A method, device, and storage medium for capturing a target unmanned aerial vehicle

Abstract

The present application relates to the technical field of unmanned aerial vehicle, and especially relates to a method and device for capturing a target unmanned aerial vehicle, equipment and storage medium; image sequence data and multi-source observation data are predicted according to a preset maneuvering tracking filter model to obtain target state estimation and future predicted trajectory; threat assessment is performed on the target state estimation and future predicted trajectory to obtain a threat assessment result; the threat assessment result is analyzed according to a preset disposal condition to obtain a threat analysis result; the target unmanned aerial vehicle is captured according to the threat analysis result, a preset capture envelope condition and a preset capture distance range; after capturing the target unmanned aerial vehicle, detection and shooting data are collected; by constructing an unmanned aerial vehicle security whole-process disposal system, multi-source sensing and collection data are accurately estimated on target state and trajectory through the maneuvering tracking filter model, disposal decisions are output after quantifying threat assessment, and low-altitude security decision efficiency is improved in all directions.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to a method, apparatus, device, and storage medium for capturing a target UAV. Background Technology

[0002] Existing technologies have significant shortcomings in the field of drone interception: target drone tracking methods lack maneuverability and adaptability, making it difficult to match the high-speed maneuverability and directional flight characteristics of target drones. This easily leads to problems such as trajectory estimation drift and target loss, making it impossible to accurately output the predicted trajectory of the target drone and making it difficult to formulate efficient interception decisions. The close-range acquisition stage only uses a single distance threshold as the trigger condition, resulting in large fluctuations in the acquisition hit rate and unstable interception effects. Ultimately, the overall interception operation suffers from insufficient accuracy, reliability, and compliance traceability, making it difficult to meet the needs of drone interception in complex scenarios. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, the purpose of this invention is to provide a method, apparatus, device and storage medium for capturing a target drone.

[0004] The first aspect of this invention provides a method for capturing a target unmanned aerial vehicle (UAV), comprising: acquiring image sequence data and multi-source observation data; predicting the image sequence data and multi-source observation data according to a preset maneuvering tracking filter model to obtain a target state estimate and a future predicted trajectory; performing a threat assessment on the target state estimate and the future predicted trajectory to obtain a threat assessment result; analyzing the threat assessment result according to preset handling conditions to obtain a threat analysis result; capturing the target UAV according to the threat analysis result, preset capture envelope conditions, and preset capture distance range; and acquiring detection and shooting data after capturing the target UAV.

[0005] Furthermore, the step of predicting the image sequence data and multi-source observation data according to the preset maneuvering tracking filtering model to obtain the target state estimate and future predicted trajectory includes: performing registration processing on the multi-source observation data to obtain registered observation data; analyzing the registered observation data according to the preset factory calibration error model to obtain the observation confidence sequence and the measurement covariance sequence; and performing weighted fusion of the observation confidence sequence and the measurement covariance sequence to obtain the fused observation value and the fused covariance. The image sequence data is detected based on a pre-set deep learning model to obtain adaptive parameters; the adaptive parameters, fusion covariance and fusion observations are predicted according to the maneuvering tracking filter model to obtain target state estimation and future predicted trajectory.

[0006] Furthermore, the step of predicting adaptive parameters, fusion covariance, and fusion observations based on the maneuvering tracking filter model to obtain target state estimation and future predicted trajectory includes: analyzing adaptive parameters based on a preset uniform velocity model and a preset transformation method to obtain switching weights, process noise parameters, scale factors, and increment terms; adjusting a preset measurement noise matrix based on the scale factors and increment terms to obtain an optimized noise matrix; and predicting the optimized noise matrix, switching weights, process noise parameters, fusion covariance, and fusion observations based on the maneuvering tracking filter model to obtain target state estimation and future predicted trajectory.

[0007] Furthermore, the step of analyzing the adaptive parameters according to the preset uniform speed model and the preset conversion method to obtain the switching weight, process noise parameter, scale factor and increment term includes: obtaining the maneuver probability and measurement noise estimate from the adaptive parameters; analyzing the maneuver probability according to the uniform speed model to obtain the switching weight and process noise parameter; and converting the measurement noise estimate according to the conversion method to obtain the scale factor and increment term.

[0008] Furthermore, the target drone capture system also includes a tracking drone. The step of capturing the target drone based on threat analysis results, preset capture envelope conditions, and preset capture distance range includes: generating an interception command based on the threat analysis results when the threat assessment results meet the handling conditions; obtaining the relative speed between the target drone and the tracking drone based on the interception command; calculating the capture distance based on preset basic engineering calculation formulas for the preset effective coverage radius of the capture net, preset net deployment time, and relative speed; and capturing the target drone based on the capture distance, capture envelope conditions, and capture distance range.

[0009] Furthermore, the step of capturing the target UAV based on the capture distance, capture envelope conditions, and capture distance range includes: determining whether the capture distance is within the capture distance range; if the capture distance is within the capture distance range, then obtaining the relative distance and relative heading angle between the target UAV and the tracking UAV; The relative distance and relative heading angle are analyzed based on the capture envelope conditions and the preset capture range to obtain the envelope analysis results; if the envelope analysis results show that the relative geometric area is within the capture range, a capture command is generated based on the capture distance; the target UAV is captured based on the capture command.

[0010] Furthermore, the step of analyzing the relative distance and relative heading angle based on the capture envelope conditions and the preset capture range to obtain the envelope analysis results includes: calculating the relative distance, relative motion speed, relative heading angle and net deployment time to obtain the relative geometric region; and analyzing the relative geometric region based on the capture envelope conditions and the capture range to obtain the envelope analysis results.

[0011] Furthermore, a target drone capture system performs a target drone capture method as described above, the target drone capture system comprising a control device, a target drone electrically connected to the control device, and a tracking drone.

[0012] Furthermore, a target drone capture device includes: a memory and at least one processor, the memory storing instructions; at least one processor invokes the instructions in the memory to cause the target drone capture device to perform the various steps of the target drone capture method described above.

[0013] Furthermore, a computer-readable storage medium stores instructions that, when executed by a processor, implement the steps of a target acquisition method for unmanned aerial vehicles as described above.

[0014] Furthermore, a target drone capture device includes: a memory and at least one processor, the memory storing instructions; at least one processor invokes the instructions in the memory to cause the target drone capture device to perform the steps of the target drone capture method described above.

[0015] Furthermore, a computer-readable storage medium stores instructions that, when executed by a processor, implement the steps of a target acquisition method for unmanned aerial vehicles as described above.

[0016] In the technical solution of this invention, image sequence data and multi-source observation data are collected simultaneously through a multi-source sensing system to ensure the comprehensiveness and temporal consistency of the data; the image sequence data and multi-source observation data are predicted based on a maneuvering tracking filtering model to accurately output target state estimates and future trajectories, adapting to various maneuvering scenarios; a quantitative threat assessment is carried out based on the target state estimate and future predicted trajectory, and precise handling decisions are output in combination with preset handling conditions; the capture stage takes the threat analysis results as the core, and achieves compliant and efficient interception by combining the capture envelope conditions and capture distance range; after capture, crash site detection data can also be collected to provide reliable basis for on-site handling and source tracing investigation, comprehensively improving the decision-making efficiency and execution standardization of low-altitude security. Attached Figure Description

[0017] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a first flowchart of a method for capturing a target drone provided by an embodiment of the present invention; Figure 2 This is a second flowchart of a method for capturing a target drone provided in an embodiment of the present invention; Figure 3 This is a third flowchart of a method for capturing a target drone provided by an embodiment of the present invention; Figure 4 This is a fourth flowchart of a method for capturing a target drone provided in an embodiment of the present invention; Figure 5 This is a fifth flowchart of a method for capturing a target UAV provided in an embodiment of the present invention; Figure 6 This is a sixth flowchart of a method for capturing a target drone provided in an embodiment of the present invention; Figure 7 This is a seventh flowchart of a method for capturing a target drone provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of a target acquisition system for a drone, provided in an embodiment of the present invention. Figure 9 This is a schematic diagram of a target acquisition device for a drone, provided in an embodiment of the present invention.

[0018] Figure Labels 1-Control device; 2-Target drone; 3-Tracking drone. Detailed Implementation

[0019] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the target drone capture method of the present invention includes: 101. Acquire image sequence data and multi-source observation data; In this embodiment, a multi-source sensing system is used to collect multi-source observation data and image sequence data. This system includes radar (preferably phased array radar), lidar, visible light camera, infrared camera, and acoustic array, with radio frequency detection being an optional configuration. The multi-source observation data covers lidar data, visible light data, infrared data, acoustic array data, and radio frequency observation data, and a unified timestamp is configured for all collected data. The image sequence data is divided into two categories: one is visible light image sequence data, which is continuous color frame image data / grayscale frame image data collected by the visible light camera, containing visible light visual features such as the shape, texture, and motion trajectory of the target UAV, as well as information about the surrounding environment, providing intuitive visual basis for target identification and positioning; the other is infrared image sequence data, which is continuous infrared thermal imaging frame image data collected by the infrared camera, forming a thermal feature frame sequence based on the temperature difference between the target and the environment, which can capture the thermal radiation characteristics of the UAV and is suitable for continuous target tracking in low visibility scenarios such as night and fog. 102. Based on the preset maneuvering tracking filter model, the image sequence data and multi-source observation data are predicted to obtain the target state estimate and future predicted trajectory. In this embodiment, prediction is carried out by relying on the maneuvering tracking filter model to fuse image sequence data and multi-source observation data. This effectively integrates the advantages of multi-dimensional perception data and accurately outputs target state estimation and future predicted trajectory. The model can adapt to different maneuvering characteristics of the target, take into account both linear and nonlinear motion scenarios, improve the estimation accuracy of the target's real-time motion state, and reliably predict its future flight trajectory. This provides accurate and quantitative data support for subsequent UAV interception and disposal, and improves the intelligence and decision-making efficiency of low-altitude security. 103. Conduct a threat assessment on the estimated target state and predicted future trajectory to obtain the threat assessment results; In this embodiment, the target's real-time three-dimensional position, flight speed, heading angle, and maneuvering characteristics are obtained based on target state estimation. Simultaneously, its subsequent flight path, potential destinations, and movement trends are determined by combining future predicted trajectories. Secondly, based on preset security protection requirements, key evaluation indicators are extracted and quantified from dimensions such as target threat potential (e.g., whether it is an unauthorized drone or has payload capacity), trajectory threat level (e.g., whether it approaches no-fly zones or key protection facilities, or whether it exhibits malicious maneuvering trajectories), and real-time state risk (e.g., excessive flight speed or abnormal flight attitude). Finally, the quantified results of each dimension are combined with preset weights for comprehensive analysis, classifying the threat level (low, medium, high) to form the final threat assessment result, providing a quantitative decision-making basis for whether to initiate capture and disposal measures. 104. Analyze the threat assessment results based on the preset handling conditions to obtain the threat analysis results; In this embodiment, the handling conditions are formulated based on the protection requirements of low-altitude security scenarios. Differentiated handling thresholds and execution criteria are set according to the low, medium, and high levels defined by the threat assessment, and include quantitative standards such as threat index thresholds and handling action trigger boundaries. Then, the threat assessment results of the target UAV (including the overall threat level and quantitative indicators of each dimension) are compared with the preset handling conditions one by one to determine whether it has reached the trigger threshold of different handling actions such as capture, expulsion, or monitoring only, and to verify whether it meets the special handling constraints of the protection scenario. Finally, a unique threat analysis result is output based on the matching result to clarify the specific handling decision for the target UAV, that is, whether to initiate the capture process or only implement expulsion / continuous monitoring, providing a standardized and implementable execution basis for whether to carry out net capture and interception in the future. 105. Capture the target UAV based on the threat analysis results, the preset capture envelope conditions, and the preset capture distance range; In this embodiment, the threat analysis results are used as the core basis for execution to ensure that the capture action is accurately matched with the target threat level and to avoid ineffective interception; the scene compliance boundary is defined based on the capture envelope conditions, and the capture distance range is limited to the hardware operation range, providing a standardized and implementable capture execution solution for low-altitude security; 106. After capturing the target drone, collect detection and shooting data; In this embodiment, after capturing the target drone, the system hovers to detect and photograph the crash site and transmits the data back, thus achieving full-process information retention after capture. The timely collected detection and photography data can completely record key information such as the target crash location, providing real and reliable visual and location evidence for subsequent on-site handling and source tracing investigation, and improving the source tracing efficiency of low-altitude security handling. In this embodiment, image sequence data and multi-source observation data are simultaneously collected through a multi-source sensing system to ensure the comprehensiveness and temporal consistency of the data. The image sequence data and multi-source observation data are predicted based on a maneuvering tracking filtering model to accurately output target state estimates and future trajectories, adapting to various maneuvering scenarios. Quantitative threat assessment is carried out based on target state estimates and future predicted trajectories, and precise handling decisions are output in combination with preset handling conditions. The capture stage takes the threat analysis results as the core, and achieves compliant and efficient interception by combining capture envelope conditions and capture distance range. After capture, crash site detection data can also be collected to provide reliable basis for on-site handling and source tracing investigation, comprehensively improving the decision-making efficiency and execution standardization of low-altitude security.

[0021] Please see Figure 2 A second embodiment of a method for capturing a target drone according to an embodiment of the present invention specifically includes: 201. Perform registration processing on the multi-source observation data to obtain registered observation data; In this embodiment, the registration processing of multi-source observation data includes asynchronous spatiotemporal registration and coordinate unification (unifying the station-centered coordinate system or geographic coordinate system). The registration processing includes asynchronous spatiotemporal registration and coordinate unification. Asynchronous spatiotemporal registration: To address the issues of asynchronous timestamps and different spatial sampling locations in the data collected by devices such as radar, lidar, and visual cameras, spatiotemporal synchronization of data is achieved through time interpolation and spatial coordinate mapping, ensuring that multi-source data from the same time and spatial location can be compared and fused. Coordinate unification: The observation data of all devices are uniformly converted to the station-centered coordinate system or geographic coordinate system to eliminate spatial deviations caused by the coordinate systems of different devices. For example, the polar coordinate data of radar and the pixel coordinates of visual cameras are uniformly converted into geographic latitude, longitude, and altitude coordinates. 202. Analyze the registration observation data according to the preset factory calibration error model to obtain the observation confidence sequence and measurement covariance sequence; In this embodiment, the factory calibration error model is the core reference for the entire analysis, and this model contains two key pieces of information: Inherent equipment errors: Basic measurement errors calibrated at the factory for equipment such as radar, vision cameras, and lidar (e.g., radar ranging error ±0.1m, vision pixel positioning error ±2 pixels); Error correlation rules: The correlation between data quality indicators (e.g., signal-to-noise ratio, occlusion rate) and error amplification (e.g., when the occlusion rate is >50%, the visual data error is amplified by 3 times). This model provides a unified judgment standard for subsequent "quality indicator extraction and error quantification", ensuring that the analysis results are consistent with the hardware characteristics of the equipment; From the registered observation data, quality index vectors that are strongly correlated with data reliability are extracted. Different types of equipment correspond to different core indicators: for radar and acoustic array equipment, the signal-to-noise ratio (SNR) is extracted to reflect the clarity of the signal. The lower the SNR, the greater the interference and the lower the reliability of the data. For visual cameras and infrared cameras, the occlusion rate (the proportion of the target that is occluded), visibility (the degree of environmental visibility), and pixel ratio (the proportion of the target in the image) are extracted. The higher the occlusion rate, the lower the visibility, and the smaller the pixel ratio, the lower the data reliability. The extracted quality index vector is normalized to map indices of different dimensions and ranges to the [0,1] interval, forming an observation confidence sequence. Set the scoring rules for the indicators: for example, a signal-to-noise ratio (SNR) ≥ 30dB scores 1 point (optimal), < 10dB scores 0 points (unusable); an occlusion rate of 0 scores 1 point, > 80% scores 0 points; based on the correlation between the indicators and data reliability, use a normalization formula (e.g., minimum normalization - maximum normalization). The original index values ​​are converted into confidence scores in the range of [0,1]. The confidence scores are then organized into a sequence according to the time / spatial dimensions, with each score corresponding to a set of registered observation data, thus intuitively quantifying the reliability of each set of data. By combining the extracted quality index vector, the error range of the registered observation data is quantified, and the measurement covariance sequence is output: the inherent measurement covariance of the corresponding device is retrieved (e.g., radar ranging covariance is 0.01m²); the basic covariance is adjusted according to the quality index vector (e.g., when the occlusion rate is 50%, the visual data covariance is amplified by 2 times), and the corrected covariance reflects the actual error fluctuation under the current environmental quality / data quality; the measurement covariance sequence is output according to the same dimension as the observation confidence sequence. 203. Perform weighted fusion of the observation confidence sequence and the measurement covariance sequence to obtain the fused observation values ​​and fused covariance; In this embodiment, the observation confidence scores in the observation confidence score sequence are first analyzed. Measurement covariance in the measurement covariance sequence Calculate weights ,in To measure covariance Based on observation confidence The modulated measurement covariance is then weighted and fused with the observation confidence sequence and the measurement covariance sequence to obtain the fused covariance. Combined observations Output ; 204. Detect image sequence data based on a preset deep learning model to obtain adaptive parameters; In this embodiment, the deep learning model targets image sequence data monitored by UAVs, employing an end-to-end integrated target perception and feature analysis detection principle. First, it utilizes a deep convolutional neural network (CNN) combined with temporal modeling (such as 3D-CNN or CNN+LSTM) to extract spatial and temporal dynamic features of the image sequence. Then, through a multi-task learning branch, it simultaneously completes target detection, category recognition, and adaptive parameter output. Firstly, the deep learning model performs inter-frame feature fusion on the input continuous image sequence, capturing the UAV's spatial morphological features (such as contour, size, and texture) and temporal motion features (such as inter-frame displacement, velocity changes, and flight attitude trends). It then locates the UAV target in the image using anchor-frame matching or no-anchor-frame detection algorithms, achieving accurate target detection. Secondly, based on the detection, the classification branch is used to analyze the extracted target features. The system performs classification and identification to distinguish the type and model of drones, while filtering out interfering targets such as birds and debris. Finally, the deep learning model, through a dedicated regression analysis branch, calculates adaptive parameters such as noise estimation and maneuver probability based on extracted temporal motion features and image quality features. Noise estimation quantifies the noise interference level of image data by analyzing features such as pixel fluctuations, background interference, and target imaging blur in the image sequence. Maneuver probability predicts the maneuver trend and probability of the drone by analyzing dynamic features such as the curvature, velocity change rate, and displacement abruptness of the drone's inter-frame motion trajectory. The overall deep learning model is trained and optimized through a large amount of labeled drone image sequence data, achieving simultaneous completion of detection, identification, and adaptive parameter output. The output parameters can accurately adapt to the dynamic adjustment requirements of subsequent tracking filtering. 205. Based on the maneuvering tracking filter model, the adaptive parameters, fusion covariance, and fusion observations are predicted to obtain the target state estimate and the future predicted trajectory. In this embodiment, the adaptive parameters provide core parameter support for dynamic adaptation of the filtering, and the fusion of covariance and observation values ​​ensures the comprehensiveness and accuracy of data input. The maneuvering tracking filtering model can accurately adapt to different maneuvering characteristics of the target, improve the accuracy of target state estimation, and reliably predict future flight trajectories and estimate errors, providing accurate quantitative data support for subsequent UAV interception and disposal, and improving the effectiveness and pertinence of low-altitude security decisions. In this embodiment, asynchronous spatiotemporal registration and coordinate unification eliminate spatiotemporal and coordinate deviations in multi-source device data, achieving data comparability and solidifying the foundation for analysis. Quality indicators are extracted and normalized using the factory calibration error model to generate a confidence score sequence. These indicators are then combined with the inherent errors of the equipment to obtain the observation confidence score sequence and the measurement covariance sequence, accurately quantifying data reliability and error boundaries. Weighted fusion of confidence scores and covariances enables complementary selection of multi-source data, improving the accuracy of observation data and avoiding interference from low-quality data. A deep learning model simultaneously completes adaptive parameter outputs for target detection and recognition, noise estimation, and maneuver probability, precisely adapting to subsequent filtering requirements. A maneuver tracking filtering model predicts adaptive parameters, fusion covariance, and fusion observations, accurately estimating the target state and reliably predicting future trajectories, adapting to complex maneuver scenarios. The entire solution improves the accuracy and robustness of target perception and prediction, providing quantitative data support for low-altitude counter-interception and enhancing the intelligence and precision of low-altitude security protection.

[0022] Please see Figure 3 A third embodiment of a method for capturing a target drone according to an embodiment of the present invention specifically includes: 301. Analyze the adaptive parameters according to the preset uniform velocity model and preset conversion method to obtain the switching weight, process noise parameter, scale factor and increment term; In this embodiment, the uniform velocity model is adapted to the target's maneuver characteristics, allowing the switching weights and process noise parameters to be dynamically adjusted according to the target's motion trend. Furthermore, the noise-related parameters are quantified through a conversion method, providing comprehensive and accurate parameter support for subsequent measurement noise matrix optimization and maneuver tracking filter iteration. This effectively solves the pain point of fixed traditional filter parameters, improves parameter adaptability, and lays a solid foundation for the accuracy of target state estimation and trajectory prediction. 302. Adjust the preset measurement noise matrix according to the scaling factor and increment term to obtain the optimized noise matrix; In this embodiment, an optimized noise matrix is ​​obtained by adjusting the preset measurement noise matrix through a scaling factor and an increment term, thereby achieving accurate dynamic correction of measurement noise. The scaling factor adapts to noise changes proportionally, and the increment term compensates for inherent system biases, making the noise matrix fit the actual observation scenario. This effectively improves the anti-interference capability of the filtering model and lays a reliable data foundation for subsequent accurate estimation of the target state. 303. Based on the maneuvering tracking filter model, the noise matrix, switching weights, process noise parameters, fusion covariance, and fusion observations are predicted to obtain the target state estimate and future predicted trajectory. In this embodiment, the maneuver tracking filtering model uses an interactive multi-model (IMM) framework as its core, integrating KF, EKF, and UKF filtering algorithms to achieve target state estimation and future trajectory prediction. The core principle is as follows: First, preset filtering sub-models adapted to different maneuver modes (uniform speed, uniform acceleration, turning, etc.). Then, update the initial weights of each sub-model based on the switching weights obtained from adaptive parameter analysis. Use fused observations as input and fused covariance to quantize the initial error. Simultaneously, incorporate the analyzed process noise parameters into the filtering process to dynamically adjust the process noise covariance. For linear / nonlinear maneuver characteristics of the target, KF (linear scenario), EKF (nonlinear scenario), and UKF (nonlinear scenario) are selected respectively to perform "prediction-update" iterations. During the iteration process, the self-adaptive parameters are used to dynamically adjust the process noise covariance. The scale factor and incremental term-optimized measurement noise matrix obtained from the adaptive parameter transformation are used to replace the original preset measurement noise matrix in the calculation, effectively offsetting the interference of data noise and inherent system bias. Then, the likelihood probability of each sub-model is calculated and the posterior probability is updated. The state estimation results of all sub-models are weighted and fused to output the globally optimal target state estimate (including three-dimensional position (latitude, longitude, altitude), flight speed, heading angle, pitch angle, acceleration and other indicators that characterize the real-time motion state). Finally, using this target state estimate as the initial value, combined with the maneuver equations of each sub-model and the posterior probability, the target's position, velocity and other states at future times are predicted in multiple steps to generate the future predicted trajectory. By fusing covariance, the trajectory error range is estimated, realizing accurate estimation of the target state and reliable prediction of the trajectory in complex maneuver scenarios. In this embodiment, the adaptive parameters are analyzed using a uniform velocity model and transformation method. The output switching weights and process noise parameters can be dynamically adjusted according to the target's motion trend. The scale factor and increment term provide precise support for noise correction, completely solving the pain point of fixed traditional filter parameters and improving the adaptability of parameters to target characteristics and observation scenarios. By adjusting the measurement noise matrix through the scale factor and increment term, an optimized noise matrix is ​​obtained, achieving proportional adaptation of noise and compensation for system bias, thus improving the anti-interference capability of the filter model. Finally, the filter model based on the IMM framework integrates multiple optimized parameters and fused observation data to adapt to the target's linear / nonlinear maneuvering characteristics, outputting a globally optimal target state estimate. It can also reliably predict future trajectories and estimate errors, making state estimation more accurate and trajectory prediction more reliable in complex maneuvering scenarios, providing precise quantitative data support for subsequent UAV interception and disposal.

[0023] Please see Figure 4 The fourth embodiment of a method for capturing a target drone according to the present invention specifically includes: 401. Obtain the maneuver probability and measurement noise estimate from the adaptive parameters; 402. Analyze the maneuver probability based on the uniform velocity model to obtain the switching weights and process noise parameters; In this embodiment, the uniform velocity model uses itself as the basic reference benchmark for the target motion, and performs quantitative analysis on the maneuver probability to output switching weights and process noise parameters. The core principle is as follows: the maneuver probability is used as the core quantitative indicator to determine the degree to which the target deviates from uniform velocity motion. When the maneuver probability is lower, it indicates that the target motion state is closer to uniform velocity, so the switching weight of the uniform velocity model in the filtering framework is increased, and at the same time, a smaller process noise parameter is matched to adapt to the stable motion state. When the maneuver probability is higher, it indicates that the target deviates from uniform velocity motion and the possibility of non-uniform velocity maneuvering is greater, so the switching weight of the uniform velocity model is reduced accordingly, and the process noise parameter is increased simultaneously to adapt to the dynamic changes in the target motion state. In this way, the two output parameters accurately match the actual motion trend of the target, providing adaptive parameter support for subsequent filtering iterations. 403. Transform the estimated measurement noise value according to the transformation method to obtain the scaling factor and increment term; In this embodiment, the conversion method involves calculating the estimated measurement noise using a scaling factor conversion formula to obtain the scaling factor; and calculating the estimated measurement noise using an increment term conversion formula to obtain the increment term. The expression for the scaling factor conversion formula is as follows: In the formula, This is the baseline coefficient for the scaling factor, determined by the fundamental characteristics of the filtering model (default is to adapt to the uniform velocity model), and is usually set to 1.0; The output scaling factor is used to adjust the measurement noise covariance of the filter model proportionally. This is a scaling factor adjustment coefficient, used to adapt the sensitivity of the noise estimate (which can be corrected according to the environment). The input measurement noise estimate (obtained from adaptive parameters, quantifying the noise level of the image data); This serves as a reference value for measuring noise, representing the basic measurement noise (an inherent property of the hardware) in a non-interference scenario. The expression for the incremental term conversion formula is as follows: In the formula, The incremental term of the output is used to compensate for the fixed bias of the measurement noise covariance; This is the proportional coefficient for the incremental term, and the compensation weight for the associated noise estimate. The incremental term is used as a baseline compensation value to correct inherent biases in the system (such as sensor zero bias and model error). In this embodiment, the maneuver probability and measurement noise estimate are accurately extracted from the adaptive parameters. The maneuver probability is analyzed based on the uniform velocity model. The output switching weights and process noise parameters can be dynamically adjusted according to the actual motion trend of the target, so that the model weights and noise parameters of the filtering framework are accurately matched with the target motion state, adapting to different scenarios of uniform and non-uniform velocity. At the same time, the measurement noise estimate is quantified by a dedicated conversion formula, and the scale factor and increment term are output. The scaling of the measurement noise covariance and the compensation for fixed deviation can be realized respectively, effectively offsetting the interference caused by image data noise and inherent system deviations. This provides accurate and reliable parameter support for subsequent filtering iterations and improves the accuracy of target state estimation and trajectory prediction.

[0024] Please see Figure 5 The fifth embodiment of a method for capturing a target drone according to the present invention specifically includes: 501. If the threat analysis result indicates that the threat assessment result meets the handling conditions, then an interception command is generated based on the threat analysis result; 502. Obtain the relative speed between the target drone and the tracking drone according to the interception command; In this embodiment, the relative motion speed between the target drone and the tracking drone is obtained based on the interception command, so as to achieve the precision of interception triggering; the relative motion speed is the resultant speed of the relative radial speed (approaching speed / moving speed) and tangential speed (lateral offset speed) of the target drone and the tracking drone, which reflects the motion trend of the target drone relative to the tracking drone and is the key to predicting the "target position shift during the deployment of the net". 503. Based on the preset basic engineering calculation formula, the effective coverage radius of the preset capture net, the preset net deployment time, and the relative motion speed are used to calculate the capture distance; In this embodiment, the expression for the basic engineering calculation formula is as follows: The net deployment time is the inherent physical time from the issuance of the trigger command to the complete deployment of the net capture device to form an effective capture surface (determined by the hardware characteristics of the device, such as the aerodynamic deployment time of a folding net and the net setting time of a catapult net, which is generally a fixed value that can be preset or modified according to the environment), and is the core constraint in the time dimension. The value is generally between 0.2m and 0.5m, and it is a key compensation parameter to adapt to actual engineering scenarios; 504. Capture the target UAV based on the capture distance, capture envelope conditions, and capture distance range; In this embodiment, network interception is carried out by combining the capture distance, capture envelope conditions and capture distance range. The physical range of capture distance is defined by hardware characteristics to improve the capture success rate and comprehensively ensure the effectiveness, security and compliance of the interception. In this embodiment, an interception command is generated when the handling conditions are met, avoiding the waste of resources caused by indiscriminate interception. At the same time, the relative motion speed between the target UAV and the tracking UAV is obtained based on the interception command. This speed accurately reflects the target's motion trend and provides a key basis for predicting the target's deviation when the net is deployed. The capture distance is calculated by combining the capture net parameters, net deployment time, and relative motion speed, so that the capture distance setting fits the actual equipment and engineering scenario. Finally, the interception is carried out by combining the capture distance, capture envelope conditions, and distance range. The physical range is defined by hardware characteristics, which not only accurately controls the timing of net capture and improves the capture success rate, but also makes the UAV interception process more operable.

[0025] Please see Figure 6 The sixth embodiment of a method for capturing a target drone according to the present invention specifically includes: 601. Determine if the capture distance is within the capture range; 602. When the acquisition distance is within the acquisition distance range, obtain the relative distance and relative heading angle between the target UAV and the tracking UAV; In this embodiment, the capture distance range is an insurmountable physical interval (e.g., 5m ± 1m) defined by the hardware characteristics of the net-catching device (maximum effective range, minimum safe trigger distance). This range simultaneously considers the collision safety of the tracking drone and the target drone, as well as the effective coverage capability of the net. The relative distance is the straight-line distance between the target drone and the tracking drone in three-dimensional space, which is the basic spatial constraint for net-catching and determines the basic trigger distance range for net-catching. The relative heading angle is the flight heading angle between the target drone and the tracking drone (including the pitch angle and azimuth angle in three-dimensional space), which primarily reflects the orientation of the target in the field of view covered by the net-catching device of the tracking drone, and determines whether the net can cover the target from a spatial perspective after it is deployed. 603. Analyze the relative distance and relative heading angle based on the capture envelope conditions and the preset capture range to obtain the envelope analysis results; In this embodiment, by combining the capture envelope conditions and the capture range to analyze the relative distance and heading angle, the physical capture boundary is defined by the inherent constraints of the hardware. This ensures that the relative geometric parameters match the coverage, speed and other capabilities of the net capture device at the hardware level, avoiding invalid triggering. It also avoids risks such as restricted areas and unsafe fall zones at the scene level, ensuring interception compliance and improving the capture success rate. This approach balances the effectiveness, safety and scene adaptability of net capture execution. 604. If the envelope analysis result indicates that the relative geometric area is within the capture range, then generate a capture command based on the capture distance; 605. Capture the target drone according to the net capture command; In this embodiment, the high-speed tracking drone adopts vertical take-off and landing for rapid response; once a certain distance is reached (the relative geometric area is within the capture range), the image guidance mode is switched. In this embodiment, the physical range of capture distance is first defined based on hardware characteristics to avoid the risk of collision between two drones, ensure effective net coverage, and filter out invalid interception scenarios. Then, the core parameters of relative distance and relative heading angle are obtained to accurately control the basic constraints of net capture space and the matching of the device's field of view. At the same time, the inherent capture range of the hardware and the scene-based strategy capture envelope conditions are verified to ensure that the parameters are compatible with the hardware capabilities of the net capture device, avoid triggering invalid capture, and ensure interception compliance. Finally, a net capture command is generated only after it is determined that capture is possible. Combined with the rapid response of the drone's vertical take-off and landing and the accurate tracking of the image guidance mode, the determination and efficient execution of net capture triggering are realized, improving the capture success rate. The effectiveness of net capture and interception, operational safety, scene adaptability, and execution accuracy are all taken into account in an all-round way.

[0026] Please see Figure 7 The seventh embodiment of a method for capturing a target drone according to the present invention specifically includes: 701. Calculate the relative distance, relative speed, relative heading angle, and net deployment time to obtain the relative geometric region; In this embodiment, the relative approach displacement during the net deployment time is first calculated. , It represents the approximate displacement of the target UAV relative to the tracking UAV during the net deployment time, in meters (m), and determines the center position of the relative geometric area; This indicates the relative speed between the target drone and the tracking drone (specifically, the "approach speed"), which is the speed at which the two drones approach each other, and is measured in meters per second (m / s). The net deployment time represents the physical time from the triggering of the net capture command to the net fully deploying into an effective capture surface (determined by hardware characteristics; for example, the deployment time of a pneumatic net is typically 0.3~0.5 seconds), measured in seconds (s). A Cartesian coordinate system is established with the tracking drone as the origin O(0,0), the front of the net capture device as the positive x-axis, and the horizontal plane perpendicular to the x-axis as the y-axis. , ( , The precise predicted position of the target UAV on the horizontal plane at the moment the net is fully deployed is also the center coordinate of the relative geometric region; in the formula... R represents the predicted relative distance between the two aircraft upon completion of deployment. The relative heading angle is calculated; finally, the relative geometric region is calculated: The circular equation represents the boundary of a relative geometric region within the horizontal plane, where the center of the circle is... (Target's predicted position upon completion of deployment); radius is (Effective capture radius and comprehensive measurement error tolerance of the net body to avoid target loss of coverage due to measurement / jitter); Inequality expression: Only when the predicted position of the target UAV is... Only when the object falls within this circular area can it be captured on a horizontal plane. The preset effective capture radius of the net. To account for the overall measurement error tolerance (including sensor measurement errors (such as radar ranging errors), UAV attitude jitter errors, airflow interference errors, etc., it is usually taken as 0.2~0.5m in engineering, but can also be dynamically corrected according to the environment); 702. Analyze the relative geometric region based on the capture envelope conditions and the capture range to obtain the envelope analysis results; In this embodiment, the capture range is an inherent constraint of the net capture device's hardware, including the maximum coverage distance, angle range, and maximum relative velocity adaptation value of the net (e.g., the net can only capture targets with a relative velocity ≤20m / s), which are insurmountable physical boundaries. The capture envelope conditions are policy constraints of the interception scenario, including geofencing (e.g., avoiding restricted areas), safe drop zones (e.g., the interception point must be in an open area), and the attitude stability of the tracking drone (e.g., the heading angle ≤45°), which are dynamically adjusted policy boundaries. It is determined whether the distance and angle of the relative geometric area are within the coverage range of the net hardware; whether the relative velocity of the target is within the carrying capacity range of the net; and whether the interception position and timing meet the scenario constraints such as geofencing and safe drop zones. Only when all three layers of verification pass will the "capable" envelope analysis result be output. In this embodiment, the precise calculation of the relative geometric region, by coupling relative distance, relative motion speed, relative heading angle, and net deployment time, achieves a technical upgrade from static distance judgment to dynamic spatiotemporal prediction, accurately locking the target position at the moment of net deployment; the circular judgment boundary incorporating comprehensive error tolerance effectively copes with interference such as sensor measurement errors and attitude jitter, improving the robustness of the capture range; the three-layer verification of the capture envelope and the capture range ensures that the interception meets the hardware limits of the net and satisfies the constraints of scenarios such as geofencing and safe fall zones, avoiding invalid triggering and illegal interception; the overall solution improves the net capture hit rate and stability, adapts to complex and dynamic interception scenarios, and provides quantitative basis for full-process traceability.

[0027] The above describes a method for capturing a target drone according to an embodiment of the present invention. The following describes a system for capturing a target drone according to an embodiment of the present invention. Please refer to [link / reference]. Figure 8 One embodiment of the target acquisition system for unmanned aerial vehicles (UAVs) according to the present invention includes: A target drone capture system performs a target drone capture method as described above. The target drone capture system includes a control device 1, a target drone 2 electrically connected to the control device 1, and a tracking drone 3.

[0028] Figure 9 This is a schematic diagram of a target drone capture device 900 provided in an embodiment of the present invention. This target drone capture device 900 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 910 (e.g., one or more processors) and a memory 920, and one or more storage media 930 (e.g., one or more mass storage devices) for storing application programs 933 or data 932. The memory 920 and storage media 930 can be temporary or persistent storage. The program stored in the storage media 930 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the target drone capture device 900. Furthermore, the processor 910 may be configured to communicate with the storage media 930 and execute the series of instruction operations in the storage media 930 on the target drone capture device 900 to implement the steps of the target drone capture method provided in the above-described method embodiments.

[0029] A target drone capture device 900 may further include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input / output interfaces 960, and / or one or more operating systems 931, such as Windows Server, MacOSX, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 9 The illustrated structure of a target-capturing drone capture device does not constitute a limitation on a target-capturing drone capture device, and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0030] A computer-readable storage medium storing instructions that, when executed by a processor, implement the steps of a target acquisition method for unmanned aerial vehicles as described above.

[0031] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0032] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0033] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for capturing a target drone, characterized in that, include: Acquire image sequence data and multi-source observation data; Based on the preset maneuvering tracking filter model, the image sequence data and multi-source observation data are predicted to obtain the target state estimate and future predicted trajectory. Threat assessment is performed on the target state estimate and future predicted trajectory to obtain the threat assessment results; The threat assessment results are analyzed based on the preset handling conditions to obtain the threat analysis results; The target drone is captured based on the threat analysis results, the preset capture envelope conditions, and the preset capture distance range; After capturing the target drone, collect detection and shooting data.

2. The method for capturing a target UAV as described in claim 1, characterized in that, The step of predicting image sequence data and multi-source observation data based on a preset maneuvering tracking filter model to obtain target state estimation and future predicted trajectory includes: The multi-source observation data is registered to obtain registered observation data; The registered observation data are analyzed according to the preset factory calibration error model to obtain the observation confidence sequence and the measurement covariance sequence; The observation confidence sequence and the measurement covariance sequence are weighted and fused to obtain the fused observations and fused covariance. The image sequence data is detected based on a pre-defined deep learning model to obtain adaptive parameters; Based on the maneuvering tracking filter model, adaptive parameters, fusion covariance, and fusion observations are predicted to obtain target state estimates and future predicted trajectories.

3. The method for capturing a target UAV as described in claim 2, characterized in that, The step of predicting adaptive parameters, fusion covariance, and fusion observations based on the maneuvering tracking filter model to obtain target state estimation and future predicted trajectory includes: The adaptive parameters are analyzed based on the preset uniform velocity model and the preset conversion method to obtain the switching weight, process noise parameter, scale factor and increment term; The preset measurement noise matrix is ​​adjusted according to the scaling factor and increment term to obtain the optimized noise matrix; Based on the maneuvering tracking filter model, the noise matrix, switching weights, process noise parameters, fusion covariance, and fusion observations are predicted to obtain the target state estimate and future predicted trajectory.

4. The method for capturing a target UAV as described in claim 3, characterized in that, The step involves analyzing the adaptive parameters based on a preset uniform velocity model and a preset conversion method to obtain switching weights, process noise parameters, scaling factors, and increment terms, including: The maneuver probability and measurement noise estimate are obtained from the adaptive parameters; The maneuver probability is analyzed based on the uniform velocity model to obtain the switching weights and process noise parameters; The measurement noise estimate is transformed using a transformation method to obtain the scaling factor and increment term.

5. The method for capturing a target UAV as described in claim 1, characterized in that, The target drone capture system also includes drone tracking. The process of capturing the target drone based on threat analysis results, preset capture envelope conditions, and preset capture distance range includes: If the threat analysis result indicates that the threat assessment result meets the handling conditions, then an interception command is generated based on the threat analysis result; The relative speeds between the target drone and the tracking drone are obtained based on the interception command. The capture distance is calculated based on the preset basic engineering calculation formula, the preset effective coverage radius of the capture net, the preset net deployment time, and the relative motion speed. The target drone is captured based on the capture distance, capture envelope conditions, and capture distance range.

6. The method for capturing a target UAV as described in claim 5, characterized in that, The method of capturing the target UAV based on the capture distance, capture envelope conditions, and capture distance range includes: Determine if the capture distance is within the capture range; When the capture distance is within the capture distance range, the relative distance and relative heading angle between the target drone and the tracking drone are obtained; The relative distance and relative heading angle are analyzed based on the capture envelope conditions and the preset capture range to obtain the envelope analysis results; If the envelope analysis result indicates that the relative geometric region is within the capture range, then a capture command is generated based on the capture distance; The target drone was captured according to the net capture command.

7. The method for capturing a target drone as described in claim 6, characterized in that, The analysis of relative distance and relative heading angle based on the capture envelope conditions and the preset capture range to obtain the envelope analysis results includes: The relative distance, relative speed, relative heading angle, and net deployment time are calculated to obtain the relative geometric region; The relative geometric region is analyzed based on the capture envelope conditions and the capture range to obtain the envelope analysis results.

8. A device for capturing a target drone, characterized in that, include: The first data acquisition module is used to acquire image sequence data and multi-source observation data; The prediction module is used to predict image sequence data and multi-source observation data based on a preset maneuvering tracking filter model in order to obtain target state estimation and future predicted trajectory. The threat assessment module is used to perform threat assessment on the target state estimate and future predicted trajectory to obtain the threat assessment results; The analysis module is used to analyze the threat assessment results based on preset handling conditions to obtain the threat analysis results; The capture module is used to capture target drones based on threat analysis results, preset capture envelope conditions, and preset capture distance range. The second data acquisition module is used to collect detection and shooting data after capturing the target drone.

9. A device for capturing a target drone, characterized in that, The target acquisition device for drones includes: a memory and at least one processor, wherein the memory stores instructions; At least one of the processors invokes the instructions in the memory to cause the target drone capture device to perform the steps of the target drone capture method as claimed in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the various steps of the target acquisition method for a UAV as described in any one of claims 1-7.

Images