CN-115758092-B - Self-adaptive radar target tracking method based on current statistical model

Abstract

The invention belongs to the technical field of radar target tracking and discloses a self-adaptive radar target tracking method based on a current statistical model, which comprises the following steps that 1, a filter is initialized; step 2, one-step filtering prediction based on CS model, step 3, target maneuver judgment, step 4, self-adaptive adjustment of model parameters and filter parameters, and step 5, filtering state update. The invention ensures the filtering precision of the non-maneuvering target state, realizes the greater utilization of the current innovation, and accelerates the speed of the filter for adjusting the maneuvering state.

Inventors

• Yan Jiangzehui
• HAN XING
• ZHANG HAOQIANG
• CHEN CHUN
• SONG WEIQIANG
• WANG YANSONG

Assignees

• 中国北方工业有限公司
• 西安电子工程研究所

Dates

Publication Date

20260508

Application Date

20221125

Claims (1)

1. 1. An adaptive radar target tracking method based on a current statistical model is characterized by comprising the following steps: step 1, initializing a filter; Step 2, one-step filtering prediction based on CS model; step 3, judging the target maneuver; step 4, self-adaptive adjustment of model parameters and filter parameters; step 5, updating a filtering state; In step 1, the following parameters are initialized before starting the filtering, i.e. at time k=1: (1) Initial maneuver frequency: α(1)=α 0 (1) wherein alpha 0 is the initial maneuvering frequency and is empirically valued; (2) Initial maximum and minimum acceleration a Min0 ,a Max0 , and taking an empirical value; (3) Initial state transition matrix F: wherein T is the sampling interval; (4) Initial input control matrix G: (5) Initial filter weighting factor u (1): u(1)=u 0 =1 (4) (6) The initial maneuver detection threshold D is obtained by inquiring a χ 2 distribution table; in step 2, the one-step filtering prediction process based on the CS model is as follows: Let the tracking filter state estimate at time k-1 be The covariance matrix is P (k-1|k-1), the measurement at k moment is Z (k), and the filtering process at k moment is as follows: (1) Performing one-step target state prediction at the moment k; (2) Calculating the current acceleration variance; (3) Calculating a process noise covariance; (4) Carrying out one-step prediction of a noise covariance matrix; (5) Calculating a innovation covariance; In step2, one-step target state prediction at time k is performed according to equation (5): Wherein, the The mean value of the maneuvering acceleration is equal to the acceleration filtering value at the moment k-1; in step 2, the current acceleration variance formula at the time k is calculated as follows: The calculation process noise covariance formula is: wherein, alpha (k) is the maneuver frequency at k moment, and each element in the right matrix of the equal sign is as follows: (4) Performing one-step prediction of the noise covariance matrix according to equation (9): P(k|k-1)=u(k-1)F(k)P(k-1|k-1)+Q(k) (9) wherein u (k-1) is a filter weighting primer at time k-1; (5) Calculating the innovation covariance: S(k)=H(k)P(k|k-1)H T (k)+R(k) (10) Wherein R (k) is a measurement noise covariance matrix; in step 3, the target maneuver determination process is: calculating a target maneuver degree d (k) and performing maneuver judgment: d(k)=v T (k)S -1 (k)v(k) (11) wherein v (k) is the innovation at time k: D (k) obeys the χ 2 distribution of the n-dimensional degrees of freedom, and when the target maneuvers, the D (k) value increases, and if D is set as a maneuver judgment threshold, the method comprises the following steps: Pr{d(k)≥D}=P false (13) P false is the maneuver judgment false alarm rate, a threshold D value meeting the system performance requirement is obtained by inquiring a χ 2 distribution table, and if D (k) is less than D, no maneuver or weak maneuver of the target is judged; in step 4, the adaptive adjustment process of the model parameters and the filter parameters under the target maneuvering condition is as follows: (1) Adjusting CS model maneuver frequency and acceleration extremum according to maneuver degree: α maneu (k)＝d(k)α 0 (14) (2) By the principle of orthogonality, the filter weighting factor u (k) is adjusted according to the degree of maneuver: N(k)=V(k)-H(k)Q(k)H T (k)+R(k) (18) L(k)=H(k)F(k)P(k|k-1)F T (k)H'(k) (19) Wherein, the Ρ is a forgetting factor, taking a value between 0 and 1; (3) Updating the noise covariance using the adaptively adjusted maneuver frequency: Each element in the matrix is correspondingly updated in a maneuvering way according to the formula (8); (4) Updating the maneuver acceleration variance using the adaptively adjusted acceleration extremum: (5) Updating state prediction covariance P (k|k-1) P(k|k-1)=u(k)F(k)P(k-1|k-1)+Q(k) (22) In step 4, the adaptive adjustment process of the model parameters and the filter parameters under the condition of no maneuver or weak maneuver of the target is as follows: If the k-1 moment judges that no maneuver exists, the step 5 is directly executed; if the maneuver is determined at time k-1, the initial parameters are used to calculate the predictive covariance matrix P (k|k-1) for the maneuver-free state: P(k|k-1)=u 0 F(k)P(k-1|k-1)+Q(k) (23) in step 5, the filtering state updating content includes: Calculating an innovation covariance matrix S (k): S(k)=H(k)P(k|k-1)H T (k)+R(k) (25) calculating a filter gain K (K): K(k)=P(k|k-1)H T (k)S -1 (k) (27) Updating a target filtering state: X(k|k)=X(k-1|k)+K(k)v(k) (29) Covariance matrix update: P(k|k)=P(k-1|k)+K(k)S -1 (k)K T (k) (31) At this point, the filtering step is completed, the target tracking filtering state X (k|k) at the k moment and relevant filtering parameters are obtained, and the target tracking filtering at the k+1 moment is continued.

Description

Self-adaptive radar target tracking method based on current statistical model Technical Field The invention belongs to the technical field of radar target tracking, and relates to a self-adaptive radar target tracking method based on a current statistical model, which uses a current statistical (CS, current Statistical) model radar tracking target to carry out tracking filtering, when a target maneuver is detected, the CS model parameter and the filter parameter are self-adaptively adjusted, the problems of tracking loss and filtering divergence caused by mismatch of the radar tracking model when the target maneuver is large are solved, and the tracking capability of a radar tracking system on the maneuver target is improved. Background Radar target tracking is a process of filtering and estimating the target state according to the measurement data of the radar at the current moment and the filtering result of the past moment. The Kalman filter is an optimal linear mean square filter, is widely used in radar tracking systems, and combines the past filtering result and a target motion model to obtain a preliminary prediction of a target state at the next moment, obtains an error of the preliminary prediction according to the current measurement, and corrects and updates the preliminary state prediction to obtain a final filtering result. The object motion model is an important precondition for performing kalman filtering, but an accurate motion model is difficult to obtain, so an approximation model is generally established to approximately describe the object motion process. Constant velocity (CV, constant Velocity) models and uniform acceleration (CA, constant Acceleration) models are common classical models, wherein the CV model models assume that the target moves linearly at a constant velocity, and a small change in velocity is regarded as process white noise for modeling, and the CA model assumes that the target moves uniformly and accelerated, and a small change in acceleration is regarded as process white noise. CV and CA models are accurate in describing the motion of a non-maneuvering or weak maneuvering target, a Kalman filter based on the two models can obtain a good filtering result, but model mismatch occurs when the target moves more, a larger model error is generated, filtering precision is reduced, even filtering divergence is caused, and a radar tracking system loses the target. To solve the problem of model mismatch when targets maneuver, a series of maneuver target models, such as Singer models and "current" statistics (CS) models, have emerged that adapt more to maneuver. Unlike the CV and CA models which simulate the target maneuver using Gaussian white noise, the Singer model assumes that the target acceleration is a zero-mean random process that is approximately uniformly distributed over a range, while the CS model improves on the basis of the Singer model, assuming that the acceleration at the next moment will only take a value in a certain neighborhood of the acceleration at the current moment, and describes the target maneuver acceleration using a modified Rayleigh distribution. The CS model assumes non-zero mean value of acceleration, and can correct the mean value and distribution of the acceleration in real time according to measurement, so that the maneuvering state of the target can be described more truly. Although modeling of the maneuver by the CS model is more practical, the Kalman filter based on the CS model still has the following defects: the cs model requires a priori information, i.e., the acceleration extremum and maneuver frequency parameters of the target are assumed in advance. The prior parameter is usually an empirical value and is fixed, if the value is too small, model mismatch can occur when the target is large maneuver, so that the problems of filtering divergence and the like are caused, and if the value is too large, the tracking precision is lower in the non-maneuver or weak maneuver stage of the target, and the performance is reduced. 2. The Kalman filter takes a long time to adjust the burst maneuver, and the tracking performance is poor. Disclosure of Invention Object of the invention Aiming at the problems and the defects, the invention provides a self-adaptive radar target tracking method based on a current statistical model, which aims to solve the problem that the tracking precision of a traditional CS model Kalman filter on the complex maneuvering condition of a target is low or lost, keeps initial parameters for high-precision filtering tracking when the target is not maneuvering or is in weak maneuvering, and self-adaptively adjusts CS model parameters and filter gain according to maneuvering degree when the target is in great maneuvering, thereby avoiding the problem of model mismatch caused by poor value of a priori parameters of a model, and simultaneously enhancing the real-time response capability of the Kalman filter so as to adapt to th