A noise adaptive estimation method and system for radar track tracking
By introducing a time-varying attenuation factor and a filtering convergence criterion into the Kalman filter, and dynamically adjusting the noise covariance matrix, the problem of poor radar track tracking caused by the difficulty in estimating noise in the Kalman filter method is solved, and accurate tracking during target maneuvering is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SICHUAN JIUZHOU AIR TRAFFIC CONTROL TECHNOLOGY CO LTD
- Filing Date
- 2023-09-28
- Publication Date
- 2026-07-14
AI Technical Summary
Existing Kalman filtering methods suffer from poor tracking performance in radar track tracking due to the difficulty in estimating noise, especially when the noise distribution deviates from the Gaussian model during target maneuvering, affecting the accuracy of target tracking.
By establishing an aircraft dynamic model, a time-varying attenuation factor is introduced to update the prediction covariance matrix of the Kalman filter, and a filter convergence criterion is added to update the noise covariance matrix only when the convergence condition is not met. The Kalman gain is then calculated to obtain the trajectory prediction value after adaptive noise filtering.
It improves the adaptability of noise estimation, reduces the time complexity of the algorithm, ensures accurate tracking of trajectory changes when the target is maneuvering, and improves the accuracy and efficiency of radar track tracking.
Smart Images
Figure CN117347993B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of radar track tracking technology, and specifically to a noise adaptive estimation method and system for radar track tracking. Background Technology
[0002] Radar track tracking refers to the process of continuously monitoring and processing radar echoes from the same target to infer and predict the target's motion state and future trajectory, and is a crucial part of radar data processing. Radar track tracking typically employs data fusion, using the target's state estimate from the previous moment and the target's observed value at the current moment to calculate the target's motion state. Currently, radar track tracking commonly uses Kalman filtering or Kalman-like methods, such as α-β filtering and α-β-γ filtering. The effectiveness of Kalman filtering in tracking radar tracks depends on the statistical characteristics of the noise. Kalman filtering uses stationary Gaussian white noise, but in reality, noise changes with target motion and is difficult to accurately estimate. When the target maneuvers or the track quality is poor, the noise distribution deviates from the Gaussian model, affecting the accuracy of target tracking. In severe cases, it may calculate the target's motion state incorrectly, adversely affecting the performance of the radar system and limiting its practical application. Therefore, it is currently necessary to address the problem of poor tracking performance caused by the difficulty in estimating noise when using Kalman filtering to track radar tracks. Summary of the Invention
[0003] The technical problem to be solved by this invention is that the tracking effect is poor when using Kalman filtering to track radar tracks due to the difficulty in estimating noise. The purpose is to provide a noise adaptive estimation method and system for radar track tracking, which solves the above-mentioned problem.
[0004] This invention is achieved through the following technical solution:
[0005] An adaptive noise estimation method for radar track tracking includes: establishing an aircraft dynamic model to obtain the track prediction value at the current moment; updating the prediction covariance matrix of the Kalman filter, and introducing a time-varying attenuation factor when calculating the prediction covariance matrix; determining whether the Kalman filter has converged according to the filtering convergence criterion, and if the convergence condition is not met, calculating the process noise covariance matrix and the measurement noise covariance matrix, then calculating the Kalman gain, and finally obtaining the track prediction value after adaptive noise filtering.
[0006] A dynamic model of the aircraft is established to obtain the predicted trajectory value at the current moment, expressed as: X(k+1)=A(k)X(k)+G(k)W(k))(1); Z(k)=H(k)X(k)+V(k))(2); In equation (1), X(k)∈R n×1 It is an n-dimensional state vector, A(k)∈Rn×n G(k) is the state transition matrix, G(k) is the transition matrix between noise and state vector, and the random vector W(k) represents Gaussian noise during state transition. The statistical properties of W(k) satisfy: E[W(k)]=0(3); E[W(k)W T (k)]=Q(k)(4); where, equation (3) indicates that the expectation of the stationary Gaussian white noise W(k) is 0, and equation (4) indicates the relationship between the noise and its process noise covariance matrix Q; in equation (2), Z(k)∈R m×1 Let H(k) ∈ R be the measurement vector. m×n Let V(k) ∈ R be the measurement matrix. m×1 The measurement noise of the secondary radar follows a Gaussian distribution with a mean of zero, and the statistical properties of V(k) satisfy: E[V(k)]=0(5); E[V(k)V T (k)]=R(k)(6); where, equation (5) indicates that the expected value of the stationary Gaussian white noise V(k) is 0, and equation (6) indicates the relationship between the noise and its covariance matrix R; the formula for calculating the predicted value of the trajectory is as follows: X(k+1)=A(k)X(k)+G(k)W(k)(7).
[0007] The prediction covariance matrix of the Kalman filter is updated by introducing a time-varying attenuation factor when calculating the prediction covariance matrix, expressed as P(k+1)=λ(k)A(k)P(k)A T (k)+Q(k)(8); In equation (8), P is the prediction covariance matrix, and λ is the time-varying decay factor; wherein, the time-varying decay factor is calculated as follows: in: N(k)=H(k)Q(k)H T (k)+R(k)(11);M(k)=H(k)A(k)PA T (k)H T (k)(12); In equation (10), D is the difference between the observed value and the predicted track value, i.e., D(k)=Z(k)-H(k)X(k), tr[N(k)] is the trace of matrix N(k), and tr[M(k)] is the trace of matrix M(k).
[0008] Determining whether the Kalman filter has converged based on the filtering convergence criterion includes: V(k)V T (k)≤H(k)P(k+1)H T (k)+R(k)(13;W(k)W T (k)≤A(k)P(k+1)A T (k)+Q(k)(14); If equations (13) and (14) are satisfied, then the Kalman filter is determined to be currently converged.
[0009] When the convergence condition is not met, the process noise covariance matrix and the measurement noise covariance matrix are calculated, including: calculating the process noise covariance matrix Q, expressed as Q(k+1)=(1-c(k))Q(k)+c(k)[K(k)D(k)D T (k)K T (k)+P(k)-A(k)P(k-1)A T (k)](15); Calculate the measurement noise covariance matrix R, expressed as R(k+1)=(1-c(k))R(k)+c(k)[D(k)D T (k)+H k+1 P(k+1)H T k+1 (16); In equations (15) and (16), c(k) is defined as: c(k) = (1-b) / (1-b) k b is the forgetting factor (0) <b<1)。
[0010] Calculate the Kalman gain K, expressed as: In equation (17), H is the identity matrix, P is the prediction covariance matrix, and R is the measurement noise covariance matrix.
[0011] The predicted trajectory value X(k+1) after adaptive noise filtering is obtained as follows: X(k+1)=X(k)+K(k+1)D(k)(18).
[0012] After obtaining the predicted trajectory value in equation (18), the predicted covariance matrix P is updated for the next iteration calculation, expressed as: P(k+1)=[IK(k+1)H]P(k)(19).
[0013] An adaptive noise estimation system for radar track tracking includes: a track prediction module that establishes an aircraft dynamic model to obtain the track prediction value at the current moment; a covariance update module that updates the prediction covariance matrix of the Kalman filter and introduces a time-varying attenuation factor when calculating the prediction covariance matrix; a convergence judgment module that determines whether the Kalman filter has converged according to the filtering convergence criterion; and an optimal track module that calculates the process noise covariance matrix and the measurement noise covariance matrix when the convergence condition is not met, then calculates the Kalman gain, and finally obtains the track prediction value after adaptive noise filtering.
[0014] A radar tracking device includes a processor and a memory, wherein the memory stores computer instructions, characterized in that the computer instructions, when executed by the processor, can implement any of the aforementioned noise adaptive estimation methods for radar track tracking.
[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0016] This invention provides an adaptive noise estimation method and system for radar track tracking, based on an improvement of the Kalman filter algorithm. This method utilizes observation data to correct the statistical characteristics of noise during recursive filtering, thereby dynamically adjusting the process noise covariance Q and the observation noise covariance R. Since the calculation of Q and R requires all observation data from the start of filtering to the current time, a time-varying attenuation factor is introduced to dynamically adjust the predicted covariance matrix, reducing the influence of historical data on the calculation of Q and R and improving the adaptability of noise estimation. Simultaneously, a filtering convergence criterion is added for judgment, updating the noise covariance matrix only when the filter fails to meet the convergence condition, reducing the algorithm's time complexity. Furthermore, it can accurately estimate the noise covariance matrix when the target maneuvers, causing changes in the statistical characteristics of the noise, thus enabling accurate tracking of targets with suddenly changing trajectories. This invention solves the problem of poor tracking performance caused by the difficulty in estimating noise when using Kalman filtering to track radar tracks. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:
[0018] Figure 1 This is a flowchart of the noise adaptive estimation method for radar track tracking according to Embodiment 1 of this application;
[0019] Figure 2 This is a target machine point trace map of the current observation values in Embodiment 1 of this application;
[0020] Figure 3 This is a Kalman filter tracking trajectory diagram from Embodiment 1 of this application;
[0021] Figure 4 The Kalman filter tracking track diagram after using the noise adaptive estimation method in Embodiment 1 of this application;
[0022] Figure 5 This is a partial enlarged view of the flight path of the target aircraft with A code 5364 in Embodiment 1 of this application. Detailed Implementation
[0023] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.
[0024] Example 1
[0025] like Figure 1 As shown in the embodiments of this application, an adaptive noise estimation method for radar track tracking is proposed. The adaptive noise estimation method includes the following steps: establishing an aircraft dynamic model to obtain the track prediction value at the current time; updating the prediction covariance matrix of the Kalman filter, and introducing a time-varying attenuation factor when calculating the prediction covariance matrix; determining whether the Kalman filter has converged according to the filtering convergence criterion; if the convergence condition is not met, calculating the process noise covariance matrix and the measurement noise covariance matrix, and then calculating the Kalman gain, finally obtaining the track prediction value after adaptive noise filtering.
[0026] This invention provides an adaptive noise estimation method for radar track tracking, which is an improvement on the Kalman filter algorithm. This method utilizes observation data to correct the statistical characteristics of noise during recursive filtering, thereby dynamically adjusting the process noise covariance Q and the observation noise covariance R. Since the calculation of Q and R requires all observation data from the start of filtering to the current time, a time-varying attenuation factor is introduced to dynamically adjust the predicted covariance matrix, reducing the influence of historical data on the calculation of Q and R and improving the adaptability of noise estimation. Simultaneously, a filtering convergence criterion is added for judgment, updating the noise covariance matrix only when the filter fails to meet the convergence condition, reducing the algorithm's time complexity. Furthermore, it can accurately estimate the noise covariance matrix when the target maneuvers, causing changes in the statistical characteristics of the noise, thus enabling accurate tracking of targets with suddenly changing trajectories. This invention solves the problem of poor tracking performance caused by the difficulty in estimating noise when using Kalman filtering to track radar tracks.
[0027] In this process, the predicted trajectory value at time is the state estimate. The prediction covariance matrix P of the Kalman filter is updated. During the calculation of the prediction covariance matrix P, a time-varying attenuation factor is first calculated to reduce the influence of historical data on the subsequent calculation of the Q and R matrices, thus improving the adaptability of the noise estimation. The convergence criterion is used to determine whether the Kalman filter has converged. If the convergence condition is not met, the Q and R matrices are calculated, followed by the Kalman gain. Finally, the predicted trajectory value after adaptive noise filtering is obtained.
[0028] A filtering convergence criterion is added to determine whether the track observations have converged. If convergence is achieved, matrix updates are unnecessary, reducing the algorithm's time complexity. If convergence fails, the difference between the track observations and the predicted track values is calculated. This estimated difference is then used to further calculate the process noise covariance matrix Q and the measurement noise covariance matrix R. The Kalman filter gain is calculated, and the predicted track value after one iteration is further calculated; the error matrix of the predicted value is then updated.
[0029] A dynamic model of the aircraft is established to obtain the predicted trajectory value at the current moment, expressed as: X(k+1)=A(k)X(k)+G(k)W(k))(1); Z(k)=H(k)X(k)+V(k))(2); In equation (1), X(k)∈R n×1 It is an n-dimensional state vector, A(k)∈R n×n G(k) is the state transition matrix, G(k) is the transition matrix between noise and state vector, and the random vector W(k) represents Gaussian noise during state transition. The statistical properties of W(k) satisfy: E[W(k)]=0(3); E[W(k)W T (k)]=Q(k)(4); where, equation (3) indicates that the expectation of the stationary Gaussian white noise W(k) is 0, and equation (4) indicates the relationship between the noise and its process noise covariance matrix Q; in equation (2), Z(k)∈R m×1 Let H(k) ∈ R be the measurement vector. m×n Let V(k) ∈ R be the measurement matrix. m×1 The measurement noise of the secondary radar follows a Gaussian distribution with a mean of zero, and the statistical properties of V(k) satisfy: E[V(k)]=0(5); E[V(k)V T (k)]=R(k)(6); where, equation (5) indicates that the expected value of the stationary Gaussian white noise V(k) is 0, and equation (6) indicates the relationship between the noise and its covariance matrix R; the formula for calculating the predicted value of the trajectory is as follows: X(k+1)=A(k)X(k)+G(k)W(k)(7).
[0030] A dynamic model of the aircraft is established to obtain the predicted trajectory value at the current moment, expressed as: X(k+1)=A(k)X(k)+G(k)W(k))(1); Z(k)=H(k)X(k)+V(k))(2); In equation (1), X(k)∈R n×1 It is an n-dimensional state vector, A(k)∈R n×n G(k) is the state transition matrix, G(k) is the transition matrix between noise and state vector, and the random vector W(k) represents Gaussian noise during state transition. The statistical properties of W(k) satisfy: E[W(k)]=0(3); E[W(k)W T(k)]=Q(k)(4); where, equation (3) indicates that the expectation of the stationary Gaussian white noise W(k) is 0, and equation (4) indicates the relationship between the noise and its process noise covariance matrix Q; in equation (2), Z(k)∈R m×1 Let H(k) ∈ R be the measurement vector. m×n Let V(k) ∈ R be the measurement matrix. m×1 The measurement noise of the secondary radar follows a Gaussian distribution with a mean of zero, and the statistical properties of V(k) satisfy: E[V(k)]=0(5); E[V(k)V T (k)]=R(k)(6); where, equation (5) indicates that the expected value of the stationary Gaussian white noise V(k) is 0, and equation (6) indicates the relationship between the noise and its covariance matrix R; the formula for calculating the predicted value of the trajectory is as follows: X(k+1)=A(k)X(k)+G(k)W(k)(7).
[0031] The prediction covariance matrix of the Kalman filter is updated by introducing a time-varying attenuation factor when calculating the prediction covariance matrix, expressed as P(k+1)=λ(k)A(k)P(k)A T (k)+Q(k)(8); In equation (8), P is the prediction covariance matrix, and λ is the time-varying decay factor; wherein, the time-varying decay factor is calculated as follows: in: N(k)=H(k)Q(k)H T (k)+R(k)(11);M(k)=H(k)A(k)PA T (k)H T (k)(12); In equation (10), D is the difference between the observed value and the predicted track value, i.e., D(k)=Z(k)-H(k)X(k), tr[N(k)] is the trace of matrix N(k), and tr[M(k)] is the trace of matrix M(k).
[0032] Using the filter convergence criterion, it is determined whether interference factors such as target maneuvering cause significant errors in target measurements at certain times, thus affecting the estimation of state variables and leading to reduced filtering accuracy and decreased trajectory tracking performance. The convergence criterion for determining whether the Kalman filter has converged includes: V(k)V T (k)≤H(k)P(k+1)H T (k)+R(k)(13;W(k)W T (k)≤A(k)P(k+1)A T (k)+Q(k)(14); If equations (13) and (14) are satisfied, then the Kalman filter is determined to be currently converged.
[0033] Equations (13) and (14) are used to determine whether the measured values have large errors at certain times due to interference factors such as target maneuvering, which affects the estimation of state variables D(k) and leads to a decrease in filtering accuracy, resulting in a decline in track tracking performance. If the system noise and measurement noise satisfy equations (13) and (14), the current Kalman filter is considered to be converged, and there is no need to update the noise covariance matrices Q and R. If they are not satisfied, the current Kalman filter is considered to be unconverged, and the Q and R matrices need to be calculated. By using the filtering convergence criterion, the time complexity of the algorithm can be effectively reduced, and additional calculations can be avoided. At the same time, when the target maneuvers and causes changes in the statistical characteristics of the noise, the noise covariance matrix can be accurately estimated to achieve accurate tracking of targets whose trajectories suddenly change.
[0034] During the optimal estimation of the predicted flight path, the mean and variance of system noise and measurement noise are calculated and corrected in real time to ensure that the system filtering is in a normal state. However, during actual flight, when the flight environment is stable and the system is in a steady state, it can be approximated that the statistical characteristics of the system noise will not change drastically. When the system filtering is normal, there is no need to correct the statistical characteristics of the system noise.
[0035] When the convergence condition is not met, the process noise covariance matrix and the measurement noise covariance matrix are calculated, including: calculating the process noise covariance matrix Q, expressed as Q(k+1)=(1-c(k))Q(k)+c(k)[K(k)D(k)D T (k)K T (k)+P(k)-A(k)P(k-1)A T (k)](15); Calculate the measurement noise covariance matrix R, expressed as R(k+1)=(1-c(k))R(k)+c(k)[D(k)D T (k)+H k+1 P(k+1)H T k+1 (16); In equations (15) and (16), c(k) is defined as: c(k) = (1-b) / (1-b) k b is the forgetting factor (0) <b<1)。
[0036] The predicted trajectory value X(k+1) after adaptive noise filtering is obtained as follows: X(k+1)=X(k)+K(k+1)D(k)(18).
[0037] After obtaining the predicted trajectory value in equation (18), the predicted covariance matrix P is updated for the next iteration calculation, expressed as: P(k+1)=[IK(k+1)H]P(k)(19).
[0038] Based on general experience, b takes a value of 0.95 to 0.99, and the smaller the value of b, the less dependent it is on historical noise. If convergence is achieved, there is no need to use equations (15) and (16) to correct the statistical characteristics of system noise. By gradually weakening the effect of data that are far from the current iteration through c(k), while strengthening the effect of data that are closer to the current iteration, the weighting coefficient of new data items is increased and the weighting coefficient of historical data items is decreased through this fading memory exponential weighting method. Finally, the calculated Q and R are used to update the Kalman filter gain, the noise V and W in the system model, and the prediction covariance matrix P used to calculate in the next iteration.
[0039] In each iteration, if the estimation error decreases, then D T (k)D(k) decreases, causing λ0 to decrease. When λ0 < 1, we take λ0 = 1. In this case, the update of the predicted covariance matrix P is not affected by the time-varying decay factor. If the estimation error increases, then D T As D(k) increases, λ0 increases. When λ0≥1, the error covariance matrix P increases, which in turn affects the increase of Kalman filter gain. Therefore, the weight of D(k) will be increased, that is, the ratio of the difference between the observed value and the track prediction value at the previous moment will be increased. This will reduce the influence of historical data on the calculation of Q and R, improve the adaptability of noise estimation, and make the filtered track prediction value closer to the optimal value.
[0040] During the filtering process, the current filter convergence criterion is used to determine whether it has converged. If the convergence condition is met, the measurement noise covariance Q and process noise covariance R are not updated. If the convergence condition is not met, the process noise covariance Q and measurement noise covariance R are updated. By using the filtering convergence criterion, the algorithm's time complexity can be effectively reduced, avoiding additional computation. Simultaneously, it can accurately estimate the noise covariance matrix when the target maneuvers, causing changes in the statistical characteristics of the noise, thus enabling accurate tracking of targets whose trajectories suddenly change.
[0041] In application, six real target points with different A codes are randomly selected within the secondary radar detection range, and their trajectories in the Cartesian coordinate system are as follows: Figure 2 As shown, the units for both the horizontal and vertical axes in the graph are meters. Figure 2 It can be seen that the targets directly detected by radar have random observation errors, and the traces are not smooth and discontinuous in some stages. Therefore, it is necessary to estimate the target state through track tracking to reduce the error.
[0042] Kalman filtering was used to track the tracks of these six target points, and the results are as follows: Figure 3As shown, when the target aircraft's trajectory remains constant, the Kalman filter tracks the target well. However, when the trajectory begins to change, such as when targets 4004, 7450, and 2070 begin to turn, the Kalman filter's tracking accuracy decreases because the process noise matrix Q and the observation noise matrix R cannot be adaptively adjusted. The track only gradually adjusts to the correct position when the target aircraft's trajectory stabilizes again, exhibiting a certain lag compared to the target's actual trajectory.
[0043] Track tracking was performed using a Kalman filter incorporating an adaptive noise estimation method, and the results are as follows: Figure 4 As shown, adaptive noise estimation outperforms the Kalman filter algorithm in both target trajectory tracking and smoothing after track formation. Furthermore, due to the addition of a convergence criterion, the noise covariance matrix is only updated when the filter fails to meet the convergence condition, significantly improving computation speed.
[0044] The Kalman filter combined with a noise adaptive estimation method adds an attenuation factor to adjust the prediction covariance matrix and gain matrix in real time. This increases the ratio of the difference between observed and predicted track values, reduces the impact of observed values on the optimal estimate, improves the adaptability of the Kalman filter process, and makes the filtered estimate closer to the optimum. Figure 5 As shown, using a Kalman filter with noise adaptive estimation to generate a track for the same target aircraft can effectively reduce the impact of observations containing large noise and accurately restore the target trajectory.
[0045] Example 2
[0046] This application provides a noise adaptive estimation system for radar track tracking. The noise adaptive estimation system includes: a track prediction module that establishes an aircraft dynamic model to obtain the track prediction value at the current moment; a covariance update module that updates the prediction covariance matrix of the Kalman filter and introduces a time-varying attenuation factor when calculating the prediction covariance matrix; a convergence judgment module that determines whether the Kalman filter has converged according to the filtering convergence criterion; and a track optimal module that calculates the process noise covariance matrix and the measurement noise covariance matrix when the convergence condition is not met, then calculates the Kalman gain, and finally obtains the track prediction value after adaptive noise filtering.
[0047] The embodiments of this application are implemented based on the method described in Embodiment 1, and the principle is the same as that of Embodiment 1, so they will not be described again here.
[0048] In summary, this application provides a noise adaptive estimation method and system for radar track tracking. By introducing a time-varying attenuation factor to dynamically adjust the prediction covariance matrix, the influence of historical data on the calculation of updated noise covariance Q and observation noise covariance R is reduced, improving the adaptability of noise estimation. Furthermore, a filter convergence criterion is added, updating the noise covariance matrix only when the filter fails to meet the convergence condition, reducing the algorithm's time complexity. Applied to radar system track tracking, this application can use Kalman filtering to track radar tracks and obtain the current optimal track prediction value, dynamically adjusting the process noise matrix Q and observation noise matrix R. This effectively reduces the problem of poor track tracking performance due to difficulty in estimating system noise, improving system reliability.
[0049] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A noise adaptive estimation method for radar track tracking, characterized in that, include: Establish an aircraft dynamic model to obtain the current trajectory prediction value; The prediction covariance matrix of the Kalman filter is updated, and a time-varying attenuation factor is introduced when calculating the prediction covariance matrix; The prediction covariance matrix is expressed as follows: (1)； In equation (1), P is the prediction covariance matrix. It is the state transition matrix, where n is the dimension of the state vector, and Q(k) is the process noise covariance matrix. The time-varying decay factor is mentioned above; The time-varying decay factor is calculated as follows: (2)； in: (3)； (4)； (5)； In equation (3), D is the difference between the observed value and the predicted trajectory value, i.e. , For matrix traces, For matrix traces, For measurement vectors, Let R(k) be the measurement matrix, and R(k) be the measurement noise covariance matrix. It is an n-dimensional state vector; The Kalman filter is judged to be converged according to the filtering convergence criterion. If the convergence condition is not met, the process noise covariance matrix and the measurement noise covariance matrix are calculated, and then the Kalman gain is calculated to finally obtain the trajectory prediction value after adaptive noise filtering. Determining whether the Kalman filter has converged according to the filtering convergence criterion includes: (6)； (7)； In the formula, W(k) is a random vector. This refers to the measurement noise of the secondary radar; If equations (6) and (7) are satisfied, then the Kalman filter is determined to be currently converged; When the convergence condition is not met, calculate the process noise covariance matrix and the measurement noise covariance matrix, including: The process noise covariance matrix Q is calculated and expressed as follows: (8)； The measurement noise covariance matrix R is calculated and expressed as follows: (9)； In equations (8) and (9), c(k) is defined as: , where b is the forgetting factor (0 < b < 1) and K(k) is the Kalman gain.
2. The noise adaptive estimation method for radar track tracking according to claim 1, characterized in that, A dynamic model of the aircraft is established to obtain the predicted trajectory value at the current moment, represented as: (10)； (11)； In equation (10), It is an n-dimensional state vector. Let G(k) be the state transition matrix, G(k) be the transition matrix between noise and the state vector, and W(k) be the random vector representing Gaussian noise during state transition. The statistical properties of W(k) satisfy the following: (12)； (13)； Equation (12) indicates that the expected value of the stationary Gaussian white noise W(k) is 0, and Equation (13) indicates the relationship between the noise and its process noise covariance matrix Q. In equation (11), For measurement vectors, For the measurement matrix, The measurement noise of the secondary radar follows a Gaussian distribution with zero mean, and the statistical properties of V(k) satisfy: (14); (15); Equation (14) indicates that the expectation of the stationary Gaussian white noise V(k) is 0, and Equation (15) indicates the relationship between the noise and its covariance matrix R. The formula for calculating the predicted trajectory value is as follows: (16)。 3. The noise adaptive estimation method for radar track tracking according to claim 1, characterized in that, Calculate the Kalman gain K, expressed as: (17)； In equation (17), H is the identity matrix, P is the prediction covariance matrix, and R is the measurement noise covariance matrix.
4. The noise adaptive estimation method for radar track tracking according to claim 3, characterized in that, The predicted trajectory value X(k+1) after adaptive noise filtering is obtained as follows: (18)。 5. The noise adaptive estimation method for radar track tracking according to claim 4, characterized in that, After obtaining the predicted trajectory value using equation (18), the predicted covariance matrix P is updated for the next iteration calculation, expressed as: (19)。 6. A noise adaptive estimation system for radar track tracking, used to execute the noise adaptive estimation method for radar track tracking as described in any one of claims 1-5, characterized in that, include: A trajectory prediction module that establishes a dynamic model of the aircraft to obtain the trajectory prediction value at the current moment; The prediction covariance matrix of the Kalman filter is updated, and a covariance update module with a time-varying decay factor is introduced when calculating the prediction covariance matrix; A convergence determination module that determines whether the Kalman filter has converged based on the filtering convergence criterion; If the convergence condition is not met, the process noise covariance matrix and the measurement noise covariance matrix are calculated, and then the Kalman gain is calculated. Finally, the optimal track module is obtained by obtaining the track prediction value after adaptive noise filtering.
7. A radar tracking device, comprising a processor and a memory, wherein the memory stores computer instructions, characterized in that, The computer instructions, when executed by the processor, can implement a noise adaptive estimation method for radar track tracking as described in any one of claims 1-5.