Coordinate system transformation fusion filtering tracking method and system for dual-base-station radar
Through the coordinate system transformation fusion filtering method, the lack of performance of traditional nonlinear filtering technology in dual-base station radar in high-dimensional and strong nonlinear systems is solved, and higher tracking accuracy and robustness are achieved.
Patent Information
- Application Number
- CN202510368574.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-27
- Publication Date
- 2025-06-20
- Estimated Expiration
- 2045-03-27
AI Technical Summary
In dual-base station radar, traditional nonlinear filtering technology has poor performance in high-dimensional and strong nonlinear systems, resulting in insufficient tracking accuracy and robustness.
The coordinate system transformation fusion filtering method is used to establish the state equation and observation equation of the dual base station radar, and the U transformation is used to generate sigma points, perform Kalman filtering fusion, and convert the updated state estimate back to the Cartesian coordinate system, and loop this process to achieve tracking.
It effectively avoids nonlinear filtering problems during the update process and improves the robustness and accuracy of tracking.
Smart Images

Figure CN120178227A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of radar target tracking, and particularly to a coordinate system transformation fusion filtering tracking method and system for a dual-base radar. Background Art
[0002] In the modern air and space defense system, continuous surveillance of stealth aircraft is an important topic for national security. With the iterative upgrade of aircraft stealth technology, its radar cross-section (RCS) has been significantly reduced, and the intensity of the secondary reflection signal has also continued to decay. Facing this challenge, traditional single-base radar systems (including pulsed Doppler radars, dual / multi-base radars) have gradually shown limitations in anti-stealth operations. Active radars need to continuously emit detection electromagnetic waves, exposing their own radiation source characteristics while searching for targets, and are easily locked and struck by anti-radiation weapons; while pure passive radars rely on receiving electromagnetic leakage signals of targets. Restricted by the improvement of stealth platform electromagnetic silence technology, their detection efficiency is increasingly limited. Coupled with the high false alarm rate in complex electromagnetic environments, it is difficult to build a reliable air situation monitoring system. Dual-base radars, through the architecture of spatially separated transmitters and receivers, organically integrate the technical advantages of active and passive radars. Its working principle is as follows: several radiation nodes are responsible for irradiating the warning airspace with electromagnetic waves, and the concealed receiving nodes form monitoring stations. Due to the spatial decoupling characteristics of the radiation source and the receiving unit, it is difficult for the target to implement effective electromagnetic evasion strategies. This new type of radar system exhibits highly flexible deployment characteristics. It can either utilize the physical platforms of existing radar facilities or build multi-base detection arrays adapted to different battlefield environments through the embedding of new sensor nodes. The comprehensive advantages of the system are mainly reflected in three aspects: significantly improving the battlefield survival ability of radiation sources, realizing the dynamic reconfiguration of sensor resources, and effectively enhancing the collaborative combat effectiveness of the regional air defense and anti-missile system.
[0003] A dual-base radar is a unified radar system composed of one or more transmitters and one or more receivers. A dual-base radar has only one set of transmitting / receiving devices and is a basic component of a multi-station radar. A dual-base radar can detect, locate, measure the speed of, and track targets. Compared with a single-base radar, a dual-base radar has the following advantages: it can detect targets from multiple directions, which is beneficial for target recognition and anti-stealth; it improves the probability of target detection and the continuity of target position and trajectory; transmitters operating in different frequency bands can work alternately, improving the anti-interference ability and reliability of the system; it enhances the survival rate of the system and will not cause the entire system to collapse due to abnormalities in the local transmitting or receiving system.
[0004] Since the Kalman filter algorithm was introduced in the 1960s, it has been widely used in fields such as communication, power, aerospace, and industrial control because it can estimate the minimum mean square error of the system state by predicting values through the state equation and combining observation data. In a linear Gaussian system, the Kalman filter performs excellently, but in a non-linear system, its direct application is limited. In dual-base radar target tracking, the target information obtained by the radar includes distance, angle, Doppler velocity, etc. These observation data not only have errors but also show a non-linear relationship with the state of the target in the Cartesian coordinate system. In order to achieve continuous tracking of the target, non-linear filtering techniques based on the Kalman filter have emerged. The Extended Kalman Filter (EKF) is a commonly used method among them. It linearizes the non-linear equation through Taylor series expansion, which is convenient for engineering implementation. However, EKF has low accuracy when dealing with high-dimensional and strongly non-linear systems, and the calculation of the Jacobian matrix is relatively complex. The Unscented Kalman Filter (UKF) and the Sigma-Point Transform Kalman Filter (SPTKF) are representatives of deterministic sampling methods. UKF predicts data through unscented transformation, and SPTKF approximates the posterior probability distribution by measuring data through unscented transformation, avoiding the local linearization problem of EKF. Nevertheless, these methods still have deficiencies in terms of accuracy, time complexity, and consistency. Although non-linear filtering techniques such as EKF and UKF have made certain progress in radar target tracking, their performance in high-dimensional and strongly non-linear scenarios is still poor. Summary of the Invention
[0005] The object of the present invention is to provide a coordinate system transformation fusion filtering tracking method and system for a dual-base radar, which can avoid the non-linear filtering problem in the update process and improve the robustness and accuracy of tracking.
[0006] To achieve the above object, the present invention provides the following solutions:
[0007] A coordinate system transformation fusion filtering tracking method for a dual-base radar, comprising:
[0008] Based on the kinematic characteristics of the target, establish the state equation and observation equation of the dual-base radar tracking system for the moving target, and determine the initial motion state and initial covariance of the target based on the prior information of the target in the Cartesian coordinate system;
[0009] Perform a one-step prediction based on the motion state and covariance of the target in the Cartesian state space at time k to obtain the motion state of the target at time k + 1; the motion state of the target at time k + 1 includes the target prediction state and the prediction covariance matrix;
[0010] Based on the U transformation, generate a set of sigma points for the target predicted state and the predicted covariance matrix, and generate 2n + 1 sigma points for the n-dimensional motion state of the target in the Cartesian coordinate system, and calculate the mean and covariance for constructing the state space prediction by weighted calculation;
[0011] Use the Kalman filter to fuse the measurement data and the prediction data to obtain the optimal state estimate and covariance matrix at the (k + 1)-th moment; the measurement data is obtained by radar measurement; the prediction data is the mean and covariance for constructing the state space prediction;
[0012] Based on the U transformation, generate 2n + 1 sigma points according to the updated optimal state estimate and the updated covariance matrix at the (k + 1)-th moment, transform each sigma point in the constructed state space to the Cartesian coordinate system using the transformation formula, and calculate the mean and covariance of the optimal state estimate in the Cartesian coordinate system by weighted calculation;
[0013] Loop the steps until the tracking ends.
[0014] Optionally, based on the kinematic characteristics of the target, establish the state equation and the observation equation of the dual-base radar tracking system for the moving target, and determine the initial motion state and the initial covariance of the target based on the prior information of the target in the Cartesian coordinate system, specifically including:
[0015] Based on the kinematic characteristics of the target, establish the state equation and the observation equation of the dual-base radar tracking system for the moving target, and the formulas are respectively:
[0016] X(k) = F(k)·X(k - 1) + G(k)·w(k)
[0017] Z(k) = h(X(k)) + v(k)
[0018] where, X z (k) = [x(k) x′(k) y(k) y′(k)] is the state vector directly constructed from the received radar observation data, and this vector is used to represent the motion state of the target at the k-th moment, that is, this vector is substituted into the state at the k-th moment in the target's state equation, namely X(k), and x(k), x′(k), y(k), y′(k) are respectively the distance in the x direction, the velocity in the x direction, the distance in the y direction, and the velocity in the y direction of the target relative to the receiving radar at the k-th moment, F(k), G(k), w(k) and h are respectively the state transition matrix, the noise coefficient matrix, the process noise and the observation function, where w(k) = [w x (k) w y (k)] T , w x (k) and w y (k) are respectively the process noise in the x direction and the y direction, Zz y(k) = [R(k) θ(k)] T is the observation vector of the receiving radar affected by noise at the k-th moment, which is used to represent the target state observed at the k-th moment, that is, this vector is substituted into the state at the k-th moment in the observation equation of the target, namely Z(k). Among them, R(k) is the sum of the distances between the transmitting radar and the receiving radar and the moving target at the k-th moment, θ(k) is the angle formed by the receiving radar and the moving target at the k-th moment, v(k) = [v r (k) v α (k)] Τ is the observation noise of the receiving radar at the k-th moment, v r (k) is the measurement distance noise at the k-th moment, v α (k) is the measurement angle noise at the k-th moment, both are Gaussian white noises with zero mean, v r (k) and v α (k) are uncorrelated, and the noise covariance matrix is The specific formula of the observation function h is:
[0019]
[0020] Among them, Z true (k) is the true observation state without noise at the k-th moment, and L is the distance between the receiving radar and the transmitting radar;
[0021] For a moving target, based on the prior information of the target in Cartesian coordinates, the initial motion state X(1) and the initial covariance P(1) of the target are obtained.
[0022] Optionally, the motion state of the target at the k + 1 moment is specifically expressed as:
[0023] X k+1|k = F(k)X(k)
[0024] P k+1|k = F(k)P(k)F(k) Τ + G(k)Q(k)G(k) Τ
[0025] Among them, X k+1|k 、P k+1|k are the predicted state and the predicted covariance matrix at the k-th moment respectively, and Q(k) represents the covariance matrix of the process noise.
[0026] Optionally, based on the U transformation, a set of sigma points are generated for the target predicted state and the predicted covariance matrix, and 2n + 1 sigma points are generated for the n-dimensional motion state of the target in the Cartesian coordinate system, and the mean and covariance of the constructed state space prediction are calculated by weighting, specifically including:
[0027] Based on the U transformation, the target predicted state X at the current time k k+1|k and the predicted covariance matrix P k+1|k Generate a set of sigma points, and generate 2n + 1 sigma points for its n-dimensional motion state in the Cartesian coordinate system. The specific formula is:
[0028] x k (0) = X k+1|k
[0029]
[0030] where i = 1, 2,..., n, λ = α 2 (n + κ) - n, α and κ are scaling parameters used to adjust the distribution of σ points; x k represents the sigma points generated in the Cartesian coordinate system, while x k (i) represents the i-th sigma point. The formula for the weight of each sigma point is reflected by the following formula:
[0031]
[0032] where β is a parameter used to fuse the observation noise, W m and W c are the weights of the mean and covariance respectively, both of which are vectors of 2n + 1;
[0033] Apply the conversion formula to each sigma point in the Cartesian coordinate system to convert it to the constructed state space. The constructed state space is specifically constructed based on the position state representation and velocity state representation of the target in the two-dimensional polar coordinate system. Take the distance r1 between the target and the receiving radar, the tangential velocity v r1 of the target, the distance r2 between the target and the transmitting radar, the normal velocity v t of the target, and the angle θ formed by the target and the receiving radar as vectors describing the target state for construction, expressed as
[0034] Calculate the mean and covariance of the prediction in the constructed state space by weighted calculation. The specific formula for the calculation is:
[0035]
[0036] where, W m (i) represents the mean weight of the i-th sigma point, W c (i) represents the covariance weight of the i-th sigma point, and f is the non-linear conversion function from the Cartesian coordinate system to the constructed state space. The specific conversion method is shown in the following formula:
[0037]
[0038] v t =-v x sinθ + v y cosθ
[0039] Where L is the distance between the receiving radar and the transmitting radar, x is the distance of the target relative to the receiving radar in the x - direction, v x is the velocity of the target relative to the receiving radar in the x - direction, y is the distance of the target relative to the receiving radar in the y - direction, v y is the velocity of the target relative to the receiving radar in the y - direction.
[0040] Optionally, the optimal state estimate and covariance matrix at the (k + 1)-th moment are expressed as:
[0041] K k+1 = pol P k+1|k T H T (H pol P k+1|k H T + R k ) -1
[0042]
[0043] pol P k+1|k+1 =(I - K k+1 H) pol P k+1|k
[0044] Where K k+1 is the Kalman gain at the (k + 1)-th moment, is the updated optimal state estimate at the (k + 1)-th moment, pol P k+1|k+1 is the updated covariance matrix, I is the identity matrix, H is the observation matrix of the receiving radar observing the target, z k+1 is the observation vector observed by the receiving radar at the (k + 1)-th moment.
[0045] Optionally, based on the U - transformation, 2n + 1 sigma points are generated according to the updated optimal state estimate and updated covariance matrix at the (k + 1)-th moment. Each sigma point in the constructed state space is transformed to the Cartesian coordinate system using the transformation formula, and the mean value and covariance matrix of the optimal state estimate in the Cartesian coordinate system are calculated by weighted calculation, specifically including:
[0046] Based on the U - transformation, through the updated optimal state estimate at the (k + 1)-th moment With the updated covariance matrix pol P k+1|k+1 Generate 2n + 1 sigma points, and the specific formula is:
[0047]
[0048] Among them, pol X k+1 is the sigma point generated in the construction space, pol X k+1 (i) is the i-th sigma point;
[0049] Apply the transformation formula to each sigma point in the construction state space to transform it into the Cartesian coordinate system, and calculate the weighted optimal state estimation mean and covariance matrix in the Cartesian coordinate system. The specific formula is:
[0050]
[0051] Among them, g is the non-linear transformation function from the state space to the Cartesian coordinate system. The specific transformation method is shown in the following formula:
[0052] R = r1 + r2
[0053]
[0054] Among them, L is the distance between the receiving radar and the transmitting radar, and R is the sum of the distances from the moving target to the transmitting radar and the receiving radar.
[0055] Optionally, the target motion state under the construction state space is expressed as:
[0056]
[0057] Among them, r1(k) represents the distance between the receiving radar and the moving target at the k-th moment, represents the radial velocity of the moving target relative to the receiving radar at the k-th moment, r2(k) represents the distance between the transmitting radar and the moving target at the k-th moment, v t (k) represents the tangential velocity of the moving target relative to the receiving radar at the k-th moment, and θ(k) represents the angle of the moving target relative to the receiving radar at the k-th moment.
[0058] The present invention also provides a coordinate system transformation fusion filtering tracking system for a dual-base radar, including:
[0059] An initial state construction module, which is used to establish the state equation and observation equation of the dual-base radar tracking system for a moving target based on the kinematic characteristics of the target, and determine the initial motion state and initial covariance of the target based on the prior information of the target in Cartesian coordinates;
[0060] A one-step prediction module, which is used to perform a one-step prediction according to the motion state and covariance of the target in the Cartesian state space at time k to obtain the motion state of the target at time k+1; the motion state of the target at time k+1 includes the target prediction state and the prediction covariance matrix;
[0061] A first U transformation module, which is used to generate a set of sigma points for the target prediction state and the prediction covariance matrix based on the U transformation, and generate 2n+1 sigma points for the n-dimensional motion state of the target in the Cartesian coordinate system, and calculate the mean and covariance of the constructed state space prediction by weighting;
[0062] A fusion module, which is used to fuse the measurement data and the prediction data by using a Kalman filter to obtain the optimal state estimate and covariance matrix at the (k+1)-th moment after fusion; the measurement data is obtained by radar measurement; the prediction data is the mean and covariance of the constructed state space prediction;
[0063] A second U transformation module, which is used to generate 2n+1 sigma points based on the U transformation according to the updated optimal state estimate and the updated covariance matrix at time k+1, convert each sigma point in the constructed state space to the Cartesian coordinate system by using the conversion formula, and calculate the mean and covariance matrix of the optimal state estimate in the Cartesian coordinate system by weighting;
[0064] A loop module, which is used to loop the steps until the tracking ends.
[0065] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:
[0066] The present invention discloses a coordinate transformation fusion filtering tracking method and system for a dual-base radar. The method includes first establishing a state and an observation equation, and determining the initial state of the target based on the prior information in Cartesian coordinates. Subsequently, the target motion state is obtained through one-step prediction, and sigma points are generated by using the U transformation to calculate the mean and covariance of the constructed state space prediction. Then, the measurement data and the prediction data are fused by using a Kalman filter to obtain the optimal state estimate and covariance matrix. Finally, the updated state estimate is converted back to the Cartesian coordinate system based on the U transformation, and this process is looped until the tracking ends. The present invention can avoid the non-linear filtering problem in the update process, and improve the robustness and accuracy of the tracking. Description of the Drawings
[0067] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.
[0068] Figure 1 This is the flowchart of the coordinate system transformation fusion filtering tracking method for the dual - base - station radar in this embodiment;
[0069] Figure 2 This is the schematic diagram of constructing the state space based on the observations of the dual - base - station radar and the target motion state in this embodiment;
[0070] Figure 3 This is the schematic diagram of the RMSE results of the target position estimation for various methods in this embodiment;
[0071] Figure 4 This is the schematic diagram of the RMSE results of the target speed estimation for various methods in this embodiment. Specific Embodiments
[0072] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, rather than all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.
[0073] The purpose of the present invention is to provide a coordinate system transformation fusion filtering tracking method and system for a dual - base - station radar, which can avoid the non - linear filtering problem in the update process and improve the robustness and accuracy of tracking.
[0074] To make the above - mentioned objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.
[0075] The present invention provides a coordinate system transformation fusion filtering tracking method for a dual - base - station radar, including:
[0076] S1: Based on the kinematic characteristics of the target, establish the state equation and observation equation of the dual - base - station radar tracking system for this moving target, and their formulas are respectively:
[0077] X(k) = F(k)·X(k - 1)+G(k)·w(k) (1)
[0078] Z(k) = h(X(k))+v(k) (2)
[0079] Among them, X z (k) = [x(k) x′(k) y(k) y′(k)] is the state vector directly constructed from the received radar observation data, which is used to represent the motion state of the target at the k-th moment. That is, this vector is substituted into the state at the k-th moment in the target state equation, namely X(k). x(k), x′(k), y(k), and y′(k) are respectively the distance in the x-direction, the velocity in the x-direction, the distance in the y-direction, and the velocity in the y-direction of the target relative to the receiving radar at the k-th moment. F(k), G(k), w(k), and h are respectively the state transition matrix, the noise coefficient matrix, the process noise, and the observation function. Among them, w(k) = [w x (k) w y (k)] T , w x (k) and w y (k) are the process noises in the x-direction and y-direction. Z z (k) = [R(k) θ(k)] T is the observation vector of the receiving radar affected by noise at the k-th moment, which is used to represent the target state observed at the k-th moment. That is, this vector is substituted into the state at the k-th moment in the target observation equation, namely Z(k). Among them, R(k) is the sum of the distances between the transmitting radar and the receiving radar and the moving target at the k-th moment, and θ(k) is the angle formed by the receiving radar and the moving target at the k-th moment. v(k) = [v r (k) v α (k)] Τ is the observation noise of the receiving radar at the k-th moment. v r (k) is the measurement distance noise at the k-th moment, and v α (k) is the measurement angle noise at the k-th moment. Both are Gaussian white noises with zero mean. v r (k) and v α (k) are uncorrelated, and the noise covariance matrix is The specific formula of the observation function h is:
[0080]
[0081] Among them, Z true (k) is the true observation state without noise at the k-th moment, and L is the distance between the receiving radar and the transmitting radar.
[0082] For this moving target, based on the prior information of the target in the Cartesian coordinate system, the initial motion state X(1) and the initial covariance P(1) of the target are obtained;
[0083] S2: According to the state and covariance of the target motion in the Cartesian state space at the k-th moment, a one-step prediction is performed to obtain the motion state at the k + 1-th moment. The specific formula is:
[0084] X k+1|k = F(k)X(k)(4)
[0085] P k+1|k = F(k)P(k)F(k) Τ + G(k)Q(k)G(k) Τ (5)
[0086] where X k+1|k , P k+1|k are the predicted state and the predicted covariance matrix at time k respectively, and Q(k) represents the covariance matrix of the process noise;
[0087] S3: Based on the U transformation, generate a set of sigma points for the target predicted state X k+1|k and the predicted covariance matrix P k+1|k at the current time k, and generate 2n + 1 sigma points for their n-dimensional motion state in the Cartesian coordinate system. The specific formula is:
[0088] x k (0) = X k+1|k (6)
[0089]
[0090] where i = 1, 2,..., n, λ = α 2 (n + κ) - n, and α and κ are scaling parameters used to adjust the distribution of σ points. x k represents the generated sigma points in the Cartesian coordinate system, while x k (i) represents the i-th sigma point. The formula for the weight of each sigma point is reflected by the following formula:
[0091]
[0092] where β is a parameter used to fuse the observation noise, and W m and W c are the weights of the mean and the covariance respectively, both being vectors of 2n + 1.
[0093] Convert each sigma point in the Cartesian coordinate system to the constructed state space using the conversion formula. The constructed state space is specifically based on the position state representation and the velocity state representation of the target in the two-dimensional polar coordinate system. That is, take the distance r1 between the target and the receiving radar, the tangential velocity v r1 , the distance r2 between the target and the transmitting radar, the normal velocity v t and the angle θ formed by the target and the receiving radar as the vector describing the target state for construction, that is, [r1v r1 r2vt θ] Τ The mean and covariance of the constructed state space prediction are calculated by weighted averaging. The specific calculation formula is as follows:
[0094]
[0095] where W m (i) represents the mean weight of the i-th sigma point, and W c (i) represents the covariance weight of the i-th sigma point. f is the non-linear conversion function from the Cartesian coordinate system to the constructed state space. The specific conversion method is shown in the following formula:
[0096]
[0097]
[0098] v t =-v x sinθ + v y cosθ (17)
[0099] In the above formula, L is the distance between the receiving radar and the transmitting radar, x is the distance of the target in the x direction relative to the receiving radar, v x is the velocity of the target in the x direction relative to the receiving radar, y is the distance of the target in the y direction relative to the receiving radar, and v y is the velocity of the target in the y direction relative to the receiving radar;
[0100] S4: Obtain the measurement data z k+1 =[R k+1 θ k+1 T , where R k+1 is the sum of the distances of the moving target from the transmitting radar and the receiving radar at the (k + 1)-th moment, and θ k+1 is the angle between the moving target and the receiving radar at the (k + 1)-th moment. The measurement data and the prediction data are fused using the Kalman filter to obtain the optimal state estimate and covariance matrix at the (k + 1)-th moment after fusion. The specific formula is as follows:
[0101] K k+1 = pol P k+1|k T H T (H pol P k+1|k H T +R k ) -1 (18)
[0102]
[0103] pol P k+1|k+1 =(I - K k+1 H) pol P k+1|k (20)
[0104] where K k+1 is the Kalman gain at time k + 1, is the optimal state estimate updated at time k + 1, pol P k+1|k+1 is the updated covariance matrix, and I is the identity matrix;
[0105] S5: Based on the U transformation, generate 2n + 1 sigma points through the optimal state estimate updated at time k + 1 and the updated covariance matrix pol P k+1|k+1 . The specific formula is:
[0106]
[0107] where pol X k+1 represents the sigma points generated in the construction space, while pol X k+1 (i) represents the i-th sigma point.
[0108] Convert each sigma point in the constructed state space to the Cartesian coordinate system using the transformation formula, and calculate the mean and covariance matrix of the optimal state estimate in the Cartesian coordinate system by weighted calculation. The specific formula is:
[0109]
[0110]
[0111] where g is the non-linear transformation function from the state space to the Cartesian coordinate system. The specific transformation method is shown in the following formula:
[0112] R = r1 + r2(26)
[0113]
[0114] In the above formula, L is the distance between the receiving radar and the transmitting radar, and R is the sum of the distances from the moving target to the transmitting radar and the receiving radar.
[0115] S6: Loop through steps S2 to S5 until the tracking ends.
[0116] Furthermore, in the present invention, the target motion state in steps S3 to S5 based on the constructed state space is as follows:
[0117]
[0118] where r1(k) represents the distance between the receiving radar and the moving target at the k-th moment, represents the radial velocity of the moving target relative to the receiving radar at the k-th moment, r2(k) represents the distance between the transmitting radar and the moving target at the k-th moment, v t (k) represents the tangential velocity of the moving target relative to the receiving radar at the k-th moment, and θ(k) represents the angle of the moving target relative to the receiving radar at the k-th moment. These numerical values represent the motion state of the target in the constructed state space.
[0119] As a specific implementation manner, an embodiment as shown in Figures 1-4 is provided.
[0120] Figure 1 is a flowchart of a coordinate system transformation fusion filtering tracking method for a dual-base radar according to the present invention.
[0121] Figure 2 is a schematic diagram of constructing a state space based on the dual-base radar observation data and the target motion state according to the present invention. The essence of the target motion is independently determined by its own dynamic characteristics, rather than being restricted by the observation parameters describing the motion state. The definitions of various physical quantities are essentially mathematical representation means for multi-dimensional observation of motion characteristics. A state vector model in the Cartesian coordinate system can be used for modeling, or an observation model can be constructed in the measurement space. However, different modeling dimensions will form different information structures. As shown in Figure 2 , this patent innovatively uses the dual-base radar to comprehensively measure the distance and angle information to locate the target, and combines the tangential and normal velocity components to construct a velocity model, which not only realizes the complete characterization of motion characteristics, but also transforms the dual-base radar observation equation into a linear function relationship. This linear mapping characteristic enables the filtering process to be completed in an ideal Gaussian linear space, thus significantly improving the tracking accuracy and stability. The core breakthrough of this solution lies in constructing a general dual-base radar state space modeling paradigm, and its tracking performance shows significant advantages compared with traditional methods, providing a new technical path for the state estimation of multi-source heterogeneous observation systems.
[0122] Figure 3The RMSE of various method target position estimations is shown. It can be seen from the comparison in the figure that the method proposed in this patent has a fast convergence speed and high accuracy. Because other methods perform filtering in the Cartesian coordinate system, the probability density will be distorted during the filtering process, resulting in information loss. However, in the present invention, both prediction and update are carried out in the adaptively constructed state space. Its filtering structure is not only linear but also ensures the Gaussianity of the state probability density. Finally, the Kalman filter is used to fuse the state vector and the observed vector values, thereby ensuring the convergence and stability of the dynamic estimation.
[0123] Figure 4 The RMSE of various method target speed estimations is shown. It can be clearly seen from the figure that the method proposed in this patent has always maintained a performance advantage in speed estimation compared with the other three methods. The reason is still that a more reasonable filtering space is constructed compared with the Cartesian coordinate system.
[0124] Figure 3 and Figure 4 are the comparison graphs of the present invention with the three internationally recognized methods, namely the Extended Kalman Filter (EKF), the Unscented Kalman Filter (UKF), and the Sigma-Point Transform Kalman Filter (SPTKF), regarding the Root Mean Square Error (RMSE) of position and the Root Mean Square Error (RMSE) of speed under the embodiments of the present invention. For convenience, the meanings of the symbols and terms not explained in this patent are summarized in Table 1.
[0125] Table 1 Explanation of Terms and Symbols
[0126]
[0127] Based on the above basic description, consider a tracking system in a typical dual-base radar where the receiving radar is fixed at the origin. At each sampling moment, the radar can obtain the distance and angle of the object being tracked. The observation noise distance error σ of the dual-base radar r = 2m, and the angle error σ θ = 1.5 deg. Consider the scenario where the radar tracks an airborne target. The initial positions of the target are all at (10 km, 10 km). The radar sampling period is T = 1 second, the simulation duration is 100 seconds, and the initial speeds of the target are all (10 m / s, 18 m / s). Assume that the process noise is Gaussian white noise with zero mean and a standard deviation of 0.1 m / s. The simulation compares the Root Mean Square Error (RMSE) of the target position and speed estimations of several currently most effective methods with the method proposed in the present invention. The smaller the Root Mean Square Error, the higher the tracking accuracy. The following embodiments all perform 50 Monte Carlo simulations.
[0128] Therefore, the present invention has the following beneficial effects:
[0129] The method of coordinate transformation fusion filtering tracking for its dual-base radar is different from all existing dual-base radar tracking fusion algorithms. This theoretical technology is different from the orthogonal decomposition motion description in the Cartesian coordinate system. Its technical route is to construct a state filtering fusion space based on the motion state of the target in the Cartesian coordinate system, deduce and obtain a linear state equation, finally use the Kalman filter to complete the tracking, and transform the tracked data to the Cartesian coordinate system through unscented transformation and output. The technical effect is to solve the non-linear problem in filtering, ensure that the radar observation and state are not distorted due to various error effects during the information fusion process, and enable the target state and radar observation information to be more effectively fused. The technical purpose is that, compared with the currently recognized and latest methods, this patent improves the tracking robustness and accuracy without increasing the computational complexity.
[0130] Currently, a large number of experiments show that the position and velocity tracking accuracies obtained by the method provided by the present invention are better than those of the currently internationally recognized most effective extended Kalman filter (EKF), unscented Kalman filter (UKF), and sigma-point transformation Kalman filter (SPTKF) methods. Especially in the medium and long-distance tracking scenarios, the use of this technology can significantly improve the tracking effect and have high stability without the need to invest a large amount of money to upgrade the hardware performance.
[0131] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.
[0132] Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the core idea of the present invention. At the same time, for those of ordinary skill in the art, there will be changes in the specific implementation manner and application scope according to the idea of the present invention. In summary, the content of this specification should not be construed as a limitation of the present invention.
Claims
1. A coordinate system transformation fusion filtering tracking method for dual-base station radar, characterized in that: include: Based on the kinematic characteristics of the target, the state equation and observation equation of the dual-base radar tracking system for the moving target are established, and the initial motion state and initial covariance of the target are determined based on the prior information of the target in Cartesian coordinates; One-step prediction is performed based on the motion state and covariance of the target at time k in the Cartesian state space to obtain the motion state of the target at time k+1; the motion state of the target at time k+1 includes the target prediction state and the prediction covariance matrix; Based on the U transformation, a set of sigma points are generated for the target prediction state and prediction covariance matrix, and 2n+1 sigma points are generated for the n-dimensional motion state of the target in the Cartesian coordinate system, and the mean and covariance of the constructed state space prediction are weightedly calculated; The measured data and the predicted data are fused by using a Kalman filter to obtain the optimal state estimation and covariance matrix at the k+1th moment after fusion; the measured data are obtained by radar measurement; the predicted data are the mean and covariance of the constructed state space prediction; Based on the U transformation, 2n+1 sigma points are generated according to the updated optimal state estimate at time k+1 and the updated covariance matrix. Each sigma point in the constructed state space is transformed to the Cartesian coordinate system using the transformation formula, and the optimal state estimate mean and covariance matrix in the Cartesian coordinate system are obtained by weighted calculation. The steps are repeated until the tracking is completed.
2. The coordinate system transformation fusion filtering tracking method for dual-base station radar according to claim 1 is characterized in that: The method of establishing the state equation and observation equation of the dual-base station radar tracking system for the moving target based on the kinematic characteristics of the target, and determining the initial motion state and initial covariance of the target based on the prior information of the target in Cartesian coordinates, specifically includes: Based on the kinematic characteristics of the target, the state equation and observation equation of the dual-base station radar tracking system for the moving target are established, and the formulas are: X(k)=F(k)·X(k-1)+G(k)·w(k) Z(k)=h(X(k))+v(k) Among them, X z (k) = [x(k) x′(k) y(k) y′(k)] is a state vector constructed directly from the receiving radar observations. This vector is used to represent the motion state of the target at the kth moment, that is, this vector is substituted into the state equation of the target at the kth moment, namely, X(k). x(k), x′(k), y(k), y′(k) are the x-direction distance, x-direction velocity, y-direction distance and y-direction velocity of the target relative to the receiving radar at the kth moment, respectively. F(k), G(k), w(k) and h are the state transfer matrix, noise coefficient matrix, process noise and observation function, respectively. Among them, w(k) = [w x (k) w y (k)] T , w x (k) and w y (k) are the process noise in the x-direction and y-direction, respectively, z (k)=[R(k) θ(k)] T is the observation vector of the receiving radar under noise interference at the kth moment. This vector is used to represent the target state observed at the kth moment, that is, this vector is substituted into the target observation equation at the kth moment, namely, Z(k), where R(k) is the sum of the distances between the transmitting radar and the receiving radar and the moving target at the kth moment, θ(k) is the angle between the receiving radar and the moving target at the kth moment, and v(k) = [v r (k) v α (k)] Τ is the observation noise of the receiving radar at the kth moment, v r (k) is the measured distance noise at the kth moment, v α (k) The measured angle noise at the kth moment is Gaussian white noise with zero mean, v r (k) and v α (k) is uncorrelated, and the noise covariance matrix is The specific formula of the observation function h is: Among them, Z true (k) is the real observation state without noise at the kth moment, and L is the distance between the receiving radar and the transmitting radar; For a moving target, the initial motion state X(1) and initial covariance P(1) of the target are obtained based on the prior information of the target in Cartesian coordinates.
3. The coordinate system transformation fusion filtering tracking method for dual-base station radar according to claim 1 is characterized in that: The motion state of the target at time k+1 is specifically expressed as: X k+1|k =F(k)X(k) P k+1|k =F(k)P(k)F(k) Τ +G(k)Q(k)G(k) Τ Among them, X k+1|k , P k+1|k are the predicted state and predicted covariance matrix at time k respectively, and Q(k) represents the covariance matrix of process noise.
4. The coordinate system transformation fusion filtering tracking method for dual-base station radar according to claim 1 is characterized in that: Based on the U transformation, a set of sigma points are generated for the target prediction state and prediction covariance matrix, and 2n+1 sigma points are generated for the n-dimensional motion state of the target in the Cartesian coordinate system, and the mean and covariance of the constructed state space prediction are calculated by weighted calculation, specifically including: Based on U transformation, the target prediction state X at the current time k k+1|k and the prediction covariance matrix P k+1|k Generate a set of sigma points, and generate 2n+1 sigma points for its n-dimensional motion state in the Cartesian coordinate system. The specific formula is: x k (0) =X k+1|k where i = 1, 2, ..., n, λ = α 2 (n+κ)-n, α and κ are scaling parameters used to adjust the distribution of σ points; x k represents the sigma point generated in the Cartesian coordinate system, and x k (i) Representing the i-th sigma point, the formula for the weight of each sigma point is expressed as follows: Where β is the parameter used to fuse the observation noise, W m and W c are the weights of the mean and covariance, respectively, both are 2n+1 vectors; Each sigma point in the Cartesian coordinate system is converted to the constructed state space using the conversion formula. The constructed state space is specifically based on the position state representation and velocity state representation of the target in the two-dimensional polar coordinate system. The distance r1 between the target and the receiving radar and the tangential velocity v of the target are r1 , the distance between the target and the transmitting radar r2, the normal velocity of the target v t The angle θ formed by the target and the receiving radar is constructed as a vector describing the target state and is expressed as The weighted calculation is used to obtain the mean and covariance of the construction state space prediction. The specific calculation formula is: Among them, W m (i) Represents the mean weight of the i-th sigma point, W c (i) represents the covariance weight of the i-th sigma point, f is the nonlinear conversion function from the Cartesian coordinate system to the constructed state space, and the specific conversion method is shown in the following formula: v r1 =v x cosθ+v y sinθ v t =-v x sinθ+v y cosθ Where L is the distance between the receiving radar and the transmitting radar, x is the x-direction distance of the target relative to the receiving radar, and v x is the x-direction speed of the target relative to the receiving radar, y is the y-direction distance of the target relative to the receiving radar, v y is the y-direction velocity of the target relative to the receiving radar.
5. The coordinate system transformation fusion filtering tracking method for dual-base station radar according to claim 1, characterized in that: The optimal state estimate and covariance matrix at the k+1th moment are expressed as: K k+1 = pol P k+1|k T H T (H pol P k+1|k H T +R k ) -1 pol P k+1|k+1 =(I-K k+1 H) pol P k+1|k Among them, K k+1 is the Kalman gain at time k+1, is the optimal state estimate after updating at time k+1, pol P k+1|k+1 is the updated covariance matrix, I is the identity matrix, H is the observation matrix of the receiving radar observation target, z k+1 is the observation vector observed by the receiving radar at time k+1.
6. The coordinate system transformation fusion filtering tracking method for dual-base station radar according to claim 1, characterized in that: The U-transformation method generates 2n+1 sigma points based on the optimal state estimate updated at time k+1 and the updated covariance matrix, transforms each sigma point in the constructed state space to the Cartesian coordinate system using a transformation formula, and obtains the optimal state estimate mean and covariance matrix in the Cartesian coordinate system by weighted calculation, specifically including: Based on the U transform, the optimal state estimate after updating at time k+1 is With the updated covariance matrix pol P k+1|k+1 Generate 2n+1 sigma points. The specific formula is: in, pol X k+1 is the sigma point generated in the construction space, pol X k+1 (i) is the i-th sigma point; Each sigma point in the constructed state space is transformed into the Cartesian coordinate system using the conversion formula, and the optimal state estimation mean and covariance matrix in the Cartesian coordinate system are obtained by weighted calculation. The specific formula is: Among them, g is the nonlinear transformation function from the state space to the Cartesian coordinate system. The specific transformation method is shown in the following formula: R=r1+r2 Wherein, L is the distance between the receiving radar and the transmitting radar, and R is the sum of the distances of the moving target from the transmitting radar to the receiving radar.
7. The coordinate system transformation fusion filtering tracking method for dual-base station radar according to claim 1, characterized in that: The target motion state in the construction state space is expressed as: Among them, r1(k) represents the distance between the receiving radar and the moving target at the kth moment, represents the radial velocity of the moving target relative to the receiving radar at the kth moment, r2(k) represents the distance between the transmitting radar and the moving target at the kth moment, and v t (k) represents the tangential velocity of the moving target relative to the receiving radar at the kth moment, and θ(k) represents the angle of the moving target relative to the receiving radar at the kth moment.
8. A coordinate system transformation fusion filtering tracking system for dual-base station radar, characterized in that: include: The initial state building module is used to establish the state equation and observation equation of the dual-base station radar tracking system for the moving target based on the kinematic characteristics of the target, and determine the initial motion state and initial covariance of the target based on the prior information of the target in Cartesian coordinates; A one-step prediction module is used to perform one-step prediction based on the motion state and covariance of the target at time k in the Cartesian state space to obtain the motion state of the target at time k+1; the motion state of the target at time k+1 includes the target prediction state and the prediction covariance matrix; The first U transformation module is used to generate a set of sigma points for the target prediction state and the prediction covariance matrix based on the U transformation, and to generate 2n+1 sigma points for the n-dimensional motion state of the target in the Cartesian coordinate system, and to weightedly calculate the mean and covariance of the constructed state space prediction; A fusion module is used to fuse the measured data with the predicted data using a Kalman filter to obtain the optimal state estimation and covariance matrix at the k+1th moment after fusion; the measured data is obtained by radar measurement; the predicted data is the mean and covariance of the constructed state space prediction; The second U transformation module is used to generate 2n+1 sigma points based on the U transformation according to the optimal state estimate updated at time k+1 and the updated covariance matrix, transform each sigma point in the constructed state space to the Cartesian coordinate system using the transformation formula, and obtain the optimal state estimate mean and covariance matrix in the Cartesian coordinate system by weighted calculation; The loop module is used to loop the steps until the tracking ends.
Citation Information
Patent Citations
Multi-sensor multi-target cooperative detection information fusion method and system
CN111860589A
State transformation fusion filtering tracking method for three-coordinate radar
CN116047495A
Low-altitude unmanned aerial vehicle multi-target reflection source separation method based on 5G network
CN116430345A
Pre-observation overall fusion tracking method for Doppler measurement
CN117784064A
Space-based networking radar multi-coverage fusion target tracking method and system
CN119511223A
Cited By
Space target multi-radar handover tracking method based on forecast error covariance transmission
CN121385866A
Multi-station cooperative ballistic estimation method, system, device and storage medium
CN122506541A
