Target damage assessment analysis reorganizing system based on large model

Abstract

The invention relates to the technical field of target damage assessment, in particular to a target damage assessment, analysis and reorganization system based on a large model. The system comprises a combat target data cleaning module, a combat target feature fusion module, a target damage evaluation model establishment module and a target damage visual reorganization module, and can acquire a corresponding combat target geographic situation image through satellite reconnaissance and integrate a combat information report and a combat instruction corresponding to a target combat unit. Obtaining combat target text data; performing data specification cleaning on the combat target geographic situation image and the combat target text data to obtain a combat target situation standard image and combat target standard text data; and constructing a corresponding target feature analysis large model, carrying out combat target driving analysis, and carrying out target damage assessment analysis and target damage visual reorganization at the same time to generate a combat target damage visual reorganization result. According to the method, the accuracy of joint combat target damage assessment can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of target damage assessment, and in particular to a target damage assessment analysis and compilation system based on a large model. Background Art

[0002] Target Damage Assessment (TDA) is a key technology in the military field, used to assess the degree of damage caused to enemy targets in combat. In recent years, new reconnaissance methods such as remote sensing technology, drone monitoring, and satellite images have been widely used in battlefield situation awareness, which provides rich real-time data for target damage assessment. At present, target damage assessment methods based on data analysis have begun to be widely used. Especially driven by big data and artificial intelligence technologies, related algorithms are gradually developing in the direction of intelligence and automation. Based on image recognition, machine learning, deep learning and other technologies, researchers have tried to improve the efficiency and accuracy of damage assessment through automation. However, existing technologies still have many limitations. Most of them rely on a specific field or a specific type of target data, making it difficult to comprehensively evaluate different types of targets and damage situations. At the same time, there is still a lack of effective compilation methods, making it impossible to quickly process massive information and generate accurate target damage assessment results. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a target damage assessment and analysis compilation system based on a large model to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a target damage assessment and analysis system based on a large model is proposed, which includes the following modules:

[0005] The combat target data cleaning module is used to obtain the corresponding combat target geographic situation image through satellite reconnaissance, and integrate the combat intelligence report and combat order corresponding to the target combat unit to obtain the combat target text data; perform data standard cleaning on the combat target geographic situation image and combat target text data to obtain the combat target situation standard image and combat target standard text data;

[0006] The combat target feature fusion module is used to build a corresponding target feature analysis model using a convolutional neural network and a Transformer architecture, and input the combat target situation standard image and combat target standard text data into the target feature analysis model to perform combat target driven analysis, so as to output the corresponding combat target image features and text semantic damage features;

[0007] The target damage assessment model establishment module is used to take the large target feature analysis model as the input layer, and based on the input layer, a preset feature enhancement layer, a fusion inference layer, and an output layer, construct a joint operation target damage assessment model based on the large model;

[0008] The target damage visualization compilation module is used to obtain the combat target data corresponding to the evaluation to be performed, including the combat target situation image and the combat target intelligence text corresponding to the evaluation to be performed; input the combat target data corresponding to the evaluation to be performed into the joint operation target damage assessment model based on the large model for target damage assessment and analysis, so as to predict and output the corresponding target damage degree, target damage range, combat effectiveness loss, and the impact of target damage on the combat situation; perform target damage visualization compilation based on the target damage degree, target damage range, combat effectiveness loss, and the impact of target damage on the combat situation, so as to generate the combat target damage visualization compilation result.

[0009] Furthermore, the combat target data cleaning module includes the following functions:

[0010] Obtain the combat target geographical situation image corresponding to the target area through satellite reconnaissance;

[0011] Integrate the combat intelligence reports and combat instructions corresponding to each target combat unit within the target area to obtain combat target text data;

[0012] Perform local pixel blurriness analysis on the combat target geographical situation image to obtain the local pixel blurriness of the combat target image;

[0013] Based on the local pixel blurriness of the combat target image, perform blurring denoising processing and standardization on the combat target geographical situation image to obtain the combat target situation standard image;

[0014] Perform text error correction and standard cleaning on the combat target text data to obtain the combat target standard text data.

[0015] Furthermore, the text error correction and standard cleaning of the combat target text data includes:

[0016] Perform part-of-speech analysis and annotation on each word in the combat target text data to obtain the combat target text part-of-speech annotation data;

[0017] Based on the combat target text part-of-speech annotation data, perform grammar error correction analysis between each word in the combat target text data to correct the corresponding misspelled words and grammar errors and unify the vocabulary expression, and at the same time remove the corresponding special characters, punctuation marks, and meaningless stop words to obtain the combat target text correction data;

[0018] Standardize the data of different sources and formats corresponding to the combat target text correction data through a general combat domain data standard and coding system to obtain combat target standard text data.

[0019] Furthermore, the combat target feature fusion module includes the following functions:

[0020] Use a convolutional neural network and combine it with the target image to construct a corresponding target image feature extraction branch to output the combat target image features corresponding to the target image;

[0021] Use the Transformer architecture and combine it with the text data to construct a corresponding target semantic damage feature extraction branch to output the combat target semantic damage features corresponding to the target text;

[0022] Connect the target image feature extraction branch and the target semantic damage feature extraction branch to the corresponding cross-modal feature association module to construct a corresponding target feature analysis large model to realize the association and fusion between the target image features and the text semantics;

[0023] Input the combat target situation standard image and the combat target standard text data into the target feature analysis large model for combat target-driven analysis to output the corresponding combat target image features and text semantic damage features.

[0024] Furthermore, the target image feature extraction branch is specifically composed of a large-scale 5x5 convolutional layer, a small-scale 3x3 convolutional layer, and a 1x1 pooling layer. Among them, in the large-scale 5x5 convolutional layer, the combat target geometric shape and texture features corresponding to the target image are extracted. In the small-scale 3x3 convolutional layer, the combat target position change and surrounding environment situation features corresponding to the target image at different time periods are extracted, and the features corresponding to the large-scale and small-scale are dimension-reduced in the 1x1 pooling layer.

[0025] Furthermore, the target semantic damage feature extraction branch is specifically as follows:

[0026] Perform combat target semantic word segmentation on the text data in the corresponding input layer within the Transformer architecture to generate a combat target semantic word segmentation set;

[0027] Use the corresponding military semantic graph in the corresponding text encoding layer within the Transformer architecture to perform semantic encoding representation on each combat target word or phrase in the combat target semantic word segmentation set, so as to find the combat domain concept corresponding to each word or phrase in the military semantic graph, and embed the relevant attributes and relationships corresponding to the combat domain concept into the corresponding combat target word or phrase to generate a combat target semantic encoding representation;

[0028] By extracting target damage features from the semantic encoding representation of combat targets in the corresponding last layer in the Transformer architecture, the target type, combat mission and logical chain of target damage description events corresponding to the combat targets are extracted. By introducing additional constraints and reward mechanisms, the feature extraction branch is strengthened to extract the corresponding semantic features of combat targets. If the corresponding combat target damage features are accurately extracted, positive rewards are given, while if wrong or missing features are extracted, penalties are given. The corresponding damage feature vector representation is obtained through the forward propagation calculation corresponding to the last layer to output the combat target image features corresponding to the target image.

[0029] Furthermore, the target damage visualization compilation module includes the following functions:

[0030] Obtain the corresponding combat target data to be evaluated, including the corresponding combat target situation image and combat target intelligence text;

[0031] The combat target data to be evaluated is input into the input layer corresponding to the joint combat target damage assessment model based on the large model to extract combat target features, so as to extract and receive the corresponding combat target image features and text semantic damage features that have been preprocessed and feature extracted;

[0032] The self-attention mechanism in the corresponding feature enhancement layer of the joint combat target damage assessment model based on the large model is used to fuse the combat target image features and the text semantic damage features, so as to fuse and highlight the relevant features corresponding to the combat target damage, and obtain the combat target damage-related features, including the combat target running speed, combat target combat capability, combat target defense level, combat target terrain concealment level and combat target position change;

[0033] Calculate the target damage degree based on the damage-related characteristics of the combat target to obtain the target damage degree; determine the damage range of the target damage degree based on the change of the combat target position to obtain the target damage range; calculate the loss of the combat mission parameters corresponding to the combat target using the combat effectiveness loss calculation formula based on the target damage degree and the target damage range to obtain the combat effectiveness loss; evaluate the impact of the combat geographical situation corresponding to the combat target based on the target damage degree to obtain the impact of the target damage on the combat situation;

[0034] Target damage visualization is performed based on the target damage degree, target damage range, combat effectiveness loss, and the corresponding impact of target damage on the combat situation to generate a combat target damage visualization compilation result.

[0035] Furthermore, the target damage degree is calculated by a target damage degree calculation formula, and the target damage degree calculation formula is specifically:

[0036]

[0037] In the formula, D is the target damage degree, v is the running speed of the combat target, k1 is the running speed weight, C is the combat capability of the combat target, k2 is the combat capability weight, and D f is the defense level of the combat target, k3 is the defense level weight, D f,max is the maximum defense degree corresponding to the combat target unit, H is the concealment degree of the combat target terrain, k4 is the concealment degree weight, S is the change in the combat target position, and k5 is the position change weight.

[0038] Furthermore, the combat effectiveness loss calculation formula is specifically as follows:

[0039]

[0040] In the formula, L is the combat effectiveness loss, D is the target damage degree, R is the real-time completion progress of the combat target task, and R max is the maximum completion progress of the combat mission corresponding to the combat target, E is the combat target environment situation, T is the duration of the combat operation, and F D is the size of the target damage range, e is the base of the natural logarithm, and η is the correction coefficient for the loss of combat effectiveness.

[0041] Furthermore, the target damage visualization compilation according to the target damage degree, target damage range, combat effectiveness loss and the impact of target damage on the combat situation includes:

[0042] Generate a heat map of target damage severity corresponding to the target area based on the target damage severity, so as to intuitively present the damage severity corresponding to different target areas;

[0043] According to the target damage range, a three-dimensional dynamic model of the target damage range corresponding to the target area is produced to clearly display the target damage status corresponding to the combat target and the surrounding environment;

[0044] Based on the impact of combat effectiveness loss on the target damage on the combat situation, the effectiveness loss impact is drawn to generate a trend chart of combat effectiveness loss situation impact corresponding to the target area, and to help accurately grasp the impact trend of target damage on the overall combat situation;

[0045] The target damage degree heat map corresponding to the target area, the three-dimensional dynamic model of the target damage range and the combat effectiveness loss situation impact trend map are visually compiled to generate a combat target damage visualization compilation result.

[0046] Beneficial effects of the present invention:

[0047] The target damage assessment analysis and compilation system based on the large model proposed by the present invention is composed of a combat target data cleaning module, a combat target feature fusion module, a target damage assessment model building module and a target damage visualization compilation module. Compared with the prior art, the beneficial effect of the present application is that satellite reconnaissance can reflect the geographical situation of the battlefield and the spatial position and movement of the combat target in real time through high-resolution images, and provide accurate geographical data for subsequent combat decisions. These satellite images can reveal the enemy's military facilities, deployment status and battlefield environment, ensuring that the commander can obtain timely and accurate intelligence. In addition, the intelligence report and instructions of the combat target are combined to provide background information at the tactical and strategic levels to help analyze the intentions and trends of the enemy's actions. The process of data standard cleaning is crucial to ensuring the consistency and accuracy of the data. The original image data and text data often contain noise, incomplete information or inconsistent formats. The data cleaning process can remove these invalid information and unify the data format to make it more suitable for subsequent model analysis. Through the cleaned combat target situation standard image and standard text data, the efficiency and accuracy of subsequent model processing data can be improved, providing a more reliable basis for target damage assessment. Secondly, by using convolutional neural networks (CNN) and Transformer architecture to build a large model for target feature analysis, standard images and standard text data of combat targets are input into the model for analysis. Convolutional neural networks (CNN) are good at processing image data. They can extract fine spatial features from the geographical situation images of combat targets, such as the shape, structure, location and other information of the target. These spatial features are very important for judging the type, importance and potential threats of the target. Through the deep learning characteristics of the CNN model, the system can automatically extract important features from the image, reducing the tedious work of manual feature design, and can discover complex patterns and potential attack targets. The Transformer architecture performs well in processing text data, especially in capturing In terms of long-distance dependencies, intelligence reports and instructions of combat targets often contain a large amount of text information, including the enemy's strategic intentions, tactical layout and current combat status. The Transformer model can effectively capture these long-distance semantic relationships through the self-attention mechanism and deeply understand the potential meaning in the text data. In this way, by combining the advantages of CNN and Transformer, the target feature analysis model can process image and text data at the same time, thereby conducting a comprehensive analysis of combat targets, and can output the image features and text semantic damage features of combat targets, providing detailed basic data for subsequent damage assessments, so that decision makers can have a more comprehensive understanding of the situation of combat targets, thereby achieving a comprehensive assessment of different types of targets and damage situations.Then, by taking the target feature analysis model as the input layer, the entire damage assessment model can receive more detailed and comprehensive target data. These data not only include the target's geographic information, but also cover the target's textual semantic information, comprehensively and deeply describing the various dimensions of the combat target. Effectively transmitting these data to the assessment model can provide strong support for subsequent damage assessment. The feature enhancement layer can enhance the model's attention to target features, improve the model's sensitivity to important features, and ensure that the model can effectively capture key data. The fusion reasoning layer can integrate data from multiple different sources (such as images and text) to form a comprehensive judgment. This multi-level processing method can enhance the model's reasoning ability and enable it to more accurately assess the damage effects of combat targets. Finally, by obtaining the situation images and intelligence texts of the combat targets, we can fully understand the current status and position of the targets, which provides the most direct input for subsequent damage assessment. The data includes the latest intelligence on enemy targets, which can ensure that the assessment model has a full understanding of the current combat situation. The joint combat target damage assessment model based on the large model can accurately perform damage assessment analysis by inputting the data to be evaluated. Through comprehensive predictions of the target damage level, damage range, and combat effectiveness loss, it can help commanders understand the impact of different tactical plans on the targets, so that the model can achieve more effective reorganization methods to enhance the accuracy and predictability of combat targets, thereby improving the accuracy of target damage assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0049] Figure 1 It is a module schematic diagram of the target damage assessment and analysis compilation system based on a large model of the present invention;

[0050] Figure 2 for Figure 1 Functional flow diagram of the combat target data cleaning module;

[0051] Figure 3 for Figure 1 Schematic diagram of the functional flow of the combat target feature fusion module. DETAILED DESCRIPTION

[0052] The technical system of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0053] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor systems and / or microcontroller systems.

[0054] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0055] To achieve this, please refer to Figures 1 to 3 The present invention provides a target damage assessment analysis and compilation system based on a large model, and the system includes the following modules:

[0056] The combat target data cleaning module is used to obtain the corresponding combat target geographic situation image through satellite reconnaissance, and integrate the combat intelligence report and combat order corresponding to the target combat unit to obtain the combat target text data; perform data standard cleaning on the combat target geographic situation image and combat target text data to obtain the combat target situation standard image and combat target standard text data;

[0057] The combat target feature fusion module is used to build a corresponding target feature analysis model using a convolutional neural network and a Transformer architecture, and input the combat target situation standard image and combat target standard text data into the target feature analysis model to perform combat target driven analysis, so as to output the corresponding combat target image features and text semantic damage features;

[0058] The target damage assessment model building module is used to use the target feature analysis large model as the input layer, and build a joint combat target damage assessment model based on the large model based on the input layer and the preset feature enhancement layer, fusion reasoning layer and output layer;

[0059] The target damage visualization compilation module is used to obtain the corresponding combat target data to be evaluated, including the corresponding combat target situation image and combat target intelligence text to be evaluated; the corresponding combat target data to be evaluated is input into the joint combat target damage assessment model based on the large model to perform target damage assessment analysis to predict and output the corresponding target damage degree, target damage range, combat effectiveness loss and the corresponding impact of target damage on the combat situation; target damage visualization compilation is performed according to the target damage degree, target damage range, combat effectiveness loss and the corresponding impact of target damage on the combat situation to generate a combat target damage visualization compilation result.

[0060] In the embodiment of the present invention, please refer to Figure 1 FIG. 1 is a schematic diagram of the modules of the target damage assessment analysis and compilation system based on a large model of the present invention. In this example, the target damage assessment analysis and compilation system based on a large model includes the following modules:

[0061] S1: Combat target data cleaning module, used to obtain the corresponding combat target geographic situation image through satellite reconnaissance, and integrate the combat intelligence report and combat order corresponding to the target combat unit to obtain combat target text data; perform data standard cleaning on the combat target geographic situation image and combat target text data to obtain the combat target situation standard image and combat target standard text data;

[0062] In the embodiment of the present invention, a satellite equipped with a variety of imaging devices is controlled to enter the orbit above the target area. The high-resolution optical camera on the satellite takes pictures of the target area in clear weather to obtain clear images of the ground objects. The synthetic aperture radar is not affected by the weather. It generates images containing hidden target information by transmitting and receiving microwave signals. These image data are transmitted to the ground control center for splicing and fusion to obtain the geographical situation image of the combat target. At the same time, a dedicated data collection network is established and connected to the communication system of the target combat unit. Each combat unit uploads the combat intelligence report (such as the location and equipment of the enemy troops) and combat instructions (such as the arrangement of offensive or defensive tasks) through an encrypted channel. The collection network aggregates these text information to form combat target text data. For image data, an image enhancement algorithm is used to remove noise, adjust brightness, contrast and resolution to a unified standard, thereby obtaining a combat target situation standard image. For text data, a natural language processing tool is used to correct typos, unify terms, remove stop words and special symbols, and finally generate combat target standard text data.

[0063] S2: Combat target feature fusion module, which is used to build a corresponding target feature analysis model using convolutional neural network and Transformer architecture, and input the combat target situation standard image and combat target standard text data into the target feature analysis model to perform combat target driven analysis, so as to output the corresponding combat target image features and text semantic damage features;

[0064] In an embodiment of the present invention, a large model for target feature analysis is built by using Python language combined with the deep learning framework TensorFlow, and a convolutional neural network branch consisting of a large-scale 5x5 convolution layer, a small-scale 3x3 convolution layer, and a 1x1 pooling layer is constructed to process standard images of combat target situations. The large-scale convolution layer extracts the geometric shape and texture features of the target, such as identifying the outline and surface camouflage of a tank; the small-scale convolution layer captures the changes in the position of the target and the situation of the surrounding environment in different periods, such as analyzing the target movement trajectory and the surrounding terrain, and the 1x1 pooling layer reduces the dimension of the extracted features. The text processing branch is built using the Transformer architecture to process the standard text data of the combat target. Through the self-attention mechanism, the semantic relationship between words in the text is analyzed, and semantic features such as combat target type, combat mission, and damage description are extracted. The image and text branches are connected to the cross-modal feature association module, and the combat target situation standard image and combat target standard text data are input. The model realizes cross-modal fusion by calculating the association between the image feature vector and the text semantic feature vector, and finally outputs the combat target image feature and the text semantic damage feature.

[0065] S3: Target damage assessment model building module, used to take the target feature analysis large model as the input layer, and build a joint combat target damage assessment model based on the large model based on the input layer and the preset feature enhancement layer, fusion reasoning layer and output layer;

[0066] In an embodiment of the present invention, the previously constructed target feature analysis large model is used as the input layer of the large-model-based joint combat target damage assessment model, which is responsible for extracting the initial image and text features of the combat target. In the feature enhancement layer, the self-attention mechanism is used to further process the features output by the input layer. For example, higher weights are given to key target parts in image features and core damage descriptions in text features to highlight features related to target damage. In the fusion reasoning layer, the enhanced image features and text features are deeply fused by designing a specific fusion algorithm, such as using matrix multiplication and nonlinear activation functions to explore potential connections between cross-modal features and infer more comprehensive target damage-related information. Finally, an output layer is set, and a suitable output function is constructed according to the results of the fusion reasoning layer to convert the target damage information into an interpretable format to prepare for subsequent evaluation and analysis, thereby completing the construction of the large-model-based joint combat target damage assessment model.

[0067] S4: Target damage visualization compilation module, used to obtain the corresponding combat target data to be evaluated, including the corresponding combat target situation image and combat target intelligence text to be evaluated; input the corresponding combat target data to be evaluated into the joint combat target damage assessment model based on the large model to perform target damage assessment analysis to predict and output the corresponding target damage degree, target damage range, combat effectiveness loss and the corresponding impact of target damage on the combat situation; perform target damage visualization compilation according to the target damage degree, target damage range, combat effectiveness loss and the corresponding impact of target damage on the combat situation to generate a combat target damage visualization compilation result.

[0068] In an embodiment of the present invention, by using a variety of reconnaissance means, such as unmanned aerial vehicle reconnaissance, ground intelligence collection stations, etc., the situation image and combat target intelligence text of the combat target to be evaluated are obtained, and these data are input into the constructed joint combat target damage assessment model based on the large model. The input layer of the model (i.e., the target feature analysis large model) extracts image and text features, and the feature enhancement layer highlights the key features. The fusion reasoning layer deeply fuses and infers the damage information, and finally predicts the target damage degree (such as quantitative scoring according to the damage of the target), the target damage range (demarcating the area according to the change of the target position and the damage degree), the combat effectiveness loss (calculated in combination with the damage degree and combat mission parameters), and the impact of the target damage on the combat situation (analyzing the changes in the damage to the geographical situation, force deployment, etc.), and uses tools such as geographic information system software and data visualization platforms to visualize and organize these evaluation results. For example, different colors and ranges are used on the map to represent the target damage degree and damage range, and charts are used to show the combat effectiveness loss. The impact of the target damage on the combat situation is marked with text and arrows, and finally a combat target damage visualization compilation result is generated to provide an intuitive basis for combat decision-making.

[0069] Furthermore, the combat target data cleaning module includes the following functions:

[0070] Obtain geographical situation images of combat targets corresponding to the target area through satellite reconnaissance;

[0071] By integrating the combat intelligence reports and combat instructions corresponding to each target combat unit in the target area, the combat target text data can be obtained;

[0072] Performing local pixel fuzziness analysis on the combat target geographic situation image to obtain the local pixel fuzziness of the combat target image;

[0073] Based on the local pixel fuzziness of the combat target image, the combat target geographic situation image is fuzzy-denoised and standardized to obtain a combat target situation standard image;

[0074] Perform text error correction, standardization, and cleaning on the combat target text data to obtain the standard combat target text data.

[0075] As an embodiment of the present invention, refer to Figure 2 shown in Figure 1 the functional flow diagram of the combat target data cleaning module in

[0076] S11: Obtain the geographical situation image of the combat target corresponding to the target area through satellite reconnaissance;

[0077] In the embodiment of the present invention, by controlling a reconnaissance satellite equipped with a high-resolution optical imaging device and a synthetic aperture radar (SAR) to fly over the target area, the optical imaging device on the satellite uses light of a specific wavelength band to photograph the target area, obtaining an optical image of the target area, which can clearly present the shape, color, etc. of ground objects, such as identifying buildings, roads, etc. The SAR emits and receives microwave signals to penetrate interference such as clouds and vegetation, generating a radar image of the target area for detecting military facilities hidden under shelters. The image data obtained by the optical imaging device and the SAR are fused and processed to generate a comprehensive and accurate geographical situation image of the combat target, and the image data is transmitted to the ground control center for storage in real time through the satellite data transmission link for subsequent analysis.

[0078] S12: Integrate the combat intelligence reports and combat instructions corresponding to each target combat unit in the target area to obtain the combat target text data;

[0079] In the embodiment of the present invention, by establishing a combat data collection platform connected to the communication systems of each target combat unit in the target area, when a combat unit generates a combat intelligence report, such as information about the location, equipment type, and quantity of enemy troops, or receives a combat instruction from a superior, such as an attack or defense mission arrangement, this text information is transmitted to the data collection platform through an encrypted communication link. The platform uses data parsing algorithms to classify and integrate various types of received text data. For example, intelligence reports on the distribution of enemy tanks from different combat groups are summarized, and the collaborative combat instructions from the superior to each combat unit are sorted together, finally forming comprehensive combat target text data and storing it in the database of the platform to provide a data basis for subsequent text processing.

[0080] S13: Analyze the local pixel blurriness of the combat target geographical situation image to obtain the local pixel blurriness of the combat target image;

[0081] In an embodiment of the present invention, by using an image analysis software, such as the MATLAB image processing toolbox, the geographical situation image of the combat target is processed, and the image is divided into multiple local regions. Each region contains a certain number of pixels. For each local region, the Laplace operator is used to calculate the second derivative of the pixels, and the edge sharpness of the pixels is measured by analyzing the magnitude of the absolute value of the second derivative. For example, in a local region containing the edge of a building, if the absolute value of the second derivative of the pixels is large, it indicates that the edge of this region is clear and the blurriness is low; on the contrary, if the absolute value of the second derivative is small, the blurriness of this region is high. Such calculations are performed on all local regions in the image to obtain the pixel blurriness values of each local region, and then the local pixel blurriness data of the combat target image is generated for evaluating the sharpness status of each part of the image.

[0082] S14: Based on the local pixel blurriness of the combat target image, perform blurring denoising processing and standardization on the geographical situation image of the combat target to obtain the standard image of the combat target situation;

[0083] In an embodiment of the present invention, according to the local pixel blurriness data of the combat target image, the adaptive median filtering algorithm is used to perform blurring denoising on the image. For local regions with high blurriness, the size of the median filtering window is increased to better remove noise and restore image details; for regions with low blurriness, a smaller filtering window size is maintained to avoid over-smoothing. For example, in a blurred forest area in the image, the median filtering window is increased from 3×3 to 5×5 for denoising processing. After denoising, according to the preset image standardization rules, the brightness and contrast of the image are adjusted so that the gray value distribution of the image is within a specific range. At the same time, the resolution of the image is uniformly adjusted to the standard resolution, such as 1024×768 pixels, and finally a standard combat target situation standard image is generated to improve the image quality for subsequent analysis.

[0084] S15: Perform text error correction, specification, and cleaning on the combat target text data to obtain the standard text data of the combat target.

[0085] In an embodiment of the present invention, combat target text data is processed by utilizing a natural language processing toolkit NLTK and a customized military text error correction rule. First, each word in the text is tagged with a part-of-speech tagging function of NLTK, and then typos and grammatical errors in the text are found and corrected based on a pre-constructed military field typo library and grammatical rules. For example, the erroneous spelling of "artillery" is corrected as "fire run", and the grammatical error of "execution mission" is corrected to "execution mission". Then, special characters, punctuation marks, and meaningless stop words in the text are removed, such as "@", "!", "of", "le", etc. Finally, according to military terminology standards, the vocabulary expressions in the text are unified, and non-standard terms are converted into standard terms, such as "chariot" is unified as "tank". After these processes, standardized and accurate combat target standard text data is obtained.

[0086] Furthermore, the text error correction and standard cleaning of the combat target text data includes:

[0087] Performing part-of-speech analysis and tagging on each word in the combat target text data to obtain combat target text part-of-speech tagging data;

[0088] In an embodiment of the present invention, the natural language processing toolkit NLTK (Natural Language Toolkit) is used to perform part-of-speech analysis and tagging operations, the combat target text data is split into sentences, and then the words in each sentence are processed one by one. For example, for the text "enemy tanks quickly drive towards our positions", the part-of-speech tagger in NLTK will mark "enemy" as a noun, "tank" as a noun, "fast" as an adverb, "drive towards" as a verb, "our side" as a pronoun, and "position" as a noun. Through such operations, each word in the entire combat target text data is tagged with part-of-speech, and finally the combat target text part-of-speech tagging data is generated, which provides a basis for subsequent grammatical analysis and text processing.

[0089] Preferably, based on the combat target text part-of-speech tagging data, grammatical error correction analysis is performed on each word in the combat target text data to correct the corresponding typos and grammatical errors and unify the vocabulary expressions, and at the same time remove the corresponding special characters, punctuation marks and meaningless stop words to obtain combat target text correction data;

[0090] In the embodiment of the present invention, by applying a grammar error correction algorithm that combines rules and statistics, based on part-of-speech tagging data, first, through a pre-constructed military domain misspelled word library, the misspelled words in the text are searched for and corrected. For example, if "tank" is misspelled as "tank shell", it can be corrected through word library matching. For grammar errors, according to the common grammar rules of military texts and combined with part-of-speech tagging information, judgment and correction are carried out. For example, for the subject-predicate-object collocation error like "Our soldiers are bravely fighting the enemy", it is adjusted to "Our soldiers are bravely fighting and resisting the enemy" according to parts of speech and grammar rules. At the same time, a preset vocabulary list is used to unify the vocabulary expression, such as unifying "battle vehicle" to "tank", and the stop word list in NLTK is used to remove meaningless stop words such as "de", "le", "zai", etc., as well as special characters and punctuation marks, and finally, purified and standardized combat target text correction data is obtained.

[0091] Preferably, data standardization processing is performed on the corresponding different sources and different formats in the combat target text correction data through a general combat domain data standard and coding system to obtain combat target standard text data.

[0092] In the embodiment of the present invention, by referring to international or industry-wide combat domain data standards, such as military symbol standards like MIL-STD-2525B and related coding systems, the combat target text correction data is processed. For data from different sources (such as intelligence reports, communication records, etc.) and different formats (such as plain text, XML format, etc.), it is first uniformly converted into a standard text format. For combat terms in the text, they are standardized according to the standard term list. For example, the non-standard "missile" is unified into the standard term "guided missile". The numerical values, units, etc. in the text are standardized, such as unifying the distance unit to "meter". According to the coding system, the key information in the text is encoded, such as assigning specific codes to different types of combat targets. Finally, the combat target text correction data is converted into combat target standard text data that conforms to the general standard, which is convenient for subsequent unified processing and analysis in target damage assessment and analysis.

[0093] Furthermore, the combat target feature fusion module includes the following functions:

[0094] Using a convolutional neural network and combining with the target image to construct a corresponding target image feature extraction branch to output the combat target image features corresponding to the target image;

[0095] Using the Transformer architecture and combining with the text data to construct a corresponding target semantic damage feature extraction branch to output the combat target semantic damage features corresponding to the target text;

[0096] Connect the target image feature extraction branch and the target semantic damage feature extraction branch to the corresponding cross-modal feature association module to construct the corresponding target feature analysis large model, so as to realize the association and fusion between the target image feature and the text semantics;

[0097] Input the combat target situation standard image and the combat target standard text data into the target feature analysis large model for combat target-driven analysis, so as to output the corresponding combat target image feature and text semantic damage feature.

[0098] As an embodiment of the present invention, refer to Figure 2 shown, for Figure 1 the functional flow diagram of the combat target feature fusion module in

[0099] S21: Use a convolutional neural network and combine it with the target image to construct the corresponding target image feature extraction branch, so as to output the combat target image feature corresponding to the target image;

[0100] In the embodiment of the present invention, a convolutional neural network is constructed by using the Python programming language combined with the deep learning framework TensorFlow. The target image is input into the network. In the large-scale 5x5 convolutional layer, multiple 5x5 convolutional kernels are used to perform convolutional operations on the image. For example, for a target image containing an enemy tank, the 5x5 convolutional kernel slides on the image. By calculating the dot product of the convolutional kernel and the local area of the image, the geometric shape features of the tank are extracted, such as contour lines, turret shape, etc., and the texture features, such as the camouflage pattern texture on the surface of the tank. In the small-scale 3x3 convolutional layer, for the target images at different time periods, through convolutional operations, the combat target position change features are extracted, such as the coordinate position change of the tank at different times, and the surrounding environment situation features, such as the surrounding terrain and the distribution of other military facilities. Finally, in the 1x1 pooling layer, the features extracted by the large-scale 5x5 convolutional layer and the small-scale 3x3 convolutional layer are dimensionally reduced to reduce the number of features and the computational complexity, and finally the combat target image feature corresponding to the target image is output.

[0101] S22: Use the Transformer architecture and combine it with the text data to construct the corresponding target semantic damage feature extraction branch, so as to output the combat target semantic damage feature corresponding to the target text;

[0102] In an embodiment of the present invention, by using the natural language processing toolkit NLTK and the deep learning framework PyTorch, a target semantic damage feature extraction branch based on the Transformer architecture is built. First, the text data is preprocessed. The tokenization tool in NLTK is used to split the target text into words or phrases, and noise such as stop words is removed. For example, for the text describing the combat situation "The enemy tank was hit by our artillery fire near the bridge and part of its armor was damaged", after tokenization, we get "enemy", "tank", "near", "bridge", "by", "our", "artillery", "fire", "hit", "part", "armor", "damaged", etc. Then, these tokenized text sequences are input into the Transformer architecture. The self-attention mechanism in the Transformer calculates the correlation weights between each word and other words, thereby capturing the semantic relationships in the text. For example, the correlation weights between "tank" and "armor", "damaged" will be relatively high, indicating that they are semantically closely related. Through multiple layers of processing in the Transformer architecture, the semantic features regarding the damage of the combat target in the text are extracted, and finally, the combat target semantic damage features corresponding to the target text are output.

[0103] S23: Connect the target image feature extraction branch and the target semantic damage feature extraction branch to the corresponding cross-modal feature association module to construct the corresponding target feature analysis large model, so as to realize the association and fusion between the target image features and the text semantics;

[0104] In an embodiment of the present invention, by adopting a custom cross-modal feature association module, which is constructed based on matrix operations and fusion algorithms, the previously constructed target image feature extraction branch and the constructed target semantic damage feature extraction branch are connected to the cross-modal feature association module. The combat target image features output by the target image feature extraction branch are represented in the form of feature vectors, and the combat target semantic damage features output by the target semantic damage feature extraction branch are also converted into vector form. The cross-modal feature association module establishes the association relationship between the two by calculating methods such as the cosine similarity between the image feature vector and the semantic feature vector. For example, when the image shows a damaged tank and the text describes "The tank armor is damaged", the cross-modal feature association module can identify the association between the features such as the geometric shape of the tank and the damaged part in the image features and the semantic features such as "tank" and "damaged" in the text, and fuse the two to construct the target feature analysis large model, enabling the model to comprehensively utilize the information of the image and the text for subsequent analysis.

[0105] S24: Input the combat target situation standard image and the combat target standard text data into the target feature analysis large model for combat target-driven analysis, so as to output the corresponding combat target image features and text semantic damage features.

[0106] In the embodiments of the present invention, by collecting a large number of standard images of combat target situations and standard text data of combat targets, these data are strictly screened and labeled to ensure accuracy and consistency. The standard images of combat target situations are input into the target image feature extraction branch of the large model for target feature analysis. In the previous manner, they sequentially pass through a large-scale 5x5 convolutional layer, a small-scale 3x3 convolutional layer, and a 1x1 pooling layer to extract the combat target image features corresponding to the image. At the same time, the standard text data of combat targets are input into the target semantic damage feature extraction branch, and after being processed by the Transformer architecture, the text semantic damage features are extracted. The cross-modal feature association module in the large model for target feature analysis performs association analysis on the extracted image features and text semantic features, further optimizing and integrating these features, and finally outputs the combat target image features and text semantic damage features after being driven by combat target analysis, providing accurate and comprehensive feature data for subsequent target damage assessment.

[0107] Further, the target image feature extraction branch is specifically composed of a large-scale 5x5 convolutional layer, a small-scale 3x3 convolutional layer, and a 1x1 pooling layer. Among them, in the large-scale 5x5 convolutional layer, the geometric shape and texture features of the combat target corresponding to the target image are extracted. In the small-scale 3x3 convolutional layer, the position change of the combat target corresponding to the target image at different time periods and the surrounding environment situation features are extracted. And in the 1x1 pooling layer, the features corresponding to the large scale and the small scale are dimensionally reduced.

[0108] Further, the target semantic damage feature extraction branch is specifically as follows:

[0109] By performing combat target semantic word segmentation on the text data in the corresponding input layer within the Transformer architecture to generate a combat target semantic word segmentation set;

[0110] In the embodiments of the present invention, in the input layer of the Transformer architecture, a professional Chinese word segmentation tool, such as Jieba word segmentation, is used to process the input text data. For the text describing combat targets, sentence-by-sentence word segmentation operations are performed. For example, for the text "The enemy tank cluster is performing an assault mission in the plain area", Jieba word segmentation cuts it into words or phrases such as "enemy", "tank cluster", "in", "plain area", "performing", "assault mission", etc. All the words or phrases after segmentation are summarized to form a combat target semantic word segmentation set. During the word segmentation process, for professional terms in the military field, a military term dictionary is constructed in advance and imported into the word segmentation tool to ensure that professional words such as "tank cluster" and "assault mission" can be accurately segmented, so as to improve the accuracy and professionalism of word segmentation, and finally generate a combat target semantic word segmentation set.

[0111] Preferably, by using the corresponding military semantic graph in the corresponding text encoding layer within the Transformer architecture, semantic encoding representations are performed on each combat target word or phrase in the combat target semantic word segmentation set, so as to find the corresponding combat domain concept in the military semantic graph for each word or phrase, and embed the relevant attributes and relationships corresponding to this combat domain concept into the corresponding combat target word or phrase, so as to generate a combat target semantic encoding representation;

[0112] In an embodiment of the present invention, by loading a pre-constructed military semantic graph in the text encoding layer of the Transformer architecture, the graph contains a large number of military domain concepts and their attributes and relationships. For each word or phrase in the combat target semantic word segmentation set, such as "tank cluster", search and match in the military semantic graph to find the corresponding combat domain concept of "tank cluster", whose attributes include "equipment type" being "armored vehicle", "combat ability" being "having powerful firepower and assault ability", etc., and the relationships include "subordinate to" a certain combat unit, etc. Embed these attributes and relationships into the representation of the word "tank cluster" in a specific encoding manner, such as vector encoding. Perform such operations on all words or phrases in the combat target semantic word segmentation set, and finally generate a combat target semantic encoding representation containing rich semantic information. For example, after encoding, "tank cluster" not only retains its own meaning but also carries relevant attribute and relationship information in the military field for subsequent analysis.

[0113] Preferably, by performing target damage feature extraction on the combat target semantic encoding representation in the corresponding last layer within the Transformer architecture, the target type, combat task, and the event logic chain of the target damage description corresponding to the combat target are extracted, and by introducing additional constraint conditions and reward mechanisms, the extraction ability of the feature extraction branch for the combat target semantic features is strengthened. If the corresponding combat target damage features are accurately extracted, a positive reward is given, while if incorrect or missing features are extracted, a penalty is given, and the corresponding damage feature vector representation is obtained through the forward propagation calculation of the corresponding last layer, so as to output the combat target image features corresponding to the target image.

[0114] In the embodiment of the present invention, the semantic coding representation of the combat target is processed by using a designed feature extraction algorithm at the last layer of the Transformer architecture. For the target type, it is judged by identifying the key information in the words or phrases. For example, "tank cluster" clearly indicates that its target type is an armored combat target. For the combat mission, the words and relationships related to the mission in the semantic coding representation are analyzed. If there is "execution assault mission", the combat mission is determined to be assault. When extracting the logical chain of the target damage description event, the logical connections such as the causal relationship between the elements in the semantic coding representation are sorted out, and additional constraints are introduced, such as stipulating the extracted target damage features. It must conform to military common sense and actual combat conditions, and establish a reward mechanism. If features such as target type, combat mission, and logical chain of target damage description events are accurately extracted, positive rewards will be given, such as increasing the weight of the feature extraction branch; if incorrect or omitted features are extracted, such as misjudging the target type of a "tank cluster" as an aerial target, penalties will be given, such as reducing the weight. Finally, through forward propagation calculations, the extracted features are converted into damage feature vector representations, and the output is used as the combat target image features corresponding to the target image. For example, after a series of calculations, a feature vector containing information such as target type and combat mission is generated, providing key data for subsequent target damage assessments.

[0115] Furthermore, the target damage visualization compilation module includes the following functions:

[0116] Obtain the corresponding combat target data to be evaluated, including the corresponding combat target situation image and combat target intelligence text;

[0117] In an embodiment of the present invention, combat target data to be evaluated are obtained through a variety of intelligence collection means, and high-definition cameras carried by drones, satellites, etc. are used to shoot combat areas to obtain combat target situation images. These images cover information such as the appearance and surrounding environment of the combat target. For example, the architectural layout, facilities and equipment of the enemy's military base are photographed. At the same time, combat target intelligence texts are obtained through intelligence personnel collection, communication monitoring, etc. The text content includes detailed information on the combat target, such as the type of combat target (tank, aircraft, etc.), combat capability parameters, running speed and other related intelligence data. The obtained combat target situation images and combat target intelligence texts are sorted and stored to finally form the combat target data to be evaluated.

[0118] Preferably, the combat target data to be evaluated is input into the input layer corresponding to the joint combat target damage assessment model based on the large model to extract combat target features, so as to extract and receive the corresponding combat target image features and text semantic damage features that have been preprocessed and feature extracted;

[0119] In an embodiment of the present invention, by inputting the previously obtained combat target data into the input layer of the joint combat target damage assessment model based on a large model, for the combat target situation image, image preprocessing techniques are used, such as normalizing the brightness, contrast, etc. of the image, and then a convolutional neural network (CNN) is used to extract features from the image. For example, features such as the shape, size, and color of the combat target in the image are extracted through multiple convolutional layers and pooling layers. For the combat target intelligence text, text preprocessing is first performed, including removing noise (such as special characters, useless punctuation, etc.), word segmentation, etc. Then, word embedding techniques in natural language processing (such as Word2Vec, GloVe, etc.) are used to convert the text into a vector representation, and then semantic damage features in the text, such as semantic information related to the weaknesses and defense capabilities of the combat target, are extracted through a recurrent neural network (RNN) or its variants (such as LSTM, GRU). Finally, the combat target image features and text semantic damage features that have undergone preprocessing and feature extraction are extracted and received.

[0120] Preferably, the self-attention mechanism in the corresponding feature enhancement layer of the joint combat target damage assessment model based on a large model is used to fuse the combat target image features and text semantic damage features to fuse and highlight the relevant features corresponding to the combat target damage, obtaining combat target damage-related features, including the running speed of the combat target, the combat ability of the combat target, the defense degree of the combat target, the terrain concealment degree of the combat target, and the position change of the combat target.

[0121] In an embodiment of the present invention, in the feature enhancement layer of the joint combat target damage assessment model based on a large model, the self-attention mechanism is used to fuse the extracted combat target image features and text semantic damage features. The image features and text features are respectively represented as feature vectors. By calculating the attention weights between different feature vectors, the model can pay attention to the important features related to the combat target damage. For example, when calculating the attention weights, higher weights are given to the text features describing the defense degree of the combat target and the defense facility features shown in the image to highlight their importance in the damage assessment. These features are fused by weighted summation to obtain combat target damage-related features. Through further analysis and extraction of the fused features, specific damage-related features such as the running speed of the combat target, the combat ability of the combat target, the defense degree of the combat target, the terrain concealment degree of the combat target, and the position change of the combat target are determined.

[0122] Preferably, calculate the target damage degree for the characteristics related to the damage of the combat target to obtain the target damage degree; determine the damage range for the target damage degree based on the change in the position of the combat target to obtain the target damage range; calculate the loss of the combat mission parameters corresponding to the combat target using the combat effectiveness loss calculation formula based on the target damage degree and the target damage range to obtain the combat effectiveness loss; evaluate the impact on the combat geographical situation corresponding to the combat target based on the target damage degree to obtain the impact of the target damage on the combat situation;

[0123] In the embodiments of the present invention, a suitable calculation formula for the target damage degree is constructed by combining the operating speed of the combat target, the combat ability of the combat target, the defense level of the combat target, the maximum defense level corresponding to the combat target unit, the terrain concealment level of the combat target, the magnitude of the position change of the combat target, and the corresponding weights for quantitative calculation. Additionally, the calculation formula for the target damage degree can also use any target damage assessment algorithm in the art to replace the process of calculating the target damage degree, and is not limited to this calculation formula for the target damage degree. For example, the weight of the combat target defense level is set to 0.4, the weight of the combat ability is set to 0.3, the weight of the operating speed is set to 0.1, and the weight of the terrain concealment level is set to 0.2. Quantitative calculation is performed based on the specific values of these characteristics. For example, the defense level is represented by the intact rate of the defense facilities, and the combat ability is represented by the comprehensive score of the performance parameters of the weapons and equipment and the quality of the personnel, so as to quantitatively obtain the target damage degree. The position change information of the combat target is obtained. This information can be obtained in real time through means such as satellite positioning systems and unmanned aerial vehicle reconnaissance. Based on the target damage degree, the influence radius corresponding to different damage degrees is set. For example, when the target damage degree is between 0 and 0.3, the influence radius is 100 meters; when it is between 0.3 and 0.6, the influence radius is 200 meters; when it is between 0.6 and 1, the influence radius is 300 meters. According to the initial position and the position change trajectory of the combat target, combined with the influence radius corresponding to the target damage degree, a circular area with the target position as the center is drawn on the map to determine the damage range. If the initial position of a combat target is at the map coordinates (100, 100) and it moves to (150, 150) after a period of time, and its target damage degree is 0.6, then a circular area with (150, 150) as the center and 300 meters as the radius is drawn, and this area is the target damage range.At the same time, by combining the target damage degree and the target damage range, the combat mission parameters corresponding to the combat target (including the real-time completion progress of the combat target task, the maximum completion progress of the combat mission corresponding to the combat target, the environmental situation of the combat target, and the duration of the combat operation) and combining the corresponding correction coefficient, a suitable combat effectiveness loss calculation formula is formed for calculation to quantify the combat effectiveness loss corresponding to the combat target. In addition, the combat effectiveness loss calculation formula can also use any combat effectiveness loss assessment method in this field to replace the loss calculation process, and is not limited to the combat effectiveness loss calculation formula. Then, by analyzing the geographical situation of the area where the combat target is located, including information such as topography (such as mountains, rivers, plains, etc.), the location of major transportation routes, etc., according to the target damage degree, its impact on the geographical situation is evaluated. For example, if the combat target is located at a traffic On important roads, when the degree of damage to the target is high, the main traffic arteries will be blocked or damaged, affecting the mobility of the troops and the transportation of materials. Assessment standards for the impact of different degrees of damage on the geographical situation are set. For example, when the degree of damage is 0-0.3, the impact on the geographical situation is small, causing only minor local obstructions; when it is 0.3-0.6, it affects the use of some traffic routes or key terrain; when it is 0.6-1, it causes traffic paralysis or the key terrain completely loses its combat value. According to these standards, combined with the specific location and degree of damage of the combat target, the combat geographical situation is evaluated to obtain the impact of the target damage on the combat situation. For example, a combat target located on an important bridge has a damage degree of 0.8. According to the standard, it can be judged that the bridge can no longer be used normally, which has seriously affected the traffic situation in the surrounding areas, and thus has a major impact on the overall combat situation. Finally, the corresponding impact of the target damage on the combat situation is obtained.

[0124] Preferably, target damage visualization compilation is performed according to the target damage degree, target damage range, combat effectiveness loss and the corresponding impact of target damage on the combat situation to generate a combat target damage visualization compilation result.

[0125] In an embodiment of the present invention, by using professional visualization tools (such as geographic information system software, data visualization platform, etc.), the target damage degree, target damage range, combat effectiveness loss and the impact of target damage on the combat situation are integrated and visualized. For the target damage degree, the damage degree of different areas is indicated on the map by color coding or numerical annotation. For the target damage range, the boundary of the damaged area is drawn on the map. The combat effectiveness loss can be displayed in the form of charts (such as bar charts, line charts, etc.) to compare the losses of different combat missions. The impact of target damage on the combat situation can be marked on the map or chart by text descriptions and arrow indications, etc., and the affected geographical factors and combat links are marked on the map or chart. These visualization elements are reasonably laid out and typeset, and necessary legends and explanations are added to finally generate a visualization compilation result of combat target damage, providing an intuitive and clear reference basis for combat command and decision-making.

[0126] Furthermore, the target damage degree is calculated by a target damage degree calculation formula, and the target damage degree calculation formula is specifically:

[0127]

[0128] In the formula, D is the target damage degree, v is the running speed of the combat target, k1 is the running speed weight, C is the combat capability of the combat target, k2 is the combat capability weight, and D f is the defense level of the combat target, k3 is the defense level weight, D f,max is the maximum defense degree corresponding to the combat target unit, H is the concealment degree of the combat target terrain, k4 is the concealment degree weight, S is the change in the combat target position, and k5 is the position change weight.

[0129] The present invention obtains a target damage degree calculation formula by using a specific mathematical model and verifying it, which is used to calculate the target damage degree based on the damage-related characteristics of the combat target. The formula fully considers the target damage degree D, the combat target running speed v, the running speed weight k1, the combat target combat capability C, the combat capability weight k2, the combat target defense degree D f , defense degree weight k3, the maximum defense degree D corresponding to the combat target unit f,max , the concealment degree of the combat target terrain H, the concealment degree weight k4, the change size of the combat target position S, the position change weight k5, according to the mutual correlation between the target damage degree D and the above parameters, a functional relationship is formed The calculation formula for the degree of target damage comprehensively considers multiple key characteristics of the combat target (such as running speed, combat ability, defense level, terrain concealment, position change, etc.) to calculate the degree of target damage. This multi-dimensional evaluation method can comprehensively consider the physical characteristics and environmental factors of the combat target, making the calculation of the damage degree more accurate and in line with the actual situation. By combining the weights of different factors, the attention to each factor can be flexibly adjusted, so as to obtain a more reasonable damage assessment. By introducing the dynamic changes of parameters such as the defense level of the combat target, the combat ability of the combat target, the concealment level and the position change, the formula can provide real-time calculation of the damage degree when the battlefield environment changes. This provides more timely and accurate decision-making support for the commander, enabling rapid response to different combat situations and evaluating the vulnerability and strike effect of the target. Each factor in the formula has a corresponding weight, and these weights can be adjusted according to different combat tasks and tactical requirements. This provides flexibility for the combat plan, and the degree of attention to each factor can be optimized according to the actual situation to ensure the most appropriate damage degree assessment under specific conditions. For example, in urban warfare, the terrain concealment level is more important, while in traditional battles, the running speed of the combat target is more critical. The calculation formula not only considers the physical characteristics of the target itself (such as speed, ability, defense, etc.), but also considers the position change of the target in the combat, which helps to evaluate the vulnerability of the target in different combat stages. This comprehensive analysis can provide more refined information for actual strike decisions, making the strike effect more accurate. Through the calculation of the degree of target damage, the formula can provide direct quantitative data for the subsequent calculation of combat effectiveness loss, making the expectation of the target destruction effect in the combat plan more specific, and can be further analyzed under multiple factors (such as mission completion progress, environmental situation, combat duration, etc.), providing support for formulating a more strategically significant combat plan.

[0130] Furthermore, the specific formula for calculating the combat effectiveness loss is as follows:

[0131]

[0132] In the formula, L is the combat effectiveness loss, D is the degree of target damage, R is the real-time completion progress of the combat target task, R max is the maximum completion progress required for the combat task corresponding to the combat target, E is the environmental situation of the combat target, T is the duration of the combat operation, F D is the size of the target damage range, e is the base of the natural logarithm, and η is the correction coefficient of the combat effectiveness loss.

[0133] The present invention obtains a calculation formula for combat effectiveness loss through the use of a specific mathematical model and verification, which is used to calculate the loss of combat mission parameters corresponding to combat targets. This calculation formula for combat effectiveness loss involves multiple key factors, such as the degree of target damage, mission progress, environmental situation, combat time, damage scope, etc. The combined effect of these factors helps to accurately reflect the combat effectiveness loss, can effectively quantify the changing situations occurring in combat, and ensures that the calculation of combat effectiveness loss is more accurate and comprehensive by comprehensively considering the damage situation of combat targets, the progress of mission execution, environmental changes, etc. The formula introduces to represent the mission progress. This ratio can reflect the completion degree of the mission at different time points and timely reflect the impact of mission progress on combat effectiveness. If the mission progress is slow or does not reach the required maximum progress, the combat effectiveness loss will increase accordingly. This design enables the formula to dynamically adapt to the changes in mission completion degree during the combat process and calculate the combat effectiveness loss in real time. By introducing the combat duration and the exponential decay e -T factor, the formula can reflect the impact of time on combat effectiveness loss. As the combat duration extends, certain loss effects will intensify, especially during continuous strikes and consumption. This exponential decay term can effectively simulate the aggravating effect of the increase in combat time on effectiveness loss. The formula introduces the environmental situation of the combat target as an important factor, which reflects the impact of the specific environmental conditions of the combat area on combat effectiveness. Different environmental conditions (such as weather, terrain, enemy situation, etc.) have different impacts on the execution and effect of combat missions. The formula realizes the adaptive analysis of the environmental situation through this factor and further improves the accuracy of the calculation of combat effectiveness loss. The formula measures the impact of the damage scope on combat effectiveness loss through the target damage scope. The larger the damage scope, the more comprehensive the destruction of the target, and thus the greater the impact on the loss of combat effectiveness. The introduction of this factor enables the formula to reflect the extensiveness and influence of target damage on the battlefield and provide a more accurate calculation of combat effectiveness loss. In addition, the correction coefficient in the formula provides a flexible adjustment space for the calculation of combat effectiveness loss, and the result can be fine-tuned according to the changes in the actual combat environment or special situations. This correction mechanism makes the formula more adaptable to different combat situations and can more accurately reflect the complex situations in actual combat. In summary, the formula fully considers the combat effectiveness loss L, the degree of target damage D, the real-time completion progress R of the combat target mission, the required maximum completion progress R max of the combat mission corresponding to the combat target, the environmental situation E of the combat target, the duration T of the combat operation, the size F of the target damage scope D , the base e of the natural logarithm, the correction coefficient η of the combat effectiveness loss, and constitutes a functional relationship according to the mutual correlation relationship between the combat effectiveness loss L and the above parameters This formula can realize the loss calculation process of combat mission parameters corresponding to combat objectives. At the same time, by introducing the correction coefficient η of combat effectiveness loss, it can be adjusted according to the errors occurring in the calculation process, thereby improving the accuracy and applicability of the combat effectiveness loss calculation formula.

[0134] Furthermore, the target damage visualization compilation according to the target damage degree, target damage range, combat effectiveness loss and the impact of target damage on the combat situation includes:

[0135] Generate a heat map of target damage severity corresponding to the target area based on the target damage severity, so as to intuitively present the damage severity corresponding to different target areas;

[0136] In an embodiment of the present invention, by collecting damage degree data of each specific location in the target area, which data comes from various sensor monitoring information, front-line reconnaissance reports, etc., the target area is divided into grids using geographic information system (GIS) technology, and each grid corresponds to a specific geographic coordinate position. According to the damage degree of the target in each grid, such as indicators such as the damage ratio of buildings and the number of casualties, corresponding values ​​are assigned to it. A heat map generation algorithm is used to set different color mapping rules. For example, areas with low damage degrees are mapped to green, areas with medium damage degrees are mapped to yellow, and areas with high damage degrees are mapped to red. The damage degree values ​​of each grid are visualized according to the color mapping rules to generate a target damage degree heat map corresponding to the target area. For example, in the target damage assessment of an urban area, by analyzing the building damage and casualty data of each block, a heat map is generated using the above method, which intuitively shows which blocks are severely damaged and which blocks are relatively lightly damaged, and finally a target damage degree heat map is generated.

[0137] Preferably, a three-dimensional dynamic model of the target damage range corresponding to the target area is produced according to the target damage range, so as to clearly display the target damage status corresponding to the combat target and the surrounding environment;

[0138] In an embodiment of the present invention, three-dimensional modeling is performed on the target area by using three-dimensional modeling software (such as 3dsMax, Maya, etc.) and geographic information data to import high-precision terrain data and construct the topography of the target area, including geographical features such as mountains, rivers, and roads. According to the collected target damage range information, such as which buildings are destroyed and which facilities are damaged, the damaged targets are marked and modeled in the three-dimensional model. In order to achieve dynamic display, the animation key frames of the model are set in combination with time series data. For example, according to the time course of the operation, the damage process of the target is gradually displayed, first showing the partial structural damage of the building, and then showing the complete collapse of the building as time goes by. By setting the camera perspective and animation path, the target damage range is displayed from different angles and distances. For example, when simulating the target damage situation of an urban operation, the damage change process of the combat target and the surrounding environment can be clearly seen through the model, and finally a three-dimensional dynamic model of the target damage range is generated.

[0139] Preferably, the effectiveness loss impact is drawn based on the impact of the combat effectiveness loss on the target damage on the combat situation, so as to draw and generate a combat effectiveness loss situation impact trend map corresponding to the target area, and help accurately grasp the impact trend of the target damage on the overall combat situation;

[0140] In the embodiment of the present invention, by clarifying various evaluation indicators of combat effectiveness, such as the proportion of troop loss, the damage rate of weapons and equipment, and the degree of task completion, the impact of the target damage on various combat effectiveness indicators is analyzed according to the target damage situation, and a mathematical model is established to quantify the relationship between the target damage degree and the combat effectiveness loss. For example, it is set that the combat effectiveness will decrease by a corresponding proportion every time a certain number of weapons and equipment are damaged. With time as the horizontal axis and the combat effectiveness loss as the vertical axis, a scatter plot is drawn according to the target damage situation at different time points and the corresponding combat effectiveness loss data. Then, a curve fitting algorithm (such as the least squares method) is used to fit the scatter plot to obtain a curve that can reflect the trend of combat effectiveness loss over time, and key time points and corresponding combat events are marked on the curve to generate a trend map of combat effectiveness loss situation corresponding to the target area. Through this map, it is possible to clearly see the trend of combat effectiveness loss with the occurrence and development of target damage, so as to accurately grasp the impact of target damage on the overall combat situation.

[0141] Preferably, the target damage degree heat map corresponding to the target area, the three-dimensional dynamic model of the target damage range, and the combat effectiveness loss situation impact trend map are visually compiled for target damage to generate a combat target damage visualization compilation result.

[0142] In an embodiment of the present invention, by using professional visualization integration software (such as Adobe After Effects, Tableau, etc.), the target damage degree heat map, the target damage range three-dimensional dynamic model and the combat effectiveness loss situation impact trend map are integrated. During the integration process, a reasonable layout and interaction method are set. For example, the heat map is placed on one side of the interface as a visual display of the overall damage situation; the three-dimensional dynamic model is placed in the center of the interface, and the user can view the target damage range from different angles through interactive operations (such as mouse dragging, zooming, etc.); the combat effectiveness loss situation impact trend map is placed on the other side to facilitate user comparison and analysis, add necessary text descriptions, legends and annotations, explain and mark each visualization element, set animation transition effects, make the switching between different visualization elements smoother and more natural, and finally generate a complete, intuitive and easy-to-understand combat target damage visualization compilation result, providing a clear reference basis for combat command and decision-making.

[0143] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0144] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A target damage assessment and analysis system based on a large model, characterized in that: Includes the following modules: The combat target data cleaning module is used to obtain the corresponding combat target geographic situation image through satellite reconnaissance, and integrate the combat intelligence report and combat order corresponding to the target combat unit to obtain the combat target text data; perform data standard cleaning on the combat target geographic situation image and combat target text data to obtain the combat target situation standard image and combat target standard text data; The combat target feature fusion module is used to build a corresponding target feature analysis model using a convolutional neural network and a Transformer architecture, and input the combat target situation standard image and combat target standard text data into the target feature analysis model to perform combat target driven analysis, so as to output the corresponding combat target image features and text semantic damage features; The target damage assessment model building module is used to use the target feature analysis large model as the input layer, and build a joint combat target damage assessment model based on the large model based on the input layer and the preset feature enhancement layer, fusion reasoning layer and output layer; The target damage visualization compilation module is used to obtain the corresponding combat target data to be evaluated, including the corresponding combat target situation image and combat target intelligence text to be evaluated; the corresponding combat target data to be evaluated is input into the joint combat target damage assessment model based on the large model to perform target damage assessment analysis, so as to predict and output the corresponding target damage degree, target damage range, combat effectiveness loss and the impact of target damage on the combat situation; Target damage visualization is performed based on the target damage degree, target damage range, combat effectiveness loss, and the corresponding impact of target damage on the combat situation to generate a combat target damage visualization compilation result.

2. The target damage assessment analysis and compilation system based on a large model according to claim 1 is characterized in that: The combat target data cleaning module includes the following functions: Obtain geographical situation images of combat targets corresponding to the target area through satellite reconnaissance; By integrating the combat intelligence reports and combat instructions corresponding to each target combat unit in the target area, the combat target text data can be obtained; Performing local pixel fuzziness analysis on the combat target geographic situation image to obtain the local pixel fuzziness of the combat target image; Based on the local pixel fuzziness of the combat target image, the combat target geographic situation image is fuzzy-denoised and standardized to obtain a combat target situation standard image; The combat target text data is cleaned according to text error correction standards to obtain standard combat target text data.

3. The target damage assessment analysis and compilation system based on a large model according to claim 2 is characterized in that: The text error correction and standard cleaning of combat target text data includes: Performing part-of-speech analysis and tagging on each word in the combat target text data to obtain combat target text part-of-speech tagging data; Based on the part-of-speech tagging data of the combat target text, grammatical error correction analysis is performed on each word in the combat target text data to correct the corresponding typos and grammatical errors and unify the vocabulary expressions, while removing the corresponding special characters, punctuation marks and meaningless stop words to obtain combat target text correction data; Through the common combat field data standards and coding system, the data corresponding to different sources and different formats in the combat target text correction data are standardized to obtain the combat target standard text data.

4. The target damage assessment analysis and compilation system based on a large model according to claim 1 is characterized in that: The combat target feature fusion module includes the following functions: The convolutional neural network is used in combination with the target image to construct a corresponding target image feature extraction branch to output the combat target image features corresponding to the target image; The Transformer architecture is used in combination with text data to construct the corresponding target semantic damage feature extraction branch to output the semantic damage features of the combat target corresponding to the target text; Connect the target image feature extraction branch and the target semantic damage feature extraction branch with the corresponding cross-modal feature association module to build a corresponding target feature analysis model to achieve the association fusion between the target image features and the text semantics; The standard image of the combat target situation and the standard text data of the combat target are input into the target feature analysis model for combat target driven analysis to output the corresponding combat target image features and text semantic damage features.

5. The target damage assessment analysis and compilation system based on a large model according to claim 4 is characterized in that: The target image feature extraction branch is specifically composed of a large-scale 5x5 convolution layer, a small-scale 3x3 convolution layer and a 1x1 pooling layer. The large-scale 5x5 convolution layer is used to extract the geometric shape and texture features of the combat target corresponding to the target image, the small-scale 3x3 convolution layer is used to extract the combat target position changes and surrounding environment situation characteristics corresponding to the target image at different time periods, and the large-scale and small-scale corresponding features are reduced in dimension in the 1x1 pooling layer.

6. The target damage assessment analysis and compilation system based on a large model according to claim 4 is characterized in that: The target semantic damage feature extraction branch is specifically: By performing combat target semantic segmentation on the text data in the corresponding input layer in the Transformer architecture, a combat target semantic segmentation set is generated; By using the corresponding military semantic graph in the corresponding text encoding layer in the Transformer architecture, each combat target word or phrase in the combat target semantic word set is semantically encoded, so as to find the corresponding combat domain concept in the military semantic graph for each word or phrase, and embed the relevant attributes and relationship codes corresponding to the combat domain concept into the corresponding combat target word or phrase to generate a combat target semantic encoding representation; By extracting target damage features from the semantic encoding representation of combat targets in the corresponding last layer in the Transformer architecture, the target type, combat mission and logical chain of target damage description events corresponding to the combat targets are extracted. By introducing additional constraints and reward mechanisms, the feature extraction branch is strengthened to extract the corresponding semantic features of combat targets. If the corresponding combat target damage features are accurately extracted, positive rewards are given, while if wrong or missing features are extracted, penalties are given. The corresponding damage feature vector representation is obtained through the forward propagation calculation corresponding to the last layer to output the combat target image features corresponding to the target image.

7. The target damage assessment analysis and compilation system based on a large model according to claim 1 is characterized in that: The target damage visualization compilation module includes the following functions: Obtain the corresponding combat target data to be evaluated, including the corresponding combat target situation image and combat target intelligence text; The combat target data to be evaluated is input into the input layer corresponding to the joint combat target damage assessment model based on the large model to extract combat target features, so as to extract and receive the corresponding combat target image features and text semantic damage features that have been preprocessed and feature extracted; The self-attention mechanism in the corresponding feature enhancement layer of the joint combat target damage assessment model based on the large model is used to fuse the combat target image features and the text semantic damage features, so as to fuse and highlight the relevant features corresponding to the combat target damage, and obtain the combat target damage-related features, including the combat target running speed, combat target combat capability, combat target defense level, combat target terrain concealment level and combat target position change; Calculate the target damage degree based on the damage-related characteristics of the combat target to obtain the target damage degree; Determine the damage range of the target based on the change in the position of the combat target to obtain the damage range of the target; Based on the target damage degree and target damage range, the combat effectiveness loss calculation formula is used to calculate the loss of combat mission parameters corresponding to the combat target to obtain the combat effectiveness loss; Based on the degree of target damage, the impact of the target damage on the combat geographical situation is evaluated to obtain the impact of the target damage on the combat situation; Target damage visualization is performed based on the target damage degree, target damage range, combat effectiveness loss, and the corresponding impact of target damage on the combat situation to generate a combat target damage visualization compilation result.

8. The target damage assessment analysis and compilation system based on a large model according to claim 7 is characterized in that: The target damage degree calculation is calculated by the target damage degree calculation formula, and the target damage degree calculation formula is specifically: In the formula, D is the target damage degree, v is the running speed of the combat target, k1 is the running speed weight, C is the combat capability of the combat target, k2 is the combat capability weight, and D f is the defense level of the combat target, k3 is the defense level weight, D f,max is the maximum defense degree corresponding to the combat target unit, H is the concealment degree of the combat target terrain, k4 is the concealment degree weight, S is the change in the combat target position, and k5 is the position change weight.

9. The target damage assessment analysis and compilation system based on a large model according to claim 7 is characterized in that: The calculation formula for combat effectiveness loss is as follows: In the formula, L is the combat effectiveness loss, D is the target damage degree, R is the real-time completion progress of the combat target task, and R max is the maximum completion progress of the combat mission corresponding to the combat target, E is the combat target environment situation, T is the duration of the combat operation, and F D is the size of the target damage range, e is the base of the natural logarithm, and η is the correction coefficient for the loss of combat effectiveness.

10. The target damage assessment analysis and compilation system based on a large model according to claim 7 is characterized in that: The target damage visualization compilation according to the target damage degree, target damage range, combat effectiveness loss and the impact of target damage on the combat situation includes: Generate a heat map of target damage severity corresponding to the target area based on the target damage severity, so as to intuitively present the damage severity corresponding to different target areas; According to the target damage range, a three-dimensional dynamic model of the target damage range corresponding to the target area is produced to clearly display the target damage status corresponding to the combat target and the surrounding environment; Based on the impact of combat effectiveness loss on the target damage on the combat situation, the effectiveness loss impact is drawn to generate a trend chart of combat effectiveness loss situation impact corresponding to the target area, and to help accurately grasp the impact trend of target damage on the overall combat situation; The target damage degree heat map corresponding to the target area, the three-dimensional dynamic model of the target damage range and the combat effectiveness loss situation impact trend map are visually compiled to generate a combat target damage visualization compilation result.

Images