CN108983215A - A kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter - Google Patents

A kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter Download PDF

Info

Publication number
CN108983215A
CN108983215A CN201810515202.8A CN201810515202A CN108983215A CN 108983215 A CN108983215 A CN 108983215A CN 201810515202 A CN201810515202 A CN 201810515202A CN 108983215 A CN108983215 A CN 108983215A
Authority
CN
China
Prior art keywords
particle
state
target
matrix
measurement
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201810515202.8A
Other languages
Chinese (zh)
Inventor
张勇刚
范颖
***
汪晓雨
李宁
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Engineering University
Original Assignee
Harbin Engineering University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Harbin Engineering University filed Critical Harbin Engineering University
Priority to CN201810515202.8A priority Critical patent/CN108983215A/en
Publication of CN108983215A publication Critical patent/CN108983215A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/66—Radar-tracking systems; Analogous systems
    • G01S13/72—Radar-tracking systems; Analogous systems for two-dimensional tracking, e.g. combination of angle and range tracking, track-while-scan radar

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The present invention provides a kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter, completes the state estimation problem in object tracking process using MCAUPF.In object tracking process, the state equation of target following and measurement equation are reconstructed into nonlinear recursion model first, then importance density function needed for particle filter is generated using maximum cross-correlation entropy Unscented kalman filtering in the frame of no mark particle filter, then resampling is carried out apart from particle of the method for resampling to generation using Kullback-Leibler, the estimation to tracking dbjective state finally is obtained according to the algorithm flow of UPF, to realize the real-time tracking to target.MCAUPF method is applied in measuring the target following that outlier occurs in noise, precision more higher than existing PF, improvement PF and robust filtering can be obtained, and computation complexity is lower than existing improved particle filter algorithm.

Description

A kind of target following based on maximum cross-correlation entropy adaptively without mark particle filter Method
Technical field
The present invention relates to a kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter, are applicable in In the nonlinear system that there is thick tail to measure noise, belong to non linear robust filtering and target following technical field.
Background technique
When being tracked using particle filter (Particle Filter, PF) to target, due to the motor-driven turning of target, The abnormal sudden change point of measurement data, sensor fault, measurement is lost and deliberately the reasons such as interference will lead to measurement noise appearance Outlier, this makes radar on the tracking accuracy of target by serious influence.For this problem, it has been proposed that using Huber without Mark particle filter (Huber Unscented Particle Filter, HRUPF) is handled, but due to influencing letter in HRUPF Number is more than that will not return drop after 1.345 in affecting parameters γ, and estimation performance is caused to decline.For this problem, there has been proposed Outlier robust volume Kalman filtering (Outlier Robust Cubature Kalman based on Student's t method Filter, ORCKF), but the method is only applicable to system noise covariance and measures the lesser situation of noise covariance, at it Performance can reduce under the conditions of him.
In order to solve to measure the Target Tracking Problem in the case of outlier occurs in noise, researcher proposes a new class of be used for The method for solving the problems, such as to measure noise outlier, the i.e. filtering method based on maximum cross-correlation entropy, such as maximum cross-correlation entropy karr Graceful filtering (Maximum Correntropy Kalman Filter, MCKF) and maximum cross-correlation entropy Unscented kalman filtering (Maximum Correntropy Unscented Kalman Filter,MCUKF).It is compared with the traditional method, cross-correlation entropy can To capture the statistical information of higher order rather than common second-order statistics information, therefore better estimated result can be obtained.
In addition, when being calculated using traditional PF, calculation amount is very big, so in field higher for requirement of real-time It closes and is no longer applicable in.For this purpose, there has been proposed based on Kullback-Leibler distance (Kullback-Leibler Distance, KLD) the PF algorithm sampled, but this method postulated particle all is from true posterior density function, in practical applications It is difficult to realize.
Summary of the invention
The purpose of the invention is to provide a kind of target based on maximum cross-correlation entropy adaptively without mark particle filter Tracking utilizes maximum cross-correlation on the basis of no mark particle filter (Unscented Particle Filter, UPF) Entropy criterion is modified the measurement renewal process of UPF, and KLD method is applied to resampling process, it is made to measure noise Occur that there is robustness in the case where outlier, and the number of particle can be adjusted in real time, reduces computation complexity, improve and calculate Efficiency.
The object of the present invention is achieved like this: steps are as follows:
Step 1: the nonlinear discrete state equation of description Target Tracking System is established:
Wherein: k-1 indicates -1 moment of kth, and k indicates kth moment, xk∈RnAnd yk∈RmRespectively indicate the n dimension at kth moment The state vector of target component and the measurement vector of m dimension tracking target component are tracked, f () and h () respectively indicate known Non-linear process function and measurement function, wk-1And vkIt is irrelevant process noise and measurement noise, covariance matrix difference For Qk-1And Rk；Assuming that system noise Gaussian distributed wk-1~N (0, Qk-1), measuring noise includes outlier, obeys mixed Gaussian It is distributed vk~μ N (0, Rk)+δN(0,λRk), it is the Gauss point that μ variance is Σ that q~N (μ, Σ), which indicates that random vector q obeys mean value, Cloth, μ, δ, λ are the parameters for characterizing outlier；
Step 2: initialization:
N number of initial particle is generated according to known prior distribution and their weight is set for 1/N, passes through theory analysis The maximum population N of suitable core width cs and Kullback-Leibler in resampling is selected with practical operating experiencesmax And support domain sizes Δ, the dimension of Δ are necessarily less than the dimension equal to state variable, the dimension of Δ is selected as 1 dimension herein, gives Original stateInitial error covariance matrixWith population N, wherein j represents j-th of particle；According to known priori It is distributed P (x0) extract N number of particleMeet:
And the weight that all particles are arranged is
Step 3: importance sampling:
According to the primary that step 2 extracts, subsequent time is obtained using maximum cross-correlation entropy Unscented kalman filtering Posteriority state estimation and corresponding posteriority state estimation variance generate new Gauss point with this state estimation and variance Cloth randomly selects N number of new particle in the Gaussian Profile new from this, and it is corresponding to calculate each particle according to the particle newly extracted Weight is simultaneously normalized；
Firstly, system state equation and measurement equation to be built into the form of nonlinear recursion equation:
It defines simultaneously:
Wherein: E is averaged to variable, Pk|k-1It is the state one-step prediction error co-variance matrix at kth moment, RkIt is The measurement noise covariance matrix at k moment, Bp,k|k-1It is to Pk|k-1The matrix obtained after Cholesky decomposition is carried out,It is Bp,k|k-1Transposed matrix, Br,kIt is to RkThe matrix obtained after Cholesky decomposition is carried out,It is Br,kTransposed matrix, BkIt is By Bp,k|k-1And Br,kThe new matrix of composition；It is the one-step prediction state value at kth moment；
?The right and left be multiplied by simultaneously:
Ck=gk(xk)+εk
Wherein:
Based on information maximum entropy criterion, cost function is defined are as follows:
Wherein: n+m is CkDimension, ci,kAnd gi,k(xk) it is C respectivelykAnd gk(xk) i-th of element；xkOptimal solution Are as follows:
Wherein: n+m is CkDimension, ci,kAnd gi,k(xk) it is C respectivelykAnd gk(xk) i-th of element, Gσ() is Gauss Kernel function, it may be assumed that
Pass through the error in measurement covariance in amendment renewal processIt obtains:
Wherein: Br,kIt is to RkThe matrix obtained after Cholesky decomposition is carried out,It is Br,kTransposed matrix, Cy,k= diag(Gσ(en+1,k),...,Gσ(en+m,k)), ei,k(i=1,2 ..., n+m) it is εkI-th of element,It is Cy,kInverse square Battle array；；The error in measurement covariance that amendment obtains is applied to obtain in the measurement renewal process of Unscented kalman filtering MCUKF；
Step 4: KLD resampling:
Required population NKLDAre as follows:
Wherein: k is the number of supporting domain, and ε is error threshold, z1-δIt is the upper quantile of standard gaussian distribution；For specific Value δ, upper quantile z corresponding with it1-δIt can be found from gaussian distribution table；
Weight based on particle seriatim chooses particle again, until reaching required population NKLDIf one new Particle is resampled, then just updating NKLDWith the number k of supporting domain；These new particles are adopted again using KLD method Sample, the particle after obtaining resampling, and resetting particle weight is 1/N；
Step 5: state filtering updates:
The filtering of the particle that applying step four obtains, completion status updates, and obtains the estimated state of target following parameter With evaluated error covariance matrix Pk|k, complete the task of target following.
The invention also includes structure features some in this way:
1. σ=10.
2. the dimension of the supporting domain is 1.
3. step 5 specifically:
Wherein:Be the corresponding weight of each particle,It is particle.
Compared with prior art, the beneficial effects of the present invention are: the present invention is by introducing maximum cross-correlation entropy criterion UPF improves the robustness of UPF；The present invention adjusts population using KLD method in the resampling stage in real time, improves calculating Efficiency；The present invention is obtained in four-dimensional situation by relevant document and practical operating experiences, outlier occurs when measuring noise When, take σ=10, Nmax=2000, when Δ=1, MCAUPF proposed by the present invention can obtain than existing PF, improve PF and The higher precision of robust filtering, and computation complexity is lower than existing improvement particle filter, improves the performance of target following.
Detailed description of the invention
Fig. 1 is the flow chart of MCAUPF method of the present invention；
Fig. 2 is the algorithm flow chart of KLD method for resampling in the present invention；
Fig. 3 be method provided by the invention with based on UPF, volume particle filter (Cubature Particle Filter, CPF), to the root-mean-square error curve of target position estimation in the method for tracking target of MCUKF, HRUPF and ORCKF；
Fig. 4 is in method provided by the invention and the method for tracking target based on UPF, CPF, MCUKF, HRUPF and ORCKF To the root-mean-square error curve of target velocity estimation；
When Fig. 5 is method provided by the invention and the operation of the method for tracking target based on UPF, CPF, MCUKF, HRUPF Between contrast table.
Specific embodiment
Present invention is further described in detail with specific embodiment with reference to the accompanying drawing.
A kind of method for tracking target based on MCAUPF of the invention, flow chart is as shown in Figure 1, include following step It is rapid:
Step 1: state equation and the observational equation for establishing description Target Tracking System are as follows:
Wherein, k-1 indicates -1 moment of kth, and k indicates kth moment, xkFor the state of the n dimension tracking target component at kth moment Vector, ykFor the measurement vector of the m dimension tracking target component at+1 moment of kth, f () and h () are known nonlinear function, wk-1System noise, v are tieed up for the n at -1 moment of kthkNoise is measured for the m dimension at kth moment, it is assumed that system noise Gaussian distributed wk-1~N (0, Qk-1), measuring noise includes outlier, obeys Gaussian mixtures vk~μ N (0, Rk)+δN(0,λRk) (q~N (μ, Σ) indicating that random vector q obeys mean value is the Gaussian Profile that μ variance is Σ, and μ, δ, λ is the parameter for characterizing outlier), and wk-1 With vkIt is uncorrelated.
Step 2: initialization, parameter required for being arranged.N number of initial particle is generated simultaneously according to known prior distribution The weight that they are arranged is 1/N, selects suitable core width cs and Kullback-by theory analysis and practical operating experiences Maximum population N in Leibler distance (Kullback-Leibler Distance, KLD) resamplingmaxAnd supporting domain Size (size) Δ gives original stateWith population N.According to known prior distribution P (x0) extract N number of particleMeet
And the weight that all particles are arranged is
Step 3: importance sampling updates N number of particle of initialization, obtains new particle assembly.Using maximum cross-correlation Under entropy Unscented kalman filtering (Maximum Correntropy Unscented Kalman Filter, MCUKF) filtering obtains The posteriority state estimation at one moment and corresponding posteriority state estimation variance generate a Gauss with this state estimation and variance Distribution, randomly selects N number of new particle from this Gaussian Profile, calculates the corresponding weight of each particle according to the particle newly extracted And it is normalized.
System state equation and measurement equation are built into the form of nonlinear recursion equation:
Such as given a definition
B hereinkIt is to decompose corresponding matrix by Cholesky to obtain.In equation (4) the right and left, colleague is multiplied by ?
Ck=gk(xk)+εk (7)
Herein
Based on information maximum entropy criterion, the cost function that is defined as follows
Here n+m is CkDimension, ci,kAnd gi,k(xk) it is C respectivelykAnd gk(xk) i-th of element；xkOptimal solution can To be obtained by following step
N+m is C hereinkDimension, ci,kAnd gi,k(xk) it is C respectivelykAnd gk(xk) i-th of element, Gσ() is Gauss Kernel function, i.e.,
In the case where only considering single-step iteration, the error in measurement covariance in amendment renewal process can be passed throughTo obtain The solution of formula (12)
C hereiny,k=diag (Gσ(en+1,k),...,Gσ(en+m,k)), ei,k(i=1,2 ..., n+m) it is εkI-th yuan Element.
The error in measurement covariance that amendment is obtained is applied to Unscented kalman filtering (Unscented Kalman Filter, UKF) measurement renewal process in can obtain MCUKF.
Step 4: N number of particle that previous step obtains is screened in KLD resampling, obtains final particle collection. Resampling is carried out to these new particles using KLD method, the particle after obtaining resampling, and resetting particle weight is 1/N.
In order to ensure KLD between the Posterior distrbutionp before resampling and the Posterior distrbutionp after resampling is given less than one Error, required population NKLDIt is approximately following value
K is the number of supporting domain, z herein1-δIt is the upper quantile of standard gaussian distribution.It is right with it for specifically value δ The upper quantile z answered1-δIt can be found from gaussian distribution table.
Weight based on particle seriatim chooses particle again, until reaching required population NKLDIf one new Particle is chosen again to be arrived, then just updating NKLDWith the size k of supporting domain.
Step 5: state filtering updates, the state and measurement one-step prediction parameter obtained using preceding step, completion status Filtering update, obtain the estimated state of target following parameterEvaluated error covariance matrixComplete target following Task.
Specifically,
The estimation to target position parameter and speed parameter can be completed at this time, realize the tracking to target.Formula (17) The target position parameter providedThe state estimation error co-variance matrix P provided with formula (18)k|kIt will be used for subsequent time mesh The estimation of cursor position parameter.
Below with reference to specifically giving one embodiment of the invention: since the exception of the motor-driven turning of target, measurement data is prominent Height, sensor fault, measurement is lost and the reasons such as intentional interference will lead to measurement noise and outlier occur, this makes radar to mesh Target tracking accuracy is by serious influence.In this case, existing based on UPF, CPF, MCUKF, HRUPF and ORCKF Method for tracking target performance can reduce.Method provided by the invention has higher robustness than existing method, makes an uproar in measurement Sound can be improved the precision of target following in the case where having outlier.It is of the invention superior to illustrate with specific embodiment below Property.It is specific as follows:
According to a typical four-dimensional radar target tracking model, the state equation of Target Tracking System described below is established And observational equation:
Wherein, k indicates the kth moment；The state vector at kth momentWithRespectively It indicates the position and speed in the direction x and y and assumes that track target is run with constant speed, radar is located at (x0,y0) at；When sampling Between be Δ T=0.02s, initial state vector X0=[0m, 0m/s, 0.4m, -0.05m/s]T。wk-1For the system noise at kth moment Sound, wk~N (0, Qk), and have Qk=diag ([0.5m2,1m2/s2,0.5m2,1m2/s2]), Qk-1Characterize the not true of target velocity It is qualitative；zkIndicate the target bearing measuring value at kth moment；vkFor the measurement noise at kth moment, and there is vk~0.8N (0, Rk)+ 0.2N(0,1000Rk), Rk=diag ([10m2,0.2rad2]), RkCharacterize the uncertainty of azimuthal observation.Radar site is (x0,y0)=(1m, 1m), initial covariance matrix P0=[0.5m2,0.1m2/s2,0.5m2,0.1m2/s2]T, P0At the beginning of characterizing target The uncertainty of beginning position.
Implementation process: using if undefined performance indicator RMSE is come more various filtering methods in simulation process:
Wherein M is Carlo number of Monte, and T is simulation time.It is smaller to the RMSE of target position estimation, characterize target Tracking accuracy is higher, and effect is better.
Simulation time 50 seconds, using based on UPF, CPF, MCUKF, HRUPF and ORCKF and this under identical simulated conditions Invent the method for tracking target based on MCAUPF proposed, wherein MCAUPF and MCUKF method center width cs all take 10, maximum Population be Nmax=7000.All methods carry out 100 Monte Carlo emulation, primary number N=5000.
Implementation result:
During Fig. 3 gives radar target tracking, method provided by the invention and UPF, CPF, MCUKF, HRUPF and Root-mean-square error curve of the ORCKF method to location estimation；It is provided by the invention during Fig. 4 gives radar target tracking Method is with UPF, CPF, MCUKF, HRUPF and ORCKF method to the root-mean-square error curve of velocity estimation.In figs. 3 and 4, Curve A represents the root-mean-square error curve of UPF state estimation, and the root-mean-square error that curve B represents CPF state estimation is bent Line, curve C represent the root-mean-square error curve of HRUPF state estimation, and curve D represents the equal of MCUKF filter state estimated value Square error curve, curve E represent the root-mean-square error curve of ORCKF filter state estimated value, and curve F represents present invention offer MCAUPF method state estimation root-mean-square error curve, the root-mean-square error of state estimation is smaller to represent estimated accuracy Higher, performance is better.From figs. 3 and 4 it can be seen that UPF, CPF do not have robustness to the outlier for measuring noise, as a result send out It dissipates；HRUPF, MCUKF, ORCKF and MCAUPF method proposed by the present invention have robustness to noise outlier is measured, but wherein MCAUPF method robustness proposed by the present invention is most strong, best performance.As it can be seen from table 1 in UPF, CPF, HRUPF and In MCAUPF, MCAUPF runing time is most short, and computation complexity is minimum.
From above embodiments, it is not difficult to find out that, relative to existing method for tracking target, method provided by the invention is being measured Noise occurs that higher precision can be obtained in the case where outlier, has stronger robustness, can obtain more accurate to target Tracking, and computation complexity is lower than existing improvement PF algorithm.
The present invention discloses one kind based on maximum cross-correlation entropy adaptively without mark particle filter (Maximum Correntropy Adaptive Unscented Particle Filter, MCAUPF) method for tracking target.This method In, the state estimation problem in object tracking process is completed using MCAUPF.In object tracking process, first by target following State equation and measurement equation be reconstructed into nonlinear recursion model, then in no mark particle filter (Unscented Particle Filter, UPF) frame in using maximum cross-correlation entropy Unscented kalman filtering (Maximum Correntropy Unscented Kalman Filter, MCUKF) it generates in particle filter (Particle Filter, PF) Required importance density function, then using Kullback-Leibler distance (Kullback-Leibler Distance, KLD) particle progress resampling of the method for resampling to generation, finally obtain according to the algorithm flow of UPF to tracking The estimation of dbjective state, to realize the real-time tracking to target.MCAUPF method is applied and outlier occurs in measurement noise In target following, precision more higher than existing PF, improvement PF and robust filtering can be obtained, and computation complexity is lower than Existing improved particle filter algorithm.

Claims (4)

1. a kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter, it is characterised in that: step It is as follows:
Step 1: the nonlinear discrete state equation of description Target Tracking System is established:
Wherein: k-1 indicates -1 moment of kth, and k indicates kth moment, xk∈RnAnd yk∈RmRespectively indicate the n dimension tracking mesh at kth moment The state vector of parameter and the measurement vector of m dimension tracking target component are marked, f () and h () respectively indicate known non-linear Procedure function and measurement function, wk-1And vkIt is irrelevant process noise and measurement noise, covariance matrix is respectively Qk-1 And Rk；Assuming that system noise Gaussian distributed wk-1~N (0, Qk-1), measuring noise includes outlier, obeys Gaussian mixtures vk ~μ N (0, Rk)+δN(0,λRk), it is Gaussian Profile that μ variance is Σ that q~N (μ, Σ), which indicates that random vector q obeys mean value, μ, δ, λ is the parameter for characterizing outlier；
Step 2: initialization:
N number of initial particle is generated according to known prior distribution and their weight is set for 1/N, passes through theory analysis and reality Border operating experience selects the maximum population N of suitable core width cs and Kullback-Leibler in resamplingmaxAnd Domain sizes Δ is supported, the dimension of Δ is necessarily less than the dimension equal to state variable, gives original stateInitial covariance square Battle arrayWith population N, wherein j represents j-th of particle；According to known prior distribution P (x0) extract N number of particleMeet:
And the weight that all particles are arranged is
Step 3: importance sampling:
According to the primary that step 2 extracts, the posteriority of subsequent time is obtained using maximum cross-correlation entropy Unscented kalman filtering State estimation and corresponding posteriority state estimation variance generate a new Gaussian Profile with this state estimation and variance, from N number of new particle is randomly selected in this new Gaussian Profile, and the corresponding weight of each particle is calculated according to the particle newly extracted And it is normalized；
Firstly, system state equation and measurement equation to be built into the form of nonlinear recursion equation:
It defines simultaneously:
Wherein: E is averaged to variable, Pk|k-1It is the state one-step prediction error co-variance matrix at kth moment, RkWhen being kth The measurement noise covariance matrix at quarter, Bp,k|k-1It is to Pk|k-1The matrix obtained after Cholesky decomposition is carried out,It is Bp,k|k-1Transposed matrix, Br,kIt is to RkThe matrix obtained after Cholesky decomposition is carried out,It is Br,kTransposed matrix, BkIt is By Bp,kk-1And Br,kThe new matrix of composition；It is the one-step prediction state value at kth moment
?The right and left be multiplied by simultaneously:
Ck=gk(xk)+εk
Wherein:
Based on information maximum entropy criterion, cost function is defined are as follows:
Wherein: n+m is CkDimension, ci,kAnd gi,k(xk) it is C respectivelykAnd gk(xk) i-th of element；xkOptimal solution are as follows:
Wherein: n+m is CkDimension, ci,kAnd gi,k(xk) it is C respectivelykAnd gk(xk) i-th of element, Gσ() is Gaussian kernel letter Number, it may be assumed that
Pass through the error in measurement covariance in amendment renewal processIt obtains:
Wherein: Br,kIt is to RkThe matrix obtained after Cholesky decomposition is carried out,It is Br,kTransposed matrix, Cy,k=diag (Gσ (en+1,k),...,Gσ(en+m,k)),It is Cy,kInverse matrix；ei,k(i=1,2 ..., n+m) it is εkI-th of element；It will The error in measurement covariance that amendment obtains is applied to that MCUKF can be obtained in the measurement renewal process of Unscented kalman filtering；
Step 4: KLD resampling:
Required population NKLDAre as follows:
Wherein: k is the number of supporting domain, and ε is error threshold, z1-δIt is the upper quantile of standard gaussian distribution；For specifically value δ, Upper quantile z corresponding with it1-δIt can be found from gaussian distribution table；
Weight based on particle seriatim chooses particle again, until reaching required population NKLDIf a new particle It is resampled, then just updating NKLDWith the number k of supporting domain；Resampling is carried out to these new particles using KLD method, is obtained Particle after to resampling, and resetting particle weight is 1/N；
Step 5: state filtering updates:
The filtering of the particle that applying step four obtains, completion status updates, and obtains the estimated state of target following parameterWith estimate Count error co-variance matrix Pk|k, complete the task of target following.
2. a kind of target following side based on maximum cross-correlation entropy adaptively without mark particle filter according to claim 1 Method, it is characterised in that: σ=10.
3. a kind of target following side based on maximum cross-correlation entropy adaptively without mark particle filter according to claim 2 Method, it is characterised in that: the dimension of the supporting domain is 1.
4. a kind of target following side based on maximum cross-correlation entropy adaptively without mark particle filter according to claim 3 Method, it is characterised in that: step 5 specifically:
Wherein:Be the corresponding weight of each particle,It is particle.
CN201810515202.8A 2018-05-25 2018-05-25 A kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter Pending CN108983215A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810515202.8A CN108983215A (en) 2018-05-25 2018-05-25 A kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810515202.8A CN108983215A (en) 2018-05-25 2018-05-25 A kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter

Publications (1)

Publication Number Publication Date
CN108983215A true CN108983215A (en) 2018-12-11

Family

ID=64542097

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810515202.8A Pending CN108983215A (en) 2018-05-25 2018-05-25 A kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter

Country Status (1)

Country Link
CN (1) CN108983215A (en)

Cited By (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109582915A (en) * 2019-01-28 2019-04-05 杭州电子科技大学 The non-linear observability degree adaptive filter method of improvement applied to bearingsonly tracking
CN109901153A (en) * 2019-03-29 2019-06-18 西安电子科技大学 Targetpath optimization method based on information entropy weight and nearest-neighbor data correlation
CN110161491A (en) * 2019-06-28 2019-08-23 电子科技大学 A kind of ranging and respiratory rate estimation method for faint life entity
CN110187336A (en) * 2019-06-28 2019-08-30 电子科技大学 A kind of positioning of multistation radar site and combined tracking method based on distributed PHD
CN110334322A (en) * 2019-06-26 2019-10-15 电子科技大学 A kind of population adaptive approach of particle filter
CN110361744A (en) * 2019-07-09 2019-10-22 哈尔滨工程大学 RBMCDA underwater multi-target tracking based on Density Clustering
CN110442941A (en) * 2019-07-25 2019-11-12 桂林电子科技大学 It is a kind of to be tracked and RUL prediction technique based on the battery status for improving particle filter and process noise features fusion algorithm
CN110492866A (en) * 2019-07-22 2019-11-22 航天东方红卫星有限公司 A kind of kalman filter method of pair of moving target
CN110532517A (en) * 2019-08-28 2019-12-03 北京理工大学 Gas pipeline method for parameter estimation based on improved ARUKF
CN110598614A (en) * 2019-09-04 2019-12-20 南京邮电大学 Related filtering target tracking method combined with particle filtering
CN111711432A (en) * 2020-06-16 2020-09-25 桂林理工大学 Target tracking algorithm based on UKF and PF hybrid filtering
CN112800889A (en) * 2021-01-18 2021-05-14 浙江工业大学 Target tracking method based on distributed matrix weighting and Gaussian filtering fusion
CN113189985A (en) * 2021-04-16 2021-07-30 南京大学 Partially observable driving planning method based on adaptive particle and belief filling
CN113189578A (en) * 2021-04-20 2021-07-30 浙江大学 Extended target tracking method
CN113258965A (en) * 2021-05-17 2021-08-13 东南大学 Millimeter wave distributed MIMO system AOA tracking method based on unscented Kalman filtering
CN113625309A (en) * 2021-07-23 2021-11-09 哈尔滨工程大学 Multipath parameter estimation method based on unscented difference particle filter algorithm
CN114088086A (en) * 2021-11-23 2022-02-25 江苏科技大学 Multi-target robust positioning method for resisting measurement outlier interference
CN114124033A (en) * 2021-10-11 2022-03-01 北京川速微波科技有限公司 Kalman filter implementation method, device, storage medium and equipment
CN114445459A (en) * 2022-01-16 2022-05-06 中国人民解放军空军工程大学 Continuous-discrete maximum correlation entropy target tracking method based on variational Bayes theory
CN115248415A (en) * 2022-06-27 2022-10-28 中国矿业大学 Target positioning and tracking method for mine non-line-of-sight scene
CN115859039A (en) * 2023-03-01 2023-03-28 南京信息工程大学 Vehicle state estimation method
CN117351042A (en) * 2023-10-08 2024-01-05 哈尔滨工业大学 Arbitrary step time delay target tracking method based on Gaussian sum filtering

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104376581A (en) * 2014-12-02 2015-02-25 北京航空航天大学 Gaussian mixture unscented particle filter algorithm employing adaptive resampling

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104376581A (en) * 2014-12-02 2015-02-25 北京航空航天大学 Gaussian mixture unscented particle filter algorithm employing adaptive resampling

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
XI LIU ET AL.: "Maximum Correntropy Unscented Kalman Filter for Spacecraft Relative State Estimation", 《SENSORS》 *
裴福俊等: "基于KLD采样的自适应UPF非线性状态估计方法", 《***仿真学报》 *

Cited By (34)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109582915B (en) * 2019-01-28 2023-08-18 杭州电子科技大学 Improved nonlinear observability self-adaptive filtering method applied to pure azimuth tracking
CN109582915A (en) * 2019-01-28 2019-04-05 杭州电子科技大学 The non-linear observability degree adaptive filter method of improvement applied to bearingsonly tracking
CN109901153A (en) * 2019-03-29 2019-06-18 西安电子科技大学 Targetpath optimization method based on information entropy weight and nearest-neighbor data correlation
CN109901153B (en) * 2019-03-29 2022-11-01 西安电子科技大学 Target track optimization method based on information entropy weight and nearest neighbor data association
CN110334322A (en) * 2019-06-26 2019-10-15 电子科技大学 A kind of population adaptive approach of particle filter
CN110161491A (en) * 2019-06-28 2019-08-23 电子科技大学 A kind of ranging and respiratory rate estimation method for faint life entity
CN110187336A (en) * 2019-06-28 2019-08-30 电子科技大学 A kind of positioning of multistation radar site and combined tracking method based on distributed PHD
CN110361744A (en) * 2019-07-09 2019-10-22 哈尔滨工程大学 RBMCDA underwater multi-target tracking based on Density Clustering
CN110361744B (en) * 2019-07-09 2022-11-01 哈尔滨工程大学 RBMCDA underwater multi-target tracking method based on density clustering
CN110492866B (en) * 2019-07-22 2022-12-09 航天东方红卫星有限公司 Kalman filtering method for moving target
CN110492866A (en) * 2019-07-22 2019-11-22 航天东方红卫星有限公司 A kind of kalman filter method of pair of moving target
CN110442941A (en) * 2019-07-25 2019-11-12 桂林电子科技大学 It is a kind of to be tracked and RUL prediction technique based on the battery status for improving particle filter and process noise features fusion algorithm
CN110532517A (en) * 2019-08-28 2019-12-03 北京理工大学 Gas pipeline method for parameter estimation based on improved ARUKF
CN110598614A (en) * 2019-09-04 2019-12-20 南京邮电大学 Related filtering target tracking method combined with particle filtering
CN111711432A (en) * 2020-06-16 2020-09-25 桂林理工大学 Target tracking algorithm based on UKF and PF hybrid filtering
CN111711432B (en) * 2020-06-16 2023-03-28 桂林理工大学 Target tracking algorithm based on UKF and PF hybrid filtering
CN112800889B (en) * 2021-01-18 2024-02-23 浙江工业大学 Target tracking method based on distributed matrix weighted fusion Gaussian filtering
CN112800889A (en) * 2021-01-18 2021-05-14 浙江工业大学 Target tracking method based on distributed matrix weighting and Gaussian filtering fusion
CN113189985A (en) * 2021-04-16 2021-07-30 南京大学 Partially observable driving planning method based on adaptive particle and belief filling
CN113189578B (en) * 2021-04-20 2022-09-16 浙江大学 Extended target tracking method
CN113189578A (en) * 2021-04-20 2021-07-30 浙江大学 Extended target tracking method
CN113258965A (en) * 2021-05-17 2021-08-13 东南大学 Millimeter wave distributed MIMO system AOA tracking method based on unscented Kalman filtering
CN113625309A (en) * 2021-07-23 2021-11-09 哈尔滨工程大学 Multipath parameter estimation method based on unscented difference particle filter algorithm
CN113625309B (en) * 2021-07-23 2023-09-29 哈尔滨工程大学 Multipath parameter estimation method based on unscented differential particle filter algorithm
CN114124033A (en) * 2021-10-11 2022-03-01 北京川速微波科技有限公司 Kalman filter implementation method, device, storage medium and equipment
CN114088086A (en) * 2021-11-23 2022-02-25 江苏科技大学 Multi-target robust positioning method for resisting measurement outlier interference
CN114088086B (en) * 2021-11-23 2023-11-24 江苏科技大学 Multi-target robust positioning method for resisting measurement wild value interference
CN114445459A (en) * 2022-01-16 2022-05-06 中国人民解放军空军工程大学 Continuous-discrete maximum correlation entropy target tracking method based on variational Bayes theory
CN114445459B (en) * 2022-01-16 2024-04-09 中国人民解放军空军工程大学 Continuous-discrete maximum correlation entropy target tracking method based on variable decibel leaf theory
CN115248415A (en) * 2022-06-27 2022-10-28 中国矿业大学 Target positioning and tracking method for mine non-line-of-sight scene
CN115248415B (en) * 2022-06-27 2024-05-14 中国矿业大学 Target positioning and tracking method for mine non-line-of-sight scene
CN115859039A (en) * 2023-03-01 2023-03-28 南京信息工程大学 Vehicle state estimation method
CN117351042A (en) * 2023-10-08 2024-01-05 哈尔滨工业大学 Arbitrary step time delay target tracking method based on Gaussian sum filtering
CN117351042B (en) * 2023-10-08 2024-05-14 哈尔滨工业大学 Arbitrary step time delay target tracking method based on Gaussian sum filtering

Similar Documents

Publication Publication Date Title
CN108983215A (en) A kind of method for tracking target based on maximum cross-correlation entropy adaptively without mark particle filter
CN108802692A (en) A kind of method for tracking target based on maximum cross-correlation entropy volume particle filter
Dini et al. Class of widely linear complex Kalman filters
CN109084767A (en) A kind of AUV collaborative navigation method of the adaptive volume particle filter of maximum cross-correlation entropy
CN108489498A (en) A kind of AUV collaborative navigation methods without mark particle filter based on maximum cross-correlation entropy
CN104376581B (en) A kind of Gaussian Mixture using adaptive resampling is without mark particle filter algorithm
Hao et al. Comparison of unscented kalman filters
CN106600557B (en) PSF estimation method based on mixed Gauss model and sparse constraint
CN108896986A (en) A kind of measurement conversion Sequential filter maneuvering target tracking method based on predicted value
CN110503071A (en) Multi-object tracking method based on the more Bernoulli Jacob's Additive Models of variation Bayes's label
CN110068805A (en) High-speed target HRRP reconstruction method based on variational Bayesian inference
CN102184408B (en) Autoregressive-model-based high range resolution profile radar target recognition method
CN107705321A (en) Moving object detection and tracking method based on embedded system
Shi et al. Adaptive UKF for target tracking with unknown process noise statistics
CN105844627A (en) Sea surface object image background inhibition method based on convolution nerve network
CN111125885A (en) ASF correction table construction method based on improved kriging interpolation algorithm
CN107843875A (en) Bayes's compressed sensing Radar Data Fusion method based on singular value decomposition noise reduction
CN103268525A (en) WD-RBF (wavelet denoising-radial basis function)-based analogue prediction method of hydrological time sequence
CN110231620A (en) A kind of noise correlation system tracking filter method
CN111105068A (en) Numerical value mode correction method based on sequence regression learning
CN110456351A (en) Based on when Variable Amplitude LFM Signal parameter estimation ISAR Imaging of Maneuvering Targets method
CN113223055A (en) Image target tracking model establishing method and image target tracking method
CN106296727A (en) A kind of resampling particle filter algorithm based on Gauss disturbance
CN111263295B (en) WLAN indoor positioning method and device
CN104463245A (en) Target recognition method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20181211