk areas. Figure 2 . Detailed implementation manners

[0014] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, those of ordinary skill in the art can propose various interchangeable structural ways and implementation ways. Therefore, the following detailed implementation manners and drawings are only exemplary descriptions of the technical solution of the present invention, and should not be regarded as all of the present invention or as a limitation or restriction on the technical solution of the present invention.

[0015] The present invention will be further described in detail below with reference to the drawings, but it is not a limitation to the present invention.

[0016] As Figure 1 shown, as an embodiment of the present invention, first proposed is an unmanned aerial vehicle (UAV) fire fighting gun fire source positioning system based on infrared thermal imaging, including: a data transmission control module, and a fire scene situation acquisition module, a fire spread model construction module, a thermal image processing module, a time series model processing module, and a UAV flight path planning module that are respectively electrically connected to the data transmission control module. The data transmission control module is used to adopt 4G and / or 5G and / or satellite communication technologies to ensure the stability and real-time performance of data transmission.

[0017] In an embodiment of the present invention, the fire scene situation acquisition module, based on the infrared thermal imager and lidar (LiDAR) equipment carried by the UAV, performs real-time scanning and situation data acquisition on the target fire scene. The situation data includes thermal image data representing the thermal radiation distribution of the target fire scene and high-precision point cloud data representing the spatial information of the fire scene where the target fire scene is located. It can be understood that the infrared thermal imager detects the thermal radiation distribution of the fire scene in real time and generates a thermal image, providing basic data for the automatic identification of the fire source. The thermal radiation distribution is judged by the radiation intensity. In the field of infrared thermal imaging, the radiation intensity is usually used to describe the thermal radiation characteristics of the object surface. The radiation intensity can provide information about the temperature distribution on the object surface. Generally, a high radiation intensity usually indicates a higher temperature on the object surface, while a low radiation intensity indicates a lower temperature. The lidar (LiDAR) performs a comprehensive scan of the fire scene and generates high-precision point cloud data, which is used to construct a digital elevation combat map and a fire spread model subsequently, realizing a comprehensive monitoring of the fire scene situation, and providing accurate support for subsequent fire source identification, path planning, and fire extinguishing strategy optimization. In an embodiment of the present invention, a fire spread model construction module, based on point cloud data, quantifies the spatial influence of a fire source in a fire scene on each point cloud data acquisition point in the target fire scene, and considers the radiation intensity and fluctuation characteristics of the fire source. The fluctuation characteristics refer to the influence of the air flow in the fire scene on the spread or propagation of the fire. That is, this module needs to evaluate the fluctuation characteristics of the fire source affected by the air flow in the fire scene, and then associate the point cloud data with the radiation intensity of the fire source to establish a radiation intensity matrix that comprehensively evaluates the full-dimensional influence relationship including the fire source in the fire scene, the point cloud data acquisition point, and the fluctuation characteristics, so as to screen out the area of the point cloud data acquisition point that is preferentially affected by the fire.

[0018] Specifically, the fire spread model construction module includes: A fire scene fluid perception unit, which quantifies the spatial influence of a fire source in the fire scene on a point cloud data acquisition point based on real-time fire scene air flow data to evaluate the influence of the air flow in the fire scene on the spread of the fire. The process is as follows: First, for any temperature measurement point i and point cloud data acquisition point j, calculate the projection factor θ of their projections in the direction of the fire scene air flow ij , and judge the direction consistency between the two: , where the fire scene air flow vector represents the current fire scene air flow direction, and the spatial vector from the temperature measurement point to the point cloud data acquisition point represents the spatial relationship between the two. represents the vector length, which is used to measure the size of the vector. Among them, if the dot product result is positive, it means that the directions of the two vectors are roughly the same; if it is negative, it means that the directions are roughly opposite; if it is zero, it means that the two vectors are perpendicular. In this embodiment, the projection factor θ ij is between -1 and 1. When θ ij is close to 1, it means that the possibility of fire spread is relatively large; when θ ij is close to -1, it means that the directions are roughly opposite and the possibility of fire spread is relatively small; when θ ij is close to 0, it means that the influence on fire spread is relatively small.

[0019] Secondly, screen out the area where the point cloud data acquisition points with the same direction and within a preset threshold distance are located as the upstream fire scene area. Specifically, when implementing, set the maximum distance threshold R max based on the thermal radiation intensity of the temperature measurement point and the fire scene environmental conditions to limit the maximum influence distance between the temperature measurement point and the point cloud data acquisition point. It is set that when the distance between the temperature measurement point and the point cloud data acquisition point <R max and the projection factor θ ijWhen >cos45°, the point cloud data acquisition point j is marked as the upstream point cloud data acquisition point of the temperature measurement point i, and the area where the acquisition point is located is recorded as the upstream fire area, so as to ensure that the subsequent fire spread model can accurately identify the area that the fire may first affect, thereby providing a data basis for subsequent fire spread prediction and the fire extinguishing strategy of the unmanned aerial vehicle fire cannon.

[0020] Finally, for each point cloud data acquisition point j, the set Ω of temperature measurement points in its upstream fire area is screened out ij , and the background concentration B of the fire heat radiation in this area is calculated j . It can be understood that the background concentration B of the fire heat radiation j is used to reflect the comprehensive influence of multiple temperature measurement points on the point cloud data acquisition point, and is generated by superimposing the radiation intensities E of each fire source Si i according to the diffusion law: . It can be understood that the fire source Si is the actual source of heat radiation in the fire field, and its position and radiation intensity determine the fluctuation characteristics and influence range of the fire. The temperature measurement point i is the point used to monitor the heat radiation intensity in the fire field, which can be the high-temperature area in the thermal image or other points with heat radiation characteristics. Each temperature measurement point i includes its coordinates (xi, yi, hi) and the heat radiation intensity. In specific implementation, through the data of the temperature measurement point i, the radiation intensity and position of the fire source Si can be deduced inversely. The heat radiation intensity of the temperature measurement point i reflects the heat release situation of the fire source Si. Generally, the stronger the heat radiation of the fire source Si, the higher the heat radiation intensity of the surrounding temperature measurement points i. By monitoring the dynamic changes of the temperature measurement points, the position of the fire source Si can be located. In the above formula, is the initial heat radiation intensity of the fire source Si at the point cloud data acquisition point j, is the set Ω of temperature measurement points of the point cloud data acquisition point j in the upstream fire area ij of the size, indicating the number of temperature measurement points in the set, is the horizontal distance between the fire source Si and the point cloud data acquisition point j, is the vertical distance between the fire source Si and the point cloud data acquisition point j, is the attenuation and / or diffusion coefficient of the radiation intensity of the current fire source Si in the horizontal direction, which is determined through experiments or numerical simulations to more accurately reflect the actual fire field conditions, is the diffusion coefficient in the current vertical direction, and exp is an exponential function used to calculate the attenuation degree of the radiation intensity of the current fire source Si with distance.

[0021] In an embodiment of the present invention, the fire spread model construction module further includes: a fire source radiation intensity calculation unit, based on the obtained background concentration B of the fire heat radiation j, the radiation intensity E of the fire source Si at each point cloud data acquisition point j within the upstream fire area is obtained through iterative optimization dynamic calculation i , so as to accurately quantify the thermal radiation impact of the fire source on the surrounding area outside the upstream fire area, and provide a data basis for fire spread prediction and high-risk area division. It should be noted that the purpose of this unit is to achieve the dynamic calculation of the fire source radiation intensity, solve the prediction deviation problem caused by traditional models ignoring the background concentration of thermal radiation and the multi-source coupling effect, and provide core parameter support for the dynamic grid division of the fire scene and the precise scheduling of fire fighting resources. The process is as follows: First, based on the background concentration B of the fire thermal radiation output by the fire scene fluid perception unit j (reflecting the comprehensive influence of multiple fire sources), by using the minimum predicted thermal radiation concentration and the measured difference, a target optimization function is constructed to solve the optimal fire source radiation intensity : , where is the sparse constraint coefficient, which is used to suppress noise interference and prevent overfitting. M is the number of temperature measurement points i, and N is the number of point cloud data acquisition points.

[0022] Secondly, based on the maximum value of the background concentration B of the fire thermal radiation at the temperature measurement point i j , the initial radiation intensity of the fire source Si is assigned, and gradient descent iteration is used to calculate the minimum predicted thermal radiation concentration and the residual . By controlling the step size of each iterative update, the radiation intensity E i is updated. When is less than the set threshold, the iteration is terminated, and the optimized radiation intensity of all fire sources Si is output. The minimum predicted thermal radiation concentration refers to the minimum predicted value of the background thermal radiation intensity calculated by the model, which is used to set the threshold or as a reference value to help identify the influence degree of the fire source on each acquisition point.

[0023] In an embodiment of the present invention, the fire spread model construction module further includes: a radiation intensity matrix construction unit, which binds the fire source radiation intensity calculation result with the spatial coordinates, establishes a radiation intensity matrix that comprehensively evaluates the full-dimensional influence relationship including the fire sources in the fire scene, the point cloud data acquisition points, and the fluctuation characteristics, directly quantifies the thermal radiation impact of the fire source Si on the upstream fire scene and the surrounding area, and screens out the point cloud data acquisition point area that is preferentially affected by the fire, so as to provide a planning basis for the subsequent strike path of the unmanned aerial vehicle fire cannon. As Figure 2 shown, in the radiation intensity matrix, the fire influence intensity is represented by a gradient color, and the influence value is mapped using transparency. The brighter it is, the greater the influence. The process is as follows: First, the convection diffusion coefficient σ preset by the K diffusion model library built by the system (such as CFD and GIS-based fire spread model) is d (k) , the number of temperature measurement points M and the number of point cloud data collection points N, define the radiation intensity matrix dimension, and then construct the radiation intensity matrix C∈R M×N×K , where the element C of the radiation intensity matrix C is ijk It represents the radiation impact value of the fire source Si on the point cloud data collection point j under the kth diffusion model library: , where are the diffusion coefficients of the kth diffusion model library in the x, y, and z directions, respectively, and d j ,d i are the coordinates of the point cloud data collection point j and the fire source Si, d j -d i is the distance difference between the cloud data collection point j and the fire source Si in all directions.

[0024] Second, for each element C of the radiation intensity matrix C ijk , superimpose the radiation impact values of K diffusion model libraries to form a complete radiation intensity matrix C. Finally, all point cloud data collection points are sorted according to the integrated radiation impact values, such as from high to low, and a radiation impact threshold is selected to sequentially screen out the point cloud data collection point areas with radiation impact values greater than the threshold, and generate a radiation intensity distribution map as a whole. It can be understood that this radiation intensity distribution map has geographic coordinates. At this point, the fire diffusion path of the target fire scene is characterized by the change in radiation intensity, and a processable fire source image can be formed.

[0025] Based on the above technical concept, it can be understood that by superimposing the radiation impact values of K diffusion model libraries, a complete radiation intensity matrix C∈R is formed. M×N×K , so that the matrix integrates the radiation effect of the fire source Si on each point cloud data collection point under different diffusion models. Each element C in the matrix ijk It represents the radiation impact value of the fire source Si on the point cloud data collection point j under the kth diffusion model, such as Figures 4 - 5As shown, it contains information such as the location of the fire source, radiation intensity, and fluctuation characteristics. During subsequent processing, by projecting the three-dimensional matrix onto a two-dimensional plane, a radiation intensity distribution map that comprehensively represents the fire source image data of the target fire scene can be generated. For example, based on this map, binary segmentation is performed to guide the adaptive focusing of the thermal image and the analysis of fire scene characteristics, extract the contour of the high-risk area, ensure that the high-radiation area is given priority treatment and clearly presented, thereby guiding the strike path of the UAV fire cannon. For ease of understanding, in this embodiment, the selected area is the "priority area" determined by the radiation intensity matrix C, aiming to first process the area in the fire scene that is greatly affected by the fire source. The active area is within the selected priority area where the radiation influence value is higher than the first threshold T fir The area represents the part with the strongest influence of the fire source. The non-active area is within the selected priority area where the radiation influence value is lower than the first threshold T fir The area represents the part with a weaker influence of the fire source, aiming to further refine the analysis of the spread trend of the fire source as a basis for fire extinguishing strategies and resource scheduling.

[0026] In an embodiment of the present invention, in the thermal image processing module of the UAV fire cannon fire source positioning system based on infrared thermal imaging, the purpose is to adjust the minimum focal length of the infrared thermal imager to ensure that the area of the point cloud data acquisition points with a radiation influence value greater than the threshold is always within the imaging range. Through morphological processing, the diffusion characteristics of the fire source image in this area are extracted, and then based on the calculation of the multi-directional radiation energy change rate, a binary fire source risk characteristic image containing high, medium, and low risk levels of the fire scene is generated, providing structured input data for subsequent fire trend analysis and fire extinguishing decisions.

[0027] Specifically, the thermal image processing module includes: The imaging focus adjustment unit calculates the minimum focal length according to the position and range of the selected area to ensure that the areas where these important fire sources are located are within the imaging range. The process is as follows: Obtain the coordinates of the fire source Si, the point cloud data acquisition point j, and the field of view angle θ of the infrared thermal imager FOV , and calculate the minimum focal length f min , , where d min is the minimum distance between the fire source Si and the point cloud data acquisition point j, and h max is the maximum height that can be accommodated on the imaging plane of the infrared thermal imager to optimize the monitoring accuracy and efficiency of the fire source Si.

[0028] In an embodiment of the present invention, the thermal image processing module further includes: a morphological processing unit that performs morphological processing on the fire source image of the selected area, as Figure 7 shown, and the process is as follows: First, through the radiation intensity matrix C, obtain the radiation influence value of each pixel (p,q) in the fire source image of the selected area and the weight of this area determined based on the actual fire scene conditions. It should be noted that the radiation influence value of each pixel directly reflects the thermal radiation intensity at that position. High radiation influence values usually correspond to the fire source or high-temperature areas, while low radiation influence values correspond to non-fire source or low-temperature areas; use the weighted average formula, and combine the radiation influence value and the area weight to calculate the basic radiation influence mean value E of the fire source image in this area. mean Based on the basic radiation influence mean value E mean and the adjustment factor r, set the first threshold T fir , T fir =E mean ×r, where the adjustment factor r is set according to the dynamic characteristics of the fire scene (such as wind speed, type of combustibles), usually r ∈ [1.2, 1.5]; compare the radiation influence value with the first threshold T fir . If the radiation influence value is greater than the first threshold T fir , then mark this selected area as an active area (output value is 1), otherwise mark it as an inactive area (value is 0). It can be understood that through the above binary processing, as Figure 3 shown, the generated complex comprehensive radiation intensity distribution can be simplified into two types of areas: active and inactive, highlighting the main affected areas by the fire source Si, which is convenient for quickly identifying the diffusion trend of the fire source Si in the subsequent process.

[0029] Secondly, define a structure kernel K to simulate the fluctuation characteristics of the fire source Si. In this embodiment, the proposed structure kernel K is an m×n matrix, and the element value k pq in its matrix is calculated by the following formula: , where Z is a normalization constant to ensure that the sum of the element values of the structure kernel K is 1, α is the sine modulation amplitude (generally taken as α = 0.3) used to simulate the fire source fluctuation, λ x and λ xThey are the sine wavelengths in the x and y directions respectively, related to the fire source diffusion scale. σ is the Gaussian attenuation coefficient (usually taken as σ = 1.5), which controls the spatial attenuation range of the structure kernel K. m is the number of rows of the structure kernel K, used to define the size of the structure kernel K in the vertical direction and affect the capture range of the vertical features of the image during the convolution operation. n is the number of columns of the structure kernel K, used to define the size of the structure kernel in the horizontal direction and affect the capture range of the horizontal features of the image during the convolution operation. p is the index of the structure kernel K in the row direction, and its value range is usually from -a to a, where a is the convolution kernel radius. q is the index of the structure kernel K in the column direction, and its value range is usually from -b to b, where b is the convolution kernel radius. It can be understood that through the sine modulation term, the structure kernel K can simulate the fluctuation characteristics of the fire source Si, which helps to capture the dynamic changes of the fire source in the morphological processing. And through the Gaussian attenuation term, it ensures that the structure kernel K has certain attenuation characteristics in space, so that the processing process can focus on the local area (active and inactive areas), while reducing the influence of distant pixels. Through the above operations, the key features in the fire source image, such as edges and diffusion directions, can be extracted.

[0030] Again, use the structure kernel K to perform convolution calculation on the image of the active area of the fire source: , where a and b are the convolution kernel radii respectively, usually taken as a = (m - 1) / 2 and b = (n - 1) / 2. is the pixel value at the corresponding position in the image of the active area of the fire source. Update the pixel value according to the convolution result. If C ij is greater than the set threshold, then set the value of this pixel point to 1, otherwise to 0. Among them, the above threshold is usually set to 50% of the total energy of the structure kernel, such as the threshold is set to 0.5.

[0031] Finally, by calculating the cumulative value of the radiation energy of each column in the image of the active area of the fire source, use the gradient change formula , e ij is the radiation influence value of the pixel at the j-th row and i-th column, k is the number of rows. Analyze the column gradient change of the fire source radiation intensity in the image of the active area of the fire source, and determine the key features of the edges and diffusion directions of the suspected fire source area in the image of the active area of the fire source accordingly, and form the fire source key feature data containing this key feature.

[0032] In an embodiment of the present invention, the thermal image processing module further includes: a multi-directional radiation energy change rate calculation unit, which calculates the radiation energy change rate of the key fire source feature data in the longitude, latitude, the main diffusion direction related to the fluctuation characteristics of the fire source (obtained by statistical analysis of the diffusion directions of each fire source in the radiation intensity matrix C), and the secondary diffusion direction (forming a certain angle with the main diffusion direction, and this angle is related to the irregularity degree of the fire source), so as to identify the change in the diffusion intensity of the fire source. Specifically, when implementing, the radiation energy change rate is obtained by calculating the energy difference between the current pixel and its right pixel or the current pixel and its lower pixel or the current pixel and the next pixel along this direction, and then dividing by the total energy respectively.

[0033] In an embodiment of the present invention, the thermal image processing module further includes: a region division unit: First, perform preprocessing operations such as denoising and enhancing contrast on the captured key fire source feature data to improve the image quality; Second, select a pixel point in the key fire source feature data as the analysis starting point, and calculate the radiation energy correlation degree between each pixel point and the starting point. , it can be understood that the weight coefficient w i is related to the distance between each pixel point and the starting point and the weight of the region where it is located, Ei is the radiation influence value of the adjacent pixel, m is the number of adjacent pixels considered when calculating the radiation energy correlation degree. Finally, take the radiation energy statistical value of the high-risk area in the historical fire field data, set the second threshold T sec , according to the proportional relationship T mid = k × T sec , determine the intermediate threshold T mid , k is an empirical coefficient, and its value is 0.7. If p ≥ T sec , then it is divided into an active fire core area in the key fire source feature data, corresponding to a high radiation intensity and a fast-spreading risk level. If p ≤ <T sec , then it is divided into a weak combustion or safe area in the key fire source feature data, corresponding to a low risk level. If T mid < p < T sec , then it is divided into a stable combustion area in the key fire source feature data, corresponding to a medium risk level of controllable spread.

[0034] In an embodiment of the present invention, the thermal image processing module further includes: a fire source risk feature image generation unit: perform binary mapping according to the risk level. In this embodiment, pixels in the high-risk level area are marked as V high-risk = 1 (such as white), and the rest of the area is marked as V low-risk = 0 (such as black), generate a binary fire source risk feature image including high, medium, and low risk levels of the fire field, and mark the position of the fire front.

[0035] Based on the above technical concept, it can be understood that in the existing fire monitoring and prediction technologies, traditional methods often rely on static fire source location and intensity data, making it difficult to capture the dynamic changes and diffusion trends of the fire. Due to the high spatio-temporal complexity of fire spread, a single static model cannot comprehensively reflect the dynamic characteristics of the fire scene, resulting in insufficient accuracy and timeliness of the prediction results.

[0036] Therefore, in an embodiment of the present invention, the LSTM time series model processing module in the drone fire fighting gun fire source positioning system based on infrared thermal imaging aims to combine the binary fire source risk feature image (providing the static risk distribution of the fire scene) and the LSTM time series model (capturing the dynamic characteristics of fire spread), and further predict the diffusion direction and speed of the fire spreading from the low-risk area to the medium-risk area and then to the high-risk area, so as to achieve more accurate fire spread simulation and fire extinguishing path planning.

[0037] Specifically, when implemented, the LSTM time series model processing module includes: An image spatio-temporal alignment and fusion unit, which eliminates the inter-frame displacement of the images sampled by the infrared thermal imager carried by the drone through the motion compensation optical flow method, reduces the influence of smoke and dust interference in the fire scene, realizes sub-pixel level alignment, ensures that the N-frame binary fire source risk feature images are accurately corresponding in space, and performs spatio-temporal domain fusion on the N-frame aligned images to form a clear and stable fused image data. , where is the time decay weight, reflecting the contribution degree of the binary fire source risk feature images at different times t to the fusion result. The images closer to the current time t have higher weights. is the Gaussian filter kernel, which is used to effectively suppress random noise and transient interference, make the fused image clearer and more stable, so as to reduce the influence of interference on fire source diffusion detection. I fuesd (x,y) is the pixel value of the fused image data at the position (x,y), λ is the decay coefficient, τ is the time step, and I t (x,y) is the pixel value of the t-th frame image at the position (x,y).

[0038] In an embodiment of the present invention, the LSTM time series model processing module further includes: a data set division unit, which divides the fused image data into a training subset, a test subset and a remaining evaluation (RDA) subset to support the training, verification and performance evaluation of the LSTM time series model. It should be noted that in this embodiment, the divided training subset and test subset at least include the following data related to the processing of binary fire source risk feature images: Image ID information: used to uniquely identify each binary fire risk feature image, facilitating data management and traceability; Spatial location information: including the coordinate position of the fire source Si in the binary fire risk feature image and the size of the fire source (such as area, perimeter), facilitating the determination of the specific location and scale of the fire source Si; Infrared thermal imager parameter information: such as focal length, aperture, exposure time; Environmental information collected based on peripheral sensors: such as environmental temperature, humidity, wind speed; The RDA subset contains the real danger degree evaluation results of the test subset, which are used to compare with the test results of the test subset to evaluate the accuracy of the model.

[0039] In an embodiment of the present invention, the LSTM time series model processing module further includes: an LSTM time series model training unit, which adopts a dual-branch YOLO-Mask architecture to improve the generalization ability and adaptability of the model. In this embodiment, the detection branch uses DarkNet-53 to extract the deep features of the image, outputs the fire source bounding box and its confidence, and the segmentation branch generates a pixel-level mask with the help of the Feature Pyramid Network (FPN) to accurately segment the fire source area, and improves the accuracy of detection and segmentation by optimizing the bounding box loss function and the Dice coefficient loss function respectively.

[0040] A time series modeling unit extracts time series features Tt such as the temperature mean, area change rate, and combustion product distribution of the low-risk area from the binary fire risk feature image, and updates the gated state: , where f t is the output of the forget gate, which determines what information to discard from the cell state C t-1 , C t is the updated cell state, which combines the outputs of the forget gate and the input gate, ot is the output of the output gate, which is used to determine the final output value, h t is the hidden state, which represents the output of the LSTM time series model unit and is used to be passed to the next time step, w f corresponds to the forget gate weight, b f is the forget gate bias vector, σ is the sigmoid function, which is used to limit the output between 0 and 1, is the element-wise multiplication, h t-1 is the hidden state of the previous moment; Output the probability of the fire spreading across risk levels , which represents the possibility of the fire spreading from the low-risk area to the high-risk area. In the formula, W p is used to represent the influence degree of each time series feature in the hidden state h t on the spreading probability, b p is the bias vector, which is used to calculate the spreading probability. When P t exceeds the set threshold (taking 0.65), a warning is triggered to indicate that the fire spreads from the low-risk area to the high-risk area.

[0041] In an embodiment of the present invention, the LSTM time series model processing module further includes: a fire spread prediction output unit, which generates a heat map of fire spread according to the spread probability. In this embodiment, the heat map of fire spread is generated through the following steps: First, based on the spread probability P calculated by the LSTM time series model t , perform spatial distribution modeling on the possibility of fire spread; secondly, use an interpolation algorithm (such as Gaussian interpolation) to smooth the probability distribution to generate a continuous grid of spread probabilities, and then combine them to generate a heat map of fire spread.

[0042] In an embodiment of the present invention, the proposed drone fire fighting and fire source positioning system based on infrared thermal imaging further includes: A combat map generation module electrically connected to the data transmission control module, the purpose of which is to mark high-risk areas (probability P t >0.65) as red and low-risk areas as green or blue based on the fire spread prediction output unit, such as through color mapping, to visually display the fire spread direction and potential risk areas. Use a GIS visualization tool (such as QGIS or ArcGIS) or a Python visualization library (such as Matplotlib or Seaborn) to overlay the heat map of fire spread on the actual geographical coordinates and indicate the spread direction with arrows to form a final visualized combat map of the spread trend. It should be noted that in this embodiment, as Figure 6 、 Figure 8 shown, through the drone flight path, the visualized combat map is realized through the following steps: a. Use the infrared thermal imager carried by the drone to capture thermal images at a speed of 30 frames per second (the drone position is as Figure 9 marked as E), and at the same time bind RTK-GPS centimeter-level positioning information and accurate timestamps to each frame of the image, and update them to the time display area on the system page according to the preset specified format, and call it once every 1 second through the setInterval function to achieve real-time time update; b. In the window.onload event: obtain <canvas>Elements and their 2D drawing context information, and define a series of combat map mapping functions. The combat map mapping functions shall at least include: drawInfraredHeatMap, which is used to construct an infrared diffusion heat map. Specifically, when implemented, assume that the infrared diffusion heat map data is a two-dimensional array, and each element represents the HSV value of a pixel. Then, the HSV value is converted to an RGB value through the hsvToRgb function and filled into a rectangle to represent the pixel. hsvToRgb, which is used to implement the conversion from the HSV color space to the RGB color space, and is used to convert the HSV data of the infrared diffusion heat map into drawable RGB colors. drawLiDARPointCloud, which is used to construct LiDAR point cloud data. Assume that the LiDAR point cloud data is a two-dimensional array, and each element represents the gray value of a point. Then, the point cloud data is drawn by filling rectangles with different gray levels. drawFire, which is used to construct the fire source area, fire fighting guns and their firing ranges, and construct safety channels. Specifically, when implemented, fire source elements are generated according to the incoming fire source position, size and temperature information. And for a fire source with a temperature greater than 800 °C, it is simulated as the display effect shown by the marker B1 in Figure 9 ; for a fire source with a temperature of 600 - 800 °C, it is simulated as the display effect shown by the marker B2 in Figure 9 ; for a fire source with a temperature less than 600 °C, it is simulated as the display effect shown by the marker B3 in Figure 9 . The fire level is rendered in real time to intuitively display the danger level of the fire source; fire fighting gun elements (such as circular icons and dotted circular icons representing the firing range) are generated; according to the incoming start and end coordinates, as well as speed and temperature information, vector arrows representing the spread of the fire are generated, where the arrow length is calculated based on the speed and temperature gradient, and the color is set according to the temperature, following the mapping basis , and a streamer guide for the safety channel is generated (such as the one shown by the marker D in Figure 9 ). Specifically, when implemented, the streamer effect is represented by setting the dotted line style. In the formula, H represents the hue, which is a parameter in the color space used to describe the type of color (such as red, green, blue, etc.). Generally, the range of the hue is usually 0 to 360, corresponding to the angle on the color wheel, and T is the temperature, which is the input temperature value here (in degrees Celsius); by creating a certain number of particles, a particle trajectory for the fire fighting gun's strike path is generated (such as the one shown by the marker A in Figure 9 ), and the particle positions are set to be continuously updated to achieve a dynamic strike trajectory effect.

[0043] c. Call the above series of combat map mapping functions to ensure that the firing range of the fire fighting gun is greater than the fire source area, that is, the strike coverage radius of the fire fighting gun meets 1.5 times the circumradius of the fire source.

[0044] In an embodiment of the present invention, for the drone flight path planning module in the drone fire fighting and fire source positioning system based on infrared thermal imaging, the purpose is to generate an optimal resource scheduling objective function based on the combat map generation module, in combination with the fire area A j and the spread probability P of the fire area t , and use a path planning algorithm (such as the Dijkstra algorithm) to ensure that the drone realizes a safe and efficient flight strike path and achieves full coverage of the fire field area for the strike path. Among them, the set objective function is as follows: , Constraint conditions: , where a ij represents the suppression efficiency or fire extinguishing effect of a unit of resource i on the fire field area j, which is used to reflect the fire extinguishing effects of different resources in different fire areas, and d j =α×A j ×P t , which is used to reflect the actual danger level and resource demand of the fire field area j. α is an adjustment coefficient used to control the comprehensive influence of the fire area and the spread probability on the demand d j , and ρ is a penalty factor used to balance the objective function. C i is the cost coefficient of a unit of resource i, X i is the delivery amount of a unit of resource i, R total is the total resource limit, m is the type or quantity of available resources, and n is the number of fire field areas.

[0045] It should be noted that a ij represents the fire extinguishing effect of using one unit of resource i on the fire field area j. The specific understanding is as follows: Define the resource type: Resource i can be a fire extinguishing agent such as water, foam, dry powder, etc., or a rescue force such as firefighters, fire trucks, drones, etc.; Fire field area characteristics: The characteristics of the fire field area j include flame temperature, fuel type, fire spread speed, etc.; Example: Suppose the fire field area j is a forest fire with a fierce fire and a flame height reaching 5 meters. There are the following two resources available: Resource 1: Water, each liter of water can reduce the flame height by 0.1 meter, then α 1j =0.1 meter / liter; Resource 2: Fire truck, each fire truck can spray 100 liters of water per minute, then α 2j =100×0.1 = 10 meters / minute.

[0046] At the same time, when evaluating the fire field area in the fire fighting field, the parameter d j is used to represent the actual danger level and resource demand of the fire field area j. In specific implementation, A j represents the area of the fire field area j, usually in square meters (m 2 is quantified in units. For example, if the fire area j is a burning area of 100 square meters, then A j = 100 m 2 . Secondly, P t represents the fire spread probability corresponding to the time step t. This is a dimensionless value, and its value range is between 0 and 1. For example, based on a comprehensive evaluation of factors such as the fire spread speed and wind force, it is predicted that within the next hour (time step t), the probability of the fire spreading from area j is 30%. Then P t = 0.3. And α is an adjustment coefficient, which is used to comprehensively consider the influence degree of the fire area and the spread probability on the demand d j . Generally, it is determined according to actual experience or experimental data, and its value is also between 0 and 1. Suppose in a certain fire assessment, α is determined to be 0.5. Substituting these quantified values into the formula, d j can be calculated. Taking the above example, d j = α × A j × P t = 0.5 × 100 m 2 × 0.3 = 15 m 2 . (In this example, for the convenience of illustration, the units are simplified. Specific implementation requires unit conversion or dimensionless processing). This d j = 15 reflects the actual danger degree of the fire area j and the demand for resources. The larger the value, the more dangerous the fire area is, and the more fire-fighting resources need to be invested, such as more firefighters, more fire extinguishing agents, and larger fire-fighting equipment, etc.

[0047] Based on the above technical concept, it can be understood that the cost of the unit resource i is mainly concentrated on the consumables. As shown in Table 1, it is the basic corresponding data table of the unit resource, the delivery amount, and the total resource.

[0048] Table 1

[0049] The technical scope of the present invention is not limited to the content described above. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.< / canvas>

Claims

1. An unmanned aerial vehicle (UAV) fire fighting gun fire source positioning system based on infrared thermal imaging, characterized in that: Including: A fire scene situation acquisition module, which acquires high-precision point cloud data representing the thermal image data of the target fire scene and the spatial information of the location; A fire spread model construction module, based on the point cloud data, quantifies the spatial relationship between the air flow direction of the fire scene and the data acquisition points, and screens out the upstream acquisition points consistent with the fire source spread trend; based on the thermal image data, considering the attenuation effect of the fire source radiation intensity at the upstream acquisition points changing with their distances, calculates the thermal radiation background concentration of each upstream acquisition point, dynamically solves the fire source radiation intensity through gradient descent optimization, fuses the preset parameters of multiple diffusion model libraries, constructs a radiation intensity matrix, generates a radiation intensity distribution map with geographical coordinates, so as to characterize the fire spread path of the target fire scene through the change of radiation intensity, and forms a processable fire source image; A thermal image processing module, performs morphological processing on the fire source image, performs convolution operation through a structural kernel simulating the fluctuation characteristics of the fire source, and extracts the key fire source feature data including the fire source edge and the spread direction in the fire source image through radiation energy gradient analysis, quantifies the fire spread trend, finally selects a pixel point in the key fire source feature data as the analysis starting point, calculates the radiation energy correlation degree between each pixel point and the starting point, conducts a fire spread risk level division, and generates a structured fire source risk feature image through binary mapping according to the risk level, clarifying the fire front position and risk distribution; An LSTM time series model processing module, updates the state using a gating mechanism, calculates the probability of cross-level fire spread based on the fire source risk feature image, and generates a smooth heat diffusion map to predict the spread direction and speed; A combat map generation module, superimposes the heat diffusion map on the actual geographical coordinates to achieve data visualization; An unmanned aerial vehicle flight path planning module, combines the fire area and the spread probability, sets an objective function, and realizes the full coverage of the fire fighting cannon strike path.

2. The drone fire source positioning system based on infrared thermal imaging according to claim 1, characterized in that: The fire spread model construction module includes: A fire scene fluid perception unit, which quantifies the spatial influence of the fire source in the fire scene on the data acquisition points of the point cloud data based on the real-time fire scene air flow data, and the process is as follows: First, for any temperature measurement point i and the point cloud data acquisition point j, calculate the spatial vector projection factor of the two in the fire scene air flow direction, and judge the consistency of their directions; Secondly, screen the area where the point cloud data acquisition points with the same direction and within the preset threshold are located as the upstream fire scene area; Finally, for each point cloud data acquisition point j, a set of temperature measurement points Ω in the upstream fire area is filtered out ij , combined with the attenuation effect of the fire source radiation intensity changing with its distance in the horizontal and vertical directions, the radiation effects of each fire source Si on the upstream acquisition points are dynamically superimposed to generate the background concentration B of the fire heat radiation in this area j : , where is the initial heat radiation intensity of the fire source Si at the point cloud data acquisition point j, is the size of the set of temperature measurement points Ω in the upstream fire area of the point cloud data acquisition point j ij , is the distance between the fire source Si and the point cloud data acquisition point j in the horizontal direction, is the distance between the fire source Si and the point cloud data acquisition point j in the vertical direction, is the attenuation and / or diffusion coefficient of the radiation intensity of the current fire source Si in the horizontal direction of the fire field, is the diffusion coefficient in the vertical direction, and exp is the exponential function.

3. The drone fire fighting fire source positioning system based on infrared thermal imaging according to claim 1 or 2, characterized in that: The fire spread model construction module also includes: Fire source radiation intensity calculation unit, based on the acquired fire heat radiation background concentration B j , through iterative optimization and dynamic calculation, obtains the fire source radiation intensity E of the fire source Si at each point cloud data acquisition point j in the upstream fire field area i , and the process is as follows: First, construct an objective optimization function based on the difference between the minimum predicted thermal radiation concentration and the measured to solve for the optimal fire source radiation intensity : , where is the sparse constraint coefficient, used to suppress noise interference and prevent overfitting, M is the number of temperature measurement points i, and N is the number of point cloud data acquisition points; Secondly, based on the background concentration B of the fire heat radiation at temperature measurement point i j the maximum value, allocate the initial radiation intensity of the heat source Si , and use gradient descent iteration to calculate the minimum predicted heat radiation concentration and the residual , update the radiation intensity E by controlling the step size updated in each iteration i , when is less than the set threshold, terminate the iteration, and output the optimized radiation intensity of all heat sources Si .

4. The drone fire fighting and fire source positioning system based on infrared thermal imaging according to claim 3, characterized in that: The fire spread model construction module also includes: A radiation intensity matrix construction unit, which quantifies the radiation influence value of the fire source on the upstream fire scene area and generates a radiation intensity distribution map representing the fire source image of the fire scene, and the process is as follows: First, combine the convection-diffusion coefficient σ preset in the K diffusion model libraries built by the system d (k) , the number of temperature measurement points M, and the number of point cloud data acquisition points N to define the dimension of the radiation intensity matrix, and then construct the radiation intensity matrix C∈R M×N×K , where the element C ijk of the radiation intensity matrix C represents the radiation influence value of the fire source Si on the point cloud data acquisition point j under the k-th diffusion model library: , in the formula, are the diffusion coefficients of the k-th diffusion model library in the x, y, and z directions respectively, d j , d i are the coordinates of the point cloud data acquisition point j and the fire source Si respectively, d j - d i is the distance difference between the point cloud data acquisition point j and the fire source Si in each direction; Secondly, for each element C of the radiation intensity matrix C ijk , the radiation influence values of K diffusion model libraries are superimposed to form a complete radiation intensity matrix C; Finally, sort all the point cloud data acquisition points according to the comprehensive radiation influence value, select a radiation influence threshold, and sequentially screen out the area of the point cloud data acquisition points whose radiation influence value is greater than the threshold, and generate a radiation intensity distribution map as a whole.

5. The drone fire source positioning system based on infrared thermal imaging according to claim 1, characterized in that: The thermal image processing module includes: A morphological processing unit, which performs morphological processing on the fire source image, and the process is as follows: First, through the radiation intensity matrix C, obtain the radiation influence value and the weight of the area where each pixel (p, q) in the fire source image is located, and calculate the basic radiation influence mean value E of the fire source image in this area mean ; Based on the mean value E of the basic radiation impact mean and the adjustment factor r, set the first threshold T fir , compare the radiation impact value with the first threshold T fir . If the radiation impact value is greater than the first threshold T fir , then mark this area as an active fire source area, otherwise mark it as a non-active fire source area; Secondly, a structural kernel K is defined to simulate the fluctuation characteristics of the fire source Si, and its element value k pq is calculated by the following formula: , where Z is the normalization constant, α is the sine modulation amplitude, λ x and λ x are the sine wavelengths in the x and y directions respectively, σ is the Gaussian attenuation coefficient, m is the number of rows of the structural kernel K, n is the number of columns of the structural kernel K, p is the index of the structural kernel K in the row direction, and q is the index of the structural kernel K in the column direction; Next, use the structural kernel K to perform convolution calculation on the image of the active fire area: , where a and b are the convolution kernel radii respectively, is the pixel value at the corresponding position in the image of the active fire area. Update the pixel value according to the convolution result. If C ij is greater than the set threshold, set the value of this pixel point to 1; otherwise, set it to 0. Finally, by calculating the cumulative value of the radiation energy of each column of the active fire source area image and using the gradient change formula , analyze the column gradient change of the fire source radiation intensity in the active fire source area image, and determine the key features of the edge and diffusion direction of the suspected fire source area in the active fire source area image based on this change, and form the fire source key feature data containing this key feature. In the formula, e ij is the radiation influence value of the pixel at the j-th row and the i-th column, and k is the number of rows.

6. The drone fire fighting and fire source positioning system based on infrared thermal imaging according to claim 1, wherein: The specific process of conducting a fire spread risk level division is as follows: First, calculate the radiation energy change rate of the key feature data of the fire source in terms of longitude, latitude, the main diffusion direction and the secondary diffusion direction related to the fluctuation characteristics of the fire source, and identify the change in the diffusion intensity of the fire source. Secondly, select a pixel point in the key feature data of the fire source as the analysis starting point, and calculate the radiation energy correlation degree p between each pixel point and the starting point. , where w i is the weight coefficient, Ei is the radiation influence value of adjacent pixels, and m is the number of adjacent pixels considered when calculating the radiation energy correlation degree. Finally, take the statistical value of the radiation energy in the high-risk area from the historical fire ground data, and set the second threshold T sec , according to the proportional relationship T mid =k×T sec , determine the intermediate threshold T mid , where k is an empirical coefficient, and If p≥T sec Then it is classified as an active fire core area in the key feature data of the fire source, corresponding to a high radiation intensity and a rapid spread risk level; If p ≤ T sec Then it is classified as a weak combustion or safe area in the key feature data of the fire source, corresponding to a low risk level; If T mid <p<T sec Then it is classified into a stable combustion zone in the key feature data of the fire source, corresponding to the medium risk level of controllable spread.

7. The drone fire-fighting gun fire source positioning system based on infrared thermal imaging according to claim 5, characterized in that: Before performing morphological processing on the fire source image, it is also necessary to calculate the minimum focal length according to the position and range of the selected area to ensure that the area where the fire source is located is always within the imaging range. The process is as follows: Obtain the coordinates of the fire source Si, the point cloud data acquisition point j, and the field of view angle θ of the imaging device FOV , calculate the minimum focal length f min , , where d min is the minimum distance between the fire source Si and the point cloud data acquisition point j, and h max is the maximum height that can be accommodated on the imaging plane of the imaging device, so as to optimize the monitoring accuracy and efficiency of the fire source Si.

8. The drone fire fighting and fire source positioning system based on infrared thermal imaging according to claim 1, characterized in that: The implementation steps of the LSTM time series model processing module are as follows: First, eliminate the inter-frame displacement of the images sampled by the imaging device carried by the drone through the motion compensation optical flow method to ensure the accurate spatial correspondence of the N-frame binary fire risk feature images, and perform spatio-temporal domain fusion on the N-frame aligned images to form the fused image data. , where is the time decay weight, is the Gaussian filter kernel, and I fuesd (x, y) is the pixel value of the fused image at the position (x, y), λ is the attenuation coefficient, τ is the time step, and I t (x, y) is the pixel value of the t-th frame image at the position (x, y); Secondly, divide the fused image data into a training set, a test set and an evaluation set, and construct a time series data set with multi-dimensional attributes. Thirdly, adopt a detection and segmentation double-branch network architecture to synchronously extract the fire source bounding box features and pixel-level masks, and optimize the adaptability of the model to complex fire scenes. Finally, dynamically update the memory unit through the LSTM gating mechanism, and calculate the probability of the fire spreading across risk levels. Generate a diffusion heat map based on the probability distribution, and after interpolation and smoothing, combine it with a threshold to trigger an alarm to generate a continuous diffusion probability grid.

9. The unmanned aerial vehicle fire fighting and fire source positioning system based on infrared thermal imaging according to claim 1, characterized in that: The set target function is the optimal resource scheduling target function. Among them, the target function set based on the UAV flight path planning module is as follows: , constraint conditions: , where a ij represents the suppression efficiency or fire extinguishing effect of unit resource i on fire area j, and d j =α×A j ×P t , which is used to reflect the actual danger degree and resource demand of fire area j. A j is the fire area, P t is the diffusion probability, α is the adjustment coefficient, ρ is the penalty factor, C i is the cost coefficient of unit resource i, X i is the delivery amount of unit resource i, R total is the total resource limit, m is the type or quantity of available resources, and n is the number of fire areas.

10. The drone fire source positioning system based on infrared thermal imaging according to claim 1, characterized in that: Use a GIS visualization tool or a Python visualization library to overlay the diffusion heat map on the actual geographical coordinates, and indicate the diffusion direction with an arrow to form the final diffusion trend and achieve data visualization.

Images