CN103809173B - Frame CFAR target detection Tracking Integrative method - Google Patents

Abstract

The invention discloses a kind of frame CFAR target detection Tracking Integrative method, mainly solve the problem that prior art target detection probability is lower, target following distance is shorter.Its implementation procedure is: the Initial state estimation value and the Initial state estimation covariance matrix that 1) are obtained target by Track initialization algorithm; 2) according to kth-1 frame Target state estimator value and kth-1 frame state estimate covariance matrix, kth frame target prediction ripple door is determined; 3) false-alarm probability and the detection threshold of each detecting unit of kth frame target prediction Bo Mennei is calculated; 4) echoed signal of target prediction Bo Mennei is detected, and estimating target parameter, survey data acquisition as kth frame amount; 5) survey data acquisition to kth frame amount to associate and filtering, obtain kth frame Target state estimator value and kth frame state estimate covariance matrix.The present invention, compared with existing detecting and tracking method, improves the detection probability of target, extends the tracking range of target.

Description

Frame CFAR target detection Tracking Integrative method

Technical field

The invention belongs to Radar Technology field, a kind of detecting and tracking method utilizing target prediction information adjustment aim to predict each detecting unit false-alarm probability of Bo Mennei specifically, target detection probability is improved, Extended target tracking range under can be used for radar target tracking state.

Background technology

Modern radar system comprises two large modules usually, i.e. signal processing module and data processing module.The target information detected, as first time process, is sent into radar data processing module and is done further process by Radar Signal Processing module.The operations such as radar data processing module carries out predicting after obtaining the estimators such as the position of target, kinematic parameter, associate, filtering, thus certain inhibiting effect is played to the stochastic error in radar measurement process, it is more accurate to make the estimation of target travel information, and forms stable objects flight path.

Target detection is the important step of Radar Signal Processing module, main task processes the echoed signal that radar receives, and judge the presence or absence of target, due to the impact of Noise and Interference, need to adopt CFAR Methods to reduce the probability of erroneous judgement, ensure that Radar Signal Detection has CFAR characteristic, conventional CFAR detection algorithm comprises CA-CFAR, order statistic CFAR, Generalized Likelihood Ratio, adaptive matched filter etc.

Target following is based on detecting the target position information obtained, being followed the tracks of out the flight path of target by filtering continuously.In target tracking algorism, mainly contain linear autoregression filtering, 2 extrapolation filtering, Wiener filtering, Weighted linear regression, alpha-beta filtering and Kalman filtering etc., wherein Kalman filtering can be used for linear time varying system, its distortion EKF, converted measurement Kalman filtering and unscented kalman filter can be used for nonlinear and time-varying system, statistical model all adopts state equation and measurement equation, and filtering equations calculates in the mode of recursion, and calculated amount is little, practical, therefore in target following theory, account for leading position.

Target following carries out on the basis of target detection, and high detection perform can ensure the initial fast of targetpath, and the detection perform of difference can cause the end of targetpath, and therefore the detection perform of target directly affects the tracking performance of target.For traditional detecting and tracking treatment scheme, first carry out target detection and estimating target motion parameter, after obtaining measurement information, send into radar data processing module, then carry out predicting, associate, the process such as filtering, realize the detection and tracking to target.When target echo signal to noise ratio (S/N ratio) is lower, target detection probability is lower, will cause the uncontinuity of targetpath, easily causes flight path to terminate prematurely, and thus target following distance is shorter.

Summary of the invention

The object of the invention is to the deficiency for above-mentioned prior art, propose a kind of frame CFAR target detection Tracking Integrative method, under guarantee does not produce the condition of false track, the false-alarm probability of each detecting unit of adjustment aim prediction Bo Mennei, thus the detection probability of target under raising tracking mode, the tracking range of Extended target.

For achieving the above object, the present invention includes following technical step:

1) initiation parameter: by targetpath start algorithm could, obtains the Initial state estimation value of targetpath and Initial state estimation covariance matrix P ₀;

2) target setting state transition equation and radar measurement equation, according to kth-1 frame Target state estimator value calculate kth frame dbjective state predicted value with the predicted value that kth frame target measures

3) according to kth-1 frame state estimate covariance matrix P _k-1with step 2) the kth frame dbjective state predicted value that obtains calculate the prediction covariance matrix D that kth frame target measures _k|k-1;

4) according to step 2) predicted value that measures of the kth frame target that obtains with the prediction covariance matrix D of the kth frame target measurement that step 3) obtains _k|k-1, determine kth frame target prediction ripple door O _k;

5) there is the probability P of false-alarm in the target prediction Bo Mennei being set at least N frame in the tracing process of continuous N frame _f, then following formula is utilized to calculate kth frame target prediction ripple door O _kinside there is the probability P of false-alarm _z:

$\underset{n=N}{\overset{M}{\mathrm{\Σ}}}\left[\frac{M!}{n!(M-n)!}{P_Z}^n{(1-P_Z)}^{M-n}\right]=P_F,$

N represents that in the tracing process of continuous N frame, the possible frame number of false-alarm appears in target prediction Bo Mennei, symbol wherein! Represent factorial computing, the value of M, N need meet M > N >=1;

6) according to the kth frame target prediction ripple door O that step 5) obtains _kinside there is the probability P of false-alarm _z, calculate kth frame target prediction ripple door O _kthe false-alarm probability of each detecting unit interior and detection threshold;

7) according to the kth frame target prediction ripple door O that step 6) obtains _kthe detection threshold of each detecting unit interior, to kth frame target prediction ripple door O _kinterior echoed signal detects, and estimating target parameter, survey data acquisition Z (k) as kth frame amount;

8) the kth frame amount obtained according to step 7) surveys data acquisition Z (k), utilizes association algorithm to filter out kth frame and effectively measures set Z _k, choose kth frame and effectively measure set Z _kin the metric data the highest with track association degree, and utilize track algorithm to calculate kth frame Target state estimator value and kth frame state estimate covariance matrix P _k, return step 2).

The present invention is due in the false-alarm probability process calculating each detecting unit of target prediction Bo Mennei, consider the information of forecasting of target and the suppression problem of false track, namely ensure that the probability P of false-alarm appears in the target prediction Bo Mennei of at least N frame in the tracing process of continuous N frame _f, there is the probability P of false-alarm in each frame constant _z, thus calculate false-alarm probability and the detection threshold of each detecting unit of target prediction Bo Mennei, therefore have the following advantages:

(1) detection threshold of target prediction Bo Mennei is lower than the detection threshold of traditional detection tracking, improves the detection probability of target;

(2) in low signal-to-noise ratio situation, still there is higher detection probability, improve the continuity of targetpath, avoid targetpath and terminate prematurely, extend the tracking range of target;

(3) when target suddenly disappears, flight path correctly can terminate with very high probability, avoids the generation of false track.

Accompanying drawing explanation

Fig. 1 is workflow diagram of the present invention;

Fig. 2 is the detection perform comparison diagram of the present invention and traditional detection tracking;

Fig. 3 is the detection range comparison diagram of the present invention and traditional detection tracking;

Fig. 4 is the target probability graph that each frame targetpath exists when the 10th frame disappears.

Embodiment

With reference to Fig. 1, performing step of the present invention is as follows:

Step 1, initiation parameter: by targetpath start algorithm could, obtains the Initial state estimation value of targetpath and Initial state estimation covariance matrix P ₀.

Step 2, target setting state transition equation and radar measurement equation, according to kth-1 frame Target state estimator value calculate kth frame dbjective state predicted value with the predicted value that kth frame target measures

2a) target setting state transition equation is:

x _k＝F _k-1x _k-1+v _k-1，

Wherein, x _krepresent the state of kth frame target, F _k-1represent the state-transition matrix of kth-1 frame target, x _k-1represent the state of kth-1 frame target, v _k-1represent the process noise of kth-1 frame, F in this example _k-1adopt following form:

$F_{k-1}=\left[\begin{array}{cccc}1& \mathrm{\ΔT}& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& \mathrm{\ΔT}\\ 0& 0& 0& 1\end{array}\right],$

Wherein, Δ T represents the radar scanning cycle, gets Δ T=2s in this example.

2b) setting radar measurement equation is:

z _k＝h _k(x _k)+w _k，

Wherein, z _krepresent the measuring value of kth frame target, h _k() represents the measurement function of kth frame target, w _krepresent the measurement noise of kth frame;

2c) according to kth-1 frame Target state estimator value calculate kth frame dbjective state predicted value

${\hat{x}}_{k|k-1}=F_{k-1}{\hat{x}}_{k-1};$

2d) according to kth frame dbjective state predicted value calculate the predicted value that kth frame target measures

${\hat{z}}_{k|k-1}=h_k\left({\hat{x}}_{k|k-1}\right).$

Step 3, according to kth-1 frame state estimate covariance matrix P _k-1with the kth frame dbjective state predicted value that step 2 obtains calculate the prediction covariance matrix D that kth frame target measures _k|k-1.

3a) according to kth frame dbjective state predicted value calculate the Jacobian matrix H that kth frame target measures function _k:

$H_k={\▿}_x\left(h_k^T\left(x\right)\right){|}_{x={\hat{x}}_{k|k-1}},$

Wherein, ▽ _x() represents vector x differentiate, and () T represents transpose operation, representative function ? the functional value at place;

3b) according to kth-1 frame Target state estimator covariance matrix P _k-1with step 3a) the kth frame target that obtains measures the Jacobian matrix H of function _k, calculate the prediction covariance matrix D that kth frame target measures _k|k-1:

$D_{k|k-1}=H_k[F_{k-1}{P}_{k-1}{F}_{k-1}^T+Q_{k-1}]H_k^T,$

Wherein, Q _k-1represent the process noise covariance matrix of kth-1 frame, in this example, adopt following form:

$Q_{k-1}={{\mathrm{\σ}}_p}^2\left[\begin{array}{cccc}{\mathrm{\ΔT}}^3/3& {\mathrm{\ΔT}}^2/2& 0& 0\\ {\mathrm{\ΔT}}^2& \mathrm{\ΔT}& 0& 0\\ 0& 0& {\mathrm{\ΔT}}^3/3& {\mathrm{\ΔT}}^2/2\\ 0& 0& {\mathrm{\ΔT}}^2/2& \mathrm{\ΔT}\end{array}\right]$

Wherein, σ _prepresent process noise standard deviation, in this example, get σ _p=0.1.

Step 4, according to the predicted value of the kth frame target measurement that step 2 obtains with the prediction covariance matrix D of the kth frame target measurement that step 3 obtains _k|k-1, determine kth frame target prediction ripple door O _k.

4a) target setting falls into kth frame target prediction ripple door O _kprobability P _g, in this example, get P _g=0.9997, be the chi-square distribution table that target measures dimension by looking into degree of freedom, obtain the thresholding γ of kth frame target prediction ripple door, wherein chi-square distribution table is the distribution function table of comparisons of the side's of card distribution variables in theory of probability;

4b) according to step 4a) the kth frame target prediction ripple door O that obtains _kthresholding γ, by following formula determination kth frame target prediction ripple door O _k:

$O_k=\left\{y\right|{(y-{\hat{z}}_{k|k-1})}^T{D}_{k|k-1}^{-1}(y-{\hat{z}}_{k|k-1})\≤\mathrm{\γ}\},$

Wherein, y represents the position that target occurs, | represent conditional code, the symbol left side is set element, and the right is the condition that element meets.

Step 5, there is the probability P of false-alarm in the target prediction Bo Mennei being set at least N frame in the tracing process of continuous N frame _f, then following formula is utilized to calculate kth frame target prediction ripple door O _kinside there is the probability P of false-alarm _z:

$\underset{n=N}{\overset{M}{\mathrm{\Σ}}}\left[\frac{M!}{n!(M-n)!}{P_Z}^n{(1-P_Z)}^{M-n}\right]=P_F,$

N represents that in the tracing process of continuous N frame, the possible frame number of false-alarm appears in target prediction Bo Mennei, symbol wherein! Represent factorial computing, the value of M, N need meet M > N>=1, gets M=100, N=3, P in this example _f=0.08;

Step 6, setting radar wave detector form, obtains kth frame target prediction ripple door O according to algorithm of target detection _kdetection statistic ξ (the i of interior i-th detecting unit; K), i=1,2 ..., N _k, wherein, N _krepresent kth frame target prediction ripple door O _kthe number of interior detecting unit;

The detection form of described radar wave detector comprises, and square law detection, linear detection etc., this example is selected but is not limited to square law wave detector.

Described algorithm of target detection comprises, CA-CFAR, order statistic CFAR, Generalized Likelihood Ratio, adaptive matched filter etc., this example is selected but is not limited to CA-CFAR detection algorithm, namely by following formulae discovery kth frame target prediction ripple door O _kdetection statistic ξ (the i of interior i-th detecting unit; K):

$\mathrm{\ξ}(i;k)=\frac{x(i;k)}{\underset{l=1}{\overset{N_r}{\mathrm{\Σ}}}y(l;i,k)},i=\mathrm{1,2},...,N_k,$

Wherein, x (i; K) kth frame target prediction ripple door O is represented _kthe wave detector of interior i-th detecting unit exports data, y (l; I, k) represent kth frame target prediction ripple door O _kin the reference window of interior i-th detecting unit, the wave detector of l reference unit exports data, N _rrepresent the number of reference window internal reference unit, in this example, get N _r=20.

Step 7, according to the kth frame target prediction ripple door O that step 6 obtains _kthe detection statistic of each detecting unit interior, calculates kth frame target prediction ripple door O _kthe false-alarm probability of each detecting unit interior and detection threshold.

7a) set kth frame target prediction ripple door O _kweight w (the i of the detection statistic of interior i-th detecting unit; K), w (i is got in this example; K)=1/N _k, i=1,2 ..., N _k, obtain kth frame target prediction ripple door O _kthe weight detection statistic ξ ' (i of interior i-th detecting unit; K):

ξ′(i;k)＝w(i;k)ξ(i;k),i＝1,2,...,N _k；

Following system of equations 7b) is utilized to calculate kth frame target prediction ripple door O _kthe false-alarm probability of each detecting unit interior and detection threshold:

$\left\{\begin{array}{c}\underset{i=1}{\overset{N_k}{\mathrm{\Π}}}[1-P_f(i;k)]=1-P_Z\\ P_f(i;k)=\mathrm{Pr}\{{\mathrm{\ξ}}^{\′}(i;k)\≥T\left(k\right)|H_0\},i=\mathrm{1,2},...N_k\end{array}\right.,$

Wherein, P _f(i; K) kth frame target prediction ripple door O is represented _kthe false-alarm probability of interior i-th detecting unit, H ₀represent the non-existent situation of target, T (k) represents kth frame target prediction ripple door O _kthe detection threshold of each detecting unit interior, Pr{ ξ ' (i; K)>=T (k) | H ₀represent kth frame target prediction ripple door O in the non-existent situation of target _kthe weight detection statistic ξ ' (i of interior i-th detecting unit; K) probability of detection threshold T (k) is exceeded.

Step 8, according to the kth frame target prediction ripple door O that step 7 obtains _kthe detection threshold T (k) of each detecting unit interior, to kth frame target prediction ripple door O _kinterior echoed signal detects, and estimating target parameter, survey data acquisition Z (k) as kth frame amount;

Step 9, surveys data acquisition Z (k) according to the kth frame amount that step 8 obtains, and utilizes association algorithm to filter out kth frame and effectively measures set Z _k, choose this and effectively measure set Z _kthe metric data that the middle flight path degree of association is the highest.

Described association algorithm comprises, nearest-neighbor algorithm, Probabilistic Data Association Algorithm, optimum Bayes's association algorithm etc., and this example is selected but is not limited to Probabilistic Data Association Algorithm, namely chooses the highest metric data of track association degree in accordance with the following steps:

9a) utilize following formula to calculate kth frame and newly cease covariance matrix S _k:

S _k＝D _k|k-1+R _k，

Wherein, D _k|k-1represent the prediction covariance matrix that kth frame target measures, R _krepresent that kth frame amount surveys covariance matrix;

9b) target setting metric data is chosen for the probability P effectively measured _g, in this example, choose P _g=0.9997, being the chi-square distribution table that target measures dimension by looking into degree of freedom, obtaining the thresholding η effectively measured, and determining that kth frame effectively measures region A _k:

$A_k=\left\{z\right|{(z-{\hat{z}}_{k|k-1})}^T{S}_k^{-1}(z-{\hat{z}}_{k|k-1})\≤\mathrm{\η}\},$

Wherein, z represents that target measures the position that may occur, represent the predicted value that kth frame target measures, | represent conditional code, the symbol left side is set element, and the right is the condition that element meets;

9c) filter out kth frame amount to survey and fall into kth frame in data acquisition Z (k) and effectively measure region A _kinterior metric data, effectively measures set Z as kth frame _k, and utilize following formula to calculate this effectively measurement set Z _kthe new breath v of a middle jth metric data _j:

$v_j=Z_k\left(j\right)-{\hat{z}}_{k|k-1},j=\mathrm{1,2},...,m_k,$

Wherein, Z _kj () represents that kth frame effectively measures set Z _ka middle jth metric data, m _krepresent that kth frame effectively measures set Z _kmiddle metric data number;

9d) according to step 9a) the kth frame that obtains newly ceases covariance matrix S _kwith step 9c) the kth frame that obtains effectively measures set Z _kthe new breath v of a middle jth metric data _j, calculate kth frame and effectively measure set Z _kin the track association degree β of each metric data _j:

Wherein, represent that average is 0, variance is new breath covariance matrix S _kgaussian random vector at v _jthe probability density value at place, P _drepresent the detection probability of kth frame target, V _krepresent that kth frame effectively measures region A _karea, get in this example wherein, | S _k| represent that kth frame newly ceases covariance matrix S _kdeterminant;

9e) choose kth frame and effectively measure set Z _kin each metric data track association degree in metric data corresponding to maximal value.

Step 10, effectively measures set Z according to the kth frame that step 9 obtains _k, utilize track algorithm to calculate kth frame Target state estimator value and kth frame state estimate covariance matrix P _k, return step 2.

Described track algorithm comprises, Kalman filtering, EKF, converted measurement Kalman filtering, unscented kalman filter, particle filters etc., this example is selected but is not limited to expanded Kalman filtration algorithm, namely calculates kth frame Target state estimator value in accordance with the following steps and kth frame state estimate covariance matrix P _k:

10a) utilize following formulae discovery filter gain matrix K _k:

$K_k=(F_{k-1}{P}_{k-1}{F}_{k-1}^T+Q_{k-1})H_k^T{S}_k^{-1},$

Wherein, F _k-1represent the state-transition matrix of kth-1 frame target, P _k-1represent kth-1 frame Target state estimator covariance matrix, Q _k-1represent the process noise covariance matrix of kth-1 frame, H _krepresent that kth frame target measures the Jacobian matrix of function;

10b) according to step 10a) the filter gain matrix K k that obtains, calculate kth frame Target state estimator value

${\hat{x}}_k={\hat{x}}_{k|k-1}+K_k\underset{j=1}{\overset{m_k}{\mathrm{\Σ}}}{\mathrm{\β}}_j{v}_j,$

Wherein, represent the predicted value of kth frame dbjective state, β _jrepresent that kth frame effectively measures set Z _kthe track association degree of a middle jth metric data, v _jrepresent that kth frame effectively measures set Z _kthe new breath of a middle jth metric data, m _krepresent that kth frame effectively measures set Z _kmiddle metric data number;

Following formula 10c) is utilized to calculate kth frame state estimate covariance matrix P _k:

$P_k=[I-\left(\underset{j=1}{\overset{m_k}{\mathrm{\Σ}}}{\mathrm{\β}}_j\right)K_k{H}_k](F_{k-1}{P}_{k-1}{F}_{k-1}^T+Q_{k-1})+K_k[\underset{j=1}{\overset{m_k}{\mathrm{\Σ}}}{\mathrm{\β}}_j{v}_j{v}_j^T-\left(\underset{j=1}{\overset{m_k}{\mathrm{\Σ}}}{\mathrm{\β}}_j{v}_j\right){\left(\underset{j=1}{\overset{m_k}{\mathrm{\Σ}}}{\mathrm{\β}}_j{v}_j\right)}^T]K_k^T,$

Wherein, I representation unit matrix.

Effect of the present invention is further illustrated by following simulation comparison test:

1. experiment scene: adopt a 2D radar being positioned at true origin, if carrier frequency f _c=3GHz, antenna aperture D=2.5m, transmitted signal bandwidth B=2MHz, sample frequency is F _s=4MHz, the radar scanning cycle is Δ T=2s, and radargrammetry parameter is the distance and bearing angle of target; If initial time target is 50km in the position of X-axis, Y-axis, and flies at a constant speed away from radar station, the speed component of X-axis, Y-axis is 300m/s, and initial signal to noise ratio (S/N ratio) is 20dB, and in traditional detection track algorithm, the false-alarm probability of target detection is 10 ^-6; Flight path termination rule is: if continuous three frames do not detect target, then flight path termination, and object tracking process terminates.

2. emulate content:

Emulation 1: adopt above experiment scene, utilize traditional detecting and tracking method and detecting and tracking method of the present invention, carry out simulation comparison to the detection perform of radar, result is as Fig. 2;

Emulation 2: adopt above experiment scene, utilize traditional detecting and tracking method and detecting and tracking method of the present invention, carry out simulation comparison to the detection range of radar, result is as Fig. 3;

Emulation 3: adopt above experiment scene, assuming that the 10th frame target suddenly disappears, emulate the probability that each frame flight path exists, result is as Fig. 4.

3. interpretation:

As seen in Figure 2, when ensureing same detection probability 0.6, echo signal to noise ratio (S/N ratio) required for traditional detection tracking is 13.28dB, the echo signal to noise ratio (S/N ratio) wanted required for the present invention is 9.533dB, compared with traditional detection tracking, the echo signal to noise ratio (S/N ratio) wanted required for the present invention can reduce 3.747dB, thus improves the detection perform of radar to target.

As seen in Figure 3, when ensureing same detection probability 0.6, traditional detection tracking is 104.4km to the BURN-THROUGH RANGE of target, the present invention is 129.3km to the BURN-THROUGH RANGE of target, 24.9km is improve compared with traditional detection tracking, improve the detection range of radar to target, thus increase the tracking range of radar to target.

As seen in Figure 4, when target is after the 10th frame disappears, should disappear at the 12nd frame under normal circumstances, can be found out by simulation result, the probability existed at the 12nd frame targetpath is 0.023, and the probability that namely flight path terminates is 0.977, after ensure that target disappearance, flight path can effectively terminate, inhibit the generation of false track, demonstrate in the present invention the validity adjusting false-alarm probability, and about the 15th frame, just can terminate flight path at the latest.

Comprehensive above-mentioned emulation experiment can be found out, in the flight course of target away from radar station, signal to noise ratio (S/N ratio) reduces gradually along with the increase of target range, the present invention is relative to traditional detecting and tracking method, owing to having considered tracker, problem is suppressed to the information of forecasting of target and false track, can ensure under the condition not producing false track, reduce the detection threshold of target, thus improve the detection perform of target, namely still can ensure the continuity of flight path when signal to noise ratio (S/N ratio) is lower, thus increase the tracking range of target.

Claims (4)

1. a frame CFAR target detection Tracking Integrative method, comprises the steps:

1) initiation parameter: by targetpath start algorithm could, obtains the Initial state estimation value of targetpath and Initial state estimation covariance matrix P ₀;

2) target setting state transition equation and radar measurement equation, according to kth-1 frame Target state estimator value calculate kth frame dbjective state predicted value with the predicted value that kth frame target measures

3) according to kth-1 frame state estimate covariance matrix P _k-1with step 2) the kth frame dbjective state predicted value that obtains calculate the prediction covariance matrix D that kth frame target measures _k|k-1;

4) according to step 2) predicted value that measures of the kth frame target that obtains with step 3) the prediction covariance matrix D that measures of the kth frame target that obtains _k|k-1, determine kth frame target prediction ripple door O _k;

5) there is the probability P of false-alarm in the target prediction Bo Mennei being set at least N frame in the tracing process of continuous N frame _f, then following formula is utilized to calculate kth frame target prediction ripple door O _kinside there is the probability P of false-alarm _z:

$\underset{n=N}{\overset{M}{\Σ}}\[\frac{M!}{n!(M-n)!}{P_Z}^n{(1-P_Z)}^{M-n}\]=P_F,$

N represents that in the tracing process of continuous N frame, the possible frame number of false-alarm appears in target prediction Bo Mennei, symbol wherein! Represent factorial computing, the value of M, N need meet M>N >=1;

6) according to step 5) the kth frame target prediction ripple door O that obtains _kinside there is the probability P of false-alarm _z, calculate kth frame target prediction ripple door O _kthe false-alarm probability of each detecting unit interior and detection threshold:

6a) set radar wave detector form, obtain kth frame target prediction ripple door O according to algorithm of target detection _kdetection statistic ξ (the i of interior i-th detecting unit; K), i=1,2 ..., N _k, wherein, N _krepresent kth frame target prediction ripple door O _kthe number of interior detecting unit;

6b) set kth frame target prediction ripple door O _kweight w (the i of the detection statistic of interior i-th detecting unit; K), i=1,2 ..., N _k, obtain kth frame target prediction ripple door O _kthe weight detection statistic ξ ' (i of interior i-th detecting unit; K):

ξ′(i；k)＝w(i；k)ξ(i；k),i＝1,2,...,N _k；

Following system of equations 6c) is utilized to calculate kth frame target prediction ripple door O _kthe false-alarm probability of each detecting unit interior and detection threshold:

$\left\{\begin{array}{c}\underset{i=1}{\overset{N_k}{\Π}}\[1-P_f(i;k)\]=1-P_Z\\ P_f(i;k)=\mathrm{Pr}\left\{{\mathrm{\ξ}}^{\′}\right(i;k)\≥T(k\left)\right|H_0\},i=1,2,...N_k\end{array}\right.,$

Wherein, P _f(i; K) kth frame target prediction ripple door O is represented _kthe false-alarm probability of interior i-th detecting unit, H ₀represent the non-existent situation of target, T (k) represents kth frame target prediction ripple door O _kthe detection threshold of each detecting unit interior, Pr{ ξ ' (i; K)>=T (k) | H ₀represent kth frame target prediction ripple door O in the non-existent situation of target _kthe weight detection statistic ξ ' (i of interior i-th detecting unit; K) probability of detection threshold T (k) is exceeded;

7) according to step 6) the kth frame target prediction ripple door O that obtains _kthe detection threshold of each detecting unit interior, to kth frame target prediction ripple door O _kinterior echoed signal detects, and estimating target parameter, survey data acquisition Z (k) as kth frame amount;

8) according to step 7) the kth frame amount that obtains surveys data acquisition Z (k), and utilize association algorithm to filter out kth frame and effectively measure set Z _k, choose kth frame and effectively measure set Z _kin the metric data the highest with track association degree, and utilize track algorithm to calculate kth frame Target state estimator value and kth frame state estimate covariance matrix P _k, return step 2).

2. frame CFAR target detection Tracking Integrative method according to claim 1, wherein step 2) described in calculate kth frame dbjective state predicted value with the predicted value that kth frame target measures carry out as follows:

2a) target setting state transition equation is:

x _k＝F _k-1x _k-1+v _k-1，

Wherein, x _krepresent the state of kth frame target, F _k-1represent the state-transition matrix of kth-1 frame target, x _k-1represent the state of kth-1 frame target, v _k-1represent the process noise of kth-1 frame;

2b) setting radar measurement equation is:

z _k＝h _k(x _k)+w _k，

Wherein, z _krepresent the measuring value of kth frame target, h _k() represents the measurement function of kth frame target, w _krepresent the measurement noise of kth frame;

2c) according to kth-1 frame Target state estimator value calculate kth frame dbjective state predicted value

${\hat{x}}_{k|k-1}=F_{k-1}{\hat{x}}_{k-1};$

2d) according to kth frame dbjective state predicted value calculate the predicted value that kth frame target measures

${\hat{z}}_{k|k-1}=h_k\left({\hat{x}}_{k|k-1}\right).$

3. frame CFAR target detection Tracking Integrative method according to claim 1, wherein step 3) described in the prediction covariance matrix D that measures of calculating kth frame target _k|k-1, carry out as follows:

3a) according to kth frame dbjective state predicted value calculate the Jacobian matrix H that kth frame target measures function _k:

$H_k={\▿}_x\left(h_k^T\right(x\left)\right){|}_{x={\hat{x}}_{k|k-1}},$

Wherein, represent vector x differentiate, () ^trepresent transpose operation, representative function ? the functional value at place;

3b) according to kth-1 frame Target state estimator covariance matrix P _k-1with step 3a) the kth frame target that obtains measures the Jacobian matrix H of function _k, calculate the prediction covariance matrix D that kth frame target measures _k|k-1:

$D_{k|k-1}=H_k\[F_{k-1}{P}_{k-1}{F}_{k-1}^T+Q_{k-1}\]H_k^T,$

Wherein, Q _k-1represent the process noise covariance matrix of kth-1 frame, F _k-1represent the state-transition matrix of kth-1 frame target.

4. frame CFAR target detection Tracking Integrative method according to claim 1, wherein step 4) described in determination kth frame target prediction ripple door O _k, carry out as follows:

4a) target setting falls into kth frame target prediction ripple door O _kprobability P _g, being the chi-square distribution table that target measures dimension by looking into degree of freedom, obtaining kth frame target prediction ripple door O _kthresholding γ, wherein chi-square distribution table is the distribution function table of comparisons of the side's of card distribution variables in theory of probability;

4b) according to step 4a) the kth frame target prediction ripple door O that obtains _kthresholding γ, by following formula determination kth frame target prediction ripple door O _k:

$O_k=\left\{y\right|{(y-{\hat{z}}_{k|k-1})}^T{D}_{k|k-1}^{-1}(y-{\hat{z}}_{k|k-1})\≤\mathrm{\γ}\},$

Wherein, y represents the position that target may occur, | represent conditional code, the symbol left side is set element, and the right is the condition that element meets.