A service-oriented unmanned equipment intelligent behavior conceptual model construction method

Abstract

The application discloses a kind of service-oriented unmanned equipment intelligent behavior conceptual model construction method, it is related to military field technical field, including the following steps: when unmanned equipment executes reconnaissance and strike task, carries optical, infrared and multispectral camera equipment, battlefield real-time video is continuously collected, through multispectral data acquisition, keep efficient target identification and tracking ability under various complex battlefield environment.The application improves target identification accuracy through multispectral data acquisition, real-time monitoring video stream extracts potential attack features, and with the help of deep learning model, attack hidden danger is efficiently identified, in high-risk environment, self-adapting adjustment detection strategy, enhance the response capability to hidden attack, ensure the accuracy and immediacy of tactical execution, to optimize identification efficiency, reduce the risk of misjudgment and missed detection, improve the utilization rate of combat resources, provide efficient, accurate, adaptive solution for the intelligent application of unmanned equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of military, in particular to a service-oriented unmanned equipment intelligent behavior conceptual model construction method. BACKGROUND

[0002] The service-oriented unmanned equipment intelligent behavior conceptual model construction refers to, in the military field, in order to meet different combat needs, based on the concept of service-oriented architecture (SOA), design and construct the intelligent behavior system of unmanned equipment. The model abstracts the functions of unmanned equipment (such as reconnaissance, strike, communication, navigation, obstacle avoidance, etc.) into independent and reusable service units, and flexibly combines and cooperates through standardized interfaces to realize the intelligent execution of tasks. Its core goal is to improve the self-adaptive ability, decision autonomy and cooperative combat effectiveness of unmanned equipment in complex battlefield environment. By introducing artificial intelligence, machine learning, data fusion and other technologies, the model can dynamically adjust strategies and optimize resource allocation during task execution, and ensure the interconnection between different platforms, providing more efficient and accurate support for combat command.

[0003] In the military field, target detection through real-time video stream refers to using optical, infrared or multispectral camera equipment carried by unmanned equipment to continuously collect real-time video of the battlefield during reconnaissance and strike tasks, and using the video as input data source for intelligent target recognition and tracking. Through embedded deep learning algorithms (such as YOLO, SSD, Faster R-CNN, etc.), unmanned equipment can detect and identify key targets such as enemy personnel, vehicles and weapons in real time in the video stream, while extracting the position information, motion trajectory and threat level of the target. This process relies on edge computing capabilities to ensure fast response in a low-latency environment, while using data links to transmit detection results to command centers or other combat units to achieve multi-platform cooperative combat and precision strike.

[0004] The existing technology has the following shortcomings:

[0005] When detecting targets through real-time video stream, the video stream may face a short and hidden attack risk, and the existing technology is usually difficult to predict and identify intelligently. The enemy may take advantage of this vulnerability by tampering or falsifying video data to interfere with the accurate judgment of the battlefield situation by unmanned equipment. For example, the enemy can disguise their targets as friendly forces or environmental backgrounds, causing serious deviations in tactical execution by unmanned equipment. Such attacks may lead to misidentification of threat targets, causing waste of combat resources and missing critical strike opportunities. In addition, misjudgment may also lead to friendly fire or damage to critical infrastructure, causing significant losses, directly affecting the accuracy of battlefield command, and thus leading to errors in overall tactical deployment, which may even change the course of the battle and cause serious strategic disadvantages to the own side.

[0006] The above information disclosed in the Background section is only for the purpose of enhancing the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0007] The purpose of the present application is to provide a service-oriented unmanned equipment intelligent behavior conceptual model construction method, which uses multispectral data acquisition technology to enhance the target recognition accuracy in different environmental conditions, and through real-time monitoring of video stream transmission information, accurately extracts potential attack features, providing a solid data foundation for subsequent intelligent attack prediction. By inputting the extracted features into a deep learning model, efficient identification of potential attack risks is achieved, enabling the system to adaptively adjust the detection strategy in high-risk environments, thereby enhancing the detection capability of short and concealed attacks and ensuring the accuracy and immediacy of tactical execution. Not only does it optimize the target recognition efficiency of unmanned equipment, effectively reducing the risk of misjudgment and missed detection caused by reasoning delay, but also ultimately improves the utilization efficiency of combat resources and the reliability of task execution, providing an efficient, accurate and adaptive solution for the intelligent application of unmanned equipment in modern warfare environments to solve the problems in the above background technology.

[0008] To achieve the above purpose, the present application provides the following technical solution: a service-oriented unmanned equipment intelligent behavior conceptual model construction method, comprising the following steps:

[0009] When performing reconnaissance and strike tasks, the unmanned equipment carries optical, infrared and multispectral camera equipment, continuously collects real-time battlefield video, and through multispectral data acquisition, maintains efficient target recognition and tracking capability in various complex battlefield environments, reducing the recognition error caused by a single spectral data source;

[0010] The collected real-time video stream is transmitted to the edge computing unit through the communication link, and in the transmission process, the transmission information of the video stream is obtained in real time for subsequent attack feature extraction and analysis;

[0011] In the monitoring window, the transmission data obtained is subjected to feature extraction, and the transmission attack features are extracted to preliminarily identify potential attack behaviors in the video stream transmission process, providing key evidence for subsequent attack prediction and response;

[0012] The extracted transmission attack features are inputted into a pre-trained deep learning model as feature vectors for analysis, and the potential attack risks in the video stream transmission are intelligently predicted to identify potential attack risks in advance;

[0013] When the deep learning model identifies potential attack risks in video stream transmission, the interval time of target recognition processing is intelligently shortened according to the degree of attack risk, the detection capability of short and concealed attacks is enhanced, and the unmanned equipment can timely identify and respond to potential threats;

[0014] While improving the detection frequency, the inference times of the target recognition algorithm in a short time are also improved, so that under high-frequency detection, the target recognition in each detection period can be quickly and accurately completed, potential threats can be discovered and responded in time, and the risk of misjudgment and missed detection caused by inference delay is reduced.

[0015] Preferably, the collected real-time video stream is transmitted to the edge computing unit by using wireless communication, satellite communication, wired network and hybrid transmission technology.

[0016] Preferably, in the monitoring window, the transmission data obtained is subjected to feature extraction, and the specific process is as follows:

[0017] The change frequency of the monitoring video stream transmission path and the distribution of the data packet size in the video stream are extracted from the obtained transmission data. After feature engineering is performed on the extracted change frequency of the monitoring video stream transmission path and the distribution of the data packet size in the video stream, transmission path change frequency quantitative values and data packet size distribution abnormal quantitative values are generated, respectively. The transmission path change frequency quantitative values and the data packet size distribution abnormal quantitative values are used to preliminarily identify potential attack behaviors in the video stream transmission process, thereby providing key basis for subsequent attack prediction and response.

[0018] Preferably, the extracted transmission path change frequency quantitative values and data packet size distribution abnormal quantitative values are inputted into a pre-trained deep learning model as feature vectors for analysis. A potential intrusion evaluation coefficient is generated by the deep learning model, and the potential attack risks in the video stream transmission are intelligently predicted based on the potential intrusion evaluation coefficient, so that the potential attack risks are identified in advance.

[0019] Preferably, after the extracted change frequency of the monitoring video stream transmission path is subjected to feature engineering in the monitoring window, the specific steps of generating the transmission path change frequency quantitative values are as follows:

[0020] In the monitoring window, the change of the video stream transmission path is sampled, the features related to the path change are extracted, and the transmission path change frequency quantitative values are calculated. First, the following parameters are defined:

[0021] represents the video stream transmission path number at the i sampling time, i represents the index of the video stream transmission path number;

[0022] N This indicates the total number of samples taken within the monitoring window;

[0023] The path change between two adjacent samples is defined as follows: Where 1 indicates that the path has changed, when When 0 indicates that the transmission path of the video stream has switched or been adjusted between two consecutive samples; 0 indicates that the path has not changed. This indicates that the transmission path of the video stream remains consistent between two consecutive samples, the transmission link is stable, and no switching has occurred;

[0024] The cumulative magnitude of path changes within the monitoring window is calculated as follows:

[0025]

[0026] in, and It is a hash function that maps path identifiers to a high-dimensional feature space to enhance the nonlinear effects of scheduling changes;

[0027] After extracting the path change factor, the quantized value of the transmission path change frequency is calculated as a key indicator for preliminary identification of attack behavior. The calculation expression is as follows:

[0028]

[0029] in: To change the frequency quantization value for the transmission path, The path change sensitivity coefficient is used to adjust the response sensitivity of the quantization value of the transmission path change frequency. This is a non-linear adjustment coefficient used to balance the response amplitude of the frequency quantization value when the transmission path changes. The time weighting for path changes reflects the impact of transmission link switching delay. To prevent smoothing factors with a denominator of zero.

[0030] Preferably, the specific steps for generating a quantification value of packet size distribution anomalies after feature engineering the distribution of packet sizes in the extracted video stream within the monitoring window are as follows:

[0031] The size distribution deviation features of each data packet during transmission are extracted. A deviation function is constructed using the cumulative distribution characteristics of data packet sizes. Through nonlinear mapping and cumulative transformation, the deviation features of the data packet size distribution are obtained. The data packet size sequence in the video stream is defined as follows: ,in Indicates the first p The size of each data packetm To determine the total number of data packets, construct a cumulative deviation function. The specific expression for the cumulative deviation function is as follows:

[0032]

[0033] in: This is the cumulative deviation function, representing the cumulative deviation value from the first data packet to the current data packet. Indicates the first k The size of each data packet k This represents a loop variable indicating the number of the currently processed data packet. During the cumulative summation process, it is used to iterate through all received data packets. This is the deviation sensitivity factor, which controls the intensity of the deviation response to large data packets. This represents the maximum size of the data packet, used for normalization.

[0034] By utilizing the cumulative deviation function and nonlinear mapping, a weighted cumulative model is used to reflect the overall anomaly of packet distribution. The quantification value of packet size distribution anomaly is defined as:

[0035]

[0036] in: This is a quantification value for packet size distribution anomalies, representing the overall degree of deviation in the packet size distribution. This is a nonlinear mapping function that nonlinearly amplifies the cumulative deviation function to highlight the contribution of abnormal data packets. The dynamic weighting function assigns anomaly weights to different data packets, depending on the median size. The relative differences are calculated, and the expression for the dynamic weighting function is as follows:

[0037]

[0038] This is a weighted sensitivity factor that controls the intensity of the response to relative differences.

[0039] Preferably, the potential intrusion assessment coefficient generated when intelligently predicting potential attack risks in video stream transmission using a pre-trained deep learning model within the monitoring window is compared and analyzed with a pre-set reference threshold to identify potential attack risks during video stream transmission. The specific process is as follows:

[0040] If the potential intrusion assessment coefficient is greater than the reference threshold, a risk signal is generated, indicating that there is a potential attack risk during video stream transmission.

[0041] If the potential intrusion assessment coefficient is less than or equal to the reference threshold, a normal signal is generated, indicating that the video stream is being transmitted efficiently.

[0042] Preferably, when the deep learning model identifies potential attack risks in video streaming, i.e., a video streaming generation risk signal, the target recognition processing interval time is intelligently shortened according to the attack risk degree, and the inference number of the target recognition algorithm in a short time is increased, and the specific steps are as follows:

[0043] When the deep learning model identifies potential attack risks in video streaming, the processing interval time of target recognition is intelligently adjusted according to the potential intrusion evaluation coefficient and the preset reference threshold. In order to enhance the detection ability of short and hidden attacks, the non-linear mapping relationship between the processing interval time and the potential intrusion evaluation coefficient is dynamically adjusted to ensure that in the case of high risk, the detection interval is shortened and the recognition frequency is increased. The expression of dynamic adjustment is:

[0044]

[0045] Wherein, is the adjusted target recognition processing interval time, is the initial target recognition processing interval time, representing the detection period under normal circumstances, is the potential intrusion evaluation coefficient, generated by the deep learning model analysis, reflecting the current intrusion risk level, is the preset reference threshold, when the potential intrusion evaluation coefficient exceeds the reference threshold, the recognition interval will be shortened, is the adjustment sensitivity coefficient, which determines the amplitude of the interval time shortening, is a smoothing factor that controls the stability of the adjustment process to avoid instability caused by sudden changes;

[0046] At the same time of shortening the target recognition processing interval time, the inference number in a short time is dynamically increased to ensure the efficient target recognition ability under higher detection frequency. The inference number is exponentially amplified according to the potential intrusion evaluation coefficient and the preset reference threshold to adapt to the detection needs of different risk levels. The specific dynamic adjustment expression is:

[0047] Wherein, is the adjusted target recognition inference number, is the initial target recognition inference number, representing the inference execution number under normal circumstances, is the inference number adjustment coefficient, which determines the inference amplification degree when the attack risk increases, is a non-linear exponential factor that determines the growth rate of the inference number to cope with different levels of threats;

[0048] To support higher frequency target detection and reasoning, reasonable allocation of computing resources is carried out to ensure real-time performance and energy consumption balance of tasks, and according to the potential intrusion evaluation coefficient and system resource availability, the resource allocation strategy is dynamically adjusted, more computing resources are preferentially allocated to the reasoning task, and the energy consumption strategy is optimized to prevent excessive calculation from causing device performance degradation, and the expression of the resource allocation strategy adjustment is:

[0049]

[0050] wherein, is the computing resource allocated to target recognition, is the maximum allocatable computing resource, is the current available computing resource ratio, which is dynamically adjusted with load changes, is the resource allocation sensitivity coefficient, which controls the resource adjustment amplitude.

[0051] In the above technical solution, the technical effects and advantages provided by the present application are:

[0052] The present application uses multispectral data acquisition technology to enhance the accuracy of target recognition under different environmental conditions, and through real-time monitoring of video stream transmission information, it accurately extracts potential attack features, providing a solid data foundation for subsequent intelligent attack prediction. By inputting the extracted features into a deep learning model, efficient identification of potential attack risks is achieved, enabling the system to adaptively adjust the detection strategy in high-risk environments, thereby enhancing the detection capability of short and concealed attacks and ensuring the accuracy and immediacy of tactical execution. Not only does it optimize the target recognition efficiency of unmanned equipment, effectively reducing the risk of misjudgment and missed detection caused by reasoning delay, but also ultimately improves the utilization efficiency of combat resources and the reliability of task execution, providing an efficient, accurate and adaptive solution for the intelligent application of unmanned equipment in modern warfare environments. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0054] Figure 1 A module schematic diagram of a service-oriented unmanned equipment intelligent behavior conceptual model construction method of the present application. DETAILED DESCRIPTION

[0055] Example implementations are now described with reference to the drawings. Example implementations can, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the example implementations to those skilled in the art.

[0056] The present application provides a service-oriented unmanned equipment intelligent behavior concept model construction method as shown in Figure 1 The present application provides a service-oriented unmanned equipment intelligent behavior concept model construction method as shown in

[0057] When performing reconnaissance and attack tasks, the unmanned equipment carries optical, infrared, and multispectral camera equipment, continuously collects real-time battlefield video, maintains efficient target recognition and tracking capabilities in various complex battlefield environments through multispectral data collection, reduces recognition errors caused by a single spectral data source, and improves the reliability and accuracy of overall reconnaissance;

[0058] These multispectral camera equipment can work in different lighting and environmental conditions, providing rich environmental information. For example, optical cameras are suitable for high-resolution image acquisition in daylight and good lighting conditions, infrared cameras can capture heat source information at night or in low light conditions, and multispectral cameras combine multiple spectral data to enhance the accuracy of target recognition.

[0059] The collected real-time video stream is transmitted to the edge computing unit through the communication link. During transmission, the transmission information of the video stream is obtained in real time, which is used for subsequent attack feature extraction and analysis;

[0060] Real-time monitoring of data transmission information during video stream transmission can timely discover abnormal situations that may exist during transmission, provide basic data support for subsequent attack detection and defense measures, and ensure the integrity and reliability of the video stream.

[0061] The collected real-time video stream can be transmitted to the edge computing unit in various ways, mainly including wireless communication, satellite communication, wired network and hybrid transmission technology. In terms of wireless communication, common methods include 5G, Wi-Fi 6, millimeter wave communication (mmWave) and mobile ad hoc network (MANET), among which the 5G network has the advantages of low latency and high bandwidth, and is suitable for high-speed data transmission in battlefield environment, while the MANET can build flexible tactical communication network in dynamic environment. Satellite communication (such as Ka band, L band) is suitable for long-range reconnaissance and complex environment, and can realize long-range real-time video backhaul in global range, with strong anti-interference ability. For wired network, optical fiber or Ethernet can be deployed in fixed tactical base station or front-line command center to ensure high reliability and large capacity of transmission demand. Hybrid transmission technology combines wireless and wired methods, for example, using wireless communication as the main link, and switching to satellite or optical fiber link when the signal is disturbed or lost, to ensure the continuity and stability of data flow. The selection of specific transmission method needs to be optimized according to the factors such as combat environment, bandwidth demand, anti-interference ability and equipment energy consumption, to ensure the efficient, stable and safe transmission of video stream to the edge computing unit, and realize real-time analysis and decision support of battlefield situation.

[0062] In the monitoring window, the transmission data obtained is subjected to feature extraction, and the potential attack behavior existing in the video stream transmission process is preliminarily identified by extracting the transmission attack features, so as to provide key basis for subsequent attack prediction and response;

[0063] In the monitoring window, the transmission data obtained is subjected to feature extraction, and the potential attack behavior existing in the video stream transmission process is preliminarily identified by extracting the transmission attack features, so as to provide key basis for subsequent attack prediction and response;

[0064] The change frequency of the monitoring video stream transmission path and the distribution of the data packet size in the video stream are extracted from the obtained transmission data. In the monitoring window, the change frequency of the monitoring video stream transmission path and the distribution of the data packet size in the video stream are subjected to feature engineering, and the transmission path change frequency quantitative value and the data packet size distribution abnormal quantitative value are generated, respectively. The potential attack behavior existing in the video stream transmission process is preliminarily identified by the transmission path change frequency quantitative value and the data packet size distribution abnormal quantitative value, so as to provide key basis for subsequent attack prediction and response.

[0065] When target detection is performed through real-time video stream, if the change frequency of the video stream transmission path is high, it usually indicates that there is a potential attack behavior in the transmission process of the video stream. Under normal circumstances, the transmission path of the video stream should remain relatively stable, and the network router will select the optimal path to ensure stable transmission with low delay and high bandwidth. However, frequent path changes may be caused by routing hijacking attacks, traffic redirection attacks, man-in-the-middle attacks (MITM) or jamming deception implemented by the enemy. For example, an attacker may cause video data to be redirected to a malicious node by tampering with the routing table, DNS spoofing or hijacking the Border Gateway Protocol (BGP), etc., resulting in data interception, tampering or delay. In addition, the enemy may use jamming technology to force the video stream to constantly switch paths, increasing the uncertainty of transmission and affecting the accuracy of battlefield situation awareness.

[0066] In the monitoring window, after feature engineering of the extracted change frequency of the monitoring video stream transmission path, the specific steps of generating the transmission path change frequency quantitative value are as follows:

[0067] In the monitoring window, the change of the video stream transmission path is sampled, the features related to path change are extracted, and the transmission path change frequency quantitative value is calculated. First, the following parameters are defined:

[0068] represents the video stream transmission path number at the i th sampling, i represents the index of the video stream transmission path number;

[0069] N represents the total number of samplings in the monitoring window;

[0070] represents the path change between adjacent two samplings, which is defined as: wherein 1 represents that the path has changed, and when , it means that the transmission path of the video stream has switched or adjusted between the two consecutive samplings (for example, the router selects a different transmission path), and such path change may be a normal behavior of network dynamic scheduling, or a result of potential attack (such as routing hijacking, traffic redirection); 0 represents that the path has not changed, and when , it means that the transmission path of the video stream remains consistent between the two consecutive samplings, and the transmission link is stable without switching;

[0071] represents the cumulative amplitude of the path change in the monitoring window, which is calculated as follows:

[0072]

[0073] wherein, and is a hash function that maps path identity to a high-dimensional feature space, used to enhance the nonlinear effect of scheduling changes;

[0074] The purpose of this step is to extract the change pattern from the original transmission path sequence, and use the hash function to introduce the nonlinear characteristics of the path, to enhance the sensitivity of the model to potential attacks. Path change cumulative amplitude Both reflect the frequency of changes and combine the amplitude of changes, providing more dynamic change information.

[0075] After completing the path change factor extraction, the transmission path change frequency quantitative value is calculated as a key indicator for preliminary identification of attack behavior, and the calculation expression is:

[0076]

[0077] Where: is the transmission path change frequency quantitative value, is the path change sensitivity coefficient, adjusting the response sensitivity of the transmission path change frequency quantitative value, is a nonlinear adjustment coefficient, used to balance the response amplitude of the transmission path change frequency quantitative value, is the time weight when the path changes, reflecting the influence of the transmission link switching delay, is a smoothing factor to prevent the denominator from being zero, usually taking a very small positive number;

[0078] The purpose of this step is to provide sensitive detection capability for potential attack behavior by constructing the transmission path change frequency quantitative value, considering the path change frequency, combining the weight and nonlinear adjustment factor.

[0079] The transmission path change frequency quantization value can be obtained by performing feature engineering on the change frequency of the extracted transmission path of the monitoring video stream in the monitoring window. The greater the performance value of the transmission path change frequency quantization value generated after the feature engineering, the higher the risk of potential attack behavior in the video stream transmission process, and vice versa. In the monitoring window, the system performs feature engineering on the change frequency of the video stream transmission path, extracts key features, and calculates the transmission path change frequency quantization value. The transmission path change frequency quantization value can quantify the frequency of path switching and the complexity of the features. Generally, in a stable network environment, the transmission path of the video stream should remain relatively fixed, with a low change frequency, and the transmission path change frequency quantization value will be at a low level. When an attacker attempts to redirect traffic, route hijacking, denial of service (DoS) or malicious interference, it may cause abnormal switching of the path and irregular frequency changes, thereby causing the transmission path change frequency quantization value to rise. Therefore, a higher transmission path change frequency quantization value usually means that the video stream is facing potential threats such as link spoofing, malicious traffic injection or dynamic redirection, while a lower transmission path change frequency quantization value indicates that the video stream is in a relatively stable transmission environment.

[0080] When performing target detection through real-time video streams, if the distribution of packet sizes in the transmission process of real-time video streams differs greatly, especially if there are abnormal-sized packets, it may indicate that there is a potential attack behavior in the video stream transmission process. Under normal circumstances, the packet size of the video stream usually follows a certain distribution pattern, which is influenced by factors such as encoding format, resolution, frame rate and network bandwidth, and presents a relatively stable fluctuation range. However, if there is a significant deviation from the regular packet size during transmission, such as an abnormal large or small packet, it may mean that it has suffered a packet insertion attack, a bandwidth depletion attack, a traffic obfuscation attack or a covert channel attack. Attackers may intentionally send oversized packets to cause buffer overflow at the receiving end, or insert tiny packets to increase network overhead and disrupt the normal transmission of the video stream, or even embed malicious code in the packets to implement further attacks. Therefore, monitoring the abnormal changes in packet size distribution can effectively identify and respond to potential attack threats, ensuring the integrity and security of the video stream.

[0081] The specific steps for generating the packet size distribution anomaly quantization value after performing feature engineering on the distribution of packet sizes in the extracted video stream in the monitoring window are as follows:

[0082] The size of each packet in the transmission process is extracted for distribution deviation feature extraction. The deviation function is constructed using the cumulative distribution characteristics of the packet size, and the deviation features of the packet size distribution are obtained through nonlinear mapping and cumulative transformation. The sequence of packet sizes in the video stream is defined as wherein denotes the size of the i-th data packet, p denotes the total number of data packets, the cumulative deviation function is constructed, and the specific expression of the cumulative deviation function is: m

[0083]

[0084] wherein: denotes the cumulative deviation function, indicating the cumulative deviation value from the first data packet to the current data packet, denotes the size of the i-th data packet, k denotes a loop variable, indicating the data packet number currently being processed, and in the cumulative summation process, it is used to traverse all received data packets, k denotes the deviation sensitivity factor, controlling the response intensity of the deviation to larger data packets, denotes the maximum value of the data packet size, used for normalization; By the cumulative deviation function, the gradual change trend of the data packet size is captured, and the sensitivity to the abnormal size of the data packet is enhanced.

[0085] By using the cumulative deviation function and the nonlinear mapping, the overall abnormality of the data packet distribution is reflected through the weighted cumulative model, and the data packet size distribution abnormal quantization value is defined as:

[0086]

[0087]

[0088] wherein: denotes the data packet size distribution abnormal quantization value, indicating the overall deviation degree of the data packet size distribution, denotes the nonlinear mapping function, which nonlinearly amplifies the cumulative deviation function, and is used to highlight the contribution of abnormal data packets, denotes the dynamic weight function, which gives different data packets the weight of abnormality, and is calculated depending on the relative difference from the median size The calculation expression of the dynamic weight function is:

[0089]

[0090] denotes the weight sensitivity factor, controlling the response intensity to the relative difference;

[0091] By integrating the abnormality of all data packets through the weighted cumulative, the quantitative evaluation of the overall distribution deviation is formed, and the dynamic weight amplifies the importance of data packets deviating from the median value, better reflecting the potential attack behavior caused by distribution abnormality.

[0092] ​​From the packet size distribution anomaly quantization value, it is known that in the monitoring window, the larger the performance value of the packet size distribution anomaly quantization value generated after the feature engineering of the distribution of the extracted packet size in the video stream, the higher the risk of potential attack behavior in the video stream transmission process, and vice versa. The packet size distribution anomaly quantization value quantifies the abnormality and distribution deviation of the packet size by performing feature engineering on the packet size distribution of the extracted real-time video stream in the monitoring window. When the packet size distribution anomaly quantization value is large, it indicates that the distribution of the packet size is significantly abnormal, such as sudden large or small packets, distribution skew, abnormal fluctuations, etc., which may be caused by malicious traffic injection, bandwidth depletion attack, covert channel attack or data tampering behavior. When the packet size distribution anomaly quantization value is small, it indicates that the distribution of the packet size is stable and conforms to the normal transmission characteristics, and no obvious attack signs are detected. Therefore, the packet size distribution anomaly quantization value can be used as an important indicator for real-time monitoring of video stream security, providing effective basis for attack prediction and response, and ensuring the stability and transmission integrity of the video stream.

[0093] The extracted transmission attack features are inputted into the pre-trained deep learning model as feature vectors for analysis, and the potential attack risks in the video stream transmission are intelligently predicted to identify potential attack risks in advance.

[0094] The extracted transmission path change frequency quantization value and packet size distribution anomaly quantization value are inputted into the pre-trained deep learning model as feature vectors for analysis, and the potential attack risks in the video stream transmission are intelligently predicted to identify potential attack risks in advance.

[0095] The pre-trained deep learning model refers to a neural network model that has been fully trained and optimized on a large number of historical data sets before the actual video stream transmission task is executed. After the model is trained, it has the ability to identify potential attack risks in the video stream transmission process, and can intelligently analyze the current transmission state according to the input feature vector (such as the transmission path change frequency quantitative value and the data packet size distribution abnormal quantitative value), and output a potential intrusion evaluation coefficient for evaluating the intrusion risk. In the process of training the model, a large amount of labeled data is usually used, including normal transmission behavior and various known attack behavior instances, and the data features are automatically extracted by deep learning algorithm to establish the mapping relationship from input features to attack risk probability. The training process usually includes data preprocessing, feature engineering, model training, hyperparameter optimization, and cross-validation steps to ensure that the model has high generalization ability and real-time prediction ability when dealing with complex and variable network environments. Using the pre-trained model helps to reduce the computational overhead in actual task execution, while ensuring that the system can quickly respond to potential threats.

[0096] In actual deployment, the pre-trained deep learning model is usually trained using supervised learning or semi-supervised learning strategies, using known attack patterns and normal transmission patterns for classification or regression analysis. For example, using a convolutional neural network (CNN) to extract the spatial pattern of data packet distribution, or using a long short-term memory network (LSTM) to analyze the time series features of transmission path change frequency. During the training process, the model learns the unique patterns of different attack types in data distribution, such as distribution deviation caused by malicious data packet insertion, frequent path switching caused by path hijacking, etc. Once the model is trained and achieves high accuracy and robustness on the test set, it can be deployed to the edge computing unit of unmanned equipment or command center to efficiently infer real-time data. When the system receives the input feature vector, the deep learning model will analyze the complex relationship between the features and generate a potential intrusion evaluation coefficient, providing an intelligent evaluation of whether the current video stream is under attack, helping combat commanders or automated systems make quick response decisions. By using a pre-trained model, it can quickly and accurately provide risk assessment when facing unknown or complex attacks, and effectively reduce the impact of attacks by adjusting the target recognition processing strategy, ensuring the integrity and security of the video stream.

[0097] The deep learning model is not specifically limited, and any deep learning model that can analyze the transmission path change frequency quantitative value and the data packet size distribution abnormal quantitative value to generate a potential intrusion evaluation coefficient can be used. In order to realize the technical scheme of the present application, a specific implementation method is provided; the potential intrusion evaluation coefficient The generated expression is:

[0098]

[0099] In the formula, , are preset proportion coefficients of the transmission path change frequency quantization value and the data packet size distribution anomaly quantization value , and , are greater than 0.

[0100] In the formula, the preset proportion coefficients and are used to balance and adjust the influence weight of different features in the calculation of the potential intrusion evaluation coefficient . Since the transmission path change frequency quantization value and the data packet size distribution anomaly quantization value may have different physical meanings and numerical ranges, directly combining them for calculation may result in the dominant effect of one feature on the result, while the contribution of the other feature is weakened. The introduction of the preset proportion coefficient helps to reasonably allocate the influence of each feature in the formula, so that it maintains a reasonable balance when evaluating the potential attack risk. In addition, the preset proportion coefficient can also be set according to historical data experience or expert knowledge, ensuring that it adapts to various network environments in different application scenarios, improving the sensitivity and robustness of the system to potential intrusions. Their values are greater than zero, indicating that all features have a positive contribution to the final result. At the same time, through the introduction of the square term, the sensitivity to extreme values is reduced, and the stability of the evaluation result is enhanced.

[0101] From the potential intrusion evaluation coefficient, it can be seen that the greater the performance value of the transmission path change frequency quantization value generated after feature engineering on the change frequency of the extracted monitoring video stream transmission path, the greater the performance value of the data packet size distribution anomaly quantization value generated after feature engineering on the distribution of the data packet size in the extracted video stream, that is, the greater the performance value of the potential intrusion evaluation coefficient generated when intelligently predicting potential attack risks of video stream transmission in the monitoring window through the pre-trained deep learning model, indicating that the risk of potential attack behavior in the video stream transmission process is higher, and vice versa, indicating that the risk of potential attack behavior in the video stream transmission process is lower.

[0102] The potential intrusion evaluation coefficient generated when intelligently predicting potential attack risks of video stream transmission in the monitoring window through the pre-trained deep learning model is compared and analyzed with the pre-set reference threshold to identify the potential attack risk during video stream transmission. The specific process is as follows:

[0103] If the potential intrusion evaluation coefficient is greater than the reference threshold, a risk signal is generated, indicating that there is a potential attack risk during video stream transmission;

[0104] If the potential intrusion assessment coefficient is less than or equal to the reference threshold, a normal signal is generated, indicating that the video stream is being transmitted efficiently.

[0105] When a deep learning model identifies a potential attack risk in a video stream, it intelligently shortens the target identification and processing interval based on the level of attack risk, enhancing the ability to detect short-lived and covert attacks, ensuring that unmanned equipment can identify and respond to potential threats in a timely manner, and avoiding missing key intelligence or tactical opportunities due to excessively long detection intervals.

[0106] While increasing the detection frequency, the number of inferences of the target recognition algorithm in a short period of time is increased to ensure that the target recognition in each detection cycle can be completed quickly and accurately under high-frequency detection, so as to detect and respond to potential threats in a timely manner and reduce the risk of misjudgment and missed detection due to inference delay.

[0107] When a deep learning model identifies a potential attack vulnerability in a video stream transmission, i.e., when the video stream transmission generates a risk signal, the interval for target identification processing is intelligently shortened based on the degree of attack risk. Simultaneously, the number of inferences performed by the target identification algorithm within a short time is increased. The specific steps are as follows:

[0108] When a deep learning model identifies a potential attack vulnerability in a video stream, it intelligently adjusts the target recognition processing interval based on the potential intrusion assessment coefficient and a preset reference threshold. To enhance the detection capability against brief and covert attacks, the nonlinear mapping relationship between the processing interval and the potential intrusion assessment coefficient is dynamically adjusted to ensure that the detection interval is shortened and the recognition frequency is increased under high-risk conditions. The expression for dynamic adjustment is:

[0109]

[0110] in, This is the adjusted target recognition processing interval. The initial target identification processing interval represents the detection cycle under normal circumstances. The potential intrusion assessment coefficient, generated by a deep learning model, reflects the current level of intrusion risk. The preset reference threshold will be used to significantly shorten the identification interval when the potential intrusion assessment coefficient exceeds the reference threshold. To adjust the sensitivity coefficient, the extent to which the interval time is shortened is determined. As a smoothing factor, it controls the stability of the adjustment process and avoids sudden changes that could lead to system instability;

[0111] This step adjusts the identification interval intelligently, allowing unmanned equipment to perform target identification more frequently in high-risk environments, reducing the likelihood of missing critical threats due to long detection intervals, and ensuring the system's real-time response capabilities.

[0112] While shortening the target identification processing interval, dynamically increase the inference frequency in a short time to ensure efficient target identification capabilities at higher detection frequencies. The inference frequency is exponentially amplified based on the potential intrusion assessment coefficient and a pre-set reference threshold to adapt to different risk levels of detection needs. The specific dynamic adjustment expression is:

[0113]

[0114] Where, is the adjusted target identification inference frequency, is the initial target identification inference frequency, representing the inference execution frequency under normal circumstances, is the inference frequency adjustment coefficient, determining the inference amplification degree when the attack risk increases, is a non-linear exponential factor that determines the rate of inference frequency growth to address different levels of threats.

[0115] This step ensures that the system can complete target identification with faster inference speed in high-frequency detection mode, reducing the risk of misjudgment and missed detection due to insufficient processing time, and improving the system's response capability in battlefield environments.

[0116] To support higher frequency target detection and inference, it is necessary to allocate computing resources reasonably to ensure real-time performance and energy consumption balance. Based on the potential intrusion assessment coefficient and system resource availability, dynamically adjust the resource allocation strategy, prioritize more computing resources for inference tasks, and optimize energy consumption strategies to prevent excessive computing from causing device performance degradation. The resource allocation strategy adjustment expression is:

[0117]

[0118] Where, is the allocated computing resource (CPU / GPU / memory, etc.) for target identification, is the maximum allocatable computing resource, is the current available computing resource proportion, dynamically adjusted with load changes, is the resource allocation sensitivity coefficient, controlling the resource adjustment amplitude.

[0119] Through the above-mentioned service-oriented unmanned equipment intelligent behavior conceptual model construction method, the autonomous reconnaissance, target identification and attack response capability of the unmanned equipment in complex battlefield environment can be significantly improved. The scheme uses multispectral data acquisition technology to enhance the target identification accuracy in different environmental conditions, and through real-time monitoring of video stream transmission information, potential attack features are accurately extracted, providing a solid data foundation for subsequent intelligent attack prediction. By inputting the extracted features into a deep learning model, efficient identification of potential attack risks is realized, enabling the system to adaptively adjust the detection strategy in high-risk environments, such as intelligently shortening the target identification processing interval time and increasing the reasoning frequency, thereby enhancing the detection capability of short and concealed attacks and ensuring the accuracy and immediacy of tactical execution. In addition, the scheme optimizes the target identification efficiency of the unmanned equipment, effectively reducing the risk of misjudgment and missed detection caused by reasoning delay, and ultimately improving the utilization efficiency of combat resources and the reliability of task execution, providing an efficient, accurate and adaptive solution for the intelligent application of unmanned equipment in modern warfare environment.

[0120] The above formulas are dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the latest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0121] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0122] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without deviating from the spirit and scope of the present application. Therefore, the above figures and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

Claims

1. A method for constructing a service-oriented intelligent behavior conceptual model for unmanned equipment, characterized in that, Includes the following steps: When performing reconnaissance and strike missions, unmanned equipment is equipped with optical, infrared and multispectral cameras to continuously collect real-time battlefield video. Through multispectral data acquisition, it maintains efficient target identification and tracking capabilities in various complex battlefield environments, reducing identification errors caused by a single spectral data source. The acquired real-time video stream is transmitted to the edge computing unit via a communication link. During the transmission process, the transmission information of the video stream is acquired in real time for subsequent attack feature extraction and analysis. Within the monitoring window, feature extraction is performed on the acquired transmission data. By extracting transmission attack features, potential attack behaviors in the video stream transmission process are initially identified, providing a key basis for subsequent attack prediction and response. The extracted transmission attack features are used as feature vectors and input into a pre-trained deep learning model for analysis. This enables intelligent prediction of potential attack risks during video stream transmission and early identification of potential attack risks. When a deep learning model identifies a potential attack risk in video stream transmission, it intelligently shortens the interval for target identification and processing based on the level of attack risk, enhancing the ability to detect short-lived and covert attacks and ensuring that unmanned equipment can identify and respond to potential threats in a timely manner. While increasing the detection frequency, the number of inferences of the target recognition algorithm in a short period of time is increased to ensure that the target recognition in each detection cycle can be completed quickly and accurately under high-frequency detection, so as to detect and respond to potential threats in a timely manner and reduce the risk of misjudgment and missed detection caused by inference delay. Within the monitoring window, feature extraction is performed on the acquired transmission data. The specific process is as follows: The frequency of changes in the transmission path of the monitored video stream and the distribution of data packet sizes in the video stream are extracted from the acquired transmission data. Within the monitoring window, feature engineering is performed on the extracted frequency of changes in the transmission path of the monitored video stream and the distribution of data packet sizes in the video stream to generate quantitative values ​​of transmission path change frequency and abnormal data packet size distribution. The quantitative values ​​of transmission path change frequency and abnormal data packet size distribution are used to initially identify potential attack behaviors in the video stream transmission process, providing key basis for subsequent attack prediction and response. Within the monitoring window, the specific steps for generating a quantified value of the transmission path change frequency after performing feature engineering on the extracted frequency of changes in the transmission path of the monitoring video stream are as follows: Within the monitoring window, changes in the video stream transmission path are sampled, features related to path changes are extracted, and the quantization value of the transmission path change frequency is calculated. First, the following parameters are defined: Indicates the first i The video stream transmission path number at the time of the next sample. i An index representing the video stream transmission path number; N This indicates the total number of samples taken within the monitoring window; The path change between two adjacent samples is defined as follows: Where 1 indicates that the path has changed, when When 0 indicates that the transmission path of the video stream has switched or been adjusted between two consecutive samples; 0 indicates that the path has not changed. This indicates that the transmission path of the video stream remains consistent between two consecutive samples, the transmission link is stable, and no switching has occurred; The cumulative magnitude of path changes within the monitoring window is calculated as follows: in, and It is a hash function that maps path identifiers to a high-dimensional feature space to enhance the nonlinear effects of scheduling changes; After extracting the path change factor, the quantized value of the transmission path change frequency is calculated as a key indicator for preliminary identification of attack behavior. The calculation expression is as follows: in: To change the frequency quantization value for the transmission path, The path change sensitivity coefficient is used to adjust the response sensitivity of the quantization value of the transmission path change frequency. This is a non-linear adjustment coefficient used to balance the response amplitude of the frequency quantization value when the transmission path changes. The time weighting for path changes reflects the impact of transmission link switching delay. To prevent smoothing factors with a denominator of zero.

2. The method for constructing a service-oriented intelligent behavior conceptual model for unmanned equipment according to claim 1, characterized in that, The acquired real-time video stream is transmitted to the edge computing unit using wireless communication, satellite communication, wired network, and hybrid transmission technologies.

3. The method for constructing a service-oriented intelligent behavior conceptual model for unmanned equipment according to claim 1, characterized in that, The extracted transmission path change frequency quantization value and data packet size distribution anomaly quantization value are used as feature vectors and input into a pre-trained deep learning model for analysis. The deep learning model generates potential intrusion assessment coefficients, and based on these coefficients, it intelligently predicts potential attack risks during video stream transmission, thus identifying potential attack risks in advance.

4. The method for constructing a service-oriented intelligent behavior conceptual model for unmanned equipment according to claim 1, characterized in that, Within the monitoring window, the specific steps for generating quantification values ​​of packet size distribution anomalies after feature engineering of the packet size distribution in the extracted video stream are as follows: The size distribution deviation features of each data packet during transmission are extracted. A deviation function is constructed using the cumulative distribution characteristics of data packet sizes. Through nonlinear mapping and cumulative transformation, the deviation features of the data packet size distribution are obtained. The data packet size sequence in the video stream is defined as follows: ,in Indicates the first p The size of each data packet m To determine the total number of data packets, construct a cumulative deviation function. The specific expression for the cumulative deviation function is as follows: in: This is the cumulative deviation function, representing the cumulative deviation value from the first data packet to the current data packet. Indicates the first k The size of each data packet k This represents a loop variable indicating the number of the currently processed data packet. During the cumulative summation process, it is used to iterate through all received data packets. This is the deviation sensitivity factor, which controls the intensity of the deviation response to large data packets. This represents the maximum size of the data packet, used for normalization. By utilizing the cumulative deviation function and nonlinear mapping, a weighted cumulative model is used to reflect the overall anomaly of packet distribution. The quantification value of packet size distribution anomaly is defined as: in: This is a quantification value for packet size distribution anomalies, representing the overall degree of deviation in the packet size distribution. This is a nonlinear mapping function that nonlinearly amplifies the cumulative deviation function to highlight the contribution of abnormal data packets. The dynamic weighting function assigns anomaly weights to different data packets, depending on the median size. The relative differences are calculated, and the expression for the dynamic weighting function is as follows: This is a weighted sensitivity factor that controls the intensity of the response to relative differences.

5. The method for constructing a service-oriented intelligent behavior conceptual model for unmanned equipment according to claim 3, characterized in that, The potential intrusion assessment coefficient generated when a pre-trained deep learning model intelligently predicts potential attack risks in video stream transmission within the monitoring window is compared and analyzed with a pre-set reference threshold to identify potential attack risks during video stream transmission. The specific process is as follows: If the potential intrusion assessment coefficient is greater than the reference threshold, a risk signal is generated, indicating that there is a potential attack risk during video stream transmission. If the potential intrusion assessment coefficient is less than or equal to the reference threshold, a normal signal is generated, indicating that the video stream is being transmitted efficiently.

6. The method for constructing a service-oriented intelligent behavior conceptual model for unmanned equipment according to claim 5, characterized in that, When a deep learning model identifies a potential attack vulnerability in a video stream transmission, i.e., when the video stream transmission generates a risk signal, the interval for target identification processing is intelligently shortened based on the degree of attack risk. Simultaneously, the number of inferences performed by the target identification algorithm within a short time is increased. The specific steps are as follows: When a deep learning model identifies a potential attack vulnerability in a video stream, it intelligently adjusts the target recognition processing interval based on the potential intrusion assessment coefficient and a preset reference threshold. To enhance the detection capability against brief and covert attacks, the nonlinear mapping relationship between the processing interval and the potential intrusion assessment coefficient is dynamically adjusted to ensure that the detection interval is shortened and the recognition frequency is increased under high-risk conditions. The expression for dynamic adjustment is: in, This is the adjusted target recognition processing interval. The initial target identification processing interval represents the detection cycle under normal circumstances. The potential intrusion assessment coefficient, generated by a deep learning model, reflects the current level of intrusion risk. The reference threshold is set. When the potential intrusion assessment coefficient exceeds the reference threshold, the identification interval will be shortened. To adjust the sensitivity coefficient, the extent to which the interval time is shortened is determined. As a smoothing factor, it controls the stability of the adjustment process and avoids instability caused by sudden changes; While shortening the target recognition processing interval, the number of inferences within a short period is dynamically increased to ensure efficient target recognition capabilities at higher detection frequencies. The number of inferences is exponentially amplified based on the potential intrusion assessment coefficient and a preset reference threshold to adapt to detection needs at different risk levels. The specific dynamic adjustment expression is as follows: in, The adjusted number of inference steps for target recognition. The number of inferences for initial target identification represents the number of inferences executed under normal circumstances. The adjustment factor for the number of inferences determines the degree of inference amplification when the attack risk increases. It is a non-linear exponential factor that determines the rate of increase in the number of inferences to deal with different levels of threats; To support higher-frequency target detection and inference, and to ensure a balance between task real-time performance and energy consumption, the resource allocation strategy is dynamically adjusted based on the potential intrusion assessment coefficient and system resource availability. Priority is given to allocating more computing resources to inference tasks, while energy consumption strategies are optimized to prevent over-computation from degrading device performance. The expression for adjusting the resource allocation strategy is as follows: in, The computing resources allocated for target identification, To maximize the allocable computing resources, This represents the proportion of currently available computing resources, which is dynamically adjusted according to changes in load. This is a sensitivity coefficient for resource allocation, used to control the magnitude of resource adjustments.

Images