Methods, devices and electronic equipment for fire rescue tactics

Abstract

This application proposes a method, device, and electronic device for acquiring fire rescue tactics. The method includes: acquiring the fire line classification features and terrain classification features of any first grid cell in a set of grid cells corresponding to the fire line; acquiring the feature vector corresponding to any first grid cell based on the first feature vector corresponding to the fire line classification features, the second feature vector corresponding to the terrain classification features, and the third feature vector corresponding to the environmental features of any first grid cell; and analyzing the feature vector corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model to acquire the fire rescue tactics corresponding to any first grid cell. This solves the technical problem in the prior art that fire relief methods cannot adapt to changes, resulting in a mismatch between the fire relief methods and the current fire situation, and poor fire extinguishing effect.

Description

Technical Field

[0001] This application relates to the field of fire protection technology, and in particular to a method, device and electronic equipment for obtaining fire rescue tactics. Background Technology

[0002] With the increasing frequency of forest and urban fires, appropriate fire suppression methods can be employed to extinguish them. These methods may be based on experience or pre-defined rules; however, they cannot adapt to changing circumstances, leading to a mismatch between the methods and the current fire situation and resulting in poor fire suppression effectiveness. Summary of the Invention

[0003] This application aims to at least partially address one of the technical problems in the related art.

[0004] Therefore, the first objective of this application is to propose a method for obtaining fire rescue tactics, so as to improve the accuracy of fire rescue tactic determination, optimize the allocation of fire rescue resources, reduce fire fighting time, and improve fire extinguishing efficiency.

[0005] The second objective of this application is to propose a fire rescue tactical acquisition device.

[0006] The third objective of this application is to propose an electronic device.

[0007] The fourth objective of this application is to provide a computer-readable storage medium.

[0008] The fifth objective of this application is to provide a computer program product.

[0009] To achieve the above objectives, the first aspect of this application proposes a method for obtaining fire rescue tactics, comprising the following steps:

[0010] Obtain the fire line classification features and terrain classification features of any first grid cell in the set of grid cells corresponding to the fire line;

[0011] Based on the first feature vector corresponding to the fire line classification feature, the second feature vector corresponding to the terrain classification feature, and the third feature vector corresponding to the environmental feature of any first grid cell, obtain the feature vector corresponding to any first grid cell;

[0012] The feature vector corresponding to any first grid cell is analyzed using a target graph convolutional network (GCN) model and a target temporal transformer model to obtain the fire rescue tactics corresponding to any first grid cell.

[0013] To achieve the above objectives, a second aspect of this application provides a fire rescue tactical acquisition device, comprising:

[0014] The feature acquisition unit is used to acquire the fire line classification feature and the terrain classification feature of any first grid cell in the set of grid cells corresponding to the fire line;

[0015] The vector acquisition unit is used to acquire the feature vector corresponding to any first grid cell based on the first feature vector corresponding to the fire line classification feature, the second feature vector corresponding to the terrain classification feature, and the third feature vector corresponding to the environmental feature of any first grid cell.

[0016] The tactical acquisition unit is used to analyze the feature vector corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model to obtain the fire rescue tactics corresponding to any first grid cell.

[0017] To achieve the above objectives, a third aspect of this application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0018] The memory stores computer-executed instructions;

[0019] The processor executes computer execution instructions stored in the memory to implement the method as described in any of the first aspects above.

[0020] To achieve the above objectives, a fourth aspect of this application provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any of the first aspects above.

[0021] To achieve the above objectives, a fifth aspect of this application provides a computer program product including a computer program that, when executed by a processor, implements the method described in any of the first aspects above.

[0022] The fire rescue tactics acquisition method, device, and electronic equipment provided in this application acquire fire line classification features and terrain classification features of any first grid cell in a set of grid cells corresponding to the fire line; based on the first feature vector corresponding to the fire line classification features, the second feature vector corresponding to the terrain classification features, and the third feature vector corresponding to the environmental features of any first grid cell, a feature vector corresponding to the first grid cell is obtained; a target graph convolutional neural network model and a target temporal Transformer model are used to analyze the feature vector corresponding to the first grid cell to obtain the fire rescue tactics corresponding to the first grid cell, thus solving the problem that fire rescue methods cannot adapt to changes. This addresses the problem of mismatched fire rescue methods with the current fire situation, resulting in poor firefighting effectiveness. Fire rescue tactics can be determined by considering fire line classification features, terrain classification features, and changing environmental characteristics. It can also consider the interaction between terrain changes and environmental factors during fire spread, allowing for adaptive adjustments to real-time changes in the fire process. Furthermore, by combining graph convolutional neural network models and temporal Transformer models, and integrating spatial and temporal information, fire rescue tactics can be determined. This improves the matching between fire rescue tactics and the current fire situation, eliminating reliance on experience or human rules, thus increasing the accuracy of fire rescue tactic determination. It can also optimize fire rescue resource allocation, reduce firefighting time, and improve firefighting efficiency and effectiveness.

[0023] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0024] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0025] Figure 1 A flowchart illustrating a method for obtaining fire rescue tactics provided in an embodiment of this application;

[0026] Figure 2 A flowchart illustrating a method for obtaining fire rescue tactics provided in an embodiment of this application;

[0027] Figure 3 This is a schematic diagram illustrating an example of a fireline classification feature acquisition method provided in an embodiment of this application.

[0028] Figure 4 A schematic diagram illustrating an example of a neighborhood voting method provided in an embodiment of this application;

[0029] Figure 5This is a schematic diagram illustrating an example of a method for determining firefighting tactics provided in an embodiment of this application.

[0030] as well as Figure 6 This is a schematic diagram of a fire rescue tactical acquisition device provided in an embodiment of this application. Detailed Implementation

[0031] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0032] The following description, with reference to the accompanying drawings, describes a method and apparatus for obtaining fire rescue tactics according to embodiments of this application.

[0033] Figure 1 This is a flowchart illustrating a method for obtaining fire rescue tactics as provided in an embodiment of this application.

[0034] To address this issue, embodiments of this application provide a method for acquiring fire rescue tactics, aiming to improve the accuracy of acquiring fire rescue tactics while reducing the cost of acquiring them. Figure 1 As shown, the method for obtaining fire rescue tactics includes the following steps:

[0035] Step 101: Obtain the fire line classification feature and the terrain classification feature of any first grid cell in the set of grid cells corresponding to the fire line;

[0036] According to some embodiments, the executing entity of the embodiments of this application may be, for example, an electronic device. This electronic device does not specifically refer to a particular fixed device, and its name is not limited. The electronic device may also be referred to as a terminal, a mobile device, etc. For example, when the device identifier of the electronic device changes, the electronic device may also change accordingly.

[0037] In some embodiments, the fire line may refer to, for example, the boundary line of fire spread, i.e., the location of open flames. Different ignition locations or ignition methods may correspond to different fire lines. For example, different ignition times may also correspond to different fire lines. That is, the fire line in this application embodiment does not specifically refer to a fixed fire line. The fire line in this application embodiment may, for example, be used to indicate the fire line corresponding to the currently determined fire rescue tactics.

[0038] According to some embodiments, a grid cell can be used to divide a fireline into fireline regions. The name of this grid cell is not limited. For example, a grid cell can also be called a study area, a fireline region, or a grid cell, etc.

[0039] In some embodiments, a set of grid cells may be a collection of at least one grid cell. This set of grid cells does not specifically refer to a fixed set. For example, the set of grid cells may change when the grid cells it comprises change. Similarly, the set of grid cells may change when the way the grid cells are divided changes.

[0040] According to some embodiments, the first grid cell can be used, for example, to indicate a grid cell that meets a density requirement. The "first" in the first grid cell is used to distinguish it from the remaining grid cells.

[0041] In some embodiments, fire line classification features can be used to indicate which region of the fire line a given grid cell belongs to. These fire line classification features may include, for example, a fire head region, a fire wing region, and a fire tail region. Each grid cell can correspond to one fire line classification feature; that is, a grid cell's fire line classification feature is a single fire line classification feature. For example, if the fire line classification feature of grid cell A is a fire head region, then grid cell A's fire line classification feature is neither a fire wing region nor a fire tail region.

[0042] In some embodiments, terrain classification features can be used to indicate the terrain of a particular grid cell. A single grid cell may correspond to at least one type of terrain, and this embodiment of the application does not limit this.

[0043] According to some embodiments, the fire line classification features and terrain classification features of any first grid cell in the set of grid cells corresponding to the fire line can be obtained.

[0044] The order in which fire line classification features and terrain classification features are acquired is not distinguished. For example, fire line classification features and terrain classification features can be acquired simultaneously, or fire line classification features can be acquired first and then terrain classification features, or terrain classification features can be acquired first and then fire line classification features.

[0045] Step 102: Based on the first feature vector corresponding to the fire line classification feature, the second feature vector corresponding to the terrain classification feature, and the third feature vector corresponding to the environmental feature of any first grid cell, obtain the feature vector corresponding to any first grid cell.

[0046] According to some embodiments, the feature vector may be, for example, a mathematical representation of the model input data. In the embodiments of this application, the feature vector may be used to indicate a mathematical identifier corresponding to the current feature.

[0047] In some embodiments, the first feature vector may be, for example, a feature vector corresponding to the fire classification feature. The "first" in the first feature vector is used to distinguish it from the other feature vectors. This first feature vector does not specifically refer to a fixed vector. For example, when the fire classification feature changes, the first feature vector may also change accordingly.

[0048] According to some embodiments, the second feature vector may be, for example, a feature vector corresponding to the terrain classification features. The "second" in the second feature vector is used to distinguish it from the other feature vectors. This second feature vector does not specifically refer to a fixed vector. For example, when the terrain classification features change, the second feature vector may also change accordingly.

[0049] According to some embodiments, environmental features may be used to indicate environmental characteristics corresponding to the fire line. These environmental features may include, for example, wind direction and wind speed.

[0050] In some embodiments, the third feature vector may be, for example, a feature vector corresponding to environmental features. The "third" in the third feature vector is used to distinguish it from the other feature vectors. This third feature vector does not specifically refer to a fixed vector. For example, when environmental features change, the third feature vector may also change accordingly.

[0051] In some embodiments, the feature vector corresponding to any first grid cell can be used to refer to the comprehensive feature vector corresponding to any first grid cell.

[0052] According to some embodiments, the feature vector corresponding to any first grid cell can be obtained based on the first feature vector corresponding to the fire line classification feature, the second feature vector corresponding to the terrain classification feature, and the third feature vector corresponding to the environmental feature of any first grid cell.

[0053] The order in which the first, second, and third feature vectors are obtained is not limited.

[0054] Step 103: Analyze the feature vector corresponding to any first grid cell using the target graph convolutional neural network model and the target temporal Transformer model to obtain the fire rescue tactics corresponding to any first grid cell.

[0055] In some embodiments, the target graph convolutional neural network model may refer, for example, to a pre-trained graph convolutional neural network model. This target graph convolutional neural network model does not specifically refer to a fixed model. For example, when the parameters corresponding to the target graph convolutional neural network model change, the target graph convolutional neural network model may also change accordingly. For example, when the training dataset corresponding to the target graph convolutional neural network model changes, the target graph convolutional neural network model may also change accordingly.

[0056] According to some embodiments, the target temporal Transformer model may refer to a pre-trained temporal Transformer model. This target temporal Transformer model does not specifically refer to a fixed model. For example, when the parameters of the target temporal Transformer model change, the target temporal Transformer model may also change accordingly. For example, when the training dataset for the target temporal Transformer model changes, the target temporal Transformer model may also change accordingly.

[0057] According to some embodiments, fire rescue tactics can be, for example, the rescue tactics corresponding to the current first grid cell. Different first grid cells can correspond to different fire rescue tactics. Alternatively, when the mapping relationship between fire rescue tactics and feature vectors changes, the fire rescue tactics corresponding to the first grid cell can also change accordingly.

[0058] According to some embodiments, a target graph convolutional neural network model and a target temporal sequence are employed.

[0059] The Transformer model analyzes the feature vector corresponding to any first grid cell to obtain the fire rescue tactics corresponding to any first grid cell.

[0060] The fire rescue tactics acquisition method, device, and electronic equipment provided in this application acquire fire line classification features and terrain classification features of any first grid cell in a set of grid cells corresponding to the fire line; based on the first feature vector corresponding to the fire line classification features, the second feature vector corresponding to the terrain classification features, and the third feature vector corresponding to the environmental features of any first grid cell, a feature vector corresponding to the first grid cell is obtained; a target graph convolutional neural network model and a target temporal Transformer model are used to analyze the feature vector corresponding to the first grid cell to obtain the fire rescue tactics corresponding to the first grid cell, thus solving the problem that fire rescue methods cannot be implemented. The problem of fire rescue methods being mismatched with the current fire situation and resulting in poor firefighting effectiveness can be addressed by adapting to changes in fire line classification features, terrain classification features, and changing environmental features. This allows for the determination of fire rescue tactics by considering the interaction between terrain changes and environmental factors during fire spread, enabling adaptive adjustments to real-time changes in the fire process. Furthermore, by combining graph convolutional neural network models and temporal Transformer models to determine fire rescue tactics, this approach not only considers the static characteristics of the fire scene but also dynamically responds to changes during fire spread. This improves the matching between fire rescue tactics and the current fire situation, increases the accuracy of fire rescue tactic determination, optimizes fire rescue resource allocation, reduces firefighting time, and enhances firefighting efficiency.

[0061] This embodiment provides another method for obtaining fire rescue tactics. Figure 2 This is a flowchart illustrating a method for obtaining fire rescue tactics as provided in an embodiment of this application.

[0062] like Figure 2 As shown, this method for obtaining fire rescue tactics may include the following steps:

[0063] Step 201: Obtain the fireline classification feature of any first grid cell in the set of grid cells corresponding to the fireline;

[0064] The specific process is as described above and will not be repeated here.

[0065] In some embodiments, the technical solutions of this application can be applied to the field of fire protection, especially in the fields of fire rescue tactics and methods matching and fire zone classification.

[0066] According to some embodiments, Figure 3 This is an example illustration of a method for obtaining fireline classification features provided in an embodiment of this application. The step of obtaining the fireline classification feature of any first grid cell in the set of grid cells corresponding to the fireline includes:

[0067] The fire line is identified using a cellular automata model, and the fire line spread data of any first grid cell in the set of grid cells corresponding to the fire line is obtained. The fire line spread data includes the spread speed and the spread direction.

[0068] Obtain the angle between the spread direction and the current environmental wind direction, and obtain the angle weight;

[0069] Based on the propagation speed and the included angle weight, obtain the feature scoring matrix corresponding to any first grid cell;

[0070] Based on the feature scoring matrix of each grid cell in the set of grid cells and the adaptive density clustering algorithm, the fire line classification features of any first grid cell are obtained. Therefore, fire line classification features can be obtained based on environmental information, and the adaptive density clustering algorithm can be used to obtain fire line classification features, which can reduce rule-based rigid classification, flexibly identify dynamic changes in the fire line, and improve classification accuracy and adaptability.

[0071] According to some embodiments, the adaptive density clustering algorithm is used to classify fire-line areas, which can automatically cope with the diversity and complexity of fire-line areas and enhance the adaptability to different fire scenarios.

[0072] According to some embodiments, the cellular automata model does not specifically refer to a particular fixed model. For example, the cellular automata model may also be called a forest fire spread model or a fire line identification model, etc.

[0073] In some embodiments, the propagation direction may be, for example, the propagation direction corresponding to the next time step. The first grid cell may, for example, be a grid cell that is currently igniting.

[0074] In some embodiments, the propagation direction of each first grid cell in the ignition state can be matched with the current environmental wind direction, and the angle θ between the propagation direction and the wind direction can be calculated using formula (1):

[0075]

[0076] Where d is the propagation direction vector, w is the wind direction vector, · represents the dot product of vectors, and |d| and |w| represent the magnitudes of the vectors, respectively.

[0077] According to some embodiments, the angle weight can be calculated using formula (2) based on a set angle weight factor. Specifically, this can be achieved by transforming the angle relationship into a smooth weight, reducing the model's sensitivity to different fire environment conditions and enhancing its adaptability. The angle weight W is calculated using the Sigmoid function. θ :

[0078]

[0079] Where θ is the angle between the spread direction and the wind direction, θ0 is the angle with the highest weight, and k is a constant that controls the rate of change of the weight.

[0080] The Sigmoid function can prevent the weight factors from fluctuating drastically when the angle changes, and convert the angle relationship into a smooth weight value, so that the change is slow at small angles and fast at large angles. By adjusting the constants θ0 and k, it can adapt to different fire environment and improve the flexibility of the model.

[0081] According to some embodiments, the propagation velocity V and the included angle weighting factor W can be combined. θ The comprehensive score S for each grid cell is calculated using formula (3). The smoothed feature score matrix is ​​then obtained and used to determine the fire classification features.

[0082] S = V·W θ (3)

[0083] According to some embodiments, an adaptive density clustering method based on the feature score matrix can automatically merge adjacent similar grid cells into firehead, fire wing, and fire tail regions. By applying the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, density clustering can be performed on the smoothed feature scores to automatically identify different types of connected regions.

[0084] According to some embodiments, where a grid cell can be, for example, a data point, the following two key parameters can be set:

[0085] Neighborhood radius ∈: Defines the neighborhood range of a data point, that is, points within this radius are considered to be nearby.

[0086] Minimum number of samples MinPts: Specifies the minimum number of neighboring points required to form a density cluster.

[0087] The selection of these two key parameters can be optimized based on the actual data distribution of the feature scoring matrix through cross-validation.

[0088] According to some embodiments, isolated points can be processed, specifically by DBSCAN marking those points that do not meet the density requirements as noise points (isolated points) and analyzing those noise points.

[0089] According to some implementations, the output of the adaptive density clustering model is as follows: the DBSCAN algorithm generates a set of connected regions, each region corresponding to a cluster, and belonging to a single category (firehead, fire wing, fire tail). That is, a grid cell belongs to a single category, and a grid cell can, for example, correspond to a cluster.

[0090] According to some embodiments, the method further includes:

[0091] If it is determined that there is a subset of grid cells in the set of grid cells that does not meet the density requirement, the neighborhood range corresponding to any second grid cell in the subset of grid cells is obtained;

[0092] Obtain the number of grid cells corresponding to any fire classification feature within the neighborhood;

[0093] If the amount of data in the grid cell is less than a first quantity threshold, at least one grid cell within the neighborhood range is ignored;

[0094] When the amount of data in a grid cell exceeds a second threshold, fire classification features corresponding to each grid cell within at least one grid cell in the neighborhood are obtained. Based on these fire classification features, a majority voting principle is used to obtain the fire classification features corresponding to any second grid cell, where the second threshold is greater than or equal to the first threshold. Therefore, a few misclassified or isolated noise points marked by the DBSCAN algorithm can be checked and adjusted, ensuring classification consistency within the region and improving the reliability of the overall classification results. Clear category labels can be obtained for each grid cell, representing different parts of the fire line and improving the accuracy of fire rescue tactics determination.

[0095] According to some embodiments, the second grid cell may be, for example, a grid cell from a subset of grid cells that do not meet the density requirement. Wherein, when a grid cell is referred to as a data point, the second grid cell may be referred to as an isolated point, for example.

[0096] In some embodiments, Figure 4 This is an example illustration of a neighborhood voting method provided in an embodiment of this application. Specifically, it may include: determining a neighborhood range, wherein the neighborhood range may, for example, have the same neighborhood radius as the DBSCAN algorithm. This neighborhood range does not specifically refer to a fixed range. For example, when a modification instruction for the neighborhood range is received, the neighborhood range may also change accordingly.

[0097] According to some embodiments, the number of grid cells corresponding to any fire line classification feature within the neighborhood can be obtained. For example, it can be to count the number of data points of each category within the neighborhood of the isolated point. Specifically, it can be to count the number of data points of the fire head, the number of data points of the fire wing, and the number of data points of the fire tail within the neighborhood of the isolated point.

[0098] According to some embodiments, the majority voting principle can be used based on statistical results to classify isolated points into the category with the largest number of points in the neighborhood, and the category label of the isolated point can be updated to the label corresponding to the voting result.

[0099] According to some embodiments, the processing strategy can be determined based on the number of grid cells corresponding to any fire classification feature. For example, if the number of grid cells is small and has little impact on the overall classification result, these points can be ignored, and their noise labels retained. If the number of grid cells is large and may affect the continuity and accuracy of the classification results, a neighborhood voting method should be used to reclassify them.

[0100] Step 202: Obtain the terrain parameters of any grid cell using a Digital Elevation Model (DEM);

[0101] According to some embodiments, a digital elevation model (DEM) can, for example, acquire DEM data of a study area for calculating terrain parameters.

[0102] In some embodiments, topographic parameters may include, for example, slope, aspect, curvature, flow direction, flow accumulation, and wind direction. Wherein:

[0103] Slope and aspect: The slope and aspect of each grid cell can be calculated using DEM data.

[0104] Curvature: Calculates the curvature of the terrain surface, including profile curvature and planar curvature, reflecting the concave and convex characteristics of the terrain.

[0105] Flow direction and flow accumulation: Flow direction and flow accumulation are calculated based on DEM data to identify valley regions.

[0106] Wind direction data: Obtain wind direction data at different times to serve as auxiliary information for terrain classification.

[0107] According to some embodiments, the terrain parameters of any grid cell can be obtained using a digital elevation model (DEM).

[0108] Step 203: Determine the order based on the terrain parameters and terrain features, and obtain the terrain classification features of any first grid cell;

[0109] According to some embodiments, the terrain feature determination order can be used, for example, to indicate a pre-set terrain feature determination order. The terrain feature determination order is not specifically defined by a fixed order. For example, when the application scenario of this application embodiment changes, the terrain feature determination order can also change accordingly.

[0110] In some embodiments, the order of determining terrain features may be, for example, ridge identification, valley identification, windward and leeward slope identification, and the order of determining flat terrain. Wherein:

[0111] 1) Ridge identification:

[0112] Ridges are typically characterized by regions with high elevation values ​​and positive curvature. To ensure classification accuracy, a dynamic threshold model is established by calculating local maxima and standard deviations. The local maximum elevation z is defined as... max and local elevation standard deviation σ z Calculate the threshold Δz for elevation difference. threshold for:

[0113] Δz threshold =k×σ z (4)

[0114] Here, k is an empirical coefficient used to adjust the sensitivity of elevation differences. A ridge is identified when the elevation value meets the following conditions:

[0115] z≥z max -Δz threshold And curvature > C positive

[0116] Among them, C positive The threshold value for positive curvature.

[0117] 2) Valley identification:

[0118] Valleys typically have low elevation values ​​and negative curvature, while also exhibiting high flow accumulation values. To eliminate the influence of noise, a threshold A for flow accumulation is used. threshold The determination of a valley is made by combining elevation and curvature.

[0119] And curvature <C negative And A>A threshold

[0120] Among them, C negative The threshold for negative curvature.

[0121] 3) Windward and leeward slopes:

[0122] The determination of windward, leeward, and crosswind slopes is based on the angular difference between the slope aspect and the wind direction. The specific process is as follows:

[0123] Calculation of slope aspect and wind direction difference:

[0124] The angle difference Δθ between the slope aspect and the wind direction is calculated using formula (5):

[0125] Δθ=|θ aspect -θ wind |(5)

[0126] Where, θ aspect It's the slope aspect, θ wind It refers to the wind direction. By setting an angle threshold, the windward and leeward slopes can be determined separately.

[0127] 4) Flat terrain:

[0128] Flat terrain is limited by slope and runoff accumulation:

[0129] S threshold And A small

[0130] It is then determined to be flat terrain.

[0131] Step 204: Based on the first feature vector corresponding to the fire line classification feature, the second feature vector corresponding to the terrain classification feature, and the third feature vector corresponding to the environmental feature of any first grid cell, obtain the feature vector corresponding to any first grid cell.

[0132] The specific process is as described above and will not be repeated here.

[0133] According to some embodiments, the grid cell can be called a node, the first feature vector is the fire line classification feature vector xifire: each fire line region is defined as a node vi, and the fire line features are the fire head, fire wings and fire tail obtained by the above extraction and classification. After generating classification labels, they are used as feature inputs.

[0134] According to some embodiments, the second feature vector, namely the terrain classification feature vector xiterrain, includes basic data on slope and aspect, as well as terrain classification features such as ridges, valleys, windward slopes, leeward slopes, and flat terrain. Since the model's research unit is the fireline area, the feature vector xiterrain contains the probability distribution of terrain classifications within the fireline area.

[0135]

[0136] According to some embodiments, the third feature vector, namely the environmental feature vector xienv, includes external environmental factors such as wind speed, wind direction, and temperature, further enriching the feature information after classification.

[0137] ​​The eigenvector corresponding to any first grid cell, i.e., the comprehensive eigenvector of each node i, is represented as:

[0138] Step 205: Analyze the feature vector corresponding to any first grid cell using the target graph convolutional neural network model and the target temporal Transformer model to obtain the fire rescue tactics corresponding to any first grid cell.

[0139] The specific process is as described above and will not be repeated here.

[0140] According to some embodiments, Figure 5 This is an example illustration of a method for determining fire rescue tactics provided in an embodiment of this application. The step of analyzing the feature vector corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model to obtain the fire rescue tactic corresponding to any first grid cell includes:

[0141] A target graph convolutional neural network model is used to propagate the feature vector corresponding to any first grid cell forward to obtain the fire space features at each time step;

[0142] The target temporal Transformer model is used to process the fire spatial features at each time step to obtain the location features at each time step;

[0143] Based on the fire space features and the location features of each time step, obtain the fire representation corresponding to any first grid cell;

[0144] Based on the mapping relationship between fire scene representation and fire rescue tactics, the fire rescue tactics corresponding to the fire scene representation are obtained, and these tactics are used as the fire rescue tactics corresponding to any first grid cell. Therefore, a combination of graph convolutional neural network (GCN) and temporal Transformer (TTR) models can be used to determine fire rescue tactics, improving the accuracy of fire rescue demonstration. Specifically, GCN effectively extracts the spatial features of the fire line area and captures the complex relationship between the fire line and the terrain, while TTR models the temporal changes in fire spread through a self-attention mechanism, thus achieving accurate modeling of dynamic changes during fire spread. This deep learning framework not only improves the accuracy and flexibility of tactical selection but also automatically selects the most suitable rescue strategy in complex fire environments, significantly improving firefighting efficiency.

[0145] In some embodiments, the fire rescue tactics corresponding to the fire scene representation are obtained according to the mapping relationship between the fire scene representation and the fire rescue tactics. For example, the probability of each fire rescue tactic in at least one fire rescue tactic corresponding to the fire scene representation can be obtained, and the fire rescue tactic with the highest probability can be taken as the fire rescue tactic corresponding to any first grid cell.

[0146] According to some embodiments, the step of employing a target graph convolutional neural network model to propagate the feature vector corresponding to any first grid cell forward to obtain the fire scene spatial features at each time step includes:

[0147] Based on the spatial proximity relationship between the grid cells, obtain the adjacency matrix corresponding to any first grid cell;

[0148] Based on the weight matrix, activation function, and adjacency matrix of the target graph convolutional neural network model, the fire spatial features at each time step are obtained. Therefore, the fire spatial features can be determined using the weight matrix, activation function, and adjacency matrix, improving the accuracy of fire spatial feature determination and thus the accuracy of fire rescue tactics.

[0149] According to some embodiments, the adjacency matrix A of the graph is constructed using the spatial proximity relationships between nodes, where:

[0150]

[0151] According to some embodiments, the method further includes:

[0152] Obtain a fire training dataset, wherein the fire training dataset includes the fire line classification of each historical grid cell in the historical grid cell set corresponding to the historical fire line;

[0153] The initial graph convolutional neural network model and the initial temporal Transformer model are trained using the fire training dataset to obtain the historical fire rescue tactics corresponding to each historical grid cell.

[0154] If the historical fire rescue tactics corresponding to each historical grid cell meet the tactical requirements, the target graph convolutional neural network model and the target temporal Transformer model are obtained.

[0155] According to some implementations, the optimal tactic label yi corresponding to each node is a label set {1,2,…,C}, where C is the number of tactic types. The optimal tactic is derived from the rescue and handling of historical fire cases.

[0156] In some embodiments, the training process for the model may include, for example:

[0157] 1. Input feature vector construction:

[0158] Historical fire line classification feature vector xifire: Each fire line region is defined as a node vi. The fire line features are the fire head, fire wings, and fire tail extracted and classified above. After generating classification labels, these are used as feature inputs.

[0159] Historical terrain classification feature vector xiterrain: This feature vector includes basic data on slope and aspect, as well as classification results for ridges, valleys, windward slopes, leeward slopes, and flat terrain generated from terrain classification labels. Since the model's research unit is the fireline area, the feature vector xiterrain contains the probability distribution of terrain classifications within that fireline area.

[0160] Historical environmental feature vector xienv: includes external environmental factors such as wind speed, wind direction, and temperature, further enriching the feature information after classification.

[0161] 2. Eigenvector representation:

[0162] The comprehensive historical feature vector of each node i is represented as:

[0163] 3. Tag Data:

[0164] The optimal tactic label yi corresponds to each node, and the label set is {1,2,…,C}, where C is the number of tactic types. The optimal tactic is derived from the rescue and handling of historical fire cases.

[0165] 4. Adjacency matrix construction:

[0166] Using the spatial proximity relationships between nodes, construct the adjacency matrix A of the graph, where:

[0167]

[0168] 5. GCN layer:

[0169] For the node features H(l) of the l-th layer, the forward propagation formula of GCN is (6):

[0170]

[0171] in: It is an adjacency matrix with self-loops added. yes The degree matrix,

[0172] W (l) σ is the weight matrix of the l-th layer. σ(·) is the activation function.

[0173] The initial input is H. (0) = [x1,x2,…,x N ] T .

[0174] 2) Temporal dynamic modeling: Transformer model, which may also be called Transformer encoder, and may include multi-head attention mechanism, residual connection and normalization processing, and feedforward neural network, etc.

[0175] During fire rescue operations, the changes in the fire scene are temporal, meaning the fire intensity changes over time and is influenced by various factors (such as wind direction and terrain). To effectively capture these temporal dynamics, the model introduces the Transformer architecture, particularly the self-attention mechanism, to handle the temporal features in the fire scene data.

[0176] 1. Input Feature Sequence: The output of the GCN layer provides the spatial features of the fire scene at each time step, and these spatial feature sequences become the input of the Transformer model.

[0177] 2. Location Encoding: To enable the model to process temporal information in time-series data, we add location encoding to the input spatial features to help the model distinguish features at different time steps. The location encoding formula (7) is as follows:

[0178]

[0179] Where t is the time step, and d model 'i' represents the dimension of the model, and 'i' is the dimension index.

[0180] 3. Attention Mechanism: The self-attention mechanism captures long-range dependencies in fire data by calculating the relationship between the features of each location and the features of all other locations.

[0181] For the input sequence X = [x1, x2, ..., x...] N ] T Self-attention is calculated as follows:

[0182] Linear change: Q = XW Q K = XW K V = XW V .

[0183] Among them, W Q W K W V It is a learnable parameter matrix.

[0184] Attention weight calculation:

[0185] Where, d k It is the dimension of the key vector.

[0186] Bullish Attention:

[0187] By computing multiple attention heads in parallel, and then concatenating and linearly transforming the results:

[0188] MultiHead(Q,K,V)=Concat(head1,…,head h W O

[0189] Each attention head i yes:

[0190] 3) Model fusion and output:

[0191] After extracting spatial features using GCN, the Transformer model further processes temporal information, combining spatial features and temporal dynamics to generate a complete fire scene representation. This representation is then mapped to the final tactical category through a fully connected layer.

[0192] Final output: The model output is a probability distribution representing the appropriate rescue tactics for the current fire situation. The output layer uses the softmax activation function for classification, and the output categories include "direct attack," "indirect attack," etc.

[0193] y = softmax(H final W O )

[0194] Among them, H final W is the final feature representation after processing by the Transformer. O It is the weight matrix of the output layer.

[0195] According to some embodiments, during model training, for example, a cross-entropy loss function such as Equation (8) can be used to optimize the prediction results:

[0196]

[0197] Among them, y i It's a real tag, p i It is the probability of the battle tactics predicted by the model.

[0198] Evaluation metrics: Accuracy, recall, F1 score and other metrics are used to evaluate model performance, and the generalization ability of the model is optimized through cross-validation and hyperparameter tuning.

[0199] In some embodiments, the historical fire rescue tactics corresponding to each historical grid cell meet the tactical requirements. For example, based on the optimal tactical label yi and label set corresponding to each node, if the consistency between the fire rescue tactics output by the model and the optimal tactical label is greater than the consistency threshold, it can be determined that the trained model can be applied to the determination of actual fire rescue tactics.

[0200] In some embodiments,

[0201] Firefighting tactics may include at least one of the following:

[0202] 1) Direct attack: Extinguish the flames directly at the front of the fire line.

[0203] 2) Indirect attack: Prevent the fire from spreading by setting up control lines and firebreaks.

[0204] 3) Firebreak: Establish a firebreak in the direction of fire advance to control the spread of the fire.

[0205] 4) Breakthrough at one point, advance to both flanks: Concentrate forces to break through one point of the fire line, and then advance to both flanks. Suitable for areas where the fire line is narrow.

[0206] 5) Flanking attack: Advance from both flanks of the fire line towards the fire front simultaneously, forming a pincer movement. Suitable for wide fire lines and large fire front areas.

[0207] 6) Multi-point breakthrough and segmented encirclement and annihilation: Simultaneous firefighting operations are launched at multiple points along the fire line to extinguish the fire in segments. This method is suitable for fires that are dispersed and have long fire lines.

[0208] 7) Flanking maneuvers and advances: Bypassing weak points in the fire line to fight the fire from the rear or flanks. This method is suitable when the fire is intense on the front of the fire line, the wind direction is unfavorable, the terrain is complex, and the weak points in the fire line are obvious.

[0209] 8) Using existing fires to fight fire: Utilize already burned areas to control the direction of new fires, creating a fire-on-fire situation. This is suitable for areas where a large area has already been burned, creating clear open spaces or low-fuel zones, where the terrain is favorable for fire-on-fire operations, and where the wind direction is favorable.

[0210] 9) Pre-set isolation zones to prevent forest fires: Pre-set isolation zones in the direction of fire spread to prevent the fire from continuing to spread. This is suitable when the protected target is in the fire path and there is sufficient time to set up the isolation zone before the fire arrives.

[0211] 10) Ground-air coordinated, three-dimensional firefighting: Ground and air forces work together to achieve all-round firefighting. Suitable for areas where ground forces cannot quickly reach the core of the fire.

[0212] 11) Do not fight uphill fires (corresponding to the slope); fight them from the rear or flanks: For areas with steep slopes, avoid frontal firefighting and fight them from the flanks or rear. This is suitable for fire areas with steep slopes and relatively gentle terrain on the flanks or rear.

[0213] In one or related embodiments, a digital elevation model (DEM) is used to obtain the terrain parameters of any grid cell; the terrain classification features of any first grid cell are obtained according to the terrain parameters and terrain features in a determined order. Therefore, the terrain classification features can be determined according to the terrain parameters and terrain features in a determined order, which can reduce the determination time of terrain classification features, improve the accuracy and efficiency of terrain classification feature determination, and improve the accuracy of front-line rescue tactics determination.

[0214] To achieve the above embodiments, this application also proposes a fire rescue tactical acquisition device.

[0215] Figure 6 This is a schematic diagram of a fire rescue tactical acquisition device provided in an embodiment of this application.

[0216] like Figure 6 As shown, the fire rescue tactical acquisition device includes:

[0217] The feature acquisition unit 601 is used to acquire the fire line classification feature and the terrain classification feature of any first grid cell in the set of grid cells corresponding to the fire line;

[0218] The vector acquisition unit 602 is used to acquire the feature vector corresponding to any first grid cell based on the first feature vector corresponding to the fire line classification feature, the second feature vector corresponding to the terrain classification feature, and the third feature vector corresponding to the environmental feature of any first grid cell.

[0219] The tactical acquisition unit 603 is used to analyze the feature vector corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model to obtain the fire rescue tactics corresponding to any first grid cell.

[0220] Furthermore, in one possible implementation of this application embodiment, when the feature acquisition unit 601 is used to acquire the fireline classification feature of any first grid cell in the grid cell set corresponding to the fireline, it is specifically used for:

[0221] The fire line is identified using a cellular automata model, and the fire line spread data of any first grid cell in the set of grid cells corresponding to the fire line is obtained. The fire line spread data includes the spread speed and the spread direction.

[0222] Obtain the angle between the spread direction and the current environmental wind direction, and obtain the angle weight;

[0223] Based on the propagation speed and the included angle weight, obtain the feature scoring matrix corresponding to any first grid cell;

[0224] Based on the feature scoring matrix of each grid cell in the set of grid cells and the adaptive density clustering algorithm, the fire classification features of any first grid cell are obtained.

[0225] Furthermore, in one possible implementation of this application embodiment, the feature acquisition unit 601 is further configured to:

[0226] If it is determined that there is a subset of grid cells in the set of grid cells that does not meet the density requirement, the neighborhood range corresponding to any second grid cell in the subset of grid cells is obtained;

[0227] Obtain the number of grid cells corresponding to any fire classification feature within the neighborhood;

[0228] If the amount of data in the grid cell is less than a first quantity threshold, at least one grid cell within the neighborhood range is ignored;

[0229] When the amount of data in the grid cell is greater than the second quantity threshold, fire classification features corresponding to each grid cell in at least one grid cell within the neighborhood are obtained, and fire classification features corresponding to any second grid cell are obtained by majority voting based on the fire classification features corresponding to each grid cell, wherein the second quantity threshold is greater than or equal to the first quantity threshold.

[0230] Furthermore, in one possible implementation of this application embodiment, when the feature acquisition unit 601 is used to acquire the terrain classification features of any first grid cell, it is specifically used for:

[0231] The terrain parameters of any grid cell are obtained using a digital elevation model (DEM).

[0232] Based on the terrain parameters and terrain features, the order is determined to obtain the terrain classification features of any first grid cell.

[0233] Furthermore, in one possible implementation of this application embodiment, the tactical acquisition unit 603, when analyzing the feature vector corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model to acquire the fire rescue tactic corresponding to any first grid cell, is specifically used for:

[0234] A target graph convolutional neural network model is used to propagate the feature vector corresponding to any first grid cell forward to obtain the fire space features at each time step;

[0235] The target temporal Transformer model is used to process the fire spatial features at each time step to obtain the location features at each time step;

[0236] Based on the fire space features and the location features of each time step, obtain the fire representation corresponding to any first grid cell;

[0237] Based on the mapping relationship between fire scene representation and fire scene rescue tactics, the fire scene rescue tactics corresponding to the fire scene representation are obtained, and the fire scene rescue tactics are used as the fire scene rescue tactics corresponding to any first grid unit.

[0238] Furthermore, in one possible implementation of this application embodiment, the tactical acquisition unit 603, when using a target graph convolutional neural network model to propagate the feature vector corresponding to any first grid cell forward to acquire the fire scene spatial features at each time step, is specifically used for:

[0239] Based on the spatial proximity relationship between the grid cells, obtain the adjacency matrix corresponding to any first grid cell;

[0240] Based on the weight matrix, activation function, and adjacency matrix of the target graph convolutional neural network model, the fire spatial features at each time step are obtained.

[0241] Furthermore, in one possible implementation of this application embodiment, the tactical acquisition unit 603 is further configured to:

[0242] Obtain a fire training dataset, wherein the fire training dataset includes the fire line classification of each historical grid cell in the historical grid cell set corresponding to the historical fire line;

[0243] The initial graph convolutional neural network model and the initial temporal Transformer model are trained using the fire training dataset to obtain the historical fire rescue tactics corresponding to each historical grid cell.

[0244] If the historical fire rescue tactics corresponding to each historical grid cell meet the tactical requirements, the target graph convolutional neural network model and the target temporal Transformer model are obtained.

[0245] It should be noted that the foregoing explanation of the embodiment of the fire rescue tactics acquisition method also applies to the fire rescue tactics acquisition device of this embodiment, and will not be repeated here.

[0246] The fire rescue tactic acquisition device provided in this application includes: a feature acquisition unit for acquiring fire line classification features and terrain classification features of any first grid cell in a set of grid cells corresponding to the fire line; a vector acquisition unit for acquiring feature vectors corresponding to any first grid cell based on a first feature vector corresponding to the fire line classification features, a second feature vector corresponding to the terrain classification features, and a third feature vector corresponding to the environmental features of any first grid cell; and a tactic acquisition unit for analyzing the feature vectors corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model to acquire the fire rescue tactics corresponding to any first grid cell. This approach addresses the problem of fire remediation methods failing to adapt to changing circumstances, resulting in mismatches between methods and the current fire situation and poor firefighting effectiveness. By leveraging fire line classification features, terrain classification features, and changing environmental characteristics to determine fire rescue tactics, and considering the interaction between terrain changes and environmental factors during fire spread, it allows for adaptive adjustments to real-time fire events. Furthermore, by combining graph convolutional neural network models and temporal Transformer models to determine fire rescue tactics, and by integrating spatial and temporal information, it improves the matching between fire rescue tactics and the current fire situation, enhances the accuracy of tactic determination, optimizes fire rescue resource allocation, reduces firefighting time, and increases firefighting efficiency.

[0247] To implement the above embodiments, this application also proposes an electronic device, including: a processor and a memory communicatively connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the method provided in the foregoing embodiments.

[0248] To implement the above embodiments, this application also proposes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the methods provided in the foregoing embodiments.

[0249] To implement the above embodiments, this application also proposes a computer program product, including a computer program that, when executed by a processor, implements the methods provided in the foregoing embodiments.

[0250] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0251] It should be noted that personal information collected from users should be used for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. Furthermore, such collection / sharing should only be conducted after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign an agreement / authorization that includes authorization of relevant user information before the user uses the function. In addition, any necessary steps must be taken to protect and safeguard access to such personal information data and ensure that others with access to personal information data comply with their privacy policies and procedures.

[0252] This application is intended to provide an implementation scheme for users to selectively prevent the use or access to their personal information data. Specifically, this application is intended to provide hardware and / or software to prevent or block access to such personal information data. Once personal information data is no longer needed, risks can be minimized by restricting data collection and deleting data. Furthermore, where applicable, such personal information is de-identified to protect user privacy.

[0253] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0254] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0255] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0256] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0257] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0258] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0259] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0260] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for acquiring fire rescue tactics, characterized in that, include: Obtain the fire line classification features and terrain classification features of any first grid cell in the set of grid cells corresponding to the fire line; Based on the first feature vector corresponding to the fire line classification feature, the second feature vector corresponding to the terrain classification feature, and the third feature vector corresponding to the environmental feature of any first grid cell, the feature vector corresponding to any first grid cell is obtained, wherein the fire line classification feature includes the fire head region, the fire wing region, or the fire tail region. The feature vector corresponding to any first grid cell is analyzed using a target graph convolutional neural network model and a target temporal Transformer model to obtain the fire rescue tactics corresponding to any first grid cell; The step of analyzing the feature vector corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model to obtain the fire rescue tactics corresponding to any first grid cell includes: A target graph convolutional neural network model is used to propagate the feature vector corresponding to any first grid cell forward to obtain the fire space features at each time step; The target temporal Transformer model is used to process the fire spatial features at each time step to obtain the location features at each time step; Based on the fire space features and location features of each time step, obtain the fire representation corresponding to any first grid cell; Based on the mapping relationship between fire scene representation and fire scene rescue tactics, the fire scene rescue tactics corresponding to the fire scene representation are obtained, and the fire scene rescue tactics are used as the fire scene rescue tactics corresponding to any first grid unit. The step of employing a target graph convolutional neural network model to propagate the feature vector corresponding to any first grid cell forward to obtain the fire scene spatial features at each time step includes: Based on the spatial proximity relationship between the grid cells, obtain the adjacency matrix corresponding to any first grid cell; Based on the weight matrix, activation function, and adjacency matrix of the target graph convolutional neural network model, the spatial features of the fire scene at each time step are obtained.

2. The method according to claim 1, characterized in that, The step of obtaining the fireline classification feature of any first grid cell in the set of grid cells corresponding to the fireline includes: The fire line is identified using a cellular automata model, and the fire line spread data of any first grid cell in the set of grid cells corresponding to the fire line is obtained. The fire line spread data includes the spread speed and the spread direction. Obtain the angle between the spread direction and the current environmental wind direction, and obtain the angle weight; Based on the propagation speed and the included angle weight, obtain the feature scoring matrix corresponding to any first grid cell; Based on the feature scoring matrix of each grid cell in the set of grid cells and the adaptive density clustering algorithm, the fire classification features of any first grid cell are obtained.

3. The method according to claim 2, characterized in that, The method further includes: If it is determined that there is a subset of grid cells in the set of grid cells that does not meet the density requirement, the neighborhood range corresponding to any second grid cell in the subset of grid cells is obtained; Obtain the number of grid cells corresponding to any fire classification feature within the neighborhood; If the amount of data in the grid cell is less than a first quantity threshold, at least one grid cell within the neighborhood range is ignored; When the amount of data in the grid cell is greater than the second quantity threshold, fire classification features corresponding to each grid cell in at least one grid cell within the neighborhood are obtained, and fire classification features corresponding to any second grid cell are obtained by majority voting based on the fire classification features corresponding to each grid cell, wherein the second quantity threshold is greater than or equal to the first quantity threshold.

4. The method according to claim 1, characterized in that, The step of obtaining the terrain classification features of any first grid cell includes: The terrain parameters of any first grid cell are obtained using a digital elevation model (DEM). Based on the terrain parameters and terrain features, the order is determined to obtain the terrain classification features of any first grid cell.

5. The method according to any one of claims 1 to 4, characterized in that, The method further includes: Obtain a fire training dataset, wherein the fire training dataset includes the fire line classification of each historical grid cell in the historical grid cell set corresponding to the historical fire line; The initial graph convolutional neural network model and the initial temporal Transformer model are trained using the fire training dataset to obtain the historical fire rescue tactics corresponding to each historical grid cell. If the historical fire rescue tactics corresponding to each historical grid cell meet the tactical requirements, the target graph convolutional neural network model and the target temporal Transformer model are obtained.

6. A fire rescue tactical acquisition device, characterized in that, include: The feature acquisition unit is used to acquire the fire line classification feature and the terrain classification feature of any first grid cell in the set of grid cells corresponding to the fire line; The vector acquisition unit is used to acquire the feature vector corresponding to any first grid cell based on the first feature vector corresponding to the fire line classification feature, the second feature vector corresponding to the terrain classification feature, and the third feature vector corresponding to the environmental feature of any first grid cell. The tactical acquisition unit is used to analyze the feature vector corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model, and to acquire the fire rescue tactics corresponding to any first grid cell. The tactical acquisition unit, used to analyze the feature vector corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model to acquire the fire rescue tactics corresponding to any first grid cell, specifically is used for: A target graph convolutional neural network model is used to propagate the feature vector corresponding to any first grid cell forward to obtain the fire space features at each time step; The target temporal Transformer model is used to process the fire spatial features at each time step to obtain the location features at each time step; Based on the fire space features and location features of each time step, obtain the fire representation corresponding to any first grid cell; Based on the mapping relationship between fire scene representation and fire scene rescue tactics, the fire scene rescue tactics corresponding to the fire scene representation are obtained, and the fire scene rescue tactics are used as the fire scene rescue tactics corresponding to any first grid unit. The tactical acquisition unit, specifically used to acquire the fire scene spatial features at each time step by employing a target image convolutional neural network model to propagate the feature vector corresponding to any first grid cell forward, is used for: Based on the spatial proximity relationship between the grid cells, obtain the adjacency matrix corresponding to any first grid cell; Based on the weight matrix, activation function, and adjacency matrix of the target graph convolutional neural network model, the spatial features of the fire scene at each time step are obtained.

7. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-5.

Images