CN104155650A - Object tracking method based on trace point quality evaluation by entropy weight method - Google Patents

Abstract

The invention relates to a nearest neighbor object tracking method based on trace point quality evaluation, belongs to the technical field of radar target tracking, and particularly relates to an object tracking method based on multiple-evaluation-index information fusion of entropy weight. The invention puts forward a method for improving the degree of nearest neighbor correlation based on comprehensive evaluation of multiple features of the trace point. In data correlation, four measurement indexes of a trace point dropping into a wave gate, including position, amplitude, Doppler and work frequency, are taken into comprehensive consideration, evaluation values under the indexes are obtained by referring to predicted values or prior information, weight values are determined by an entropy weight method so as to obtain a comprehensive evaluation value, and the optimal measurement in quality evaluation instead of the measurement nearest to a predication point in a traditional method is used as track update. Therefore, the sensor information utilization rate is improved, and the performance of weak target tracking under complex environments is improved.

Description

A kind of method for tracking target based on the quality evaluation of entropy weight method point mark

Technical field

The invention belongs to radar target tracking technical field, particularly a kind of method for tracking target of the many evaluation indexes information fusion based on entropy weights.

Technical background

At present, due to the strong clutter of the formation such as the mountain region in radar surveillance scene, city, ocean, the interference of strong noise background, motor-driven, the stealthy impact of target itself, radar target tracking technology is faced with huge challenge.Under complex background, some clutter is all similar to target on time domain or frequency domain, and CFAR detection technology is difficult to effectively distinguish clutter and target, and a large amount of False Intersection Points marks bring huge challenge to rear end target following, data correlation.

The core of Target Tracking Problem is data correlation.Common data correlation method mainly contains arest neighbors method (NN) and probabilistic data association method (PDA) etc., they are all to utilize sensor to obtain the information directly related with dbjective state vector calculation, as document " Tracking in clutter with nearest neighbor filters:Analysis and performance.IEEE Trans.Aerospace Electron.Syst.32:995-1010 " has proposed a kind of data association algorithm of arest neighbors, use from the nearest point of predicted value statistical distance and upgrade flight path, it is little and to the good feature of clutter changes in distribution robustness that this algorithm has calculated amount, but in the higher environment of clutter density, only consider that positional information makes hard decision, probably cause mistake with or the situation of lose objects.In fact, it is the positional information of acquisition point mark that sensor has more than, and can also obtain the more further feature data about target.In passive multiple target tracking process, the metric data that conventionally can be directly used in calculating dbjective state comprises direction of arrival of signal (DOA), time of arrival (toa) (TOA) and Doppler frequency, amplitude etc.The obtainable radiation source characteristic of passive location system comprises: polarization phases, frequency of operation, pulsewidth, pulse amplitude, pulse repetition time etc.Document " Wang Jiegui, Luo Jingqing. based on the Passive Tracking [J] of multiple goal multicharacteristic information fused data association. electronic letters, vol, 2004, 32 (6): 1013-1016.2004 (06) " in, in the time following the tracks of, utilize the frequency of target, pulsewidth, pulse repetition time three kinds of knowledge assistance features, and regard respectively different evidence sources as, utilize Dempste-Shafer evidence theory to carry out multicharacteristic information fusion, obtain the degree of association of each effective observation and real goal, but the method does not take into full account the mutual relationship between feature, divide timing " to be determined on a case-by-case basis " at the synthetic weights of information, adopt experience weights, cause so in actual applications with certain subjectivity, even same operator, in different time and environment, same target is also tended to draw inconsistent subjective judgement.This must make information building-up process with getting sth into one's head property significantly, thereby the confidence level of the synthetic result of information is declined, and is difficult to ensure tracking performance, lacks versatility.

Summary of the invention

The object of the invention is to improve a kind of method for tracking target based on the quality evaluation of entropy weight method point mark of design for the weak point of background technology, the method is on nearest-neighbor basis, model assessment models, utilize the many indexes that the entropy weight method in operational research is ordered to Bo Mennei to carry out comprehensive assessment, upgrade flight path by quality evaluation optimum point, thereby improve sensor information utilization factor, improve weak target tracking performance under complex environment.

A kind of method for tracking target based on the quality evaluation of entropy weight method point mark of the present invention, the method comprises:

Step 1, initialization dbjective state X (k) and target covariance matrix P (k), k=0;

The state transition equation of target is:

X(k+1)＝FX(k)+ω(k)

Wherein, F is state-transition matrix, the Gaussian process noise that ω (k) is zero-mean, and its covariance matrix is Q, measurement model is:

Z(k)＝HX(k)+v(k)

Z (k) is that target measures, and H is observing matrix, Gauss's observation noise that v (k) is zero-mean, and its covariance matrix is R;

Step 2, according to the state X (k-1) of k-1 moment target t and covariance matrix P (k-1) thereof, obtain the status predication value of k moment target t:

X(k|k-1)＝FX(k-1)

Measurement predictor:

Z(k|k-1)＝HX(k|k-1)

Covariance one-step prediction value P (k|k-1):

P(k|k-1)＝FP(k-1)F ^T+Q；

Step 3, from radar receiver, read k frame position measure set z _i(k) be i the measurement in k moment, meanwhile, and obtain the set of k frame measurement amplitude doppler's set frequency of operation set and corresponding three knowledge assistances of each position quantity measured value measure index a (k), d (k), f (k),

Whether calculated amount measured value z (k) meets following formula, if meet as candidate's echo, and adds up the echo number meeting the demands,

[z(k)-Z(k|k-1)] ^TS ^-1(k)×[z(k)-Z(k|k-1)]≤γ

Wherein S (k) is new breath covariance S (k)=HP (k|k-1) H in k moment ^t+ R, γ is predefined ripple door size;

Step 4, obtain falling into the candidate target number of Bo Mennei according to step 3 statistics, upgrade targetpath,

If candidate's echo number is zero, do not measure and fall into relevant Bo Mennei, upgrade flight path by the predicted value of step 2, proceed to step 8.1;

Only have one if fall into the measuring value of relevant Bo Mennei, this measuring value can be directly used in flight path renewal, proceeds to step 8.2;

If there is more than one echo to fall into the relevant Bo Mennei of tracked target, Bo Mennei point mark is carried out to quality evaluation, choose optimum point, proceed to step 5;

Step 5, the locus assessed value of calculating each candidate's echo, amplitude assessed value, Doppler's assessed value, frequency of operation assessed value, obtain the assessment vector under each evaluation index, then assessment Vector Groups is combined into an evaluating matrix A;

Step 6, calculate the entropy power method weights of every evaluation index, composition weight vector;

Step 6.1 is established x _ijfor the element in a mark mass matrix A, i=1,2,3 ..., m _k, m _kfor falling into the total number of effective measurement of Bo Mennei, j=1,2,3,4 represent that respectively four measure index, then carry out data translation x _ij=x _ij+ 1, calculate with logarithm when utilizing entropy to ask flexible strategy, meaningless when carrying out data translation and can avoiding taking the logarithm;

The value that step 6.2 is calculated i the factor under j item index shared proportion in this index

Step 6.3 is calculated the entropy of j item index

Step 6.4 is calculated the coefficient of variation of j item index:

Step 6.5 is calculated weights obtain four evaluation index weight vectors;

The weight vector that step 7, the evaluating matrix that step 5 is obtained and step 6 obtain multiplies each other, and obtains final mass assessed value V=Aw, finds position i corresponding to maximal value in assessed value V _op, obtain the optimum z of measurement of Bo Mennei _op(k);

Filter gain matrix K (k)=P (k|k-1) H (k) S (k) of step 8, calculating target t ^-1, the state that obtains upgrades expression formula x (k|k), carries out flight path renewal;

If step 8.1 candidate echo number is zero, do not measure and fall into relevant Bo Mennei, utilize future position to upgrade flight path;

x(k|k)＝x(k|k-1)

P(k|k)＝P(k|k-1)

Only have one if step 8.2 falls into the measuring value of relevant Bo Mennei, this measuring value can be directly used in flight path renewal;

x(k|k)＝x(k|k-1)+K(z)[z(k)-HX(k|k-1)]

P(k|k)＝P(k|k-1)-K(k)HP(k|k-1)

Step 8.3 Ruo Bomennei has multiple points, chooses after a mark quality evaluation and upgrades flight path by optimum point

x(k|k)＝x(k|k-1)+K(z)[z _op(k)-Z(k|k-1)]

P(k|k)＝P(k|k-1)-K(k)HP(k|k-1)。

The concrete steps of described step 5 are:

Step 5.1 computer memory position assessed value:

In the k moment, m in ripple door _kindividual effective status measures as z _i(k), i=1,2,3 ..., m _k, the filtering residuals vector of i state measurement is v _i(k)=z _i(k)-Z (k|k-1), the statistical distance of itself and predicted position is i the locus assessed value effectively measuring is:

$m_s\left(i\right)=\left\{\begin{array}{cc}\frac{\min (D_1,D_2,...,D_{m_k})}{D_i}& ,D_i\≠0\\ 1& ,D_i=0\end{array}\right.,i=\mathrm{1,2,3},...,m_k;$

Step 5.2 is calculated amplitude assessed value:

In the k moment, m in ripple door _kindividual effective observation amplitude information is a _i(k), i=1,2,3 ..., m _k, definition amplitude assessed value m _a:

$m_a\left(a_i\left(k\right)\right)=\frac{p_1^T\left(a_i\left(k\right)\right)}{p_1^T\left(a_i\left(k\right)\right)+p_0^T\left(a_i\left(k\right)\right)}$

Wherein, p ₀(a) probability density function of amplitude when echoed signal only comes from noise or clutter, p ₁(a) probability density function while comprising echo signal for echoed signal, T is detection threshold, p _fa, p _dbe respectively false-alarm probability, detection probability;

Step 5.3 is calculated Doppler's assessed value:

In the k moment, m in ripple door _kthe doppler information of individual effective observation is d _i(k), i=1,2,3 ..., m _k, the Doppler frequency observed reading of establishing the k-1 time corresponding target of when sampling this tracking gate institute is d (k-1), note observation Doppler differs from Δ d _i(k)=| d _i(k)-d (k-1) |, the assessed value of Doppler frequency is designated as m _d:

$m_d\left(i\right)=\left\{\begin{array}{cc}\frac{\min (\mathrm{\Δ}{d}_1,\mathrm{\Δ}{d}_2,...,\mathrm{\Δ}{d}_{m_k})}{\mathrm{\Δ}{d}_i}& ,\mathrm{\Δ}{d}_i\≠0\\ 1& ,\mathrm{\Δ}{d}_i=0\end{array}\right.;$

The assessed value of step 5.4 frequency of operation:

In the k moment, m in ripple door _kindividual effective observation working frequency information is f _i(k), i=1,2,3 ..., m _k, the frequency of operation observed reading of establishing the k-1 time corresponding target of when sampling this tracking gate institute is f (k-1), remembers Δ f _i(k)=| f _i(k)-f (k-1) |, be observation frequency difference, the assessed value of frequency of operation is designated as m _f, only consider the target of fixed frequency (FIX) type, i the effective m of observation _ffor:

$m_f\left(i\right)=\left\{\begin{array}{cc}1& ,\mathrm{\Δ}{f}_i\≤\mathrm{\ϵf}\\ \frac{{(\mathrm{\Δ}{f}_i-2\mathrm{\ϵf})}^2}{\mathrm{\ϵ}{f}^2}& ,\mathrm{\ϵf}\≤\mathrm{\Δ}{f}_i\≤2\mathrm{\ϵf}\\ 0& ,\mathrm{\Δ}{f}_i\≥2\mathrm{\ϵf}\end{array}\right.$

Wherein ε f is by system noise and the determined frequency measurement tolerance of measurement noise;

Each evaluation index composition point mark quality evaluation matrix A that step 5.5 obtains step 5.1,5.2,5.3,5.4, i.e. A=[m _s; m _a; m _d; m _f] ^t.

A kind of method for tracking target based on the quality evaluation of entropy weight method point mark of the present invention, utilize information of forecasting to be derived from the possibility of target from each this mark of feature angle analysis, and use that the comprehensive estimation method in operational research---entropy power method is determined the weights of many index evaluations, and then a mark quality is had to objective comprehensive assessment, sequence, utilize optimal amount in the quality evaluation of some mark to measure for measuring and upgrade flight path recently from future position in nearest-neighbor method, thereby the Evaluation Model on Quality with foundation is applicable to multiple different knowledge assistance information fusion, solution procedure is simple, improve sensor information utilization factor, improve the effect of the tracking performance of weak target.

Brief description of the drawings

Fig. 1 is process flow diagram of the present invention;

Fig. 2 is the present invention and the schematic diagram of traditional nearest-neighbor track algorithm in the time of data correlation;

Fig. 3 is at P _faafter within=0.01 o'clock, improving, nearest-neighbor is followed the tracks of with traditional nearest-neighbor and is followed the tracks of 200 times, Monte Carlo square error curve comparison curve map.

Embodiment

The present invention mainly adopts the method for Computer Simulation to verify, institute in steps, conclusion all on MATLAB-R2010b checking correct.Concrete implementation step is as follows:

Step 1, input flight path, and calculate the predicted value in k moment:

Consider a linear uniform motion target in rectangular coordinate system, the original state of establishing target is X (0)=[80m, 6m/s, 100m, 0m/s], and flight path front cross frame is initial.Design parameter arranges as table 1, table 2.In the k moment, obtain the status predication value X (k|k-1) of target: X (k|k-1)=FX (k-1), and covariance one-step prediction value P (k|k-1): P (k|k-1)=FP (k-1) F ^t+ Q.

Wherein $F=\left[\begin{array}{cccc}1& T& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& T\\ 0& 0& 0& 0\end{array}\right],H=\left[\begin{array}{cccc}1& 0& 0& 0\\ 0& 0& 1& 0\end{array}\right]$

Step 2, calculating fall into candidate's echo of Bo Mennei, and add up its number:

From radar receiver, read k frame and measure location sets z _i(k) be i the measurement in k moment, meanwhile, also obtained k frame and measured amplitude set doppler's set frequency of operation set $F\left(k\right)={\left\{f_i\left(k\right)\right\}}_{i=1}^n.$

When aim parameter measured value z (k) meets formula [z (k)-Z (k|k-1)] ^ts ^-1(k) × [z (k)-Z (k|k-1)] when≤γ, as candidate's echo, the number of statistics candidate echo;

If 2.1 candidate's echo numbers are zero, do not measure and fall into relevant Bo Mennei, upgrade flight path with future position, proceed to step 6.1;

If 2.2 measuring values that fall into relevant Bo Mennei only have one, this measuring value can be directly used in flight path renewal, proceeds to step 6.2;

If 2.3 have more than one echo to fall into the relevant Bo Mennei of tracked target, Bo Mennei point mark is carried out to quality evaluation, choose optimum point, proceed to step 3

The foundation of step 3, assessment models, the assessment of feature:

The assessed value of 3.1 locus

In the k moment, the effective observer state of Bo Mennei measures as z _i(k), i=1,2,3 ..., m _k, the filtering residuals vector that the state of i observation measures is v _i(k)=z _i(k)-x _i(k|k-1), the statistical distance of itself and predicted position defining i the quantity of state test and appraisal valuation of effectively observing is:

$m_s\left(i\right)=\left\{\begin{array}{cc}\frac{\min (D_1,D_2,...,D_{m_k})}{D_i}& ,D_i\≠0\\ 1& ,D_i=0\end{array}\right.,i=\mathrm{1,2,3},...,m_k.$

The assessed value of 3.2 amplitudes

In the k moment, m in ripple door _kindividual effective observation amplitude information is a _i(k), i=1,2,3 ..., m _k.Assessed value m under definition amplitude index _a, $m_a\left(a_i\left(k\right)\right)=\frac{p_1^T\left(a_i\left(k\right)\right)}{p_1^T\left(a_i\left(k\right)\right)+p_0^T\left(a_i\left(k\right)\right)}.$ Wherein, $p_0^T\left(a\right)=\frac{1}{P_{\mathrm{fa}}}{p}_0\left(a\right),p_1^T\left(a\right)=\frac{1}{P_d}{p}_1\left(a\right);$ Wherein $p_0\left(a\right)=\frac{a}{{\mathrm{\σ}}^2}\exp (-\frac{a^2}{2{\mathrm{\σ}}^2}),p_1\left(a\right)=\frac{a}{{\mathrm{\σ}}^2+{{\mathrm{\σ}}_s}^2}\exp (-\frac{a^2}{2({\mathrm{\σ}}^2+{{\mathrm{\σ}}_s}^2)}),$ Wherein σ ²the power spectrum density of multiple Gaussian noise, for the power spectrum density of multiple gaussian signal, and signal to noise ratio snr=10lg (σ _s ²/ σ ²).

3.3 Doppler's assessed value

In the k moment, the Bo Mennei effectively doppler information of observation is d _i(k), i=1,2,3 ..., m _k, the Doppler frequency observed reading of establishing the k-1 time corresponding target of when sampling this tracking gate institute is d (k-1), note observation Doppler differs from Δ d _i(k)=| d _i(k)-d (k-1) |, the assessed value of definition Doppler frequency is designated as m _d:

$m_d\left(i\right)=\left\{\begin{array}{cc}\frac{\min (\mathrm{\Δ}{d}_1,\mathrm{\Δ}{d}_2,...,\mathrm{\Δ}{d}_{m_k})}{\mathrm{\Δ}{d}_i}& ,\mathrm{\Δ}{d}_i\≠0\\ 1& ,\mathrm{\Δ}{d}_i=0\end{array}\right.$

The assessed value of 3.4 frequency of operation

In the k moment, it is f that Bo Mennei effectively observes working frequency information _i(k), i=1,2,3 ..., m _k.If the frequency of operation observed reading of the k-1 time corresponding target of when sampling this tracking gate institute is f (k-1), remember Δ f _i(k)=| f _i(k)-f (k-1) |, be observation frequency difference.The assessed value of definition frequency of operation is designated as m _f, only consider the target of fixed frequency (FIX) type, i the effective m of observation _fbe defined as:

$m_f\left(i\right)=\left\{\begin{array}{cc}1& ,\mathrm{\Δ}{f}_i\≤\mathrm{\ϵf}\\ \frac{{(\mathrm{\Δ}{f}_i-2\mathrm{\ϵf})}^2}{\mathrm{\ϵ}{f}^2}& ,\mathrm{\ϵf}\≤\mathrm{\Δ}{f}_i\≤2\mathrm{\ϵf}\\ 0& ,\mathrm{\Δ}{f}_i\≥2\mathrm{\ϵf}\end{array}\right.$

Wherein ε f is by system noise and the determined frequency measurement tolerance of measurement noise.

3.5 each evaluation index composition point mark quality evaluation matrix A that step 3.1,3.2,3.3,3.4 is obtained, i.e. A=[m _s; m _a; m _d; m _f] ^t.

The calculating of step 4, entropy power method weights:

4.1 establish x _ijfor element in a mark mass matrix A, i=1,2,3 ..., m _k, j=1,2,3,4, calculate with logarithm when utilizing entropy to ask flexible strategy, carry out data translation x _ij=x _ij+ 1, meaningless can avoid taking the logarithm time;

4.2 values of calculating i the factor under j item indexs shared proportion in this index

4.3 calculate the entropy of j item index

4.4 calculate the coefficient of variation of j item index.For j item index, desired value x _ijdifference larger, entropy is just less.The definition of difference value coefficient: $g_i=\frac{1-e_j}{\underset{j=1}{\overset{n}{\mathrm{\Σ}}}(1-e_j)};$

4.5 ask weights $w_j=\frac{g_j}{\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{g}_j},(1\≤j\≤n),$ Obtain weight vector.

Step 5, overall quality assessment

Assessed value matrix and weight vector are multiplied each other, obtain final mass assessed value V=Aw.Find the position i that in assessed value V, greatest member is corresponding _op, obtain the optimum z of measurement of Bo Mennei _op(k).

Step 6, flight path upgrade

Calculate the filter gain matrix K (k) of target t, K (k)=P (k|k-1) H (k) S (k) ^-1, the state that obtains upgrades expression formula x (k|k):

If 6.1 candidate's echo numbers are zero, do not measure and fall into relevant Bo Mennei, utilize future position to upgrade flight path

x(k|k)＝x(k|k-1)

P(k|k)＝P(k|k-1)

If 6.2 measuring values that fall into relevant Bo Mennei only have one, this measuring value can be directly used in flight path renewal

x(k|k)＝x(k|k-1)+K(z)[z(k)-HX(k|k-1)]

P(k|k)＝P(k|k-1)-K(k)HP(k|k-1)

6.3 Ruo Bomennei have multiple points, choose after a mark quality evaluation and upgrade flight path by optimum point

x(k|k)＝x(k|k-1)+K(z)[z _op(k)-Hx(k|k-1)]

P(k|k)＝P(k|k-1)-K(k)HP(k|k-1)

6.4 repeat to return step 1, finish until follow the tracks of.

Table 1 basic parameter table

Parameter	Symbol	Numerical value
			Sweep spacing	T	1s
Totalframes	N	25
			Datum plane	D_P	300×200
False-alarm probability	P fa	0.001
			Detection probability	P d	0.7
Process noise spectral density	q	0.01
			Measure noise covariance	R	diag(2,2)
The first thresholding	c 1	3
			The second thresholding	c 2	8
Track loss is differentiated	m	6
			Signal to noise ratio	SCR	10dB
Monte Carlo number of times	M T	200

Table 2 knowledge assistance parameter list

Claims (2)

1. the method for tracking target based on the quality evaluation of entropy weight method point mark, the method comprises:

Step 1, initialization dbjective state X (k) and target covariance matrix P (k), k=0;

The state transition equation of target is:

X(k+1)＝FX(k)+ω(k)

Wherein, F is state-transition matrix, the Gaussian process noise that ω (k) is zero-mean, and its covariance matrix is Q, measurement model is:

Z(k)＝HX(k)+v(k)

Z (k) is that target measures, and H is observing matrix, Gauss's observation noise that v (k) is zero-mean, and its covariance matrix is R;

Step 2, according to the state X (k-1) of k-1 moment target t and covariance matrix P (k-1) thereof, obtain the status predication value of k moment target t:

X(k|k-1)＝FX(k-1)

Measurement predictor:

Z(k|k-1)＝HX(k|k-1)

Covariance one-step prediction value P (k|k-1):

P(k|k-1)＝FP(k-1)F ^T+Q；

Step 3, from radar receiver, read k frame position measure set z _i(k) be i the measurement in k moment, meanwhile, and obtain the set of k frame measurement amplitude doppler's set frequency of operation set and corresponding three knowledge assistances of each position quantity measured value measure index a (k), d (k), f (k),

Whether calculated amount measured value z (k) meets following formula, if meet as candidate's echo, and adds up the echo number meeting the demands,

[z(k)-Z(k|k-1)] ^TS ^-1(k)×[z(k)-Z(k|k-1)]≤γ

Wherein S (k) is new breath covariance S (k)=HP (k|k-1) H in k moment ^t+ R, γ is predefined ripple door size;

Step 4, obtain falling into the candidate target number of Bo Mennei according to step 3 statistics, upgrade targetpath,

If candidate's echo number is zero, do not measure and fall into relevant Bo Mennei, upgrade flight path by the predicted value of step 2, proceed to step 8.1;

Only have one if fall into the measuring value of relevant Bo Mennei, this measuring value can be directly used in flight path renewal, proceeds to step 8.2;

If there is more than one echo to fall into the relevant Bo Mennei of tracked target, Bo Mennei point mark is carried out to quality evaluation, choose optimum point, proceed to step 5;

Step 5, the locus assessed value of calculating each candidate's echo, amplitude assessed value, Doppler's assessed value, frequency of operation assessed value, obtain the assessment vector under each evaluation index, then assessment Vector Groups is combined into an evaluating matrix A;

Step 6, calculate the entropy power method weights of every evaluation index, composition weight vector;

Step 6.1 is established x _ijfor the element in a mark mass matrix A, i=1,2,3 ..., m _k, m _kfor falling into the total number of effective measurement of Bo Mennei, j=1,2,3,4 represent that respectively four measure index, then carry out data translation x _ij=x _ij+ 1, calculate with logarithm when utilizing entropy to ask flexible strategy, meaningless when carrying out data translation and can avoiding taking the logarithm;

The value that step 6.2 is calculated i the factor under j item index shared proportion in this index

Step 6.3 is calculated the entropy of j item index

Step 6.4 is calculated the coefficient of variation of j item index:

Step 6.5 is calculated weights obtain four evaluation index weight vectors;

The weight vector that step 7, the evaluating matrix that step 5 is obtained and step 6 obtain multiplies each other, and obtains final mass assessed value V=Aw, finds position i corresponding to maximal value in assessed value V _op, obtain the optimum z of measurement of Bo Mennei _op(k);

Filter gain matrix K (k)=P (k|k-1) H (k) S (k) of step 8, calculating target t ^-1, the state that obtains upgrades expression formula x (k|k), carries out flight path renewal;

If step 8.1 candidate echo number is zero, do not measure and fall into relevant Bo Mennei, utilize future position to upgrade flight path;

x(k|k)＝x(k|k-1)

P(k|k)＝P(k|k-1)

Only have one if step 8.2 falls into the measuring value of relevant Bo Mennei, this measuring value can be directly used in flight path renewal;

x(k|k)＝x(k|k-1)+K(z)[z(k)-HX(k|k-1)]

P(k|k)＝P(k|k-1)-K(k)HP(k|k-1)

Step 8.3 Ruo Bomennei has multiple points, chooses after a mark quality evaluation and upgrades flight path by optimum point

x(k|k)＝x(k|k-1)+K(z)[z _op(k)-Z(k|k-1)]

P(k|k)＝P(k|k-1)-K(k)HP(k|k-1)。

2. a kind of method for tracking target based on the quality evaluation of entropy weight method point mark as claimed in claim 1, is characterized in that the concrete steps of step 5 are:

Step 5.1 computer memory position assessed value:

In the k moment, m in ripple door _kindividual effective status measures as z _i(k), i=1,2,3 ..., m _k, the filtering residuals vector of i state measurement is v _i(k)=z _i(k)-Z (k|k-1), the statistical distance of itself and predicted position is i the locus assessed value effectively measuring is:

$m_s\left(i\right)=\left\{\begin{array}{cc}\frac{\min (D_1,D_2,...,D_{m_k})}{D_i}& ,D_i\≠0\\ 1& ,D_i=0\end{array}\right.,i=\mathrm{1,2,3},...,m_k;$

Step 5.2 is calculated amplitude assessed value:

In the k moment, m in ripple door _kindividual effective observation amplitude information is a _i(k), i=1,2,3 ..., m _k, definition amplitude assessed value m _a:

$m_a\left(a_i\left(k\right)\right)=\frac{p_1^T\left(a_i\left(k\right)\right)}{p_1^T\left(a_i\left(k\right)\right)+p_0^T\left(a_i\left(k\right)\right)}$

Wherein, p ₀(a) probability density function of amplitude when echoed signal only comes from noise or clutter, p ₁(a) probability density function while comprising echo signal for echoed signal, T is detection threshold, p _fa, p _dbe respectively false-alarm probability, detection probability;

Step 5.3 is calculated Doppler's assessed value:

In the k moment, m in ripple door _kthe doppler information of individual effective observation is d _i(k), i=1,2,3 ..., m _k, the Doppler frequency observed reading of establishing the k-1 time corresponding target of when sampling this tracking gate institute is d (k-1), note observation Doppler differs from Δ d _i(k)=| d _i(k)-d (k-1) |, the assessed value of Doppler frequency is designated as m _d:

$m_d\left(i\right)=\left\{\begin{array}{cc}\frac{\min (\mathrm{\Δ}{d}_1,\mathrm{\Δ}{d}_2,...,\mathrm{\Δ}{d}_{m_k})}{\mathrm{\Δ}{d}_i}& ,\mathrm{\Δ}{d}_i\≠0\\ 1& ,\mathrm{\Δ}{d}_i=0\end{array}\right.;$

The assessed value of step 5.4 frequency of operation:

In the k moment, m in ripple door _kindividual effective observation working frequency information is f _i(k), i=1,2,3 ..., m _k, the frequency of operation observed reading of establishing the k-1 time corresponding target of when sampling this tracking gate institute is f (k-1), remembers Δ f _i(k)=| f _i(k)-f (k-1) |, be observation frequency difference, the assessed value of frequency of operation is designated as m _f, only consider the target of fixed frequency (FIX) type, i the effective m of observation _ffor:

$m_f\left(i\right)=\left\{\begin{array}{cc}1& ,\mathrm{\Δ}{f}_i\≤\mathrm{\ϵf}\\ \frac{{(\mathrm{\Δ}{f}_i-2\mathrm{\ϵf})}^2}{\mathrm{\ϵ}{f}^2}& ,\mathrm{\ϵf}\≤\mathrm{\Δ}{f}_i\≤2\mathrm{\ϵf}\\ 0& ,\mathrm{\Δ}{f}_i\≥2\mathrm{\ϵf}\end{array}\right.$

Wherein ε f is by system noise and the determined frequency measurement tolerance of measurement noise;

Each evaluation index composition point mark quality evaluation matrix A that step 5.5 obtains step 5.1,5.2,5.3,5.4, i.e. A=[m _s; m _a; m _d; m _f] ^t.