CN104199022A

CN104199022A - Target modal estimation based near-space hypersonic velocity target tracking method

Info

Publication number: CN104199022A
Application number: CN201410439348.0A
Authority: CN
Inventors: 易伟; 董天发; 苟清松; 郝凯利; 崔国龙; 孔令讲; 杨建宇; 李小龙; 夏玫
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2014-08-30
Filing date: 2014-08-30
Publication date: 2014-12-10
Anticipated expiration: 2034-08-30
Also published as: CN104199022B

Abstract

The invention provides a target modal estimation based near-space hypersonic velocity target tracking method, which comprises the steps of: tracking a target by utilizing an interactive multi-model tracking algorithm, estimating a target movement mode dynamically and in real time, judging the target movement mode by counting target characteristics, and finally turning to corresponding single-mode matched tracking based on the estimated target movement mode. Competition among multiple models is avoided, and the problems of low tracking precision and the like caused by complex calculation and great model competition of an existing near-space hypersonic velocity tracking algorithm are solved. The target modal estimation based near-space hypersonic velocity target tracking method is low in calculation amount and high in tracking precision, and effectively improves the integral performance of the tracking system.

Description

Near space hypersonic target tracking method based on target modal estimation

Technical Field

The invention belongs to a radar signal processing technology, and particularly relates to a technology for tracking a hypersonic typical moving target in an adjacent space.

Background

The near space refers to an airspace with the height of 20-100 Km from the ground, and the speed of the near space hypersonic aerocraft can reach Mach 4-20. Currently, near space hypersonic target tracking is a hot spot in the tracking field. The complexity of the movement of the maneuvering target in the adjacent space and the variability of the operation environment cause that an accurate tracking model is difficult to establish for the targets, and the development of the stealth technology further increases the difficulty of tracking the hypersonic target in the adjacent space.

At present, the tracking algorithm based on an interactive multi-model (IMM) structure is a tracking algorithm which is recognized as the most effective near space hypersonic speed target. In the documents "Research of Method for Tracking High Speed and High throughput target, International Conference on ITS Telecommunications projects, 1236-1239, 2006", a Tracking algorithm based on an interactive multimode structure is proposed, which utilizes a plurality of different maneuvering models to interactively track a hypersonic target in an adjacent space, and has a wider coverage range and greater maneuvering adaptability compared with single mode Tracking; however, the mutual competition among the algorithm models results in poor tracking accuracy and large calculation amount.

Disclosure of Invention

The invention aims to solve the technical problem of providing a method for tracking a hypersonic target in a near space, which has high tracking precision and small calculation amount.

The invention adopts the technical scheme that a near space hypersonic target tracking method based on target mode estimation comprises the following steps:

step 1, carrying out target track initiation on measurement data acquired by a radar;

step 2, estimating the state of a next frame target, a state covariance matrix and corresponding model transition probabilities corresponding to a uniform motion model, a uniform accelerated motion model and a turning motion model by utilizing an interactive multimode tracking IMM algorithm; the target state comprises a target position, a speed and an acceleration; when the transition probability, the target speed and the acceleration corresponding to the uniform motion model, the uniform acceleration motion model and the turning motion model under the L frames are continuously counted, the current motion mode of the target can be determined, the step 3 is carried out, and otherwise, the step 2 is returned;

if in the time of L frames, the transition probability u of the uniform velocity model_cvAlways keeping maximum speed variation quantity delta v_k≤τ_vAcceleration change amount Δ a_k≤τ_aJudging that the target is in a uniform motion stage;

if in the L frame time, the transition probability u of the uniform acceleration model_caAlways keeping maximum speed variation quantity delta v_k≥τ_vAcceleration change amount Δ a_k≤τ_aJudging that the target is in a uniform motion stage;

if in the L frame time, the transition probability u of the uniform acceleration model_ctAlways keeping maximum speed variation quantity delta v_k≥τ_vAcceleration Δ a_k≥τ_aJudging that the target is in a jumping motion stage;

wherein, tau_vRepresenting the speed fluctuation threshold, tau_aRepresenting an acceleration fluctuation threshold;

step 3, single-mode matching tracking:

3-1, after determining a target motion mode, switching to single-mode matching tracking; initializing single mode parameters and a target initial state;

3-2, estimating the target state of the next frame by utilizing a single-mode matching tracking algorithm;

3-3 calculating the normalized residual squared epsilon for the current time k_vComprises the following steps:v_xfor the current measurement residual, S_xRepresenting the current residual covariance matrix, the subscript x representing the motion model matching the true motion of the target, i.e., x ∈ { CV, CA, CT }, CV representing uniform motion, CA representing uniform acceleration motion, CT representing jerking motion, (·)^TRepresenting a matrix transposition; normalized residual squared epsilon of Angelica_vWhen the target motion state is less than or equal to the threshold tau, the target motion state is considered to be unchanged, the step 3-2 is returned, and the normalized residual error squared epsilon is obtained_vAnd when the motion state of the target is larger than the threshold tau, the motion state of the target is considered to be changed, and the step 2 is returned.

The method comprises the steps of tracking a target by utilizing an interactive multi-model tracking algorithm, dynamically estimating a target motion mode in real time, and judging the target motion mode by counting target characteristics; and finally, switching to corresponding single-mode matching tracking according to the estimated target motion mode, avoiding competition among multiple models, and solving the problems of complex calculation, high model competition, low tracking precision and the like of the conventional near space hypersonic tracking algorithm.

The invention has the advantages of small calculation amount and high tracking precision, and effectively improves the overall performance of the tracking system.

Drawings

FIG. 1 is a flow chart of a tracking algorithm based on interactive multiple models and dynamic estimation of target motion modalities.

FIG. 2 is a graph of tracking error for the new tracking algorithm and the interactive multi-mode tracking algorithm.

FIG. 3 shows that the new tracking algorithm and the interactive multi-mode tracking algorithm track success probability curves under different detection probabilities.

FIG. 4 shows that the new tracking algorithm and the interactive multi-mode tracking algorithm track success probability curves at different sampling rates.

Detailed Description

The effect of the invention is verified by adopting a computer simulation method, and the implementation steps are as shown in fig. 1 through MATLAB-R2010 b:

step 1, radar data conversion

The tracked target enters a ground radar scanning area, and data collected by the radar is measured dataAnd Doppler information (Doppler) f of the target_dR represents a target distance, θ represents a target azimuth,Representing the target pitch angle. Measuring data of polar coordinate system by using correction measurement deflection-free conversion MUCMKFConverting the measured data into a rectangular coordinate system to obtain converted measured dataWhereinRespectively representing the target positions measured in the rectangular coordinate system after conversion;

wherein the compensation factorMean square error, θ, representing the azimuth and pitch angles, respectively_m,ε_mRespectively, azimuth and pitch.

Step 2, target track initiation

2.1 store the measurement data after 3 frames conversion, and the k-th time data z (k) ═ z₁(k),z₂(k),z₃(k),...z_m1(k) The data set obtained at the k +1 th time is Z (k +1) ═ Z₁(k+1),z₂(k+1),z₃(k+1),...z_m2(k +1) }, and the data set obtained at the k +2 th time is Z (k +2) ═ Z₁(k+2),z₂(k+2),z₃(k+2),...z_m3(k +2) }, traversing and associating continuous three-frame data, and utilizing height constraint h_min<h<h_maxAnd velocity constraint v_min<v<v_maxAnd Doppler information f_dAnd removing partial clutter traces. Wherein z is_i(k) I-th measurement at the k-th time, m1, m2, m3 respectively represent the number of measurements at 3 times, h_min,h_maxRespectively representing the minimum and maximum flying heights, v_min,v_maxMinimum and maximum flight speeds.

2.2 projecting the rest continuous 3 frames of data to two planes (x-z and y-z planes), and transforming the two planes of data to a parameter space, wherein rho is xcos theta + ysin theta, wherein (theta, rho) are coordinates in the parameter space, and (x, y) are observation data in a rectangular coordinate system; and respectively carrying out track initiation on the two-plane measurement data by using modified Hough transform, and comparing and associating the two-plane initiated tracks so as to determine the target track initiation. In order to suppress false tracks as much as possible, a modified Hough transform is preferably used here. Other track initiation methods in the art, such as direct track initiation, correction logic, for target track initiation in the near space, can result in an increase in false tracks, which is not conducive to target tracking.

Step 3, tracking the target by utilizing an interactive multimode tracking IMM algorithm

3.1 after the target track is initiated, the target state is initialized first

Each model transition probability u ═ u_cv u_ca u_ct]And each model state covariance matrix P_cv,P_ca,P_ct. Assume that the measurement values at the first 3 times are Z (1), Z (2), Z (3), and Z (i) ([ x (i), y (i), Z (i))]^TObserving a noise covariance matrix R;

wherein R is directly obtained through a measurement conversion equation;

initializing a system state:

X = [\begin{matrix} x (3) \\ (x (3) - x (2)) / T \\ ((x (3) - x (2)) / T - (x (2) - x (1)) / T) / T \\ y (3) \\ (y (3) - y (2)) / T \\ ((y (3) - y (2)) / T - (y (2) - (1)) / T) / T \\ z (3) \\ (z (3) - z (2)) / T \\ ((z (3) - z (2)) / T - (z (2) - z (1))) \end{matrix}]

wherein,

representing position, velocity and acceleration on the x, y and z axes, respectively; [ u ] of_cv,u_ca,u_ct],[P_cv,P_ca,P_ct]Are respectively at uniform speedModel probabilities and state covariance matrices corresponding to the model, the uniform acceleration model, and the cornering model, and u_cv+u_ca+u_ct＝1；

3.2 predicting the target states of the next frame corresponding to the uniform velocity motion CV model, the uniform acceleration motion CA model and the turning motion CT model respectively by utilizing an interactive multi-model tracking algorithmThe corresponding state covariance matrix is

3.3 estimating the target states corresponding to the models by using the predicted valuesTarget state covariance matrixAnd updating transition probabilities for each model

Updating the states of the models:

updating covariance of each model state:

estimating the transition probability of each model:

wherein N represents the number of interaction models, u_i(K-1) represents the transition probability of the model i in the previous frame, p is the known model transition probability matrix, K_i、v_iAnd S_iRespectively representing a gain matrix, a residual error matrix and a residual error covariance matrix of a Kalman filter corresponding to the model i, wherein a subscript i belongs to { CV, CA, CT };

3.4 transition probability according to model

u = [\begin{matrix} {\hat{u}}_{cv} & {\hat{u}}_{ca} & {\hat{u}}_{ct} \end{matrix}]

And the estimated target states of the modelsAnd interactively outputting the estimated target state X.

Step 4, estimating the motion mode of the target

When the target is tracked by using the interactive multi-mode tracking algorithm, the model transition probability and the target state at the k-th moment are assumed to be u (k) ═ u respectively_cv(k) u_ca(k) u_ct(k)]And

velocity and acceleration are respectively v_k,a_k(ii) a Counting the transition probability u, the target speed and the acceleration information of the continuous L-frame model, counting the motion characteristics of the target in the L-frame time, and judging the current motion mode of the target:

the target motion mode identification method comprises the following steps:

if in the time of L frames, the probability u of the uniform model_cvAlways keeping maximum speed variation quantity delta v_k≤τ_vAcceleration change amount Δ a_k≤τ_aAnd judging that the target is in a uniform motion stage.

If in the L frame time, the probability u of the uniform acceleration model_caAlways keeping maximum speed variation quantity delta v_k≥τ_vAcceleration change amount Δ a_k≤τ_aAnd judging that the target is in a uniform motion stage.

If in the L frame time, the probability u of the uniform acceleration model_ctAlways keeping maximum speed variation quantity delta v_k≥τ_vAcceleration Δ a_k≥τ_aThen, the target is judged to be in the jumping motion stage.

Wherein, tau_vRepresenting the speed fluctuation threshold, tau_aRepresenting an acceleration fluctuation threshold.

Step 5, single-mode matching tracking algorithm

5.1, after determining a target motion mode, switching to single-mode matching tracking; initializing single mode parameters, target initial state

And the object transition probability matrix P_o。

5.2 predicting the next frame state of the target by utilizing a single-mode matching tracking algorithmAnd corresponding error covariance matrix

Where the subscript x denotes the motion model that matches the true motion of the target, i.e., x ∈ { CV, CA, CT }. CV represents uniform motion; CA represents uniform acceleration motion; CT denotes curve motion.

5.3 estimating the target state and the corresponding state covariance matrix by using the predicted values:

wherein, K_xIs the Kalman filter gain, v_xTo measure residual errors, S_xRepresenting the residual covariance matrix.

5.4, residual errors and residual error covariance of the estimated state of the target at each moment are evaluated, and the normalized residual error square at the k moment is as follows:

wherein epsilon_vObeying with a degree of freedom of n_zChi of²Distribution, n_zRepresenting the measurement dimension.

When epsilon_vAnd when the threshold tau is exceeded, the motion state of the target is considered to be changed, and the step 3 is returned.

The simulation results are shown in fig. 2, fig. 3 and fig. 4, compared with the tracking of the existing IMM algorithm, the tracking algorithm of the invention has the advantages of smaller calculated amount, higher tracking precision, higher tracking success probability under different detection probabilities and higher tracking success probability under different sampling rates, thereby improving the overall tracking performance.

Claims

1. A near space hypersonic target tracking method based on target mode estimation is characterized by comprising the following steps:

step 2, estimating model transition probabilities and target states of a next frame corresponding to the uniform motion model, the uniform accelerated motion model and the turning motion model by utilizing an interactive multimode tracking IMM algorithm; the target state comprises a target position, a speed and an acceleration; when the transition probability, the target speed and the acceleration corresponding to the uniform motion model, the uniform acceleration motion model and the turning motion model under the L frames are continuously counted, the current motion mode of the target can be determined, the step 3 is carried out, and otherwise, the step 2 is returned;

if in the L frame time, the transition probability u of the turning model_ctAlways keeping maximum speed variation quantity delta v_k≥τ_vAcceleration Δ a_k≥τ_aJudging that the target is in a jumping motion stage;

step 3, single-mode matching tracking:

3-2, predicting the target state of the next frame by utilizing a single-mode matching tracking algorithm;

3-3 calculating the normalized residual squared epsilon for the current time k_vComprises the following steps:v_xfor the current measurement residual, S_xRepresenting the current residual covariance matrix, the subscript x representing the motion model matching the true motion of the target, i.e., x ∈ { CV, CA, CT }, CV representing uniform motion, CA representing uniform acceleration motion, CT representing jerking motion, (·)^TRepresenting a matrix transposition; normalized residual squared epsilon of Angelica_vWhen the target motion state is less than or equal to the threshold tau, the target motion state is considered to be unchanged, the step 3-2 is returned, and the normalized residual error squared epsilon is obtained_vGreater than threshold tauAnd (4) considering that the motion state of the target is changed, and returning to the step 2.

2. The near space hypersonic target tracking method based on target mode estimation as claimed in claim 1, characterized in that the specific method for estimating the target state of the next frame corresponding to the uniform velocity motion model, the uniform acceleration motion model and the turning motion model by using the interactive multimode tracking IMM algorithm is as follows: obtaining a next frame uniform motion model, a uniform accelerated motion model and a target prediction state value under a turning motion model by utilizing an interactive multimode tracking IMM algorithmAnd corresponding state covariance matrixThen, the target state corresponding to each model is updatedAnd corresponding covariance matrixRe-estimating each model transition probability matrixAnd finally outputting a target estimation state X:

the specific method for estimating the next frame target state by the single-mode matching tracking algorithm comprises the following steps: obtaining the next frame target prediction state value by utilizing a single-mode matching tracking algorithmSum state covariance matrixRe-updating target statesSum state covariance matrix K_x、v_xAnd S_xAnd respectively representing a gain matrix, a residual error matrix and a residual error covariance matrix of the Kalman filter corresponding to the model x, and the subscript x belongs to { CV, CA, CT }.