Target tracking method of low-altitude strong-maneuvering unmanned aerial vehicle

Abstract

The embodiment of the invention discloses a target tracking method for a low-altitude strong-maneuvering unmanned aerial vehicle, and relates to the technical field of target tracking, and the method comprises the steps: obtaining a measurement point set, and a state estimation value and a state covariance matrix of a previous moment, so as to obtain a state estimation value of the previous moment through a plurality of preset strong-maneuvering models; obtaining a current prediction index of the current moment corresponding to each strong maneuvering model; determining an elliptical correlation gate corresponding to each current prediction index, so as to determine a candidate measurement point set falling into the elliptical correlation gate according to the measurement point set; performing intra-model association and inter-model association through a preset hierarchical association strategy, and screening a global optimal measurement corresponding to the current moment and a corresponding matching strong maneuvering model; and according to the global optimal measurement corresponding to the current moment, performing Kalman filtering updating on the current prediction index of the matched strong maneuvering model to obtain an updated prediction index of the target at the current moment so as to realize target tracking.

Description

TECHNICAL FIELD

[0001] The present specification relates to the technical field of target tracking, and in particular to a target tracking method for low-altitude strong maneuvering unmanned aerial vehicles. BACKGROUND

[0002] With the rapid development and wide application of unmanned aerial vehicles, reliable tracking of low-altitude flying targets has become a focus in the field of radar surveillance. The low-altitude environment is complex, not only existing ground strong clutter caused by buildings, terrain, etc., but also often accompanied by complex electromagnetic interference, resulting in a large number of false alarms and interference points in radar measurement. At the same time, unmanned aerial vehicles have strong maneuvering capabilities such as hovering, rapid turning, and U-turn, and their movement patterns show rapid switching and strong nonlinearity. This specific scenario of coupling of low-altitude environment and target strong maneuvering poses a severe challenge to traditional tracking methods.

[0003] In the prior art, interactive multiple models (IMM) or variable structure multiple models (VSMM) and their improved algorithms are the mainstream solutions to deal with target maneuvering. These methods run multiple motion models (such as uniform speed, uniform acceleration, coordinated turning model) in parallel and fuse the estimates based on model probabilities, which to some extent improves the tracking adaptability to maneuvering targets. However, when dealing with low-altitude strong maneuvering unmanned aerial vehicles, the preset model set is mostly based on idealized kinematic assumptions, which is difficult to accurately describe the strong nonlinear maneuvering patterns such as sudden stop and sharp turning specific to unmanned aerial vehicles, resulting in model mismatch and increased deviation between predicted trajectory and real motion. More critically, existing methods mostly focus on model set optimization and state estimation algorithm improvement, and fail to fully consider the reliability of measurement association in the low-altitude dense clutter environment. Traditional association methods (such as global nearest neighbor method) are prone to misassociate high-intensity clutter points as target measurements or produce association ambiguity among multiple candidate points in a strong clutter background.

[0004] Therefore, in the process of tracking low-altitude strong maneuvering unmanned aerial vehicles, due to the difficulty of the general motion model set used by existing multiple model methods to match the actual strong maneuvering patterns of the target, and the insufficient reliability of its association strategy in the low-altitude dense clutter environment, there are problems of large model prediction deviation and high measurement association error rate, resulting in discontinuous target tracking trajectory, and even target loss under complex interference. SUMMARY

[0005] One or more embodiments of the present specification provide a target tracking method of a low-altitude strong maneuvering unmanned aerial vehicle, which is used to solve the technical problem that in the process of tracking the low-altitude strong maneuvering unmanned aerial vehicle, the general motion model set used by the existing multi-model method is difficult to match the actual strong maneuvering mode of the target, and the association strategy thereof is insufficient in reliability in a low-altitude dense clutter environment, and there are problems of large model prediction deviation and high measurement association error rate, resulting in discontinuous target tracking trajectory, and even target loss under complex interference.

[0006] One or more embodiments of the present specification adopt the following technical solutions: One or more embodiments of the present specification provide a target tracking method of a low-altitude strong maneuvering unmanned aerial vehicle, the method comprising: acquiring a measurement point set output by a radar in a current detection period, and a state estimation value and a state covariance matrix of a target at a previous time, to perform one-step prediction on the previous time through a plurality of preset strong maneuvering models, to obtain a current prediction index corresponding to the current time of each strong maneuvering model, wherein the current prediction index comprises a predicted state and a predicted covariance matrix; determining an elliptical association gate corresponding to each current prediction index, to determine a candidate measurement point set falling into the elliptical association gate according to the measurement point set; performing intra-model association and inter-model association in the candidate measurement point set through a preset hierarchical association strategy, to screen a global optimal measurement corresponding to the current time and a matching strong maneuvering model; performing Kalman filter updating on the current prediction index of the matching strong maneuvering model according to the global optimal measurement corresponding to the current time, to obtain an updated prediction index of the target at the current time, to realize target tracking, wherein the updated prediction index comprises an updated state estimation value and an updated state covariance matrix.

[0007] The above at least one technical scheme adopted by the embodiments of the present specification can achieve the following beneficial effects: through the technical scheme of the embodiments of the present specification, in the prediction stage, a plurality of preset strong maneuvering models are introduced in parallel, through the model of preset low-altitude unmanned aerial vehicle hovering, turning around, sharp turning and other maneuvering modes, state deduction is carried out along multiple high-probability maneuvering directions at each moment, which greatly reduces the systematic risk that the prediction point deviates from the real target position due to model mismatch. No matter what maneuvering the target actually performs, the prediction direction of one or several models will generally coincide with it, thereby reserving the possibility for subsequent data association and avoiding the fatal weakness of traditional single model or limited model set that completely fails to predict when the target suddenly maneuvers, directly leading to tracking loss. Secondly, the strategy of establishing an elliptical association gate for each model is implemented, realizing personalized management of prediction uncertainty and refinement of spatial search. The uncertainty of different maneuvering models has anisotropic characteristics, and according to the dynamic characteristics of each model, the most likely search area that can capture the real measurement in a statistical sense is determined, which can form a complementary and efficient distributed search network in the clutter environment, significantly improving the capture rate of the first discovery of real target measurement in dense false alarms, while avoiding the introduction of too much clutter due to a too large gate, or missing the target due to a too small gate. The preset hierarchical association strategy converts the complex global optimal search problem into local optimal judgment and high-level arbitration through two-level progression of model internal screening and inter-model screening, so that the association decision has environment perception and target cognition ability. In an environment full of uncertainty, a high-confidence global optimal measurement and its corresponding matching model are output. Based on the Kalman filter update of a single matching model and a global optimal measurement, the update path is clear, avoiding potential mutual interference and information dilution in multi-model soft fusion, ensuring the theoretical optimality of state estimation under a given decision. Since the measurement driving the update is the optimal measurement strictly selected, the innovation item is highly reliable, so that the state correction direction is accurate and the amplitude is reasonable, which can effectively pull the estimated trajectory back to the real path. Through multi-model active coverage to solve the unpredictability of maneuvering, through independent gate fine search to solve the spatial interference of clutter, through hierarchical information fusion decision to solve the association ambiguity, and finally through optimal estimation based on clear decision to realize accurate state correction and tracking closed loop, the tracking system can maintain strong robustness of continuous track and stable state estimation under the dual pressure of low-altitude dense clutter and target strong maneuvering. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to make the technical solutions in the specification clearer, the following will briefly introduce the drawings needed to be used in the description of the embodiments or prior art. Obviously, the drawings in the following description are only some embodiments described in the specification, and other drawings can be obtained by those skilled in the art without creative labor. In the drawings: Figure 1 A flowchart of a target tracking method of a low-altitude strong maneuvering unmanned aerial vehicle provided by the embodiments of the present specification is shown in the figure. Figure 2 A tracking result display example diagram of a tracking method using a traditional single model correlation provided by the embodiments of the present specification is shown in the figure. Figure 3 A tracking result display example diagram of a tracking method using a multi-model and hierarchical correlation strategy provided by the embodiments of the present specification is shown in the figure. Figure 4 A structural diagram of a target tracking device of a low-altitude strong maneuvering unmanned aerial vehicle provided by the embodiments of the present specification is shown in the figure. DETAILED DESCRIPTION

[0009] In order to make those skilled in the art better understand the technical solutions in the specification, the technical solutions in the embodiments of the present specification will be described clearly and completely in the following with reference to the drawings in the embodiments of the present specification. Obviously, the described embodiments are only some embodiments of the specification, not all embodiments. Based on the embodiments of the present specification, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present specification.

[0010] The embodiments of the present specification provide a target tracking method of a low-altitude strong maneuvering unmanned aerial vehicle. It should be noted that the execution subject in the embodiments of the present specification can be a server or any device with data processing capability. Figure 1 A flowchart of a target tracking method of a low-altitude strong maneuvering unmanned aerial vehicle provided by the embodiments of the present specification is shown in the figure. Figure 1 As shown in the figure, the method mainly includes the following steps: In step S101, a set of measurement points output by the radar in the current detection period and the state estimation value and state covariance matrix of the target at the last time are obtained, so as to perform one-step prediction on the last time through a plurality of preset strong maneuvering models, to obtain a current prediction index corresponding to each strong maneuvering model at the current time.

[0011] The current prediction index includes a predicted state and a predicted covariance matrix. The common model set, such as constant velocity (CV), constant acceleration (CA), coordinated turn (CT) and the like, is generally used in the conventional multi-model tracking method, is derived from the general abstraction of the motion of a traditional aircraft, and has relatively mild and continuous assumed motion modes (such as constant acceleration and constant turn rate), which cannot effectively cover the discrete, sudden and high dynamic maneuvering characteristics of the unmanned aerial vehicle in a low-altitude complex environment. For example, the CA model can describe acceleration but cannot describe sudden stop, and the CT model describes smooth arc turn instead of right-angle sharp turn. When the target performs such strong maneuvering, the predicted trajectory based on the common model will deviate from the real path of the target systematically and greatly, and the model mismatch directly leads to the fact that the subsequent association gate cannot cover the real measurement point, causing association failure and tracking interruption.

[0012] In an embodiment of the present specification, the strong maneuvering model includes a constant velocity model, a hovering model, a U-turn model, a left tangent maneuvering model and a right tangent maneuvering model; wherein the state transition matrix of the constant velocity model is used to describe the motion of constant target velocity; the transition coefficient corresponding to the velocity component in the state transition matrix of the hovering model is less than 1, which is used to describe the motion of the target whose velocity decays to near static; the transition coefficient corresponding to the position component in the state transition matrix of the U-turn model is opposite in sign to the transition coefficient corresponding to the velocity component, which is used to describe the motion of the target whose motion direction is reversed; the state transition matrix of the left tangent maneuvering model is used to describe the motion of the target whose velocity direction deflects a preset angle counterclockwise in a two-dimensional plane; and the state transition matrix of the right tangent maneuvering model is used to describe the motion of the target whose velocity direction deflects a preset angle clockwise in a two-dimensional plane.

[0013] In an embodiment of the present specification, after the radar system completes a complete scanning cycle, the signal processing unit outputs a measurement point set containing all the detected point tracks at the current time, wherein each measurement point contains its position information, echo amplitude information and time stamp in the radar coordinate system. At the same time, the tracking processing unit reads the state estimation value and the state covariance matrix of the target after the Kalman filter update from its own cache at the last tracking cycle (k-1 time). The state estimation value is a vector containing the position and velocity of the target, and the state covariance matrix is used to describe the uncertainty of the state estimation.

[0014] Subsequently, the preset five strong maneuvering models start working in parallel, and the five strong maneuvering models of uniform speed CV, hovering, turning around, left tangential, and right tangential respectively describe the future behavior of the target with different kinematic assumptions. Each model has a predefined state transition matrix, and the mathematical form of the state transition matrix directly encodes the motion assumption of the model. The state transition matrix of the uniform speed model assumes that the speed is constant; the state transition matrix of the hovering model introduces a decay factor in the speed dimension, so that the predicted speed tends to zero to simulate stagnation; the state transition matrix of the turning around model makes the predicted position continue in the old speed direction while reversing the speed direction, thereby describing the sharp turning maneuver; and the state transition matrices of the left and right tangential models respectively make the predicted speed direction produce a fixed angle of counterclockwise or clockwise deflection to describe the rapid right-angle turning maneuver.

[0015] Based on the five different state transition matrices and the shared state estimate and covariance matrix at the last time, five independent Kalman prediction steps are performed in parallel. Specifically, for each model, the prediction operation is to multiply the state estimate vector at the last time by the state transition matrix corresponding to the model, thereby obtaining the predicted state at the current time (k time) based on the assumption of the model. At the same time, the state covariance matrix at the last time is also transformed by the state transition matrix, and the process noise covariance matrix representing the uncertainty of the model itself is superimposed, and finally the predicted covariance matrix at the current time is obtained. At this point, five different current predicted indicators at the current time are generated for the same target, i.e., five groups of predicted states and their uncertainty measures, i.e., predicted covariance matrices, derived from different motion assumptions, which provide a complete prediction candidate set for subsequent measurement association in multiple possible motion trajectory directions.

[0016] Specifically, the algorithm input is the motion state estimate value of the target at k-1 time and the covariance , and the measurement point set detected by the radar working in search mode at the k screen . Among them, represents the th measurement, and there are measurements. First, based on the defined five types of strong maneuvering models, the Kalman prediction step is used to predict and for one step,

[0017] where , and are the predicted state, predicted covariance, and state transition matrix of the th maneuvering model. In addition, the state transition matrices of the five types of maneuvering models are defined as follows:

[0018] wherein, F1, F2, F3, F4 and F5 are state transition matrices of uniform CV, hover, U-turn, left tangential and right tangential five strong maneuver models respectively.

[0019] Based on the above five maneuver models, the state prediction is carried out, and the specific method is as follows: The uniform CV model is defined as the speed direction and size of the target do not change. Based on the uniform CV model, the target motion state and the covariance are one-step state prediction according to the Kalman prediction formula. are the positions and velocities of the target in the x-axis and y-axis respectively, is a 4*4-dimensional covariance matrix describing uncertainty. The update formula is as follows:

[0020] wherein, and are the predicted state and predicted covariance based on the uniform CV model, is the state transition matrix of the CV model. represents the data rate of the measurement.

[0021] The hover model is defined as the target suddenly stops, i.e. the size of the target velocity in each dimension becomes zero. Based on the hover model, the target motion state and the covariance are one-step state prediction according to the Kalman prediction formula. are the positions and velocities of the target in the x-axis and y-axis respectively, is a 4*4-dimensional covariance matrix describing uncertainty. The update formula is as follows:

[0022] wherein, and are the predicted state and predicted covariance based on the hover model, is the state transition matrix of the hover model.

[0023] The U-turn model is defined as the target suddenly reverses the motion within a scan period, which is specifically manifested as the size of the target velocity does not change, but the direction of the velocity is reversed. Based on the U-turn model, the target motion state and the covariance are one-step state prediction according to the Kalman prediction formula. are the positions and velocities of the target in the x-axis and y-axis respectively, is a 4*4 dimensional covariance matrix describing the uncertainty. The update formula is as follows:

[0024] where, and are the predicted state and predicted covariance based on the U-turn model, is the state transition matrix of the U-turn model.

[0025] Define the left-side tangent model as the target has made a sharp turn to the left of the moving direction, which is specifically manifested as the speed of the target remains unchanged, but the speed direction has rotated counterclockwise by 90 degrees in the horizontal plane. Based on the left-side tangent model, make one-step state prediction on the target motion state and its covariance according to the Kalman prediction formula. are the position and speed of the target in the x-axis and y-axis respectively, is a 4*4 dimensional covariance matrix describing the uncertainty. The update formula is as follows:

[0026] where, and are the predicted state and predicted covariance based on the U-turn model, is the state transition matrix of the U-turn model.

[0027] Define the right-side tangent model as the target has made a sharp turn to the right of the moving direction, which is specifically manifested as the speed of the target remains unchanged, but the speed direction has rotated counterclockwise by 90 degrees in the horizontal plane. Based on the right-side tangent model, make one-step state prediction on the target motion state and its covariance according to the Kalman prediction formula. are the position and speed of the target in the x-axis and y-axis respectively, is a 4*4 dimensional covariance matrix describing the uncertainty. The update formula is as follows:

[0028] where, and are the predicted state and predicted covariance based on the U-turn model, is the state transition matrix of the U-turn model.

[0029] Compared with the conventional interactive multi-model (IMM) method which usually adopts uniform speed, uniform acceleration and other basic models, the step closely matches the unique and severe nonlinear motion mode of the low-altitude unmanned aerial vehicle by presetting hovering, turning around, tangential turning and other strong maneuvering models at the prediction level. When facing the target's sudden stop and sharp turning behaviors, the prediction trajectory of the basic model of the conventional method will deviate significantly from the real maneuvering of the target, resulting in that the subsequent correlation gate cannot effectively cover the real measurement. However, the special strong maneuvering model running in parallel in the technical solution can cover a variety of extreme maneuvering situations that the target may occur when deducing the state at the last moment, so that at least one prediction direction of the model can approach the real motion trend of the target, which is equivalent to expanding the search range to multiple maneuvering directions with high probability before data association, rather than relying on a single prediction point that may have deviated seriously, laying a solid foundation for robustly capturing real target measurements in a low-altitude dense clutter environment. When the target suddenly maneuvers, the corresponding real measurement may be far from the predicted position based on the assumption of uniform speed, but it may fall within the predicted gate based on the turning or tangential model, thereby avoiding the risk of immediate loss caused by model mismatch. Through the scene-based customization of the front-end prediction model set, the pre-matching ability of the system to the motion mode of the low-altitude strong maneuvering target is significantly improved, and the pressure of coping with maneuvering is moved from the rear-end association and filtering algorithm to the prediction stage, thereby reducing a series of subsequent association failures and tracking divergence problems caused by inaccurate prediction from the source, and enhancing the foresight and fault tolerance of the entire tracking system in complex environments.

[0030] In step S102, an elliptical association gate corresponding to each current prediction index is determined to determine a candidate measurement point set falling into the elliptical association gate according to the measurement point set.

[0031] The elliptical association gate corresponding to each current prediction index is determined to determine a candidate measurement point set falling into the elliptical association gate according to the measurement point set, specifically including: calculating the innovation covariance matrix corresponding to each strong maneuvering model according to the prediction covariance matrix in the current prediction index of each strong maneuvering model and the preset measurement noise covariance matrix; determining the boundary condition of the elliptical association gate of each strong maneuvering model based on the innovation covariance matrix and the preset threshold parameter; calculating the normalized distance of each measurement point with respect to the predicted state according to the position information of each measurement point in the measurement point set, the preset observation matrix and the predicted state of the strong maneuvering model; judging whether the normalized distance of each measurement point satisfies the boundary condition of the elliptical association gate, and if so, the measurement point is classified into the candidate measurement point set of the strong maneuvering model.

[0032] Conventional methods usually adopt a uniform or optimal model prediction based gate shape, which is single and cannot adapt to the uncertainty difference caused by various possible target maneuvers. For example, when the target flies at a constant velocity, the prediction uncertainty mainly expands along the velocity direction; but when the target makes a sharp turn, the uncertainty increases dramatically and changes direction. The single gate is difficult to dynamically adapt to such changes, resulting in either too large a gate that introduces too many false alarms or too small a gate that misses real maneuvering targets.

[0033] In an embodiment of the present disclosure, an elliptical association gate is established for the predicted state of each strong maneuver model, and the elliptical association gate is used to preliminarily screen out candidate point tracks that may match the prediction of each model from all the measurement points provided by the radar.

[0034] The prediction covariance matrix in the five current prediction indicators calculated in the previous step is used, which quantifies the uncertainty of the real state of the target relative to the predicted state after one-step prediction based on the model. This uncertainty is caused by process noise (i.e., the incompleteness of the model itself in describing the random maneuver of the target) and the error transmission of the state estimation at the previous time. However, in order to determine whether a radar measurement is from the target corresponding to the predicted state, not only the prediction uncertainty needs to be considered, but also the observation error of the sensor itself must be taken into account. Therefore, first, for each strong maneuver model, the innovation covariance matrix corresponding to the model is calculated according to the prediction covariance matrix and the preset measurement noise covariance matrix. The innovation covariance matrix essentially describes the statistical characteristics of the difference between the predicted measurement and the real measurement, and its calculation formula directly adds the uncertainty of the predicted state (projected into the measurement space by the observation matrix) and the inherent measurement noise of the sensor. The innovation covariance matrix reflects the total uncertainty in the entire chain from model prediction to sensor observation.

[0035] Then, based on the innovation covariance matrix and a preset threshold parameter, the precise mathematical boundary conditions of the elliptical association gate of each strong maneuver model are determined. The threshold parameter is usually selected according to the chi-square distribution and is used to control the size of the gate, i.e., the confidence level of accepting a measurement as a candidate point. The boundary condition is specifically represented by a quadratic inequality: if a measurement point falls into the elliptical gate, its distance relative to the predicted position must be less than or equal to the threshold value. The distance here is not the ordinary Euclidean distance, but a statistical distance that takes into account the uncertainty. Then, the original measurement point set output by the radar in the current period is scanned and judged one by one. For each measurement point in the set, its position information (such as distance, azimuth, or x, y coordinates converted to the rectangular coordinate system) is extracted. A preset observation matrix is used, which defines how to extract directly observable quantities from the target state vector (including position, velocity).

[0036] For a given strong maneuvering model, the specific process for calculating the normalized distance of a measurement point relative to its predicted state is as follows: First, calculate the difference vector between the measurement point's position and the predicted measurement position obtained by projecting the model's predicted state through the observation matrix; this difference is called the innovation. Then, using the inverse of the previously calculated innovation covariance matrix as the weight matrix, calculate the weighted inner product of the innovation vectors, i.e., multiply the innovation vector on the left by its transpose and then on the right by the inverse of the innovation covariance matrix. This scalar result is the required normalized distance (or Mahalanobis distance). By whitening the difference vector using the inverse of the innovation covariance matrix, elliptical regions with different uncertainties in different directions are equivalent to an isotropic sphere when calculating the distance. Therefore, even if the prediction uncertainty is large in a certain direction (e.g., the velocity direction), causing the gate to be elongated in that direction, as long as the measurement point falls within this elongated ellipse, its calculated normalized distance will still be small. Finally, the normalized distance of each measurement point relative to each model is calculated and compared with the boundary conditions determined by the model based on threshold parameters. These boundary conditions are the threshold values. If the normalized distance of a measurement point relative to a model is less than or equal to the threshold value, the measurement point is determined to satisfy the boundary conditions of the elliptic correlation gate of the model. This means that, statistically speaking, the measurement point is highly likely to be an observation result generated by the target state predicted by the model, and therefore, the measurement point is added to the candidate measurement point set of this specific strong maneuvering model. After completing the above traversal calculation and judgment for all measurement points and all models, each strong maneuvering model obtains its own exclusive candidate measurement point set, which contains all radar traces that are statistically compatible with its predictions.

[0037] The following is a formulaic expression of the above process, based on the predicted state of each of the above strong maneuvering models. and , , used to represent the i-th strong maneuvering model, establishes an elliptical correlation gate:

[0038] in The sensor at time k obtained the first time Individual measurement, For the observation matrix, For the preset threshold parameters, measurements that satisfy the conditions of the above formula are considered to fall into the threshold. Measurement points within the elliptic correlated gate established by the prediction model.

[0039] Compared to the fixed-size gates or gates calculated based on the covariance of a single prediction model commonly used in conventional tracking methods, the technical solution in this specification establishes an independent elliptical associated gate for each strong maneuvering model. Since the dynamic characteristics of each strong maneuvering model (such as hovering, turning, and tangential maneuvering) are different, the uncertainties introduced by one-step prediction are also drastically different. By independently calculating the innovation covariance matrix based on its own prediction covariance for each model, the shape and orientation of the elliptical gate are determined, ensuring that each gate can most accurately reflect the statistical distribution area where the measurement is most likely to occur if the target does indeed maneuver according to this model. The gate for the turning model may be elongated in the reverse velocity direction, while the gate for the tangential model may expand in the turning tangential direction. This precise matching greatly improves the probability that the gate spatially covers the actual target measurement. Secondly, a parallel and complementary search space was constructed. Five different models generated five elliptical gates with different shapes, positions, and orientations. Together, they formed a complementary search network covering multiple high-probability maneuver directions in the space at the current moment. Regardless of which of the five typical maneuvers the target actually performed, or even the transition between them, its true measurement point has a high probability of being captured by one or more correct gates. Compared with the traditional method that relies on a single gate for correlation, the parallel search mechanism significantly enhances the robustness and fault tolerance when the target performs unexpected maneuvers, and effectively reduces the risk of tracking loss caused by instantaneous mismatch of the prediction model.

[0040] Step S103: Through a preset hierarchical association strategy, perform intra-model association and inter-model association within the candidate measurement point set, and filter the global optimal measurement and the corresponding matching strong maneuver model at the current moment.

[0041] Using a pre-defined hierarchical association strategy, intra-model and inter-model associations are performed within the candidate measurement point set to filter the globally optimal measurement and the corresponding matching strong maneuver model at the current moment. Specifically, this includes: within the candidate measurement point set corresponding to each strong maneuver model, calculating the Euclidean distance between each candidate measurement point and the predicted state, and selecting the candidate measurement point with the smallest Euclidean distance as the candidate associated measurement of the strong maneuver model; if multiple candidate measurement points have the same minimum Euclidean distance, then based on the amplitude information of each candidate measurement point and the historical amplitude information of the target, the candidate associated measurement corresponding to the strong maneuver model is filtered from the multiple candidate measurement points; based on the amplitude information of each candidate associated measurement, the globally optimal measurement at the current moment is determined, and the matching strong maneuver model at the current moment is determined based on the strong maneuver model that generates the globally optimal measurement.

[0042] The candidate associated measurement corresponding to the strong maneuvering model is determined from the candidate associated measurements based on the amplitude information of each candidate associated measurement and historical amplitude information of the target, and specifically includes: obtaining current flight height information of the target, and determining whether the current flight height information of the target is lower than a preset height threshold; if the current flight height information is lower than the preset height threshold, historical track data of the target is obtained, the historical track data including amplitude values of at least one historically associated successful measurement point; an arithmetic mean value of the amplitude values of the at least one historically associated successful measurement point is calculated to determine a historical amplitude reference value corresponding to the target, so as to calculate a difference value between the amplitude information of each candidate measurement point and the historical amplitude reference value; and the candidate measurement point with the smallest difference value is determined as the candidate associated measurement of the current strong maneuvering model; if the current flight height information is not lower than the preset height threshold, the amplitude information of the candidate measurement points is compared, and the candidate measurement point with the largest amplitude information is determined as the candidate associated measurement of the current strong maneuvering model.

[0043] The global optimal measurement at the current time is determined according to the amplitude information of each candidate associated measurement, and specifically includes: comparing the amplitude information of each candidate associated measurement, and determining the candidate associated measurement with the largest amplitude information as the global optimal measurement at the current time. After comparing the amplitude information of each candidate associated measurement, the method further includes: if there are at least two candidate associated measurements with the same maximum amplitude information, a specified strong maneuvering model corresponding to each specified candidate associated measurement with the same maximum amplitude information is obtained; a new information distance between the maximum amplitude information of each specified candidate associated measurement and a predicted state of the specified strong maneuvering model is calculated according to the maximum amplitude information and the predicted state; and the specified candidate associated measurement with the smallest new information distance is determined as the global optimal measurement at the current time.

[0044] The global nearest neighbor (GNN) association commonly used in conventional tracking methods or the soft decision based on probability weighting in the interactive multiple model (IMM) method, the conventional GNN method performs global minimum distance search between all measurements and all model predictions (or a comprehensive prediction), and is easily misled by the nearest clutter point under dense clutter and cannot distinguish the difference in association quality under different maneuvering assumptions. Although the conventional IMM method maintains multiple models, the association is essentially performed independently for each model, and the dominant model is implicitly selected through model probability updating, which is sensitive to association errors in a single frame of measurement, and the convergence of model probability needs time, and the reaction is lagged when facing the rapid and sudden maneuvering of a UAV.

[0045] In an embodiment of the present disclosure, in order to accurately select a unique and reliable target measurement at the current time from a plurality of candidate measurements corresponding to each model gate, and determine the most matching maneuver mode, a hierarchical association strategy of intra-model association and inter-model association is used to determine the globally optimal measurement at the current time.

[0046] In the intra-model association stage, for each strong maneuver model running in parallel, a candidate association measurement is selected from the corresponding candidate measurement point set. The specific implementation process is as follows: First, for a certain strong maneuver model (for example, a U-turn model), the candidate measurement point set is read, which contains all the tracks statistically compatible with the prediction. Then, for each candidate measurement point in the candidate measurement point set, the Euclidean distance between the position of the candidate measurement point and the position component contained in the predicted state of the model is calculated. The Euclidean distance is the straight-line distance between two points, which is direct and fast to calculate and reflects the proximity of the measurement point and the predicted point in the geometric space. After the calculation is completed, the Euclidean distances of all candidate measurement points are compared, and the measurement point with the smallest Euclidean distance is preliminarily marked as the potential candidate association measurement of the model. It is assumed here that under the premise that the model prediction is basically correct, the measurement of the real target should be closest to the predicted position.

[0047] The premise of the prediction model-based association is that the predicted maneuver model matches the actual maneuver form of the target, in other words, if the actual motion mode of the target matches a certain maneuver model, then the measurement track of the target should be near the predicted point. Therefore, in the association gate of each model, the nearest neighbor association criterion is used as the primary criterion to select the locally optimal measurement. Specifically, the distance between all measurements falling within the predicted association gate of the th maneuver model and the predicted position is calculated using the following formula

[0048] wherein the symbol represents the two-norm of a vector. The measurement corresponding to the smallest is selected as the measurement associated with the th maneuver model.

[0049] However, the low altitude dense clutter environment introduces uncertainty, and multiple clutter points may be exactly equidistant from the predicted position, causing the nearest neighbor criterion to fail. That is, after filtering by the nearest neighbor criterion, multiple measurements are equidistant from the predicted position, that is, the nearest neighbor criterion cannot determine the locally optimal measurement, and the maximum amplitude criterion is used as a supplement. Determine whether there is only one measurement point with the minimum Euclidean distance; if there are two or more measurement points with the same minimum Euclidean distance, a refined screening sub-process is performed based on amplitude information and historical information. Obtain the current flight height estimate of the target, which is derived from the height component in the last state estimation, or by filtering the height information of consecutive measurements. Compare this height value with the preset height threshold. This threshold physically distinguishes two completely different interference environments, for example, it can be set to 100 meters, below this threshold represents the very low altitude, the radar beam depression angle is very small, and the strong ground stationary clutter (such as buildings) dominates, whose return amplitude may be much higher than that of the UAV target; above this threshold represents the medium-low altitude, the ground clutter influence is weakened, and the main interference may become flying birds and other small scattering cross-section targets.

[0050] If the current height is below the threshold, enter the low altitude anti-strong clutter mode, retrieve the amplitude value sequence of the measurement points from the target's historical flight record archive in the last few successful associations and used for state update. This sequence reflects the statistical characteristics of the radar cross section (RCS) of this particular target. Calculate the arithmetic mean of this amplitude sequence as the historical amplitude reference value representing the typical return strength of the target. Then, calculate the absolute difference between the amplitude information of each candidate measurement point with the same minimum Euclidean distance (i.e. the amplitude value of this detection) and the historical amplitude reference value, and select the candidate point with the amplitude value closest to the historical reference value, i.e. the candidate measurement point with the smallest absolute difference. The principle here is that the amplitude of strong ground clutter (such as buildings) may be abnormally high and unstable, while the amplitude of the real UAV target is relatively stable over time, fluctuating around its mean value. Therefore, in the case where distance cannot be distinguished, selecting the measurement with the most consistent amplitude behavior with the target's historical characteristics can effectively filter out strong stationary clutter false alarms with a large difference in RCS characteristics from the target. Conversely, if the current height is not below the threshold, enter the medium-high altitude anti-weak interference mode. At this height, the RCS of the main false alarm source (such as birds) is usually smaller than that of the UAV, so the real target return amplitude has an advantage.

[0051] For multiple measurements with the same distance from the predicted position, obtain their detection amplitude information, taking two measurements as an example, denoted as and When applying the maximum amplitude screening criterion, the altitude layer of the predicted target location needs to be considered, and different maximum amplitude screening criteria are applied to different altitude layers. This fully considers the limitations of low-altitude detection scenarios. When the target altitude is below 100m, the radar detection beam needs to be configured with an extremely low elevation angle to detect the target. Due to the certain beamwidth of the radar beam, stationary targets on the ground, such as buildings, can be easily detected, resulting in false alarms with large amplitudes. Therefore, when the altitude is below 100m, we use the amplitude sequence of historically associated successful points stored in the track and calculate its average. ,choose and Closest in numerical terms The measurement point is selected as the locally optimal measurement. When the target altitude is greater than 100m, the influence of ground clutter is relatively small, the detection scene is relatively clean, and UAV targets are usually larger than low-flying birds in terms of radar cross-section (RCS), so the optimal measurement point can be directly selected. and The measurement corresponding to the larger of the two values ​​is used as the local optimum measurement to avoid repeatedly calculating the mean of the amplitude sequence of the track-related points, thus reducing the computational burden. Therefore, after determining the values ​​using the nearest neighbor criterion and the maximum amplitude criterion, at most one local optimum measurement can be selected for each maneuver model.

[0052] Subsequently, inter-model correlation is performed, with the goal of determining a unique globally optimal measurement from the candidate correlation measurements of each model, and thereby identifying the most likely maneuver pattern of the target. After intra-model screening, at most one locally optimal measurement has been selected for each maneuver model, avoiding the influence of strong ground clutter. The remaining rank only needs to consider UAV targets and bird targets in low-altitude detection scenarios. Furthermore, since the radar echo intensity of UAV targets is greater than that of low-altitude birds, the globally optimal measurement is selected primarily based on the criterion of maximizing the measurement amplitude among all locally optimal measurements of the models.

[0053] Firstly, the candidate measurements of five strong maneuver models are obtained. The amplitude information of each candidate measurement is read and compared directly, and the measurement point with the largest amplitude is preliminarily determined as the global optimal measurement at the current time. The strong maneuver model that produces the measurement is preliminarily determined as the matching model. The assumption of this selection is that after the model internal screening, especially the highly adaptive screening, the echo strength of the real UAV target should be the most competitive among the remaining candidate measurements. However, an additional special case needs to be considered, that is, due to the possible overlap of the association gates set by different maneuver models, there is a situation where the same measurement is selected by multiple model association gates. In this case, the amplitude maximum screening criterion cannot directly determine which maneuver model conforms to the true motion form of the target. At this time, only the amplitude cannot distinguish which model is more accurate. In the above case, that is, when the amplitude maximum screening criterion cannot be determined, the innovation distance minimum criterion is used as the final decision. When it is detected that there are two or more candidate associated measurements with the same maximum amplitude value, screening is performed by the innovation distance minimum criterion. For each candidate measurement with the same amplitude and its corresponding strong maneuver model, the innovation distance of the measurement point relative to the predicted state of the corresponding model is calculated. The innovation distance here is the aforementioned normalized distance (Mahalanobis distance) used to construct the elliptical gate, which takes into account the geometric deviation of the measurement and the prediction and the uncertainty of the model prediction (innovation covariance). After the calculation is completed, the innovation distances are compared, and the candidate measurement point with the smallest innovation distance is determined as the global optimal measurement, and its corresponding model is the final matching strong maneuver model.

[0054] The innovation distance is the optimal statistical measure to measure the degree of fit between the measurement and the prediction of a specific model. In the case of the same amplitude, the prediction of which model can make the measurement point appear with a higher probability density (i.e., a smaller statistical distance) indicates that the motion hypothesis of which model is more matched to the actual maneuver of the target.

[0055] Firstly, the local optimal decision is made in five different motion hypothesis spaces in parallel and independently, so that even if the prediction factor of a certain model does not match the actual maneuver and deviates seriously, the associated error is limited within the model and does not contaminate the global decision, improving the survival ability of the system when part of the model fails. Secondly, the geometric distance (Euclidean distance), target feature information (radar amplitude and its historical statistics), environmental context (flight height) and prediction consistency (innovation distance) and other heterogeneous information are organically integrated. The adaptive amplitude screening based on flight height is not a fixed logic, but an intelligent strategy that dynamically switches according to the type of environmental threat, so that the association criterion can closely match the actual constraints of the physical world. In the very low altitude, it effectively suppresses the strong stationary clutter which is extremely deceptive, and in the medium and high altitude, it uses the RCS advantage of the unmanned aerial vehicle to the birds for rapid screening. Finally, the model inter-association uses the amplitude information with the most discriminant degree for rapid preliminary judgment, and returns to the statistical distance for final arbitration when there is competition, which ensures the processing efficiency under normal circumstances and ensures the decision optimality in extreme conflict situations.

[0056] In step S104, the current prediction index of the matching strong maneuver model is updated by Kalman filtering according to the global optimal measurement corresponding to the current time, to obtain an updated prediction index of the target at the current time, so as to realize target tracking.

[0057] The updated prediction index includes an updated state estimation value and an updated state covariance matrix.

[0058] The current prediction index of the matching strong maneuver model is updated by Kalman filtering according to the global optimal measurement corresponding to the current time, and specifically includes: calculating a Kalman gain according to the prediction state and the prediction covariance matrix of the matching strong maneuver model; correcting the prediction state of the matching strong maneuver model according to the Kalman gain and the difference between the global optimal measurement and the prediction state of the strong maneuver model, to obtain the updated state estimation value; and updating the updated state covariance matrix according to the Kalman gain and the prediction covariance matrix of the matching strong maneuver model.

[0059] In an embodiment of the present specification, the global optimal measurement and the matching strong maneuver model are received. At the same time, the current prediction index of the matching strong maneuver model calculated in the prediction stage of the last step is read, that is, the prediction state vector and the prediction covariance matrix of the model. Kalman filtering update is based on this prediction value as the prior basis, and the global optimal measurement as the innovation (i.e. new information brought by observation) for data fusion.

[0060] Firstly, the predicted covariance matrix is projected to the measurement space by the observation matrix and its transpose, and then the measurement noise covariance matrix is added to obtain the inverse or correlation form of the innovation covariance matrix. Finally, the predicted covariance matrix and the transpose of the observation matrix are multiplied by the inverse of this intermediate result to obtain the Kalman gain matrix. The calculation principle of this gain matrix is based on the minimum mean square error estimation criterion, and the mathematical result guarantees that the final state correction can effectively reduce the estimation error in a statistical sense. If the model prediction is very uncertain in a certain state component (corresponding to a large predicted covariance), and the sensor observation in that direction is relatively accurate (small measurement noise), then the gain value corresponding to that component will be large, meaning that more measurement information will be adopted during updating; otherwise, more trust will be placed in the predicted value.

[0061] After obtaining the Kalman gain, the innovation, which is the difference vector between the actual position of the globally optimal measurement and the predicted measurement position obtained by projecting the predicted state of the matching model through the observation matrix, is calculated. This difference directly reflects the inconsistency between the sensor observation and the model prediction. Subsequently, the calculated Kalman gain matrix is multiplied by this innovation vector, and the Kalman gain is used as the optimal weight to reasonably distribute and feedback the position difference between the observation and the prediction to the correction of the complete state vector (including position, velocity, etc.), and the result of the multiplication is the state correction. Finally, this correction is added to the original predicted state vector of the matching model to obtain the updated state estimation at the current time. The final state estimation not only contains the dynamic recursion of the model based on historical information, but also incorporates the latest evidence provided by the current observation, theoretically achieving the optimal estimation under given information.

[0062] After completing the correction of the state estimation, the state covariance matrix is updated synchronously using the obtained Kalman gain and the original predicted covariance matrix of the matching model. The standard formula subtracts a term related to the Kalman gain and the innovation covariance from the predicted covariance through specific matrix operations. Due to the injection of observation information, part of the uncertainty in the original state is eliminated. The updated state covariance matrix obtained by calculation will be smaller in value than the predicted covariance matrix. The updated covariance matrix serves as one of the inputs for the next round of one-step prediction of the five strong maneuvering models at the start of the next tracking period, thereby starting the next prediction, association, and update recursive cycle. By then, not only is the most accurate state estimation of the target at the current time output, but also an optimized uncertainty measure for its future motion is provided, realizing the closed loop and continuation of the tracking process.

[0063] Specifically, using the globally optimal measurement associated, the predicted state , is corrected using the Kalman update formula to obtain the state update result of the maneuvering target at the current time and So far, the above steps are repeated to iteratively solve the motion state of the maneuvering target at each time, and the robust tracking of the maneuvering target is realized. Figure 2 A tracking result display example graph of a tracking method using a traditional single model association provided by an embodiment of the present specification, Figure 3 A tracking result display example graph of a tracking method using a multi-model and hierarchical association strategy provided by an embodiment of the present specification, as shown in Figure 2 and Figure 3 As shown in the above figures, the method in the present specification can effectively improve the trajectory continuity of the maneuvering unmanned aerial vehicle target in a complex low-altitude scene.

[0064] After determining the candidate measurement point set falling within the elliptical association gate based on the measurement point set, if the candidate measurement point set of all strong maneuvering models is empty, a specified model is selected from the plurality of strong maneuvering models based on a preset rule, and the prediction state and the prediction covariance matrix of the specified model are directly used as the updated state estimation value and the updated state covariance matrix of the target at the current time, respectively.

[0065] The specified model is selected from the plurality of strong maneuvering models based on a preset rule, and specifically includes: obtaining an association success flag sequence of each strong maneuvering model within a plurality of preset historical tracking periods, wherein each element in the association success flag sequence is used to indicate whether the strong maneuvering model is successfully associated to a candidate association measurement within the corresponding historical tracking period; based on the association success flag sequence of each strong maneuvering model, the cumulative association success number of each strong maneuvering model within the plurality of historical tracking periods is obtained; the cumulative association success number of each strong maneuvering model is used to calculate the association success weight corresponding to the strong maneuvering model, wherein the higher the cumulative association success number, the greater the corresponding association success weight; and the specified model is selected from the plurality of strong maneuvering models according to the association success weight corresponding to each strong maneuvering model.

[0066] In an embodiment of the present specification, fault tolerance and state maintenance in extreme cases are provided to solve the technical problem of how to continue tracking when all the elliptical association gate of the preset strong maneuvering model fails to capture any candidate measurement due to strong maneuvering, temporary occlusion or extremely dense clutter.

[0067] After the judgment of the candidate measurement set of all strong maneuver models is completed, the global detection logic is started to check the candidate set of the five models one by one, i.e. uniform motion, hovering, U-turn, left tangential and right tangential. If the detection finds that all the five sets are empty sets, i.e. no radar measurement point falls into the statistical confidence region of any model in the current period, it is determined that the unassociated measurement event is triggered, and then the model selection rule is performed to avoid the interruption of the update of the tracking filter. First, a sequence of association success flags is recorded for each strong maneuver model (such as the U-turn model). The sequence is a fixed-length first-in-first-out queue, and the length corresponds to a plurality of preset historical tracking periods, for example, the last 30 radar scanning periods. Each element in the sequence corresponds to a specific historical period, for example, time k-1, time k-2, etc., and its value is a binary flag. The value of 1 indicates that in the historical period, the model successfully outputs the candidate associated measurement in the hierarchical association process at that time, i.e. the candidate set is not empty and is preferred in the internal association stage. The value of 0 indicates that the model fails to successfully associate at that time, i.e. the candidate set is empty or is not selected in the internal screening. When a new tracking period is completed, whether the above process is triggered or not, the sequence will be updated, the earliest historical flag will be removed, and the association success or failure result of the current period will be added as a new flag to the sequence. Therefore, the sequence reflects the activity and effectiveness of each model in the recent history in real time.

[0068] For each strong maneuver model, the complete association success flag sequence is read, and the sum of all flag values (1 and 0) in the sequence is calculated. The sum is the cumulative number of association successes of the model in a plurality of historical tracking periods, which intuitively quantifies the effectiveness of the model in matching the target motion pattern in the recent past. For example, a model (such as the left tangential model) that frequently describes target maneuvers (such as a target performing a series of evasive turns) in the past will have a significantly higher cumulative number than a model (such as the hovering model) that has not been matched for a long time. The higher the cumulative number, the stronger the reliability and relevance of the model in the recent tracking context.

[0069] Based on the above cumulative association success number, the association success weight of each model is calculated. Through a preset normalization function, the cumulative association success numbers of all five models are added to obtain a total number, and then the number of each model is divided by the total number to obtain a normalized weight value between 0 and 1, and the sum of the weights of all models is 1. A nonlinear mapping function can also be introduced to square or exponentiate the number before normalization to amplify the weight advantage of high-number models. The higher the cumulative association success number, the greater the association success weight corresponding to it, ensuring that the decision basis for model selection is objective and data-driven historical performance, rather than fixed or random selection.

[0070] Finally, according to the calculated correlation success weight of each model, a preset strategy is used to select a unique designated model from the five models, and the strong maneuvering model with the maximum correlation success weight value is directly selected as the designated model. When the weight calculation uses the aforementioned normalization method, it is equivalent to selecting the model with the highest recent cumulative correlation success number. After selecting the designated model, Kalman filter update is no longer needed (because there is no new measurement available), and the predicted state and predicted covariance matrix of the designated model at the current time are directly assigned to the updated state estimation value and updated state covariance matrix of the target at the current time. In the absence of observation information, the most reliable motion model in the near future is selected for prediction, which enables the target state to continue to extrapolate forward along the most likely motion pattern in the near future, while the state covariance increases reasonably due to the lack of new observations, which is included in the prediction covariance, reflecting the growth of estimation uncertainty. This state will serve as the starting point for the next tracking iteration, thereby maintaining a reasonable and consistent temporary estimate of the motion trend in the near future when the target temporarily disappears in clutter or extreme unmodeled maneuvers, greatly improving the survivability and continuity of the track in harsh environments.

[0071] The model selection based on historical performance adaptation can dynamically determine which model to trust when there is no observation by relying on the recent correlation success history of each model when dealing with the frequent short data loss problem in low-altitude strong maneuvering UAV tracking. In essence, it uses the target maneuvering mode preference learned from the tracking history in the past. If the target is frequently performing tangential maneuvers before losing the signal, the weight of the tangential model will be high, and after losing the signal, the model will naturally be selected for state extrapolation. The turning trend of its prediction is more likely to be consistent with the target's real behavior than blindly selecting the constant speed model; during the signal interruption period, the state estimation will not deviate seriously, but will be extrapolated along a trajectory consistent with the recent behavior logic, leaving more space for re-association after signal recovery; by introducing learning and decision-making based on historical performance in extreme cases, the fault tolerance, adaptability, and track maintenance ability of the entire tracking system in high-uncertainty scenarios such as low-altitude complex electromagnetic environments are significantly improved.

[0072] The technical solution of the embodiments in this specification introduces multiple preset strong maneuvering models in parallel during the prediction stage. By using models of maneuvering modes such as hovering, turning, and sharp turns of low-altitude UAVs, state inference is performed simultaneously along multiple high-probability maneuvering directions at each moment. This greatly reduces the systemic risk of the predicted point deviating significantly from the actual target position due to model mismatch. Regardless of what kind of maneuver the target actually performs, there will always be one or more models whose predicted direction roughly matches it, thus preserving the possibility for subsequent data association. This avoids the fatal weakness of traditional single models or limited model sets where predictions completely fail when the target suddenly maneuvers, directly leading to tracking loss. Secondly, by independently establishing an elliptical correlation gate strategy for each model, personalized management of prediction uncertainty and refined spatial search are achieved. The uncertainties of different maneuvering models exhibit anisotropic characteristics. Based on the dynamic characteristics of each model, the statistically optimal search region most likely to capture the true measurement is defined. This forms a complementary and efficient distributed search network in cluttered environments, significantly improving the capture score rate for the first detection of the true target measurement in dense false alarms, while avoiding excessive clutter due to an overly large gate or missed targets due to an overly small gate. The pre-defined hierarchical correlation strategy, through a two-level progression of initial screening within the model and screening between models, transforms the complex global optimal search problem into local optimal judgment and high-level arbitration. This enables the correlation decision to possess environmental awareness and target recognition capabilities, outputting high-confidence globally optimal measurements and their corresponding matching models in environments full of uncertainty. Kalman filter updates based on a single matching model and globally optimal measurements offer a clear update path, avoiding potential mutual interference and information dilution in multi-model soft fusion. This ensures the theoretical optimality of state estimation under a given decision. Since the measurements driving the update are rigorously selected optimal measurements, their innovation terms are highly reliable, resulting in accurate state correction direction and reasonable magnitude, effectively pulling the estimated trajectory back to the true path. Multi-model active coverage addresses maneuver unpredictability, independent gate fine-grained search resolves clutter spatial interference, and hierarchical information fusion decision-making resolves correlation ambiguity. Finally, optimal estimation based on explicit decisions achieves precise state correction and tracking closure, enabling the tracking system to maintain strong robustness in maintaining track continuity and stable state estimation under the dual pressure of low-altitude dense clutter and strong target maneuvering.

[0073] This specification also provides a target tracking device for a low-altitude, highly maneuverable unmanned aerial vehicle (UAV), such as... Figure 4 As shown, the device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described method.

[0074] The embodiments of the present specification further provide a non-volatile computer storage medium, which stores computer executable instructions, and the computer executable instructions are configured to execute the above method.

[0075] Each of the embodiments in the present specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each of the embodiments focuses on the difference from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiments.

[0076] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article, or equipment. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of other identical elements in the process, method, article, or equipment including the element.

[0077] The above only describes one or more embodiments of the present specification and is not intended to limit the present specification. One or more embodiments of the present specification can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of one or more embodiments of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. A target tracking method of a low-altitude strong maneuvering unmanned aerial vehicle, characterized in that, The method comprises: acquiring a set of measurement points output by a radar in a current detection period, and state estimation value and state covariance matrix of a target at a previous time, to perform one-step prediction on the previous time through a plurality of preset strong maneuvering models, to obtain current prediction indexes corresponding to the current time for each strong maneuvering model, wherein the current prediction indexes comprise predicted state and predicted covariance matrix; determining an elliptical association gate corresponding to each current prediction index, to determine a set of candidate measurement points falling within the elliptical association gate according to the set of measurement points; performing intra-model association and inter-model association in the set of candidate measurement points through a preset hierarchical association strategy, to screen a globally optimal measurement corresponding to the current time and a matching strong maneuvering model; performing Kalman filter updating on the current prediction index of the matching strong maneuvering model according to the globally optimal measurement corresponding to the current time, to obtain updated prediction indexes of the target at the current time, to realize target tracking, wherein the updated prediction indexes comprise updated state estimation value and updated state covariance matrix. 2.The target tracking method of a low-altitude strong maneuvering unmanned aerial vehicle according to claim 1, characterized in that, The strong maneuvering model comprises a uniform speed model, a hovering model, a U-turn model, a left tangential maneuvering model and a right tangential maneuvering model; wherein the state transition matrix of the uniform speed model is used to describe motion of the target with constant speed; the transition coefficient corresponding to the speed component in the state transition matrix of the hovering model is less than 1, and is used to describe motion of the target with speed decaying to near static; the transition coefficient corresponding to the position component in the state transition matrix of the U-turn model is opposite in sign to the transition coefficient corresponding to the speed component, and is used to describe motion of the target with motion direction reversing; the state transition matrix of the left tangential maneuvering model is used to describe motion of the target with speed direction deflecting by a preset angle counterclockwise in a two-dimensional plane; the state transition matrix of the right tangential maneuvering model is used to describe motion of the target with speed direction deflecting by a preset angle clockwise in a two-dimensional plane.

3. The target tracking method of a low-altitude high-maneuverable unmanned aerial vehicle according to claim 1, wherein, The method comprises: calculating an innovation covariance matrix corresponding to each strong maneuvering model according to the predicted covariance matrix in the current prediction index of each strong maneuvering model and a preset measurement noise covariance matrix; determining boundary conditions of the elliptical association gate of each strong maneuvering model based on the innovation covariance matrix and a preset threshold parameter; calculating a normalized distance of each measurement point relative to the predicted state according to position information of each measurement point in the set of measurement points, a preset observation matrix and the predicted state of the strong maneuvering model; judging whether the normalized distance of each measurement point satisfies the boundary conditions of the elliptical association gate, and if yes, the measurement point is classified into the set of candidate measurement points of the strong maneuvering model.

4. The target tracking method of a low-altitude high-maneuverable unmanned aerial vehicle according to claim 1, wherein, The method comprises: performing intra-model association and inter-model association in the set of candidate measurement points through a preset hierarchical association strategy, to screen a globally optimal measurement corresponding to the current time and a matching strong maneuvering model, specifically comprising: calculating the Euclidean distance between each candidate measurement point and the predicted state in each candidate measurement point set corresponding to the strong maneuvering model, and selecting a candidate measurement point with the minimum Euclidean distance as the candidate associated measurement of the strong maneuvering model; if there are multiple candidate measurement points with the same minimum Euclidean distance, screening the candidate associated measurement corresponding to the strong maneuvering model from the multiple candidate measurement points based on the amplitude information of each candidate measurement point and the historical amplitude information of the target; determining the globally optimal measurement at the current time according to the amplitude information of each candidate associated measurement, and determining the matching strong maneuvering model at the current time with the strong maneuvering model that produces the globally optimal measurement.

5. The target tracking method of a low-altitude high-maneuverable unmanned aerial vehicle according to claim 4, wherein, screening the candidate associated measurement corresponding to the strong maneuvering model from the multiple candidate measurement points based on the amplitude information of each candidate measurement point and the historical amplitude information of the target, specifically including: obtaining the current flight height information of the target, and judging whether the current flight height information of the target is lower than a preset height threshold; if the current flight height information is lower than the preset height threshold, obtaining the historical track data of the target, the historical track data containing the amplitude values of at least one historically associated successful measurement point; calculating the arithmetic mean of the amplitude values of the at least one historically associated successful measurement point to determine the historical amplitude reference value corresponding to the target, and calculating the difference between the amplitude information of each candidate measurement point and the historical amplitude reference value; determining the candidate associated measurement of the current strong maneuvering model as the candidate measurement point with the minimum difference; if the current flight height information is not lower than the preset height threshold, comparing the amplitude information of the multiple candidate measurement points, and determining the candidate associated measurement of the current strong maneuvering model as the candidate measurement point with the maximum amplitude information.

6. The target tracking method of a low-altitude high-maneuverable unmanned aerial vehicle according to claim 4, wherein, determining the globally optimal measurement at the current time according to the amplitude information of each candidate associated measurement, specifically including: comparing the amplitude information of each candidate associated measurement, and determining the candidate associated measurement with the maximum amplitude information as the globally optimal measurement at the current time.

7. The target tracking method of a low-altitude high-maneuverable unmanned aerial vehicle according to claim 6, wherein, After comparing the amplitude information of each candidate associated measurement, the method further includes: if there are at least two candidate associated measurements with the same maximum amplitude information, obtaining the specified strong maneuvering model corresponding to each specified candidate associated measurement with the same maximum amplitude information; calculating the innovation distance between the maximum amplitude information of each specified candidate associated measurement and the predicted state of the specified strong maneuvering model according to the maximum amplitude information of each specified candidate associated measurement and the predicted state of the specified strong maneuvering model; determining the globally optimal measurement at the current time as the specified candidate associated measurement with the minimum innovation distance.

8. The target tracking method of a low-altitude high-maneuverable unmanned aerial vehicle according to claim 1, wherein, performing Kalman filter update on the current prediction index of the matching strong maneuvering model according to the globally optimal measurement corresponding to the current time, specifically including: calculating the Kalman gain according to the predicted state and the predicted covariance matrix of the matching strong maneuvering model; According to the Kalman gain, a difference between the global optimal measurement and a predicted state of the matched strong maneuver model, the predicted state of the matched strong maneuver model is corrected to obtain the updated state estimation value; According to the Kalman gain and a predicted covariance matrix of the matched strong maneuver model, the updated state covariance matrix is obtained.

9. The target tracking method of a low-altitude high-maneuverable unmanned aerial vehicle according to claim 1, wherein, After determining the candidate measurement point set falling into the elliptical association gate according to the measurement point set, the method further comprises: If all the candidate measurement point sets of the strong maneuver models are empty, a specified model is selected from the plurality of strong maneuver models based on a preset rule, and a predicted state and a predicted covariance matrix of the specified model are directly taken as the updated state estimation value and the updated state covariance matrix of the target at the current time, respectively.

10. The target tracking method of a low-altitude high-maneuverable unmanned aerial vehicle according to claim 9, wherein, The method for selecting the specified model from the plurality of strong maneuver models based on the preset rule comprises: An association success flag sequence of each strong maneuver model in a plurality of preset historical tracking periods is obtained, wherein each element in the association success flag sequence is used to indicate whether the strong maneuver model is successfully associated to a candidate association measurement in a corresponding historical tracking period; Based on the association success flag sequence of each strong maneuver model, a cumulative association success number of each strong maneuver model in the plurality of historical tracking periods is obtained; Based on the cumulative association success number of each strong maneuver model, a corresponding association success weight of the strong maneuver model is calculated, wherein the higher the cumulative association success number is, the greater the corresponding association success weight is; According to the corresponding association success weight of each strong maneuver model, the specified model is selected from the plurality of strong maneuver models.

Images