Unmanned aerial vehicle countering method and system based on multi-platform cooperation

Abstract

The invention discloses an unmanned aerial vehicle countering method and system based on multi-platform cooperation, and relates to the technical field of unmanned aerial vehicle countering, and the method comprises the steps: carrying out the data collection based on a multi-unmanned aerial vehicle platform, including radio frequency signals, radar scanning and image data; and extracting a motion feature vector, calculating instantaneous frequency drift according to instantaneous frequency fluctuation, analyzing and calculating distortion deviation through a radio frequency signal, and fitting a response curve of the power amplifier through a logarithmic power curve. According to the method, through joint characterization of constellation diagram distortion and a power amplifier response curve, the distinguishing capability of radio frequency characteristics is enhanced, the omnibearing characterization capability from signal waveform details to power degradation behaviors is provided, and through a closed-loop mechanism of coupling characteristics, abnormal point detection and position inversion, the power degradation behavior of the power amplifier is improved. The real-time performance and precision supported by the countering decision are remarkably enhanced, the abnormal analysis of sample points is directly linked to target space positioning, and target position estimation is gradually optimized.

Description

Technical Field

[0001] This invention relates to the field of drone countermeasures technology, and in particular to a drone countermeasures method and system based on multi-platform collaboration. Background Technology

[0002] To counter the potential security threats posed by drones, existing drone countermeasures mainly focus on several aspects, including radio frequency jamming, physical interception, and electronic warfare interference. For example, monitoring the drone's communication signals and using radio frequency jamming equipment to cut off its communication with the operator is currently the most mainstream technical means; physical interception equipment such as net-catching drones and laser strike devices have also been deployed in practice.

[0003] However, when facing intelligent drones in complex and dynamic environments, existing technologies rely heavily on the single analysis of radio frequency signal features for target recognition, while ignoring the interaction between the target's kinematic features and radio frequency features. This results in poor detection performance for drones with complex evasion strategies or highly flexible behavior. Secondly, most existing countermeasures use static matching rules or preset models, lacking the ability to interact with the target's actions in real time. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention provides a multi-platform collaborative method and system for countering unmanned aerial vehicles (UAVs), which aims to solve the shortcomings of existing technologies in target recognition, which rely heavily on the single analysis of radio frequency signal features, resulting in poor detection performance for UAVs with complex evasion strategies or highly flexible behaviors; and the problem of using static matching rules, which lacks the ability to interact with target actions in real time.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for countering unmanned aerial vehicles (UAVs) based on multi-platform collaboration, comprising: Data acquisition is conducted based on multiple UAV platforms, including radio frequency signals, radar scans, and image data. Motion feature vectors are extracted, instantaneous frequency drift is calculated based on instantaneous frequency fluctuations, distortion deviation is calculated through radio frequency signal analysis, and a power amplifier response curve is formed by fitting a logarithmic power curve. The attenuation coefficient and the logarithmic value of the initial power intensity are extracted and combined with the instantaneous frequency drift and distortion deviation to form a radio frequency feature vector. The optimal projection direction of kinematic and radio frequency features is analyzed through canonical correlation, and the final canonical correlation is calculated. Projecting is performed according to the optimal projection direction, forming a coupling feature matrix and calculating the coupling weight factor based on canonical correlation. The weighted coupling feature matrix is ​​used to construct an isolated forest model, and the path length in each isolated tree is calculated for anomaly detection, forming an anomaly sample point set. The signal timestamp is extracted and the TDOA method is used to back-calculate the target spatial location. The target location estimation model is constructed by combining the propagation distance difference to locate the location coordinates. Define a non-cooperative game model for reconnaissance and avoidance, calculate the field-of-view maintenance cost, resolution cost, and observation angle cost of UAV reconnaissance of threat sources, analyze the impact of target field-of-view deviation on reconnaissance equipment, the sensitivity impact of target image quality, and the degree of deviation between the imaging angle and the target's optimal imaging direction, and perform weighted processing to calculate the composite observation cost; The optimization objective is determined and iterative optimization is carried out until Nash equilibrium is reached and convergence is achieved. The trajectory optimization result and action sequence of the UAV are output. Combined with high-resolution images and radio frequency features, a high-definition evidence chain vector is formed, which is then used to form a shared data packet. The data is transmitted to the countermeasure module to perform collaborative data estimation and data synchronization.

[0006] As a preferred embodiment of the multi-platform collaborative UAV countermeasure method described in this invention, the extraction of the attenuation coefficient and the logarithm of the initial power intensity, combined with instantaneous frequency drift and distortion deviation, forms a radio frequency feature vector. The optimal projection direction of the kinematic and radio frequency features is analyzed through canonical correlation, and the final canonical correlation is calculated, including... The target velocity and acceleration are calculated by measuring the distance the object moves within the video frame interval using the object image data. The target angular velocity is calculated by measuring the rotation angle of the object image data within the video frame interval using the position information collected by the radar module. This angular velocity is then combined with the target velocity and acceleration to form a motion feature vector. The instantaneous phase is calculated by capturing the target radio frequency signal using an SDR sensor, and the instantaneous frequency is calculated by numerically differentiating the instantaneous phase sequence using the center difference method. The instantaneous frequency drift is then calculated based on the standard deviation of the instantaneous frequency fluctuation. The quadrature and in-phase components of the normalized radio frequency signal are used to calculate the distortion based on the IQ constellation diagram and the distortion deviation is calculated based on the IQ components corresponding to the ideal constellation points. Power is calculated based on the quadrature and in-phase signal components, and the logarithmic change of the amplifier response is represented by fitting a logarithmic power curve to form the power amplifier response curve. The radio frequency characteristic vector is composed of instantaneous frequency drift, distortion deviation, attenuation coefficient, and logarithmic value of initial power intensity; The motion feature vector and radio frequency feature vector are processed by standard deviation. Correlation analysis is performed by canonical correlation analysis (CCA). The covariance matrix of the motion feature vector and radio frequency feature vector after standard deviation processing is calculated. Eigenvalue decomposition is performed to obtain projection vectors a and b, which correspond to the optimal projection directions of the kinematic feature and radio frequency feature, respectively. Finally, the canonical correlation is calculated.

[0007] As a preferred embodiment of the multi-platform collaborative UAV countermeasure method described in this invention, the following steps are included: forming an abnormal sample point set, extracting signal timestamps, using the TDOA method to inversely calculate the target's spatial location, constructing a target location estimation model based on propagation distance difference, and locating the position coordinates. The motion feature vector and the radio frequency feature vector are projected separately according to the projection vector, and then merged and spliced ​​into a coupled feature vector. The coupling feature vector is used to form a coupling feature matrix based on the number of sample points, motion features, and radio frequency features. The coupling weight factor is calculated based on the variance and canonical correlation value of the coupling feature matrix, and the coupling feature matrix is ​​adjusted. An isolated forest model is constructed using a weighted coupled feature matrix. A subset of samples is randomly selected, and features are randomly selected. A segmentation threshold is randomly set in the selected feature dimensions. The subset samples are recursively split according to the segmentation conditions. The isolated forest model is iterated repeatedly until the maximum split depth is reached. The isolated forest is used to calculate the path length of each sample point in each isolated tree and to calculate the anomaly score for anomaly detection. The anomaly threshold is based on the sum of the mean and twice the standard deviation of historical outliers. If the calculated outlier is greater than or equal to the anomaly threshold, the sample point is judged to be an anomaly and a set of anomaly sample points is formed. Based on the set of abnormal sample points, the timestamps of the signals received by the sensors in different UAVs are obtained and used as threat sources. The TDOA method is used to calculate the target spatial location, which calculates the time difference between each pair of sensors and considers the signal propagation speed to calculate the propagation distance difference between the threat source and the sensor. A target position estimation model is constructed by combining the propagation distance difference, and the target position is solved by nonlinear least squares method; The gradient descent method is used to iteratively solve the problem until the target iterative value converges and the target location estimation function value is minimized, thus obtaining the location coordinates of the threat source.

[0008] As a preferred embodiment of the multi-platform collaborative UAV countermeasure method described in this invention, the step of analyzing the impact of target field-of-view deviation on reconnaissance equipment, the sensitivity impact of target image quality, and the degree of deviation between the imaging angle and the target's optimal imaging direction, and then performing weighted processing to calculate the composite observation cost, includes: A non-cooperative game model of reconnaissance and avoidance is defined, in which a composite observation cost function is defined. Based on the position and velocity of the threat source and the angle between it and the reconnaissance drone, the cost of maintaining the field of view, the cost of resolution, and the cost of observation angle of the drone reconnaissance of the threat source are calculated. We weighted the costs of maintaining the field of view, resolution, and observation angle for UAV reconnaissance threat sources, and analyzed the impact of target field of view deviation on reconnaissance equipment, the sensitivity-related impact on target image quality, and the degree of deviation between the current image imaging angle and the target's optimal imaging direction. The composite observation cost is calculated by weighting the costs of maintaining the field of view, resolution, and observation angle for drone-based threat source reconnaissance.

[0009] As a preferred embodiment of the multi-platform collaborative drone countermeasure method of the present invention, wherein: the output drone trajectory optimization results and action sequence, combined with high-resolution images and radio frequency features, form a high-definition evidence chain vector, including: A two-layer non-cooperative game based on composite observation cost is used. The optimization objective is to minimize the dynamic observation cost function by adjusting the UAV's observation position, velocity, and acceleration. This includes determining the optimal action of the reconnaissance UAV, minimizing the dynamic observation cost function based on the dynamic avoidance strategy of the threat target, and performing inner-layer optimization, namely, seeking to maximize the composite observation cost for reconnaissance behavior avoidance actions. The reconnaissance behavior avoidance actions are updated using the gradient ascent method, and the UAV's observation position, velocity, and acceleration are updated using the gradient descent method. The game converges after reaching Nash equilibrium and outputs the trajectory optimization results and action sequence of the UAV. Based on the trajectory optimization results of the reconnaissance drone and the action sequence data collected, including the real-time trajectory of the threat target, high-resolution images, and radio frequency characteristics, a high-definition evidence chain vector is formed.

[0010] As a preferred embodiment of the multi-platform collaborative UAV countermeasure method of the present invention, the step of forming a shared data packet, transmitting the data to the countermeasure module for collaborative data estimation, and performing data synchronization includes: The high-definition evidence chain vector data is composed into a shared data packet, which is transmitted to the countermeasure module via the MQTT message queue protocol. A distributed filtering algorithm is then used to perform collaborative data estimation on multiple drones and the countermeasure module to synchronize the data.

[0011] As a preferred embodiment of the multi-platform collaborative UAV countermeasure method described in this invention, the data acquisition based on multiple UAV platforms includes radio frequency signals, radar scanning, and image data, including... The drone is equipped with optical and infrared cameras, radar modules, and SDR sensors. The optical and infrared cameras acquire high-resolution image data of the object, the SDR sensors acquire radio frequency signals of the drone, including quadrature and in-phase signal components, and the radar modules acquire position information.

[0012] Secondly, the present invention provides a multi-platform collaborative unmanned aerial vehicle (UAV) countermeasure system, comprising: The data acquisition module uses multiple UAV platforms to collect radio frequency signals and image data of threat targets, and extracts motion feature vectors and radio frequency feature vectors; The feature fusion module calculates instantaneous frequency drift, distortion deviation, and power characteristics based on the radio frequency signal, and obtains the optimal projection direction by canonical correlation analysis with the motion features, thus constructing a coupled feature matrix. The anomaly detection module trains an isolated forest model based on a weighted coupled feature matrix to detect a set of anomalous sample points and locates the spatial location of threat targets from the timestamps of anomalous signals. The reconnaissance optimization module outputs the optimal trajectory and action sequence of the UAV through weighted optimization of composite observation costs and iterative non-cooperative game model. The evidence chain generation module generates high-definition evidence chain vectors based on optimized trajectory acquisition high-resolution images and radio frequency features, providing comprehensive data characterization of threat targets; The shared collaboration module packages high-definition evidence chain vectors into shared data packets, which are then transmitted to the countermeasure module via a message queue protocol for collaborative data estimation and dynamic decision synchronization.

[0013] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein the computer program, when executed by the processor, implements any step of the multi-platform collaborative UAV countermeasure method described in the first aspect of the present invention.

[0014] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the multi-platform collaborative UAV countermeasure method as described in the first aspect of the present invention.

[0015] The beneficial effects of this invention are as follows: by jointly characterizing constellation diagram distortion and power amplifier response curves, the distinguishability of radio frequency features is enhanced, providing a comprehensive characterization capability from signal waveform details to power degradation behavior. Through a closed-loop mechanism of coupled features, anomaly detection, and position inversion, the real-time performance and accuracy of countermeasure decision support are significantly enhanced. Anomaly analysis of sample points is directly linked to target spatial positioning and target position estimation is gradually optimized. By collecting real-time trajectories, high-resolution images, and radio frequency features of threat targets, a full-dimensional high-definition evidence chain is constructed, which precisely controls the distance, observation angle, and tracking speed of UAVs to threat targets, ensuring that the collected target data is always kept within a high availability range. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the UAV countermeasure method based on multi-platform collaboration in Example 1.

[0018] Figure 2 This is a schematic diagram of the structure of the UAV countermeasure system based on multi-platform collaboration in Example 1. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Example 1, referring to Figures 1 to 2 This is the first embodiment of the present invention, which provides a method for countering unmanned aerial vehicles (UAVs) based on multi-platform collaboration, including the following steps: S1 collects data based on multiple UAV platforms, including radio frequency signals, radar scans, and image data; extracts motion feature vectors, calculates instantaneous frequency drift based on instantaneous frequency fluctuations, calculates distortion deviation through radio frequency signal analysis, and forms a power amplifier response curve by fitting a logarithmic power curve. It extracts the attenuation coefficient and the logarithmic value of the initial power intensity, combines the instantaneous frequency drift and distortion deviation to form a radio frequency feature vector, analyzes the optimal projection direction of kinematic and radio frequency features through canonical correlation, and calculates the final canonical correlation. Preferably, data acquisition based on multiple UAV platforms includes radio frequency signals, radar scans, and image data, including: The drone is equipped with optical and infrared cameras, radar modules, and SDR sensors. The optical and infrared cameras acquire high-resolution image data of the object, the SDR sensors acquire radio frequency signals of the drone, including quadrature and in-phase signal components, covering frequency bands such as communication, control, and telemetry. The SDR extracts physical layer features of the radio frequency signals, such as instantaneous frequency drift and amplitude modulation, and the radar module acquires position information.

[0023] Furthermore, the attenuation coefficient and the logarithm of the initial power intensity are extracted and combined with instantaneous frequency drift and distortion bias to form an RF feature vector. The optimal projection direction of the kinematic and RF features is analyzed through canonical correlation analysis, and the final canonical correlation is calculated, including... The target velocity and acceleration are calculated by measuring the distance the object moves within the video frame time interval using the object image data. A second velocity and a second acceleration are calculated by measuring the distance the object moves within the time interval using the position information collected by the radar module. The difference between these two values ​​is calculated and used as the verification error. The sum of the mean and standard deviation of the historical error values ​​is used as the verification threshold. If the verification error is less than or equal to the verification threshold, the verification is considered successful. The target angular velocity is calculated by measuring the rotation angle of the object image data within the video frame time interval and is combined with the target velocity and acceleration to form a motion feature vector. The instantaneous phase is calculated using the target RF signal captured by the SDR sensor, and the instantaneous frequency is calculated by numerically differentiating the instantaneous phase sequence using the center difference method. The instantaneous frequency drift is then calculated based on the standard deviation of the instantaneous frequency fluctuation, expressed as:

[0024]

[0025]

[0026]

[0027] in, This represents the instantaneous phase of the signal at time t. The quadrature component represents the amplitude of the radio frequency signal in the quadrature channel, and is usually obtained from the demodulator output of the receiver. This represents the in-phase signal component of the signal, indicating the amplitude of the radio frequency signal in the in-phase channel. The frequency of the signal at time t is represented by k, and the index of the time point is k. express, Indicates at a point in time instantaneous frequency, Indicates instantaneous frequency drift (unit: Hz), and represents the standard deviation of signal frequency fluctuations; it is a key indicator of hardware modulation stability. This represents the total number of sampling points in the instantaneous frequency sequence. The average value representing the instantaneous frequency; The quadrature and in-phase components of the normalized radio frequency signal are used, and the distortion is calculated based on the IQ constellation diagram. The distortion deviation is expressed as follows:

[0028] in, Distortion bias represents the distortion of the constellation diagram, and Euclidean distance represents the dimensionless Euclidean distance. Represents the time point of normalization The in-phase signal component, Represents the time point of normalization orthogonal signal components, and These represent the IQ components corresponding to the ideal constellation points. Based on historical sample points, the sample points with the minimum distortion deviation are calculated using the nearest neighbor matching method and used as the ideal constellation points to calculate the distortion deviation for the newly acquired IQ components. Power is calculated based on the quadrature and in-phase signal components, and the logarithmic change of the amplifier response is represented by a logarithmic power curve fitting, forming the power amplifier response curve, as shown below:

[0029]

[0030]

[0031]

[0032] in, Represents the power over time t. The logarithm of the initial power intensity can be determined from the intercept of the response curve at time t=0. Represents the natural logarithm of power, and represents the logarithmic change in the amplifier's response. The attenuation coefficient represents the rate of decrease of the logarithmic power, which can be obtained through... And the corresponding time points are determined using the least squares method; The radio frequency characteristic vector is composed of instantaneous frequency drift, distortion deviation, attenuation coefficient, and logarithmic value of initial power intensity; Standard deviation processing is applied to the motion feature vectors and radio frequency feature vectors. Correlation analysis is performed using the canonical correlation analysis (CCA) method. The covariance matrices of the motion and radio frequency feature vectors after standard deviation processing are calculated. Eigenvalue decomposition is then performed to obtain projection vectors a and b, which correspond to the optimal projection directions of the kinematic and radio frequency features, respectively. The final canonical correlation is then calculated and expressed as:

[0033]

[0034]

[0035]

[0036]

[0037]

[0038] in, The covariance matrix representing the kinematic characteristics, The covariance matrix representing the radio frequency characteristics. This represents the covariance matrix between kinematic features and radio frequency features. This represents the preprocessed motion feature vector. Let represent the preprocessed radio frequency (RF) eigenvectors, T represent transpose calculation, n represent the total number of kinematic eigenvectors, m represent the total number of RF eigenvectors, a represent the kinematic projection vector, and b represent the RF projection vector. This indicates the calculation of covariance. This indicates canonical correlation.

[0039] By acquiring the target's motion feature vector and radio frequency feature vector in real time, a multi-dimensional characteristic description of the target is formed. After extracting and vectorizing the target's motion characteristics (velocity, acceleration, angular velocity) and radio frequency signal characteristics (instantaneous frequency drift, constellation diagram distortion, and power amplifier response) in real time, a multi-dimensional data representation of the target in dynamic space and signal space is provided. By combining the calculation of instantaneous frequency drift with radio frequency signal fluctuation analysis, the accuracy of target hardware characteristic inversion is improved, and the identification capability of target equipment type and design characteristics is enhanced. The generated radio frequency feature vector can effectively distinguish the differences between threat drones and ordinary equipment communication, and achieve efficient reverse analysis of communication equipment characteristics. By jointly characterizing constellation diagram distortion and power amplifier response curves, the ability to distinguish RF features is enhanced, providing a comprehensive characterization capability from signal waveform details to power degradation behavior. This significantly improves the comprehensiveness and accuracy of target characteristic modeling. Through canonical correlation analysis (CCA), the optimal linear projection direction of kinematic and RF feature spaces is constructed. The final calculated canonical correlation reflects the degree of correlation between kinematic and RF features. Furthermore, enhanced detection or specific pattern recognition can be achieved through strongly correlated signal patterns.

[0040] S2, project according to the optimal projection direction, form a coupling feature matrix and calculate the coupling weight factor based on canonical correlation, use the weighted coupling feature matrix to construct an isolated forest model, calculate the path length in each isolated tree to perform anomaly detection, form an anomaly sample point set, extract the signal timestamp, use the TDOA method to back calculate the target spatial position, combine the propagation distance difference to construct the target position estimation model, and locate the position coordinates; Preferably, an abnormal sample point set is formed, the signal timestamp is extracted, the TDOA method is used to inversely calculate the target's spatial location, and a target location estimation model is constructed by combining the propagation distance difference to locate the position coordinates, including... The motion feature vector and the radio frequency feature vector are projected onto the projection vector respectively, and then merged and concatenated into a coupled feature vector, which is represented as follows:

[0041]

[0042]

[0043] in, Represents the motion projection vector. V represents the radio frequency projection vector, and V represents the coupling characteristic vector. The coupling feature vector is used to construct a coupling feature matrix based on the number of sample points, motion features, and radio frequency features. The coupling weight factor is calculated based on the variance and canonical correlation values ​​of the coupling feature matrix, and the coupling feature matrix is ​​adjusted, as shown below:

[0044]

[0045] in, The j-th feature weight factor represents the motion and radio frequency characteristics. This indicates variance calculation. Let represent the variance of the j-th column feature of the coupling feature matrix. This represents the number of feature dimensions related to motion. This represents the number of motion-related feature dimensions and the number of radio frequency-related feature dimensions. This represents the weighted coupling characteristic matrix; An isolated forest model is constructed using a weighted coupled feature matrix. A subset of samples is randomly selected, and features are randomly selected. A segmentation threshold is randomly set in the selected feature dimensions. The subset samples are recursively split according to the segmentation conditions. The isolated forest model is iterated repeatedly until the maximum split depth is reached. For each sample point, the path length in each isolated tree is calculated using an isolation forest, and an anomaly score is calculated for anomaly detection. An anomaly threshold is set based on the sum of the mean and two standard deviations of historical outliers. If the calculated outlier is greater than or equal to the threshold, the sample point is considered an anomaly and a set of anomaly sample points is created, represented as follows:

[0046]

[0047] in, This represents the anomaly score of the i-th sample point. Represents the i-th sample point The average of all path lengths Indicates the expected path length. The harmonic number is approximately represented as: in Represents Euler's constant. This represents the size of the subset of samples in each isolated tree, which is the number of sampling points in the isolated tree; Based on the set of abnormal sample points, the timestamps of the signals received by the sensors of different UAVs are obtained and used as threat sources. The TDOA method is used to calculate the target's spatial location. This involves calculating the time difference between every two sensors and considering the signal propagation speed to calculate the propagation distance difference between the threat source and the sensor, expressed as:

[0048] in, and Let f and g represent the Euclidean distances from the threat source to the f-th and g-th sensors, respectively, and c represent the signal propagation speed. This represents the time difference of signal arrival between the f-th and g-th sensors; A target position estimation model is constructed by combining the propagation distance difference, and the target position is solved by the nonlinear least squares method, expressed as:

[0049] in, This represents the target location estimation function value, the error in estimating the threat source location, and ncd represents the total number of drone sensors; The gradient descent method is used to iteratively solve the problem until the target iterative value converges and the target location estimation function value is minimized, thus obtaining the location coordinates of the threat source.

[0050] By projecting and merging motion features and radio frequency features, a coupled feature vector is constructed, which effectively integrates multi-dimensional features to enhance significance. By optimizing the weight factors based on the variance of coupled features and canonical correlation, the concentration and discriminative power of the coupled feature matrix are improved. The contribution of each feature is quantified and assigned a dynamic weight. This weight allocation takes into account the differences in importance of multimodal features and can optimize the feature discriminative power in subsequent data processing. By training and segmenting weighted coupling features using the Isolation Forest model, abnormal behavior of dynamic targets can be detected efficiently. In particular, the weighted coupling feature matrix integrates information from multimodal projection features into a new dimension, enabling the Isolation Forest to capture anomalies more quickly in high-dimensional space. By extracting sample points based on anomaly detection and combining them with the TDOA method, higher accuracy in target spatial location analysis is achieved. The set of anomalous sample points is dynamically generated based on the previous isolated forest model and can directly characterize targets whose behavior or signals deviate significantly from the conventional pattern. This anomaly extraction provides highly reliable samples for the location inversion of threatening targets. Through a closed-loop mechanism of coupled features, anomaly detection, and location inversion, the real-time performance and accuracy of countermeasure decision support are significantly enhanced. Anomaly analysis of sample points is directly linked to target spatial positioning, and target location estimation is gradually optimized.

[0051] S3 defines a non-cooperative game model for reconnaissance and avoidance, calculates the field-of-view maintenance cost, resolution cost, and observation angle cost of UAV reconnaissance of threat sources, analyzes the impact of target field-of-view deviation on reconnaissance equipment, the sensitivity impact of target image quality, and the degree of deviation between the imaging angle and the target's optimal imaging direction, and performs weighted processing to calculate the composite observation cost. Preferably, the impact of target field-of-view deviation on reconnaissance equipment, the sensitivity impact on target image quality, and the degree of deviation between the imaging angle and the target's optimal imaging direction are analyzed separately, and weighted processing is performed to calculate the composite observation cost, including: A non-cooperative game model of reconnaissance and avoidance is defined, in which a composite observation cost function is defined. Based on the position and velocity of the threat source and the angle between it and the reconnaissance drone, the cost of maintaining the field of view, the cost of resolution, and the cost of observation angle for the drone to reconnoiter the threat source are calculated, expressed as:

[0052]

[0053]

[0054] in, Indicates the cost of maintaining the field of view. Indicates the current location of the threat source. Indicates the current location of the reconnaissance drone. This indicates the redundant field-of-view radius, ensuring that the reconnaissance drone maintains a safe field of view within the target's maneuvering range. This represents the projection point of the target threat source onto the imaging plane of the reconnaissance drone. These represent the boundary points of the safe field of view. Indicates the cost of resolution. Indicates the distance between the surveillance drone and the threat source. This represents the optimal observation distance calculated based on camera parameters and target sharpness requirements. Indicates the camera performance correction factor. This indicates the angle between the target's attitude angle and the reconnaissance aircraft's line of sight. Indicates the cost of the observation angle. This represents the angle between the reconnaissance aircraft's line of sight to the target and the target's velocity vector normal plane. Indicates the speed of the reconnaissance drone. Indicates the speed of the threat source; We weight the costs of maintaining the field of view, resolution, and observation angle for UAV-based threat sources, and analyze the impact of target field of view deviation on reconnaissance equipment, the sensitivity-related impact on target image quality (ground resolution), and the degree of deviation between the current image imaging angle and the target's optimal imaging direction. These are expressed as follows:

[0055]

[0056]

[0057]

[0058]

[0059]

[0060]

[0061] in, This indicates the cost weighting for maintaining the field of view. The risk quantification value representing the current deviation of the target's field of view can be calculated and determined by the ratio of the distance of the target image from the center of the redundant field of view to the radius of the redundant field of view. This indicates the time during which the reconnaissance drone can safely maintain the center of its field of view. Indicates resolution cost weighting. This represents the current target resolution sensitivity quantization value. This indicates the real-time distance between the reconnaissance drone and the target. This indicates the working distance that reconnaissance drones should be adjusted to optimize target imaging clarity. This represents the actual deviation from the optimal ground sampling resolution. This indicates the minimum required image sharpness for threatening targets. This indicates the optimal distance between the reconnaissance drone and the target. Calculated resolution, This indicates the spatial size corresponding to a single pixel in the camera. Indicates the focal length of the camera system. Indicates the cost weight of the observation angle. This indicates that the current imaging angle deviates from the quantization value. This represents the probability value of the optimal imaging angle under the current configuration, which is directly quantified from the historical data of the probability distribution of target position and angle. The optimal imaging angle is determined based on target identification experimental data. The composite observation cost is calculated by weighting the costs of maintaining the field of view, resolution, and observation angle for drone-based threat source reconnaissance, and is expressed as follows:

[0062] in, This represents the cost of combined observations.

[0063] By defining and calculating the field-of-view maintenance cost in real time, the system ensures that the target is always within the safe field of view. The field-of-view maintenance cost function considers the degree to which the projection of the threat target deviates from the redundant field of view box. The degree of deviation is quantified by an exponential decay function, ensuring that the model can dynamically assess whether the reconnaissance UAV can always lock the target within the safe field of view. The redundant field of view radius introduces dynamic modeling of the "emergency deviation buffer", which allows the UAV to have time to adjust its path when the target moves rapidly, rather than immediately losing the target. This combination enables the reconnaissance UAV to adaptively optimize the dynamic "field of view locking" of the threat target, improving the continuous distance monitoring capability in complex scenarios. By dynamically coupling the weighted key factors, the priority of cost items can be flexibly adjusted. The importance of maintaining the field of view is enhanced in extreme evasion scenarios because it is directly related to the continuity of reconnaissance. When the target imaging clarity requirements change dynamically, the real-time response of the resolution cost weight significantly improves the ability to capture details of different targets. By coupling the target's dynamic characteristics with the imaging state, the multi-dimensional reinforcement capability of game decision-making is enhanced. The cost of maintaining the field of view and the cost of resolution emphasize spatial distribution optimization, while the cost of observation angle integrates dynamic geometric deviation. Their organic integration constructs a weighted mechanism for the optimal selection of dynamic scenarios and target characteristics. The UAV not only has the dynamic capability of real-time cost adaptation, but also enhances the rationality of reconnaissance decision-making under the target avoidance action mode. In the non-cooperative game model, the action optimization of the reconnaissance UAV not only has higher global adaptability to location, speed, and acceleration, but also can make accurate decisions under the dynamic influence of multiple variables, ensuring that the game reaches the optimal solution in terms of performance and robustness.

[0064] S4. Determine the optimization objective and perform iterative optimization until Nash equilibrium is reached and convergence is achieved. Output the trajectory optimization result and action sequence of the UAV. Combine with high-resolution images and radio frequency features to form a high-definition evidence chain vector, form a shared data packet, and transmit the data to the countermeasure module to perform collaborative data estimation and data synchronization. Preferably, the output drone trajectory optimization results and action sequences, combined with high-resolution images and radio frequency features, form a high-definition evidence vector, including: A two-layer non-cooperative game based on composite observation cost is used. The optimization objective is to minimize the dynamic observation cost function by adjusting the UAV's observation position, velocity, and acceleration. This includes determining the optimal actions of the reconnaissance UAV, minimizing the dynamic observation cost function based on the dynamic avoidance strategy of the threat target, and simultaneously performing inner-layer optimization, namely, seeking to maximize the composite observation cost for reconnaissance behavior avoidance actions (where the reconnaissance UAV's action commands include acceleration, turning rate, and trajectory update, and the threat source's avoidance commands include velocity and trajectory bias parameters). The reconnaissance behavior avoidance actions are updated using gradient ascent, and the UAV's observation position, velocity, and acceleration are updated using gradient descent. The game converges after reaching Nash equilibrium, and the trajectory optimization results and action sequence of the UAV are output. Based on the trajectory optimization results of the reconnaissance drone and the action sequence data collected, including the real-time trajectory of the threat target, high-resolution images, and radio frequency characteristics, a high-definition evidence chain vector is formed.

[0065] By employing a two-layer optimization game based on composite observation costs, and dynamically balancing the adversarial logic of reconnaissance and evasion, the combination of gradient ascent and descent methods in the two-layer optimization game achieves dynamic adaptation of path planning and adversarial actions. The action sequence generated by the trajectory optimization results ensures the dynamic adaptability of data acquisition quality. The reconnaissance UAV trajectory optimized by the two-layer game contains a set of optimal action sequences, precisely controlling the distance, observation angle, and tracking speed of the UAV to the threat target, so that the collected target data is always kept within a high availability range. By collecting real-time trajectories, high-resolution images, and radio frequency features of threat targets, a full-dimensional high-definition evidence chain is constructed.

[0066] Furthermore, shared data packets are formed and transmitted to the countermeasure module for collaborative data estimation and data synchronization, including: The high-definition evidence chain vector data is composed into a shared data packet, which is transmitted to the countermeasure module via the MQTT message queue protocol. A distributed filtering algorithm is then used to perform collaborative data estimation on multiple drones and the countermeasure module to synchronize the data.

[0067] By constructing shared data packets using high-definition evidence chain vector data, a multi-dimensional representation of the dynamic characteristics of threat targets is achieved. The MQTT message queue protocol ensures real-time and efficient data transmission. Through a data synchronization mechanism, the dynamic collaborative effectiveness between multiple UAVs and countermeasure modules is enhanced, enabling multi-platform collaborative countermeasures.

[0068] This embodiment also provides a multi-platform collaborative drone countermeasure system, including, The data acquisition module uses multiple UAV platforms to collect radio frequency signals and image data of threat targets, and extracts motion feature vectors and radio frequency feature vectors; The feature fusion module calculates instantaneous frequency drift, distortion deviation, and power characteristics based on the radio frequency signal, and obtains the optimal projection direction by canonical correlation analysis with the motion features, thus constructing a coupled feature matrix. The anomaly detection module trains an isolated forest model based on a weighted coupled feature matrix to detect a set of anomalous sample points and locates the spatial location of threat targets from the timestamps of anomalous signals. The reconnaissance optimization module outputs the optimal trajectory and action sequence of the UAV through weighted optimization of composite observation costs and iterative non-cooperative game model. The evidence chain generation module generates high-definition evidence chain vectors based on optimized trajectory acquisition high-resolution images and radio frequency features, providing comprehensive data characterization of threat targets; The shared collaboration module packages high-definition evidence chain vectors into shared data packets, which are then transmitted to the countermeasure module via a message queue protocol for collaborative data estimation and dynamic decision synchronization.

[0069] This embodiment also provides a computer device applicable to the multi-platform collaborative drone countermeasure method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the multi-platform collaborative drone countermeasure method proposed in the above embodiment.

[0070] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0071] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the UAV countermeasure method based on multi-platform collaboration as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0072] In summary, this invention enhances the distinguishability of radio frequency features by jointly characterizing constellation diagram distortion and power amplifier response curves, providing a comprehensive characterization capability from signal waveform details to power degradation behavior. Through a closed-loop mechanism of coupled features, anomaly detection, and position inversion, it significantly enhances the real-time performance and accuracy of countermeasure decision support. Anomaly analysis of sample points is directly linked to target spatial positioning and gradually optimizes target position estimation. By collecting real-time trajectories, high-resolution images, and radio frequency features of threat targets, a full-dimensional high-definition evidence chain is constructed, precisely controlling the distance, observation angle, and tracking speed of UAVs towards threat targets, ensuring that the collected target data is always kept within a high availability range.

[0073] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for countering unmanned aerial vehicles (UAVs) based on multi-platform collaboration, characterized in that, include: Data acquisition is performed using multiple UAV platforms, including radio frequency signals, radar scans, and image data. Motion feature vectors are extracted, instantaneous frequency drift is calculated based on instantaneous frequency fluctuations, and distortion deviation is calculated through radio frequency signal analysis. A power amplifier response curve is formed by fitting a logarithmic power curve, and the attenuation coefficient and the logarithm of the initial power intensity are extracted and combined with instantaneous frequency drift and distortion deviation to form a radio frequency feature vector. The optimal projection direction of kinematic and radio frequency features is analyzed through canonical correlation, and the final canonical correlation is calculated. Projecting is performed according to the optimal projection direction, forming a coupling feature matrix and calculating the coupling weight factor based on canonical correlation. The weighted coupling feature matrix is ​​used to construct an isolated forest model, and the path length in each isolated tree is calculated for anomaly detection, forming an anomaly sample point set. The signal timestamp is extracted and the TDOA method is used to back-calculate the target spatial location. The target location estimation model is constructed by combining the propagation distance difference to locate the location coordinates. Define a non-cooperative game model for reconnaissance and avoidance, calculate the field-of-view maintenance cost, resolution cost, and observation angle cost of UAV reconnaissance of threat sources, analyze the impact of target field-of-view deviation on reconnaissance equipment, the sensitivity impact of target image quality, and the degree of deviation between the imaging angle and the target's optimal imaging direction, and perform weighted processing to calculate the composite observation cost; The optimization objective is determined and iterative optimization is carried out until Nash equilibrium is reached and convergence is achieved. The trajectory optimization result and action sequence of the UAV are output. Combined with high-resolution images and radio frequency features, a high-definition evidence chain vector is formed, which is then used to form a shared data packet. The data is transmitted to the countermeasure module to perform collaborative data estimation and data synchronization.

2. The UAV countermeasure method based on multi-platform collaboration as described in claim 1, characterized in that: The extracted attenuation coefficient and the logarithm of the initial power intensity, combined with instantaneous frequency drift and distortion deviation, form an RF feature vector. The optimal projection direction of the kinematic and RF features is analyzed using canonical correlation, and the final canonical correlation is calculated, including... The target velocity and acceleration are calculated by measuring the distance the object moves within the video frame interval using the object image data. The target angular velocity is calculated by measuring the rotation angle of the object image data within the video frame interval using the position information collected by the radar module. This angular velocity is then combined with the target velocity and acceleration to form a motion feature vector. The instantaneous phase is calculated by capturing the target radio frequency signal using an SDR sensor, and the instantaneous frequency is calculated by numerically differentiating the instantaneous phase sequence using the center difference method. The instantaneous frequency drift is then calculated based on the standard deviation of the instantaneous frequency fluctuation. The quadrature and in-phase components of the normalized radio frequency signal are used to calculate the distortion based on the IQ constellation diagram and the distortion deviation is calculated based on the IQ components corresponding to the ideal constellation points. Power is calculated based on the quadrature and in-phase signal components, and the logarithmic change of the amplifier response is represented by fitting a logarithmic power curve to form the power amplifier response curve. The radio frequency characteristic vector is composed of instantaneous frequency drift, distortion deviation, attenuation coefficient, and logarithmic value of initial power intensity; The motion feature vector and radio frequency feature vector are processed by standard deviation. Correlation analysis is performed by canonical correlation analysis (CCA). The covariance matrix of the motion feature vector and radio frequency feature vector after standard deviation processing is calculated. Eigenvalue decomposition is performed to obtain projection vectors a and b, which correspond to the optimal projection directions of the kinematic feature and radio frequency feature, respectively. Finally, the canonical correlation is calculated.

3. The UAV countermeasure method based on multi-platform collaboration as described in claim 2, characterized in that: The set of anomalous sample points is used to extract signal timestamps, and the TDOA method is used to inversely calculate the target's spatial location. A target location estimation model is constructed by combining the propagation distance difference to pinpoint the location coordinates. The motion feature vector and the radio frequency feature vector are projected separately according to the projection vector, and then merged and spliced ​​into a coupled feature vector. The coupling feature vector is used to form a coupling feature matrix based on the number of sample points, motion features, and radio frequency features. The coupling weight factor is calculated based on the variance and canonical correlation value of the coupling feature matrix, and the coupling feature matrix is ​​adjusted. An isolated forest model is constructed using a weighted coupled feature matrix. A subset of samples is randomly selected, and features are randomly selected. A segmentation threshold is randomly set in the selected feature dimensions. The subset samples are recursively split according to the segmentation conditions. The isolated forest model is iterated repeatedly until the maximum split depth is reached. The isolated forest is used to calculate the path length of each sample point in each isolated tree and to calculate the anomaly score for anomaly detection. The anomaly threshold is based on the sum of the mean and twice the standard deviation of historical outliers. If the calculated outlier is greater than or equal to the anomaly threshold, the sample point is judged to be an anomaly and a set of anomaly sample points is formed. Based on the set of abnormal sample points, the timestamps of the signals received by the sensors in different UAVs are obtained and used as threat sources. The TDOA method is used to calculate the target spatial location, which calculates the time difference between each pair of sensors and considers the signal propagation speed to calculate the propagation distance difference between the threat source and the sensor. A target position estimation model is constructed by combining the propagation distance difference, and the target position is solved by nonlinear least squares method; The gradient descent method is used to iteratively solve the problem until the target iterative value converges and the target location estimation function value is minimized, thus obtaining the location coordinates of the threat source.

4. The UAV countermeasure method based on multi-platform collaboration as described in claim 3, characterized in that: The analysis separately examines the impact of target field-of-view deviation on reconnaissance equipment, the sensitivity impact on target image quality, and the degree of deviation between the imaging angle and the target's optimal imaging direction. Weighted processing is then used to calculate the composite observation cost, including... A non-cooperative game model of reconnaissance and avoidance is defined, in which a composite observation cost function is defined. Based on the position and velocity of the threat source and the angle between it and the reconnaissance drone, the cost of maintaining the field of view, the resolution cost, and the observation angle cost of the drone reconnaissance of the threat source are calculated. We weighted the costs of maintaining the field of view, resolution, and observation angle for UAV reconnaissance threat sources, and analyzed the impact of target field of view deviation on reconnaissance equipment, the sensitivity-related impact on target image quality, and the degree of deviation between the current image imaging angle and the target's optimal imaging direction. The composite observation cost is calculated by weighting the costs of maintaining the field of view, resolution, and observation angle for drone-based threat source reconnaissance.

5. The UAV countermeasure method based on multi-platform collaboration as described in claim 4, characterized in that: The output drone trajectory optimization results and action sequence, combined with high-resolution images and radio frequency features, form a high-definition evidence vector, including: A two-layer non-cooperative game based on composite observation cost is used. The optimization objective is to minimize the dynamic observation cost function by adjusting the UAV's observation position, velocity, and acceleration. This includes determining the optimal action of the reconnaissance UAV, minimizing the dynamic observation cost function based on the dynamic avoidance strategy of the threat target, and performing inner-layer optimization, namely, seeking to maximize the composite observation cost for reconnaissance behavior avoidance actions. The reconnaissance behavior avoidance actions are updated using the gradient ascent method, and the UAV's observation position, velocity, and acceleration are updated using the gradient descent method. The game converges after reaching Nash equilibrium and outputs the trajectory optimization results and action sequence of the UAV. Based on the trajectory optimization results of the reconnaissance drone and the action sequence data collected, including the real-time trajectory of the threat target, high-resolution images, and radio frequency characteristics, a high-definition evidence chain vector is formed.

6. The UAV countermeasure method based on multi-platform collaboration as described in claim 5, characterized in that: The shared data packet is transmitted to the countermeasure module for collaborative data estimation and data synchronization, including: The high-definition evidence chain vector data is composed into a shared data packet, which is transmitted to the countermeasure module via the MQTT message queue protocol. A distributed filtering algorithm is then used to perform collaborative data estimation on multiple drones and the countermeasure module to synchronize the data.

7. The UAV countermeasure method based on multi-platform collaboration as described in claim 6, characterized in that: The data acquisition based on multiple UAV platforms includes radio frequency signals, radar scans, and image data, including... The drone is equipped with optical and infrared cameras, radar modules, and SDR sensors. The optical and infrared cameras acquire high-resolution image data of the object, the SDR sensors acquire radio frequency signals of the drone, including quadrature and in-phase signal components, and the radar modules acquire position information.

8. A multi-platform collaborative drone countermeasure system, based on the multi-platform collaborative drone countermeasure method according to any one of claims 1 to 7, characterized in that: include, The data acquisition module uses multiple UAV platforms to collect radio frequency signals and image data of threat targets, and extracts motion feature vectors and radio frequency feature vectors; The feature fusion module calculates instantaneous frequency drift, distortion deviation, and power characteristics based on the radio frequency signal, and obtains the optimal projection direction by canonical correlation analysis with the motion features, thus constructing a coupled feature matrix. The anomaly detection module trains an isolated forest model based on a weighted coupled feature matrix to detect a set of anomalous sample points and locates the spatial location of threat targets from the timestamps of anomalous signals. The reconnaissance optimization module outputs the optimal trajectory and action sequence of the UAV through weighted optimization of composite observation costs and iterative non-cooperative game model. The evidence chain generation module generates high-definition evidence chain vectors based on optimized trajectory acquisition high-resolution images and radio frequency features. The shared collaboration module packages high-definition evidence chain vectors into shared data packets, which are then transmitted to the countermeasure module via a message queue protocol for collaborative data estimation and dynamic decision synchronization.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the multi-platform collaborative UAV countermeasure method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the multi-platform collaborative UAV countermeasure method according to any one of claims 1 to 7.

Images