A knowledge graph-based method and system for battlefield situation semantic understanding

Abstract

This application provides a knowledge graph-based method and system for battlefield situation semantic understanding. The method acquires multi-dimensional visual and electromagnetic wave data from the battlefield, extracts enemy equipment outlines and environmental occlusion features through distributed processing; optimizes the capture range of thermal radiation features using vegetation density, and identifies equipment types using an equipment database; integrates terrain, thermal radiation features, and equipment location to establish an entity relationship model and combat unit network; and finally, dynamically adjusts spatial weights based on equipment type feedback to establish an equipment-terrain adaptation relationship to optimize network semantic expression and output the understanding results. The technical solution provided in this application significantly improves the accuracy and semantic level of battlefield situation understanding through multi-source data fusion, environmentally adaptive optimization of feature extraction and equipment identification, and dynamic construction and optimization of the equipment-terrain association model, providing precise support for decision-making.

Description

Technical Field

[0001] This application relates to the field of intelligent battlefield situational awareness and understanding technology, and in particular to a method and system for battlefield situational semantic understanding based on knowledge graphs. Background Technology

[0002] In modern, highly competitive and complex environments, there is an urgent technical need to effectively address the challenges of rapid location of concealed targets and semantic construction of dynamic situational hotspots. This requires the system to be able to integrate multi-source heterogeneous perception data to penetrate camouflage or concealment and capture targets, while deeply analyzing the inherent semantic relationships between targets, environment, and tactical elements, automatically generating high-value threat hotspot areas with tactical interpretability, and meeting the stringent requirements of command and decision-making for timeliness down to the second, thus achieving rapid transformation from raw data to an understandable semantic situational awareness.

[0003] A representative technical solution for addressing the above needs is a method based on multimodal perception fusion and deep learning heatmap generation. This method integrates multispectral data from satellites, UAVs, and other sources, as well as electronic reconnaissance signals, through spatiotemporal registration, and processes them by inputting them into a cascaded deep neural network. The front-end network is responsible for extracting the local features of the target, while the back-end network generates a battlefield heatmap representing the spatial distribution density of the target. Based on the clustering results of the target density and motion trajectory, potential threat areas are marked, and finally, a heatmap with probability weights is output as the situational awareness result.

[0004] However, this existing solution has certain limitations. It relies too heavily on data-driven models and lacks explicit modeling of physical constraints and tactical rules in the natural environment, resulting in a high false alarm rate in complex scenarios. The generated heatmaps can only reflect the spatial clustering of targets and cannot reveal the deep semantic relationships between equipment type, terrain obstacles, and combat capabilities, making the marked hotspot areas lack tactical interpretability. At the same time, the solution adopts a static weight fusion strategy, which is difficult to dynamically respond to sudden variables such as battlefield electromagnetic interference or sudden weather changes, requiring manual readjustment of parameters and causing a significant delay in the critical decision window. Summary of the Invention

[0005] This application provides a knowledge graph-based battlefield situation semantic understanding method and system to solve the problem that existing technologies cannot effectively integrate environmental physical constraints and tactical rules knowledge, dynamically construct interpretable equipment terrain semantic associations, and thus achieve accurate and understandable battlefield situation semantic expression.

[0006] Firstly, this application provides a battlefield situation semantic understanding method based on knowledge graphs, including:

[0007] Acquire multi-dimensional visual data of the battlefield area and electromagnetic wave reflection data of fixed bands to form a raw dataset containing terrain texture features and thermal radiation features;

[0008] Based on the spatiotemporal coordinate encoding rules, the original dataset is transmitted to a distributed processing node, the original dataset is processed in a structured manner, and the processed dataset is subjected to pattern recognition to extract the contour parameters of enemy equipment and environmental occlusion features.

[0009] Based on the vegetation density distribution data in the environmental shading features, the effective capture range of the thermal radiation features in the vegetation-covered area is optimized to determine the optimized thermal radiation features. The corresponding equipment type of the enemy equipment is determined by comparing the contour parameters of the enemy equipment with the contour parameters corresponding to each equipment type in the pre-stored equipment database.

[0010] By utilizing the terrain texture features, optimized thermal radiation features, and spatial correlation with the distribution of enemy equipment, a battlefield entity relationship model is established, and the location information of enemy equipment is logically correlated with the distribution data of terrain obstacles to form a combat unit network.

[0011] The equipment type is fed back to the battlefield entity relationship model to adjust the weight parameters in the spatial topology analysis and to establish the adaptation relationship between equipment type and terrain obstacles.

[0012] The semantic representation of the combat unit network is optimized through the adaptation relationship, and the semantic understanding results are output.

[0013] Optionally, optimizing the semantic representation of the combat unit network through the adaptation relationship and outputting the semantic understanding result includes:

[0014] Based on the adaptation relationship, the adaptation coefficients are mapped to the blocking nodes in the combat unit network to generate semantic tags containing obstacle types and the adaptation coefficients;

[0015] The equipment type, the adjusted path travel weight parameters, and the line-of-sight observation weight parameters are integrated into a multi-dimensional semantic descriptor for network connection edges;

[0016] Based on the semantic tags and multi-dimensional semantic descriptors, the combat unit network is reorganized into three layers of semantics. The first layer constructs the mapping relationship between the equipment type and spatial coordinates. The second layer integrates the set of passage weight parameters and observation weight parameters of all network connection edges. The third layer associates the adaptation relationship chain between the equipment nodes and the blocking nodes in the combat unit network and outputs the reorganized network structure as the semantic understanding result.

[0017] Optionally, feeding back the equipment type to the battlefield entity relationship model to adjust the weight parameters in the spatial topology analysis and establish an adaptation relationship between the equipment type and terrain obstacles includes:

[0018] The equipment type is fed back to the battlefield entity relationship model, and the mobility and observation capabilities of the corresponding equipment are obtained based on the attribute table in the pre-stored equipment database.

[0019] The mobility capability level is applied to the connecting edges in the combat unit network, and the path passage weight parameters in the spatial topology analysis are adjusted. At the same time, the observation capability level is applied to the connecting edges in the combat unit network, and the line-of-sight observation weight parameters in the spatial topology analysis are adjusted.

[0020] By combining the terrain obstacle type identifiers in the equipment type and battlefield entity relationship model, an adaptation relationship between equipment type and terrain obstacle is established, and a passability adaptation coefficient for various types of equipment in different terrain obstacle environments is generated.

[0021] Optionally, the step of establishing a battlefield entity relationship model by utilizing the terrain texture features, optimized thermal radiation features, and spatial correlation with the distribution of enemy equipment, and logically associating the enemy equipment location information with terrain obstacle distribution data to form a combat unit network, includes:

[0022] The location coordinates and type identifiers of obstacles are extracted from the terrain texture features, and the spatial coordinate set of enemy equipment is located based on the optimized thermal radiation features.

[0023] Based on the spatial correlation of enemy equipment distribution reflected by the set of spatial coordinates, the straight-line distance between each pair of enemy equipment is calculated. The location coordinates and type identifiers of the obstacles are combined to determine whether there is visual obstruction. The judgment result, which includes direct visibility relationship and indirect connectivity relationship, is output. Based on the judgment result and the pre-stored equipment database, a battlefield entity relationship model is formed.

[0024] Based on the battlefield entity relationship model, the spatial coordinates of enemy equipment are transformed into network nodes, the directly visible relationships are transformed into unweighted connecting edges, and the indirect connected relationships are transformed into connecting edges with detour path length weights. At the same time, obstacles are mapped as blocking nodes and associated edges are established with the affected equipment nodes, generating a combat unit network containing the equipment nodes, blocking nodes, and connecting edges.

[0025] Optionally, the step of optimizing the effective capture range of the thermal radiation feature in the vegetation-covered area based on the vegetation density distribution data in the environmental shading features to determine the optimized thermal radiation feature, and determining the corresponding equipment type of the enemy equipment by comparing the contour parameters of the enemy equipment with the contour parameters corresponding to each equipment type in the pre-stored equipment database, includes:

[0026] Based on the vegetation density distribution data in the environmental shading features, the corresponding signal attenuation coefficient is obtained through a preset mapping relationship between density values ​​and signal attenuation coefficients.

[0027] The coordinate range of the environmental shading feature indicated by the vegetation-covered area is located, and numerical enhancement calculation is performed according to the signal attenuation coefficient. The enhanced thermal radiation feature value is merged with the original thermal radiation feature value of the non-vegetation-covered area, and the optimized thermal radiation feature is output.

[0028] Based on the contour parameters of the enemy equipment, the coordinates of the boundary points are extracted. Based on the distance and angle changes of adjacent boundary points, a contour shape descriptor is generated. At the same time, the standard contour templates of all equipment types in the pre-stored equipment database are traversed to extract the template shape descriptor from the standard contour template.

[0029] Calculate the matching degree between the contour shape descriptor and the template shape descriptor, and select the equipment type corresponding to the standard contour template with the highest matching degree score as the judgment result.

[0030] Optionally, the step of transmitting the original dataset to a distributed processing node based on spatiotemporal coordinate encoding rules, performing structured processing on the original dataset, and performing pattern recognition on the processed dataset to extract the contour parameters of enemy equipment and environmental occlusion features includes:

[0031] Based on the spatiotemporal coordinate encoding rules, a unique spatiotemporal identifier is generated for each data unit of the original dataset. The data units are then distributed to distributed processing nodes according to the spatiotemporal identifier, and the nodes are sorted by spatial unit coordinates to form a structured dataset.

[0032] In the structured dataset, a continuous region with thermal radiation characteristics significantly higher than the surrounding average is located and marked as a potential equipment region. The change points of continuous feature change on the boundary of the potential equipment region are tracked, and the contour parameters of the enemy equipment are generated based on the change points.

[0033] Simultaneously, the density and vertical distribution characteristics of vegetation in the terrain texture features are analyzed to generate environmental occlusion features that quantify the degree of obstruction to vision and energy propagation.

[0034] Optionally, the acquisition of multi-dimensional visual data of the battlefield area and electromagnetic wave reflection data of a fixed band to form a raw dataset containing terrain texture features and thermal radiation features includes:

[0035] Visible light imaging equipment deployed in the battlefield area collects multi-dimensional visual data covering the morphology of surface materials, uneven structure, and surface details of man-made objects. Infrared detection equipment receives electromagnetic wave reflection data radiated by objects on the battlefield in a fixed band.

[0036] The multi-dimensional visual data is divided into multiple spatial units, and the geometric morphological features of the land surface structure are extracted as terrain texture features for each spatial unit.

[0037] The electromagnetic wave reflection data is divided into the same spatial units, and the average thermal radiation intensity value of each spatial unit is extracted as the thermal radiation feature.

[0038] The terrain texture features and thermal radiation features of each spatial unit are combined into a data unit, and all data units are arranged according to their spatial location to form the original dataset.

[0039] Secondly, this application provides a battlefield situation semantic understanding system based on knowledge graphs, including:

[0040] The acquisition module is used to acquire multi-dimensional visual data of the battlefield area and electromagnetic wave reflection data of a fixed band, forming a raw dataset containing terrain texture features and thermal radiation features.

[0041] The recognition module is used to transmit the original dataset to the distributed processing node based on the spatiotemporal coordinate encoding rules, perform structured processing on the original dataset, and perform pattern recognition on the processed dataset to extract the contour parameters of enemy equipment and environmental occlusion features.

[0042] The processing module is used to optimize the effective capture range of the thermal radiation feature in the vegetation-covered area based on the vegetation density distribution data in the environmental shading feature, so as to determine the optimized thermal radiation feature, and to determine the corresponding equipment type of the enemy equipment by comparing the contour parameters of the enemy equipment with the contour parameters corresponding to each equipment type in the pre-stored equipment database.

[0043] A module is established to utilize the terrain texture features, optimized thermal radiation features, and spatial correlation of enemy equipment distribution to establish a battlefield entity relationship model, and to logically associate enemy equipment location information with terrain obstacle distribution data to form a combat unit network.

[0044] The feedback module is used to feed back the equipment type to the battlefield entity relationship model in order to adjust the weight parameters in the spatial topology analysis and establish the adaptation relationship between the equipment type and the terrain obstacles.

[0045] The output module is used to optimize the semantic representation of the combat unit network through the adaptation relationship and output the semantic understanding result.

[0046] Thirdly, this application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a knowledge graph-based battlefield situation semantic understanding method as described in the first aspect above.

[0047] Fourthly, this application provides a computer storage medium storing a computer program, which, when executed by a computer, implements a knowledge graph-based battlefield situation semantic understanding method as described in the first aspect.

[0048] In this application example, multi-dimensional visual data and fixed-band electromagnetic wave reflection data of the battlefield area are acquired to form an original dataset containing terrain texture features and thermal radiation features. Based on spatiotemporal coordinate encoding rules, the original dataset is transmitted to a distributed processing node for structured processing. Pattern recognition is then performed on the processed dataset to extract the contour parameters of enemy equipment and environmental occlusion features. Based on the vegetation density distribution data in the environmental occlusion features, the effective capture range of the thermal radiation features in the vegetation-covered area is optimized to determine the optimized thermal radiation features. The optimized features are then compared with the contours of the enemy equipment. The parameters are compared with the contour parameters corresponding to each equipment type in the pre-stored equipment database to determine the corresponding equipment type of the enemy equipment; a battlefield entity relationship model is established by using the terrain texture features, optimized thermal radiation features and spatial correlation of enemy equipment distribution, and the enemy equipment location information is logically correlated with terrain obstacle distribution data to form a combat unit network; the equipment type is fed back to the battlefield entity relationship model to adjust the weight parameters in spatial topology analysis and establish the adaptation relationship between equipment type and terrain obstacles; the semantic expression of the combat unit network is optimized through the adaptation relationship and the semantic understanding result is output.

[0049] The technical solution of this application has the following beneficial effects:

[0050] This application constructs an original dataset by fusing multi-dimensional battlefield visual data and electromagnetic wave reflection data. Through distributed node-based structured processing and pattern recognition, it extracts enemy equipment outlines and environmental occlusion features. It dynamically optimizes the effective capture range of thermal radiation features in occluded areas using vegetation density distribution, and accurately identifies enemy equipment types by combining this with an equipment database. Furthermore, it integrates terrain texture, optimizes thermal radiation features, and establishes a battlefield entity relationship model by incorporating the spatial distribution of equipment. A combat unit network is formed through the logical association between equipment location and terrain obstacles. Finally, based on equipment type feedback, it dynamically adjusts spatial topology weights to establish equipment-terrain adaptation relationships, significantly improving the accuracy and interpretability of battlefield situational semantics. This enables rapid location of concealed targets and semantic construction of high-value threat hotspots, providing precise tactical understanding support for battlefield decision-making.

[0051] Furthermore, the adaptation coefficients of equipment types and terrain obstacles are mapped to blocking nodes in the network, generating semantic labels containing obstacle types and adaptation coefficients. Simultaneously, equipment types, adjusted path passage weights, and line-of-sight observation weights are integrated into multi-dimensional semantic descriptors for network connections. Finally, based on these semantic elements, the network is restructured in three layers: the first layer establishes a mapping between equipment types and spatial coordinates; the second layer integrates the passage and observation weight sets of all connections; and the third layer constructs adaptation relationship chains between equipment nodes and blocking nodes, outputting the restructured network structure as the semantic understanding result. This scheme, by injecting structured semantic labels and multi-dimensional descriptors into network nodes and connections, and implementing three-layer semantic restructuring, achieves a deep structured expression of battlefield situation elements and their adaptation relationships, generating a quantifiable, analyzable, and clearly tactical semantic networked situation map. This significantly improves the accuracy and interpretability of threat assessment and tactical decision support in battlefield hotspot areas.

[0052] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 A flowchart of a battlefield situation semantic understanding method based on knowledge graphs provided in this application is shown;

[0055] Figure 2 A scene diagram is shown, illustrating a battlefield situation semantic understanding method based on knowledge graphs provided in this application.

[0056] Figure 3 A schematic diagram of the structure of a knowledge graph-based battlefield situation semantic understanding system provided in this application is shown.

[0057] Figure 4 A schematic diagram of the structure of a computing device provided in this application is shown. Detailed Implementation

[0058] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0059] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0060] Research indicates that current battlefield concealment target location and situational awareness hotspot construction primarily rely on multimodal perception fusion and deep learning heatmap technology, which suffers from three fundamental limitations. First, the purely data-driven model lacks explicit modeling of environmental physical constraints and tactical rules, leading to a surge in false alarm rates in complex battlefield scenarios. Second, the generated heatmaps can only characterize the spatial density distribution of targets and cannot construct deep semantic relationships between equipment type, terrain obstacles, and combat effectiveness, causing hotspot areas to lose tactical interpretability. Third, the static weighted fusion mechanism struggles to dynamically respond to battlefield electromagnetic interference or sudden weather changes, requiring manual parameter readjustment and causing delays in critical decisions, severely restricting the effectiveness of situational awareness.

[0061] To address the aforementioned issues, this invention proposes a battlefield situational semantic understanding method based on knowledge graphs. Its core lies in fusing multi-dimensional visual and electromagnetic wave data, extracting environmental occlusion features in distributed processing, and dynamically optimizing the thermal radiation capture range using vegetation density. This, combined with an equipment database, enables high-precision target identification. Furthermore, based on terrain texture, optimized thermal radiation features, and equipment spatial distribution, an entity relationship model embedding tactical rules is established, logically linking equipment locations with terrain obstacles to form a combat unit network. Finally, spatial topology weights are dynamically adjusted through equipment type feedback to establish an equipment-terrain adaptation relationship chain. This scheme solves the high false alarm rate problem through environmental constraint quantitative modeling, achieves tactical semantic interpretation of hotspot areas through equipment-terrain adaptation relationships, and overcomes the bottleneck of static modeling through a dynamic weight adjustment mechanism. Target positioning accuracy is significantly improved in complex battlefield environments, and the semantic situational awareness generation time is reduced to the second level, providing command and decision-making with accurate, interpretable, and real-time situational understanding support.

[0062] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] Figure 1 A flowchart of a battlefield situation semantic understanding method based on knowledge graphs is provided in this application embodiment, such as... Figure 1 As shown, the method includes:

[0064] 101. Acquire multi-dimensional visual data and fixed-band electromagnetic wave reflection data of the battlefield area to form a raw dataset containing terrain texture features and thermal radiation features;

[0065] Optionally, step 101 may specifically include the following steps:

[0066] 1011. By deploying visible light imaging equipment in the battlefield area, collect multi-dimensional visual data covering the morphology of surface materials, uneven structure and surface details of man-made objects, and receive electromagnetic wave reflection data of fixed bands radiated by battlefield objects through infrared detection equipment.

[0067] 1012. Divide the multi-dimensional visual data into multiple spatial units, and extract the geometric morphological features of the surface structure as terrain texture features for each spatial unit;

[0068] 1013. Divide the electromagnetic wave reflection data into the same spatial units, and extract the average thermal radiation intensity value of each spatial unit as the thermal radiation feature;

[0069] 1014. Combine the terrain texture features and thermal radiation features of each spatial unit into a data unit, and arrange all data units according to their spatial location to form an original dataset.

[0070] In the above scheme, multi-dimensional visual data refers to the set of surface material morphology information captured by visible light imaging equipment, including the outlines of surface objects, surface unevenness and texture details of man-made objects, which can be used to reconstruct a three-dimensional model of battlefield terrain. Fixed-band electromagnetic wave reflection data refers to the energy signals radiated by objects within a specific wavelength range received by infrared detection equipment, which can be used to identify the spatial location of heat-generating targets. Terrain texture features refer to the set of quantitative parameters describing the geometry of the ground surface, which can be used to analyze the degree to which the battlefield environment hinders equipment movement. Thermal radiation features refer to physical quantities reflecting the thermodynamic state of an object's surface, which can be used to detect heat-generating targets hidden behind vegetation or bunkers. The raw dataset refers to a spatially aligned set of multi-source sensing data, containing paired units of terrain texture features and thermal radiation features indexed by grid coordinates, which can be used for subsequent distributed environmental analysis.

[0071] In this embodiment, firstly, multi-dimensional visual data is obtained by a visible light imaging device deployed in the battlefield area in step 1011. Simultaneously, a fixed-band electromagnetic wave reflection data of the same area is collected by an infrared detection device, achieving spatiotemporal synchronous capture of optical details and thermal radiation signals. For example, when conducting reconnaissance in area A, the visible light camera captures metallic reflections in bushes at a resolution of 0.5 meters, while the infrared device collects radiation in the 8-14 micrometer band at the same frame rate and detects a thermal signal of 102°C at that location. The two types of data achieve millisecond-level time synchronization and coordinate alignment through a GPS clock and a geographic information system.

[0072] Next, in step 1012, the acquired multi-dimensional visual data is divided into equally sized square grids. Stereo imaging technology is used on each grid, and by comparing the perspective differences between adjacent photos, the changes in ground elevation are calculated as terrain texture features. The slope calculation formula is as follows: in This represents the elevation change of the terrain per meter traveled in the east-west direction. This represents the change in terrain slope along the north-south direction. For example, the slope calculated for a certain grid... That is, for every 1 meter you move eastward, the terrain rises by 0.3 meters. That is, substituting into the formula to obtain the slope angle. At the same time, the roughness is marked as "high" based on the degree of unevenness of the ground surface.

[0073] Then, in step 1013, the electromagnetic wave reflection data generated by the infrared detection device is divided according to the same grid boundary. The average thermal radiation intensity value is calculated by summing the radiation values ​​of all pixels in each cell and dividing by the total number of pixels. The calculation formula is as follows: For example, grid C contains 400 pixels, and three heat sources (115°C, 120°C, 118°C) and background temperature were detected. The average thermal radiation value was calculated to be 117.7 watts per square meter.

[0074] Finally, in step 1014, the terrain texture features and thermal radiation features of the same grid are bound into data units, and an indexed dataset is constructed according to the grid coordinates. For example, the output data unit {slope: 12°, roughness: medium, thermal radiation: 93W / m²} of grid D (coordinates X: 200, Y: 300) is sorted with adjacent grids according to latitude and longitude to form a matrix table, realizing the structured integration of multi-source data to form the original dataset.

[0075] In practical applications, during a reconnaissance mission, a visible light imaging device deployed on a UAV first captured surface images at a resolution of 0.6 meters. Simultaneously, an infrared detection device on the same platform collected electromagnetic wave data in the 8-14 micrometer band. When the UAV flew over the coordinate zone X1200-Y3500, the visible light device captured regular metallic reflections of approximately 2×3 meters in size from gaps in the pine forest. At the same time, the infrared device detected a continuous 105°C thermal signal emitted from this point. The two types of data were synchronized within 5 milliseconds via a satellite clock and bound to a unified geographic coordinate system. Subsequently, the visible light image was divided into a 100-meter × 100-meter grid. Stereoscopic imaging analysis was applied to the G7 grid where the target was located: by comparing the viewing angle difference of consecutive frames, the terrain elevation per meter in the east-west direction was calculated to be 0.28 meters, and the elevation per meter in the north-south direction was 0.15 meters. Substituting these values ​​into the slope formula, the slope was calculated to be approximately 18.2°. Simultaneously, the standard deviation of the surface roughness was detected to be 0.4 meters, which was marked as "medium" roughness. Next, the infrared data of the same area was processed: the G7 grid contains 625 pixels, of which 120 pixels in the metallic target area have an average temperature of 105°C, and 505 pixels in the pine forest area have an average temperature of 22°C. The average thermal radiation intensity was calculated to be approximately 35.6 W / m² after weighted calculation. Finally, the G7 grid data unit {slope: 18.2°, roughness: medium, thermal radiation: 35.6 W / m²} was generated and sorted with the adjacent G6 grid {slope: 8.1°, roughness: low, thermal radiation: 28.3 W / m²} and H7 grid {slope: 24.7°, roughness: high, thermal radiation: 41.2 W / m²} according to latitude and longitude coordinates to construct the original structured dataset covering a 3km × 3km battlefield.

[0076] The aforementioned overall solution 101 transforms raw battlefield imagery into a structured, quantifiable, and spatially precisely aligned environmental feature dataset through standardized spatial grid division and multi-source data fusion processing. Among these features, terrain texture objectively characterizes the intensity of surface access obstruction, while thermal radiation features penetrate vegetation cover to reveal the distribution of heat sources of concealed targets. The binding relationship between these two types of features within a unified spatial unit eliminates the risk of misjudgment caused by the misalignment of traditional multi-sensor data. This provides highly consistent and traceable underlying data support for subsequent distributed environmental occlusion analysis, equipment identification, and battlefield semantic modeling, significantly improving the accuracy and automation level of battlefield situational awareness.

[0077] 102. Based on the spatiotemporal coordinate encoding rules, the original dataset is transmitted to the distributed processing node, the original dataset is processed in a structured manner, and the processed dataset is subjected to pattern recognition to extract the contour parameters of enemy equipment and environmental occlusion features.

[0078] Optionally, step 102 may specifically include the following steps:

[0079] 1021. Based on the spatiotemporal coordinate encoding rules, a unique spatiotemporal identifier code is generated for each data unit of the original dataset. The data units are distributed to distributed processing nodes according to the spatiotemporal identifier code, and the nodes are sorted according to the spatial unit coordinates to form a structured dataset.

[0080] 1022. Locate the continuous region in the structured dataset where the thermal radiation characteristics are significantly higher than the surrounding average and mark it as a potential equipment region. Track the points of continuous change of characteristics on the boundary of the potential equipment region and generate the contour parameters of the enemy equipment based on the points of change.

[0081] 1023. Simultaneously analyze the density and vertical distribution characteristics of vegetation in the terrain texture features to generate environmental occlusion features that quantify the degree of obstruction to vision and energy propagation.

[0082] In the above scheme, the spatiotemporal coordinate encoding rule refers to an indexing mechanism that binds battlefield data with precise spatiotemporal location, used to achieve spatial synchronization and source tracing of multi-source heterogeneous data. The spatiotemporal identifier code is a unique spatiotemporal identity assigned to a data unit. Its physical basis is the absolute time of sensor acquisition and the three-dimensional position in the geodetic coordinate system. A fixed-length string is used to achieve precise positioning and routing of data in the distributed system. Potential equipment areas refer to continuous spatial domains where thermal radiation intensity significantly exceeds the local background, indicating the possible presence of high-heat-source targets. Contour parameters are key indicators describing the geometric characteristics of enemy equipment. Essentially, they are a set of polygon vertices formed by connecting the coordinates of abrupt changes in the physical properties of the target's edges, used to quantify the length, width, and topological structure of the equipment. Vegetation density is a spatial statistic of the number of plants per unit area of ​​the ground surface, directly determining the physical probability of electromagnetic wave scattering and line-of-sight obstruction. Vertical distribution characteristics describe the three-dimensional structural characteristics of vegetation height stratification, affecting the penetration depth of signals in different frequency bands. Environmental obscuring features are dynamic coefficients that characterize the degree to which natural obstacles inhibit battlefield perception. They are essentially based on modeling the physical blocking effect of vegetation density and vertical structure, and are used to assess the attenuation intensity of terrain on reconnaissance line of sight and energy propagation.

[0083] In this embodiment, after obtaining the original dataset, step 1021 performs spatiotemporal coordinate encoding to generate a 20-bit unique label containing precise time and geographical location for each data point. For example, data collected on a certain day in 2023 is spatiotemporally encoded, and the corresponding label generated by time (e.g., August 1, 10:30:15), latitude and longitude (e.g., 113.423 degrees east longitude and 22.315 degrees north latitude; it should be noted that this latitude and longitude is only for illustrative purposes and does not represent the actual latitude and longitude), and altitude (e.g., 50 meters) is "20230801103015_113423_22315_050". The data is then distributed to the corresponding distributed computer nodes according to the longitude range in the label. After receiving the data, the nodes arrange the longitude values ​​in ascending order into table columns and the latitude values ​​in descending order into table rows. They also assign rows and columns to the longitude and latitude values ​​in the corresponding labels. For example, they put the longitude data in the 4th column and the latitude data in the 5th row, forming a structured grid dataset similar to a chessboard.

[0084] Then, in step 1022, the computer scans the temperature values ​​of each cell in the structured dataset grid. When a 3x3 area is found to have an average temperature significantly higher than the average of its eight neighboring areas by more than 20%, it is marked as a potential equipment area. The system automatically moves along the boundary of this area to detect and locate points where the temperature drops sharply by more than 15°C per pixel. All points of change are connected to form a polygon, and its maximum length and width are calculated, outputting the contour parameters of the enemy equipment. For example, if the average temperature of the central area of ​​a 3x3 area centered at row 5 and column 6 is 62°C, which is more than 20% higher than the average temperature of its eight neighboring areas, it is marked as a potential equipment area; subsequently, it detects that the temperature at coordinate point (5,6) is 58°C and the temperature at adjacent point (5,7) is 36°C, resulting in a gradient value of 22°C pixels, which is determined to be a contour change point.

[0085] Finally, environmental occlusion analysis is performed in step 1023. For the optical image of the same grid, the number of plant trunks within a 1-square-meter area is counted. The height of each plant is measured and classified into three layers: 0-1 meters corresponds to the herbaceous layer, 1-3 meters to the shrub layer, and above 3 meters to the tree layer. For example, a 2.5-meter-tall tree is included in the shrub layer. The occlusion coefficient is calculated using the following formula: ,in Let i be the percentage of vegetation in the i-th layer. The base coefficients for the corresponding layers are 0.2 for herbaceous plants, 0.5 for shrubs, and 0.8 for trees. For example, in the optical image analysis of the grid in the 2nd row and 8th column of region B, 3 trees 2.5m tall were identified as belonging to the tree layer and 5 shrubs 0.8m tall as belonging to the shrub layer within a 1㎡ area; the tree layer accounts for 37.5% and the shrub layer accounts for 62.5%; the calculated occlusion coefficient is 0.6125, and the output value is 0.61. This parameter is ultimately output to quantify the environmental obstruction intensity as an environmental occlusion feature.

[0086] The data, carrying a contour parameter of 6.2 meters in length and 3.1 meters in width, and an occlusion coefficient of 0.24, enters the subsequent processing flow.

[0087] The aforementioned overall solution (102) achieves precise positioning and distributed, efficient processing of massive battlefield data through spatiotemporal coordinate encoding rules. It transforms multi-source, heterogeneous raw data into an ordered, gridded structure, providing a rapidly searchable spatiotemporal reference for subsequent analysis. Based on spatial comparative analysis of thermal radiation characteristics, it automatically identifies potential equipment areas and generates precise enemy equipment contour parameters by tracking abrupt changes in physical boundaries, effectively overcoming edge blurring caused by natural background interference. Simultaneously, it integrates optical data to quantify vegetation distribution density and vertical structure, transforming environmental occlusion effects into calculable standardized coefficients, forming an objective assessment of the reconnaissance field of view and the intensity of energy propagation obstruction. These three core capabilities work synergistically to construct a multi-dimensional feature system for battlefield entity identification, laying a high-precision data foundation for subsequent equipment type determination, battlefield relationship modeling, and semantic understanding.

[0088] 103. Based on the vegetation density distribution data in the environmental shading features, optimize the effective capture range of the thermal radiation features in the vegetation-covered area to determine the optimized thermal radiation features, and determine the corresponding equipment type of the enemy equipment by comparing the contour parameters of the enemy equipment with the contour parameters corresponding to each equipment type in the pre-stored equipment database.

[0089] Optionally, step 103 may specifically include the following steps:

[0090] 1031. Based on the vegetation density distribution data in the environmental shading features, obtain the corresponding signal attenuation coefficient through a preset mapping relationship between density value and signal attenuation coefficient;

[0091] 1032. Locate the coordinate range of the environmental shading feature indicated by the vegetation coverage area, perform numerical enhancement calculation according to the signal attenuation coefficient, merge the enhanced thermal radiation feature value with the original thermal radiation feature value of the non-vegetation coverage area, and output the optimized thermal radiation feature.

[0092] 1033. Extract the boundary point coordinates based on the outline parameters of the enemy equipment, generate an outline shape descriptor based on the distance and angle changes of adjacent boundary points, and simultaneously traverse the standard outline templates of all equipment types in the pre-stored equipment database to extract the template shape descriptor from the standard outline templates.

[0093] 1034. Calculate the matching degree between the contour shape descriptor and the template shape descriptor, and select the equipment type corresponding to the standard contour template with the highest matching degree score as the judgment result.

[0094] In the above scheme, vegetation density distribution data refers to a quantitative index of the density of vegetation cover within a unit area obtained through remote sensing image analysis, which can be used to assess the shielding strength of infrared detection signals by the battlefield environment. Signal attenuation coefficient refers to a physical parameter characterizing the degree of energy attenuation when infrared radiation of a specific band passes through the vegetation layer, which can be used to calculate the compensation level of thermal radiation values ​​in the shielded area. Numerical enhancement calculation refers to the mathematical processing of energy compensation for the original thermal radiation values ​​of the vegetation-shielded area, which can be used to restore the true thermal radiation intensity before absorption or scattering by vegetation. Contour shape descriptor refers to a digital representation of the equipment's shape constructed through the geometric relationships of boundary points, which can be used to quantitatively compare the morphological similarity between the measured target and a standard template. Template shape descriptor refers to a digital template of typical equipment contours pre-existing in the equipment database, which can serve as a benchmark reference for equipment type identification. Matching score refers to an index of the degree of fit between the measured contour and the template contour calculated through a vector space similarity algorithm, which can be used to objectively determine the type of enemy equipment.

[0095] In this embodiment, step 1031 first reads vegetation density distribution data representing plant density in the battlefield environment, and then finds the signal attenuation coefficient of the current vegetation area according to a pre-set rule table. In the rule table, low density corresponds to a small attenuation value, and high density corresponds to a large attenuation value. For example, if a vegetation coverage rate of 85% is detected in a jungle area, the table shows that the thermal radiation signal attenuation coefficient for that area is 0.88, meaning the sensor can only capture 12% of the actual heat signal.

[0096] Next, by locating the coordinate range of all vegetation-covered areas in step 1032, compensation calculations are performed on the original thermal radiation values ​​within the vegetation-covered areas. The calculation formula is as follows: Assuming a tank's original thermal signal in a jungle is 100°C with an attenuation coefficient of 0.75, the compensated true value is 400°C. The compensated thermal radiation data from the vegetated area is then combined with the thermal radiation data from the open area to output the optimized thermal radiation characteristics.

[0097] Subsequently, in step 1033, continuous boundary points of its outline are extracted, for example, by obtaining the coordinates of 150 key points of the tank outline using image recognition technology. Following the connection order of the boundary points, the straight-line distance and directional turning angle between adjacent points are calculated segment by segment to form a digital feature sequence describing the shape of the object, i.e., the outline shape descriptor. Simultaneously, a preset standard template is retrieved from the equipment feature library. For example, the distance between point 1 and point 2 is 1.82 meters, with a clockwise rotation of 31 degrees; the distance between point 2 and point 3 is 1.63 meters, with a counter-clockwise rotation of 18 degrees. The resulting outline shape descriptor sequence is [1.82, 31, 1.63, -18, ...]. The main battle tank template [1.75, 30, 1.60, -20, ...] and the Type B equipment template [2.10, 45, 1.90, -35, ...] are retrieved from the equipment feature library.

[0098] Finally, the matching degree between the measured sequence and each template is calculated in step 1034, treating the measured sequence and the template sequence as two sets of numerical vectors; the sum of the products of the corresponding values ​​in the two sets of vectors is calculated; the square root of the sum of the squares of the measured vector and the template vector is calculated separately; finally, according to the formula: The matching score is calculated, and the equipment type corresponding to the standard profile template with the highest score is selected as the determination result. For example, if a tank template gets a high matching score of 0.98, while other equipment templates are all below 0.6, then the target is determined to be a tank type.

[0099] In practical applications, within a certain area, multi-dimensional visual data acquired by drones and satellites detected dense vegetation cover in the coordinate grid region E7-F9. Image analysis calculated that the vegetation density in this area reached 80%, and based on a pre-stored density attenuation mapping table, the infrared signal attenuation coefficient for this area was determined to be 0.85. Subsequently, the specific coordinate range of this vegetation-covered area was located from E7:1200,450 to F9:1350,600, and the original thermal radiation value within the area was dynamically compensated. For example, the original sensor reading of a hidden armored vehicle was only 150°C, but the actual thermal radiation value was calculated to be 1000°C using a formula. Simultaneously, the system extracted the coordinates of 180 consecutive boundary points of the target outline, and sequentially calculated the characteristics of adjacent points, such as point... arrive The straight-line distance is 1.8 meters, and the direction angle is offset by 32 degrees clockwise. arrive At a distance of 1.6 meters and a counterclockwise offset of 18 degrees, a sequence of outline shape descriptors [1.8, 32, 1.6, -18, ...] was generated. Then, standard templates were retrieved from the equipment database: the main battle tank template [1.75, 30, 1.62, -20, ...] and the Type B equipment template [2.2, 45, 1.9, -35, ...]. By calculating the matching degree, the measured sequence showed a matching degree of 0.95 with the tank template and only 0.41 with the Type B equipment template. Based on this, the target was determined to be of the "main battle tank" type.

[0100] The aforementioned overall solution 103 achieves precise detection and intelligent identification of concealed targets on the battlefield by dynamically fusing environmental concealment features with equipment morphology recognition technology. Based on adaptive optimization of the thermal radiation capture range according to vegetation density distribution, it significantly improves target detection capabilities in vegetation-covered areas and eliminates thermal signal distortion caused by natural environmental interference. Simultaneously, through digital extraction of contour features and an intelligent matching mechanism with the equipment database, it achieves automated and accurate identification of enemy equipment types in complex battlefield environments.

[0101] 104. Utilize the aforementioned terrain texture features, optimized thermal radiation features, and spatial correlation with the distribution of enemy equipment to establish a battlefield entity relationship model, and logically correlate the enemy equipment location information with terrain obstacle distribution data to form a combat unit network;

[0102] Optionally, step 104 may specifically include the following steps:

[0103] 1041. Extract the location coordinates and type identifiers of obstacles from the terrain texture features, and locate the spatial coordinate set of enemy equipment based on the optimized thermal radiation features;

[0104] 1042. Based on the spatial correlation of enemy equipment distribution reflected by the set of spatial coordinates, calculate the straight-line distance between every two enemy equipment, determine whether there is line of sight obstruction by combining the position coordinates and type identifier of the obstacle, output the judgment result including direct visibility relationship and indirect connectivity relationship, and form a battlefield entity relationship model based on the judgment result and the pre-stored equipment database.

[0105] 1043. Based on the battlefield entity relationship model, the spatial coordinates of enemy equipment are transformed into network nodes, the directly visible relationships are transformed into unweighted connecting edges, the indirect connected relationships are transformed into connecting edges with detour path length weights, and the obstacles are mapped as blocking nodes and associated edges are established with the affected equipment nodes to generate a combat unit network containing the equipment nodes, blocking nodes and connecting edges.

[0106] In the above scheme, type identification refers to the classification labels of terrain features that obstruct passage or observation in the battlefield environment, including standardized codes for natural landforms and man-made structures, which can be used to quantify the constraint intensity of obstacles on operations. Optimized thermal radiation characteristics refer to infrared radiation intensity distribution data corrected for vegetation cover effects, which can be used to penetrate cover to accurately locate enemy equipment. Spatial coordinate sets refer to the clusters of enemy equipment positions in a three-dimensional battlefield coordinate system, including numerical representations of longitude, latitude, and altitude, which can be used to construct spatial topological relationships between equipment. Direct visibility relationships refer to a straight-line line-of-sight state between two pieces of equipment without physical obstruction, which can be used to determine the feasibility of direct-fire attacks or optical reconnaissance. Indirect connectivity relationships refer to a connection state between two pieces of equipment where there is a detour path but no straight-line line of sight, which can be used to assess the maneuver route planning requirements of mechanized forces. The battlefield entity relationship model refers to a dynamic topological structure describing equipment, obstacles, and their interactions, including entity position attributes, connection states, and tactical rule constraints, which can be used to deduce the evolution path of the battlefield situation. A combat unit network refers to a topological structure that represents the spatial relationships between battlefield entities. It includes enemy equipment nodes, terrain obstruction nodes, and connecting edges, used to visualize the constraints between battlefield elements. Connecting edges are the carriers of relationships between nodes in the combat unit network, and can be used to quantitatively analyze the intensity of interactions between battlefield elements. Related edges are the direct influence links between equipment nodes and obstruction nodes, and can be used to calculate concealed attack routes or blind spots in frontal attacks.

[0107] In this embodiment, step 1041 first analyzes the battlefield surface image to identify the location coordinates and type of obstacles, while simultaneously processing thermal radiation data to locate the clustering positions of enemy equipment. This process uses image outlining technology to delineate obstacle boundaries and a high-temperature point clustering method to group nearby heat sources into the same equipment group. For example, during a battlefield area scan, the system delineates the outline of a rock at coordinates (102, 58) in the image and labels it as "Rock"; at the same time, it finds a dense cluster of high-temperature points at position (110, 62) on the thermal radiation map and identifies it as a tank.

[0108] Next, the Euclidean distance formula is used in step 1042. The straight-line distance between any two pieces of equipment is calculated, followed by occlusion detection. The line connecting the two points is divided into equal segments, and each point is checked to see if it falls within the obstacle coverage area. If occlusion exists, an indirect connectivity relationship is marked; otherwise, a direct visibility relationship is marked. Finally, the equipment database is queried to add tactical relationships, generating a battlefield entity relationship model that integrates spatial and tactical elements. For example, the straight-line distance between a tank (110,62) and an armored vehicle (198,160) is 132 meters. During occlusion detection, a detection point is set every 10 meters. If the detection point (125,85) is within the 30-meter influence range of the rock (102,58), then an indirect connectivity relationship is determined, and the database displays "The tank can provide area cover for the armored vehicle within 200 meters," adding a tactical tag.

[0109] Finally, in step 1043, the equipment positions are transformed into circular equipment nodes, and obstacles are transformed into triangular blocking nodes; green solid lines are drawn to connect directly visible equipment nodes, and yellow dashed lines are drawn to connect indirectly connected nodes, and the passage weight is calculated using the following formula: Finally, draw a red connection edge from the blocking node to the affected equipment node and label the occlusion angle parameters. For example, convert the tank (110,62) to a Tank node and the rock (102,58) to a Rock node; establish a yellow dashed line connecting the Tank and armored vehicle nodes; draw a red connection edge from the Rock to the Tank and label it "Northwest 30° View Occlusion".

[0110] In practical applications, a rock obstacle at coordinates (102, 58) was identified from the terrain imagery and labeled "Rock". Simultaneously, three high-temperature clusters were located on the thermal radiation map: the tank group at (110, 62), the armored vehicle at (198, 160), and the radar vehicle at (215, 170). The distance between the equipment was calculated to be 132 meters in a straight line from the tank to the armored vehicle. When checking the connecting path, coordinates (125, 85) were found to fall within a rock-covered area, indicating an indirect connection. The distance between the armored vehicle and the radar vehicle was 20 meters and the path was unobstructed, indicating direct visibility, and a tactical tag "Radar vehicle provides reconnaissance support to armored vehicle" was added. When generating the combat unit network based on the battlefield entity relationship model, the spatial coordinates of the equipment were transformed into visual nodes. The tank position (110, 62) was mapped to a circular node "Tank" with a diameter of 8 units, and the armored vehicle position (198, 160) was mapped to... The radar vehicle is transformed into an Armor node, and the radar vehicle's position (215, 170) becomes a Radar node. At the same time, the rock obstacle (102, 58) is transformed into a triangle blocking node, Rock. Since the armored vehicle and the radar vehicle are directly visible to each other, a solid green line is drawn between the Armor and Radar nodes. The tank and the armored vehicle are determined to be indirectly connected, and a dashed yellow line is generated between the Tank and Armor nodes. The passage weight is calculated to be 1.98 based on the rock obstacle type and a detour coefficient of 1.5. Finally, an obstacle association network is established, extending red association edges from the Rock node to the Tank node directly affected by it. The "Northwest 30° View Obstruction" label is calculated based on the rock outline and the tank's azimuth angle. At the same time, an association edge is drawn to the Armor node labeled "Southeast 45° Maneuver Obstruction". The final output is a computable combat network containing node attributes, passage weights, and obstruction semantics.

[0111] The aforementioned overall solution 104, through a three-stage process of precise extraction of environmental elements, dynamic modeling of entity relationships, and structural transformation of topological networks, transforms multi-source battlefield perception data into a computable combat unit network that combines spatial constraints and tactical semantics. It not only intuitively presents the deployment locations of enemy equipment, the distribution of obstacles, and the connectivity relationships, but also quantifies the cost of passage through weight parameters and reveals the impact of obscuring by labeling associated edges. This achieves a semantic situational expression that can be quantified, analyzed, and dynamically deduced from raw data, significantly improving the depth of battlefield environment understanding and the accuracy of decision support.

[0112] 105. Feed the equipment type back to the battlefield entity relationship model to adjust the weight parameters in the spatial topology analysis and establish the adaptation relationship between equipment type and terrain obstacles;

[0113] Optionally, step 105 may specifically include the following steps:

[0114] 1051. Feed back the equipment type to the battlefield entity relationship model, and obtain the mobility level and observation level of the corresponding equipment based on the attribute table in the pre-stored equipment database.

[0115] 1052. Apply the mobility capability level to the connecting edges in the combat unit network, adjust the path passage weight parameters in the spatial topology analysis, and simultaneously apply the observation capability level to the connecting edges in the combat unit network, adjust the line-of-sight observation weight parameters in the spatial topology analysis.

[0116] 1053. By combining the terrain obstacle type identifiers in the equipment type and battlefield entity relationship model, establish the adaptation relationship between equipment type and terrain obstacle, and generate the accessibility adaptation coefficients of various equipment in different terrain obstacle environments.

[0117] In the above scheme, the mobility capability level refers to a numerical parameter that quantifies the off-road mobility effectiveness of equipment, used to calculate the attenuation coefficient of travel efficiency under specific terrain. The observation capability level refers to a numerical parameter that characterizes the equipment's reconnaissance and counter-reconnaissance capabilities, used to assess the effective detection range in an obstacle environment. The path travel weight parameter refers to a dynamic variable in the combat unit network connection edges that expresses the degree of terrain travel obstruction, including the fusion value of spatial attributes such as slope angle, surface hardness, and vegetation cover density, used to generate a feasibility assessment of equipment maneuver paths. The line-of-sight observation weight parameter refers to a dynamic variable in the combat unit network connection edges that describes the intensity of visual obstruction by obstacles, including obstacle height, distribution dispersion, and material spectral reflectance characteristics, used to predict the effective detection area of ​​reconnaissance equipment. The adaptation relationship refers to the set of matching rules between equipment type and terrain obstacle type, used to construct a suppression map of the battlefield environment on equipment combat effectiveness. The travel adaptation coefficient refers to the probability value that a specific piece of equipment retains a certain percentage of mobility effectiveness in a specific terrain, including the coupling calculation result of equipment power redundancy and terrain resistance parameters, used to generate the decision priority of tactical maneuver routes.

[0118] In this embodiment, step 1051 first identifies the type of enemy equipment in the battlefield entity relationship model and queries the characteristic record table in the pre-stored equipment database to extract the two key capability values ​​of the equipment: mobility level and observation level. These level values ​​are derived from the equipment design parameters and historical combat data. For example, if the enemy equipment is identified as a tracked engineering vehicle, querying the characteristic record table in the equipment database yields a mobility level of 6 and an observation level of 2, providing core input for subsequent dynamic adjustments, much like equipping the vehicle with a dedicated "terrain access passport".

[0119] Next, after obtaining the equipment's mobility and observation capabilities, a dynamic weight adjustment mechanism is activated. This involves reading the currently stored original access weight and observation weight values ​​for each terrain connection route in the combat unit network, and then converting the mobility capability level into an access adjustment coefficient. The calculation formula is as follows: The denominator 10 represents the highest level. For example, if a certain type of armored vehicle has a mobility level of 7, then the passage adjustment coefficient is... Then, this coefficient is multiplied by the original passage weight to generate a new passage weight. Taking the vehicle traversing a 35° steep slope as an example, the original passage weight of 0.6 is updated to 0.42 after calculation. Simultaneously, the observation weight is processed, converting the observation capability level into an observation adjustment coefficient. The calculation formula is as follows: If the vehicle's observation capability level is 4, the coefficient is 0.4. This coefficient is then multiplied by the original observation weight to obtain the new observation weight. For example, the original observation weight for the bush area is 0.8, which is updated to 0.32. Finally, all the updated weight values ​​are written into the combat unit network to complete the equipment-specific adaptation of terrain route action constraints.

[0120] Finally, in step 1053, the equipment type and current terrain identifier are jointly queried in the historical combat experience database. If a direct matching rule exists, the compatibility coefficient is directly applied; otherwise, similar terrain calculation is initiated. For example, if a tracked engineering vehicle encounters a new type of electromagnetic interference mud, and no historical data is found in the historical combat experience database, the system refers to the "wetland" coefficient of 0.3, and further adjusts it by 20% based on the conductivity characteristics of the mud, calculating a compatibility coefficient of [missing value]. This indicates that the vehicle retains only 24% of its maneuverability in this area. This step constructs a quantified "equipment-terrain compatibility dictionary," providing crucial information for tactical decision-making.

[0121] In practical applications, after identifying an enemy wheeled armored reconnaissance vehicle (VBL) in a mountainous battlefield, the system queries the equipment database to extract its mobility level (Level 8, indicating high mobility) and observation level (Level 9, indicating advanced reconnaissance capability). The system then dynamically adjusts the network weights of the combat units: for the vehicle's route ahead, the original passage weight of 0.7 for a 30° scree slope is multiplied by a movement adjustment coefficient of 0.8, resulting in a new passage weight of 0.56, indicating that the vehicle retains 56% of its mobility when traversing the scree slope. Simultaneously, the original observation weight of 0.5 for the pine forest area is multiplied by an observation adjustment coefficient of 0.9, generating a new observation weight of 0.45, indicating that the forest provides relatively weak obstruction to the reconnaissance equipment. Considering the current terrain features, when the vehicle encounters a new type of strong electromagnetic interference canyon with no direct historical records, the system first matches the most similar terrain, "granite canyon," with an adaptation coefficient of 0.6. Then, based on the electromagnetic interference characteristics, a 20% performance attenuation is added, ultimately generating the passage adaptation coefficient. This indicates that the vehicle maintains only 48% of its reconnaissance mobility in this special environment. This coefficient is then incorporated into the combat unit network to provide a quantitative basis for subsequent tactical path planning.

[0122] The aforementioned overall scheme 105 dynamically feeds back the identified equipment types to the battlefield entity relationship model, extracts mobility and observation capability levels based on the equipment database, adjusts the access weight and observation weight of terrain routes in the combat unit network, and simultaneously establishes the adaptation relationship between equipment types and terrain obstacles. This process breaks through the limitations of static environment modeling, enabling terrain obstacle assessment to accurately match specific equipment characteristics. At the same time, by quantifying equipment-terrain interaction constraints, it constructs an analyzable tactical semantic network, significantly improving the accuracy of concealed target action prediction and the efficiency of battlefield hotspot evolution simulation, providing a situational understanding kernel for command and decision-making that combines real-time performance and interpretability.

[0123] 106. Optimize the semantic representation of the combat unit network through the adaptation relationship and output the semantic understanding result.

[0124] Optionally, step 106 may specifically include the following steps:

[0125] 1061. Map the adaptation coefficients to the blocking nodes in the combat unit network according to the adaptation relationship, and generate semantic tags containing obstacle types and the adaptation coefficients;

[0126] 1062. Integrate equipment type, adjusted path travel weight parameters, and line-of-sight observation weight parameters into a multi-dimensional semantic descriptor for network connection edges;

[0127] 1063. Based on the semantic tags and multi-dimensional semantic descriptors, the combat unit network is reorganized into three layers of semantics. The first layer constructs the mapping relationship between the equipment type and spatial coordinates. The second layer integrates the set of passage weight parameters and observation weight parameters of all network connection edges. The third layer associates the adaptation relationship chain between the equipment nodes and the blocking nodes in the combat unit network and outputs the reorganized network structure as the semantic understanding result.

[0128] In the above scheme, semantic tags refer to the structured attribute set attached to terrain-blocking nodes, including the physical type of the obstacle and its adaptation coefficients for equipment, used to identify the differentiated constraint strength of terrain on different combat units. Multi-dimensional semantic descriptors are data containers that define the tactical characteristics of network connection edges, including the current equipment type, dynamically updated path access weights, and line-of-sight observation weights, used to quantify the operational applicability of paths. Three-layer semantic reorganization refers to the progressive structural transformation process of the combat unit network, including a location mapping layer, a capability integration layer, and a relationship chain layer, used to generate a machine-parseable battlefield knowledge graph.

[0129] In this embodiment, step 1061 first maps the coefficients to the corresponding terrain obstruction nodes in the combat unit network. By calling the obstacle type encoding library in the pre-stored equipment database, a structured semantic tag is generated and bound to the node. For example, if an enemy tank company is detected located on the eastern slope of Hill B with a slope of 25°, the system matches the "tank-steep slope" coefficient of 0.2 from the adaptation relationship library, generating the tag {obstacle type: steep slope, tank adaptation coefficient: 0.2} for the "eastern slope node" in the Hill B network, indicating that the terrain strongly restricts tank movement.

[0130] Next, using the semantic tags generated by 1061, combined with dynamically adjusted path access weights and line-of-sight observation weights, the equipment type, adjusted access weights, and observation weights are integrated into a multi-dimensional semantic descriptor for the connecting edge. For example, the road from the eastern slope of Hill B to the summit, i.e., connecting edge E12, has reduced tank accessibility due to its slope. The system updates its descriptor to {Equipment Type: Tank, Access Weight: 0.5, Observation Weight: 0.4}, quantifying the practical limitations of this path.

[0131] Finally, through step 1063, a direct mapping relationship between equipment and spatial coordinates is constructed based on the first-layer reorganization, binding the identified equipment nodes to their actual geographical coordinates to form a precisely locatable "equipment-location" layer. Next, the second-layer reorganization is performed, traversing the multi-dimensional semantic descriptors of all connecting edges in the combat unit network, extracting and integrating the passage weight parameters and observation weight parameters to generate a global path capability set, quantifying the mobility and reconnaissance conditions across the entire battlefield. Finally, the third-layer reorganization is executed, associating the adaptation relationship chain between equipment nodes and terrain obstruction nodes, explicitly expressing the constraints between equipment and multiple terrain obstacles through a chain structure. Taking the B highland battlefield as an example, the tank company is bound to coordinates (X:135, Y:278). Statistics of the five paths across the entire domain show an average passage weight of 0.55, indicating overall passage is obstructed, and simultaneously constructing... The relational chain is ultimately output as a reconstructed network as a semantic understanding result, directly revealing the core tactical semantics of "steep slopes and minefields co-restraining tank mobility".

[0132] In practical application, in the jungle swamp battlefield of Area C, the system detected three wheeled armored vehicles concealed at the edge of the northern wetland through multispectral reconnaissance. Based on the pre-stored equipment database, a "wheeled armored vehicle-swamp" adaptation coefficient of 0.3 indicates severely restricted mobility. A semantic label {obstacle type: flooded swamp, armored vehicle adaptation coefficient: 0.3} was generated for the "northern wetland node" in the combat unit network. Combining this semantic label, the connection edge parameters of the armored vehicle's movement path were dynamically adjusted. Due to swamp infiltration, the original passage weight of the muddy path on the east side (connection edge E08) was reduced from 0.6 to 0.2, and the observation weight was reduced from 0.9 to 0.4 due to reed obstruction. This process was then used to generate a multi-dimensional... The semantic descriptor is defined as follows: {Equipment Type: Wheeled Armored Vehicle, Passage Weight: 0.2, Observation Weight: 0.4}. Finally, a three-layer reorganization is performed. The first layer binds the three armored vehicles to their actual coordinates (X:320, Y:115), (X:325, Y:118), and (X:318, Y:122). The second layer integrates the descriptors of 12 paths within the area, calculating a global passage weight average of 0.25, indicating that muddy conditions cause difficulty in passage throughout the region, and an observation weight average of 0.45, indicating severe vegetation obstruction. The third layer constructs a relational chain. The final output reconstructs the network and generates core semantics: "The armored vehicle detachment is doubly restricted by the swamp wetlands and hilly terrain, and there is only a possibility of breaking through on the dirt road in the northwest direction." This provides the command end with a feasible dynamic decision map that includes spatial positioning, path capability assessment and terrain constraints.

[0133] The above-mentioned overall solution 106 transforms raw battlefield data into a structured knowledge graph by injecting semantic tags into network nodes, assigning multi-dimensional descriptors to connecting edges, and implementing three-layer semantic reorganization: equipment locations can be accurately traced, path accessibility and observation capabilities can be quantified globally, and the tactical association between equipment and terrain obstacles is made explicit. Ultimately, it generates semantic understanding results that combine spatial accuracy and tactical interpretability, directly supporting efficient decision-making.

[0134] The following is a complete example for steps 101-106, such as Figure 2 As shown, firstly, a swarm of drones and reconnaissance satellites collaboratively acquire multi-dimensional visual data of the entire battlefield, including 0.5m resolution visible light images and lidar terrain scanning data, as well as fixed-band electromagnetic wave data including Ku-band SAR radar reflection signals and 8-14μm infrared thermal spectra. These data are then fused to generate a spatiotemporally aligned raw dataset. The terrain texture features accurately mark the distribution of roads and rocks, while the thermal radiation features capture 17 anomalous heat sources, providing a multi-source fusion basis for subsequent analysis.

[0135] Then, based on geographic grid spatiotemporal coding, the original dataset is distributed to four distributed nodes for parallel processing. Node 1 generates a digital elevation model by structuring the lidar point cloud. Node 2 identifies equipment outline parameters in visible light images through the YOLOv5 network and extracts a rectangular target with a length of 7.2m and a width of 3.1m. Node 3 analyzes SAR images to extract environmental occlusion features and extracts a shrub density of 68% in the northeast region. Node 4 integrates and outputs the enemy equipment outline parameter set and the environmental occlusion feature matrix.

[0136] Next, taking advantage of the environmental characteristic of 68% shrub density, the thermal signal of the shielded area was optimized using a thermal radiation attenuation model, reducing the original value of 850 to 210 and eliminating false alarm targets. Simultaneously, the contour parameters were matched with the equipment database, and the matching degree with the wheeled armored vehicle was calculated based on the Euclidean distance, achieving a matching degree of 98.2%, confirming that the wheeled armored vehicle was accurately identified as a concealed target.

[0137] Secondly, by integrating terrain texture features (road hardness coefficient 0.8), optimized thermal radiation features (armor vehicle thermal signal value 210), and equipment spatial distribution, a battlefield entity relationship model is constructed. Then, the armor vehicle location nodes are logically associated with terrain obstacle nodes to generate a combat unit network containing 12 equipment nodes, 9 terrain nodes, and 27 connecting edges, forming a battlefield element topology framework.

[0138] Next, the identified equipment type, wheeled armored vehicle, is fed back to the entity relation model to dynamically adjust the network parameters. Based on the off-road capability of the armored vehicle, the weight of steep slope path passage is reduced from 0.7 to 0.4, and the equipment terrain adaptation rules in the pre-stored equipment database are activated. The adaptation coefficient of armored vehicles with rocky terrain is 0.3, and the adaptation coefficient with shrub terrain is 0.6, so that the network has tactical semantic expression capabilities.

[0139] Finally, semantic injection is performed based on the adaptation relationship, adding labels {type: granite, armored vehicle adaptation coefficient: 0.3} to the rock node, and generating road connection edge descriptors {equipment: armored vehicle, traffic weight: 0.4, observation weight: 0.6}; finally, through three-layer reorganization, the equipment coordinates (X:135, Y:278) are bound, the average global traffic weight is calculated to be 0.52, and a road connection edge descriptor is constructed. The relational chain outputs the semantic understanding result: "The armored convoy is constrained by the strong mobility of the granite terrain, but it can take advantage of the observation cover of the bushes, and the northwest road is the breakout route."

[0140] Figure 3 This application provides a schematic diagram of the structure of a battlefield situation semantic understanding system based on knowledge graphs, as shown in the embodiments below. Figure 3 As shown, the system includes:

[0141] The acquisition module 31 is used to acquire multi-dimensional visual data of the battlefield area and electromagnetic wave reflection data of a fixed band to form an original dataset containing terrain texture features and thermal radiation features.

[0142] The identification module 32 is used to transmit the original dataset to the distributed processing node based on the spatiotemporal coordinate encoding rules, perform structured processing on the original dataset, and perform pattern recognition on the processed dataset to extract the contour parameters of enemy equipment and environmental occlusion features.

[0143] Processing module 33 is used to optimize the effective capture range of the thermal radiation feature in the vegetation-covered area based on the vegetation density distribution data in the environmental shading feature, so as to determine the optimized thermal radiation feature, and to determine the corresponding equipment type of the enemy equipment by comparing the contour parameters of the enemy equipment with the contour parameters corresponding to each equipment type in the pre-stored equipment database.

[0144] Module 34 is established to utilize the terrain texture features, optimized thermal radiation features and spatial correlation of enemy equipment distribution to establish a battlefield entity relationship model, and to logically associate enemy equipment location information with terrain obstacle distribution data to form a combat unit network.

[0145] Feedback module 35 is used to feed back the equipment type to the battlefield entity relationship model in order to adjust the weight parameters in the spatial topology analysis and establish the adaptation relationship between equipment type and terrain obstacles.

[0146] Output module 36 is used to optimize the semantic expression of the combat unit network through the adaptation relationship and output the semantic understanding result.

[0147] Figure 3 The aforementioned knowledge graph-based battlefield situation semantic understanding system can execute... Figure 1 The implementation principle and technical effects of the knowledge graph-based battlefield situation semantic understanding method described in the illustrated embodiment will not be repeated here. The specific methods by which each module and unit of the knowledge graph-based battlefield situation semantic understanding system in the above embodiments perform operations have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0148] In one possible design, Figure 3 The battlefield situation semantic understanding system based on knowledge graphs shown in the embodiment can be implemented as a computing device, such as... Figure 4 As shown, the computing device may include a storage component 41 and a processing component 42;

[0149] The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 42.

[0150] The processing component 42 is used for the above Figure 1 The embodiment describes a battlefield situation semantic understanding method based on knowledge graphs.

[0151] The processing component 42 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0152] Storage component 41 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0153] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0154] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0155] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0156] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0157] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown is a battlefield situation semantic understanding method based on knowledge graph.

[0158] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0159] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0160] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0161] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A battlefield situation semantic understanding method based on knowledge graphs, characterized in that, include: Acquire multi-dimensional visual data of the battlefield area and electromagnetic wave reflection data of fixed bands to form a raw dataset containing terrain texture features and thermal radiation features; Based on the spatiotemporal coordinate encoding rules, the original dataset is transmitted to a distributed processing node, the original dataset is processed in a structured manner, and the processed dataset is subjected to pattern recognition to extract the contour parameters of enemy equipment and environmental occlusion features. Based on the vegetation density distribution data in the environmental shading features, the effective capture range of the thermal radiation features in the vegetation-covered area is optimized to determine the optimized thermal radiation features. The corresponding equipment type of the enemy equipment is determined by comparing the contour parameters of the enemy equipment with the contour parameters corresponding to each equipment type in the pre-stored equipment database. By utilizing the terrain texture features, optimized thermal radiation features, and spatial correlation with the distribution of enemy equipment, a battlefield entity relationship model is established, and the location information of enemy equipment is logically correlated with the distribution data of terrain obstacles to form a combat unit network. The equipment type is fed back to the battlefield entity relationship model to adjust the weight parameters in the spatial topology analysis and to establish the adaptation relationship between equipment type and terrain obstacles. The semantic representation of the combat unit network is optimized through the adaptation relationship, and the semantic understanding results are output.

2. The method according to claim 1, characterized in that, The step of optimizing the semantic representation of the combat unit network through the adaptation relationship and outputting semantic understanding results includes: Based on the adaptation relationship, the adaptation coefficients are mapped to the blocking nodes in the combat unit network to generate semantic tags containing obstacle types and the adaptation coefficients; The equipment type, the adjusted path travel weight parameters, and the line-of-sight observation weight parameters are integrated into a multi-dimensional semantic descriptor for network connection edges; Based on the semantic tags and multi-dimensional semantic descriptors, the combat unit network is reorganized into three layers of semantics. The first layer constructs the mapping relationship between the equipment type and spatial coordinates. The second layer integrates the set of passage weight parameters and observation weight parameters of all network connection edges. The third layer associates the adaptation relationship chain between the equipment nodes and the blocking nodes in the combat unit network and outputs the reorganized network structure as the semantic understanding result.

3. The method according to claim 1, characterized in that, The step of feeding back the equipment type to the battlefield entity relationship model to adjust the weight parameters in the spatial topology analysis and establish the adaptation relationship between equipment type and terrain obstacles includes: The equipment type is fed back to the battlefield entity relationship model, and the mobility and observation capabilities of the corresponding equipment are obtained based on the attribute table in the pre-stored equipment database. The mobility capability level is applied to the connecting edges in the combat unit network, and the path passage weight parameters in the spatial topology analysis are adjusted. At the same time, the observation capability level is applied to the connecting edges in the combat unit network, and the line-of-sight observation weight parameters in the spatial topology analysis are adjusted. By combining the terrain obstacle type identifiers in the equipment type and battlefield entity relationship model, an adaptation relationship between equipment type and terrain obstacle is established, and a passability adaptation coefficient for various types of equipment in different terrain obstacle environments is generated.

4. The method according to claim 1, characterized in that, The process involves establishing a battlefield entity relationship model by utilizing the terrain texture features, optimized thermal radiation features, and spatial correlation with the distribution of enemy equipment, and logically associating enemy equipment location information with terrain obstacle distribution data to form a combat unit network, including: The location coordinates and type identifiers of obstacles are extracted from the terrain texture features, and the spatial coordinate set of enemy equipment is located based on the optimized thermal radiation features. Based on the spatial correlation of enemy equipment distribution reflected by the set of spatial coordinates, the straight-line distance between each pair of enemy equipment is calculated. The location coordinates and type identifiers of the obstacles are combined to determine whether there is visual obstruction. The judgment result, which includes direct visibility relationship and indirect connectivity relationship, is output. Based on the judgment result and the pre-stored equipment database, a battlefield entity relationship model is formed. Based on the battlefield entity relationship model, the spatial coordinates of enemy equipment are transformed into network nodes, the directly visible relationships are transformed into unweighted connecting edges, and the indirect connected relationships are transformed into connecting edges with detour path length weights. At the same time, obstacles are mapped as blocking nodes and associated edges are established with the affected equipment nodes, generating a combat unit network containing the equipment nodes, blocking nodes, and connecting edges.

5. The method according to claim 1, characterized in that, Based on the vegetation density distribution data in the environmental shading features, the effective capture range of the thermal radiation features in the vegetation-covered area is optimized to determine the optimized thermal radiation features. The corresponding equipment type of the enemy equipment is determined by comparing the contour parameters of the enemy equipment with the contour parameters corresponding to each equipment type in a pre-stored equipment database. This includes: Based on the vegetation density distribution data in the environmental shading features, the corresponding signal attenuation coefficient is obtained through a preset mapping relationship between density values ​​and signal attenuation coefficients. The coordinate range of the environmental shading feature indicated by the vegetation-covered area is located, and numerical enhancement calculation is performed according to the signal attenuation coefficient. The enhanced thermal radiation feature value is merged with the original thermal radiation feature value of the non-vegetation-covered area, and the optimized thermal radiation feature is output. Based on the contour parameters of the enemy equipment, the coordinates of the boundary points are extracted. Based on the distance and angle changes of adjacent boundary points, a contour shape descriptor is generated. At the same time, the standard contour templates of all equipment types in the pre-stored equipment database are traversed to extract the template shape descriptor from the standard contour template. Calculate the matching degree between the contour shape descriptor and the template shape descriptor, and select the equipment type corresponding to the standard contour template with the highest matching degree score as the judgment result.

6. The method according to claim 1, characterized in that, The method based on spatiotemporal coordinate encoding rules transmits the original dataset to distributed processing nodes, performs structured processing on the original dataset, and performs pattern recognition on the processed dataset to extract the contour parameters of enemy equipment and environmental occlusion features, including: Based on the spatiotemporal coordinate encoding rules, a unique spatiotemporal identifier is generated for each data unit of the original dataset. The data units are then distributed to distributed processing nodes according to the spatiotemporal identifier, and the nodes are sorted by spatial unit coordinates to form a structured dataset. In the structured dataset, a continuous region with thermal radiation characteristics significantly higher than the surrounding average is located and marked as a potential equipment region. The change points of continuous feature change on the boundary of the potential equipment region are tracked, and the contour parameters of the enemy equipment are generated based on the change points. Simultaneously, the density and vertical distribution characteristics of vegetation in the terrain texture features are analyzed to generate environmental occlusion features that quantify the degree of obstruction to vision and energy propagation.

7. The method according to claim 1, characterized in that, The acquisition of multi-dimensional visual data and fixed-band electromagnetic wave reflection data of the battlefield area forms a raw dataset containing terrain texture features and thermal radiation features, including: Visible light imaging equipment deployed in the battlefield area collects multi-dimensional visual data covering the morphology of surface materials, uneven structure, and surface details of man-made objects. Infrared detection equipment receives electromagnetic wave reflection data radiated by objects on the battlefield in a fixed band. The multi-dimensional visual data is divided into multiple spatial units, and the geometric morphological features of the land surface structure are extracted as terrain texture features for each spatial unit. The electromagnetic wave reflection data is divided into the same spatial units, and the average thermal radiation intensity value of each spatial unit is extracted as the thermal radiation feature. The terrain texture features and thermal radiation features of each spatial unit are combined into a data unit, and all data units are arranged according to their spatial location to form the original dataset.

8. A battlefield situation semantic understanding system based on knowledge graphs, characterized in that, include: The acquisition module is used to acquire multi-dimensional visual data of the battlefield area and electromagnetic wave reflection data of a fixed band, forming a raw dataset containing terrain texture features and thermal radiation features. The recognition module is used to transmit the original dataset to the distributed processing node based on the spatiotemporal coordinate encoding rules, perform structured processing on the original dataset, and perform pattern recognition on the processed dataset to extract the contour parameters of enemy equipment and environmental occlusion features. The processing module is used to optimize the effective capture range of the thermal radiation feature in the vegetation-covered area based on the vegetation density distribution data in the environmental shading feature, so as to determine the optimized thermal radiation feature, and to determine the corresponding equipment type of the enemy equipment by comparing the contour parameters of the enemy equipment with the contour parameters corresponding to each equipment type in the pre-stored equipment database. A module is established to utilize the terrain texture features, optimized thermal radiation features, and spatial correlation of enemy equipment distribution to establish a battlefield entity relationship model, and to logically associate enemy equipment location information with terrain obstacle distribution data to form a combat unit network. The feedback module is used to feed back the equipment type to the battlefield entity relationship model in order to adjust the weight parameters in the spatial topology analysis and establish the adaptation relationship between the equipment type and the terrain obstacles. The output module is used to optimize the semantic representation of the combat unit network through the adaptation relationship and output the semantic understanding result.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a knowledge graph-based battlefield situation semantic understanding method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements a battlefield situation semantic understanding method based on a knowledge graph as described in any one of claims 1 to 7.

Images