CN109991597A - Weak-expansion-target-oriented tracking-before-detection method - Google Patents
Weak-expansion-target-oriented tracking-before-detection method Download PDFInfo
- Publication number
- CN109991597A CN109991597A CN201910273270.2A CN201910273270A CN109991597A CN 109991597 A CN109991597 A CN 109991597A CN 201910273270 A CN201910273270 A CN 201910273270A CN 109991597 A CN109991597 A CN 109991597A
- Authority
- CN
- China
- Prior art keywords
- target
- particle
- moment
- state
- probability
- 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
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
-
- 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
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of target detection and tracking, and particularly discloses a tracking-before-detection method for a weakly extended target. In the invention, under a Bayes filtering framework, a particle approximation means is used for recursively predicting and updating the target state, thereby realizing the detection and tracking of the weakly expanded target, wherein in the updating step, the intensity measurement data accumulation of the resolution units occupied by the target is completed through the particles, and when the accumulation length reaches a set value, the power of each resolution unit occupied by the target is estimated, and the power is used as a parameter of a likelihood function to update the particle weight. The method can realize the detection and tracking of the weakly extended target without target prior information, avoids the complicated nonlinear parameter estimation process of the traditional method, has no specific requirement on the type of the intensity measurement data, can be widely applied to various civil and military systems such as a video monitoring system, a robot navigation system, a military target detection and tracking system and the like, and has wide market prospect and application value.
Description
Technical field
The present invention relates to target detection and tracking technique field, more specifically to a kind of towards weak extension target
Tracking before detecting.
Background technique
Under the background that civilian monitoring and military surveillance demand are increasingly deepened, it is desirable that sensor has to weak extension target
Quickly detection and robust tracking ability.Under Low SNR, compared with traditional Detect before Track (TAD) technology, detection
Preceding tracking (TBD) technology can more effectively realize the joint-detection and tracking of target.TBD technology earliest by Maybeck in
Nineteen eighty-three is initially applied to the target detection of infrared image sequence, is hereafter widely used in radar detection field.The base of TBD technology
This thought is before doing detection judgement, by the echo-signal energy accumulation on target trajectory, to improve target
Detection signal-to-noise ratio realizes the joint-detection and tracking of weak signal target.Since each resolution that TBD technology directly applies sensor to export is single
First intensity measurement data avoids threshold process to loss caused by echo information.
Two classes can be generally divided into the implementation method of TBD technology at present: one kind is advised with three-dimensional matched filtering, dynamic
It draws, Harvard transformation, the batch processing method that the methods of multistage hypothesis testing is representative, it is difficult to guarantee real-time；Another kind of is to pass
Returning Bayesian filter is the Recursion process method of representative, and wherein success of the particle filter approximation in nonlinear application, undoubtedly mentions
The high application value of this method.Research for single goal TBD technology, most of achievements are still to be assumed based on point target.With
The extensive use of high-resolution sensor, point target hypothesis be difficult to meet engineering application requirement.It is existing for weak extension target
There is research still based on TAD technology thinking, and the TBD technological achievement based on intensity measurement data realization joint-detection and tracking is not
It is more.Currently invention addresses trackings before the detection towards weak extension target.
Existing weak extension target TBD algorithm few in number is done under Bayesian filter frame when updating, hypothesis target
Power parameter is a priori known, or measurement collection signal-to-noise ratio is known.It is this kind of since detection target often has Non-synergic
Assuming that being generally difficult to meet.Therefore scholar introduces parameter Estimation thought, by the method for maximal possibility estimation (MLE) to target
Estimated with noise power, but at least there are problems that following two in the algorithm for estimating: first is that target strength obey Lay this
Under conditions of distribution, parameter Estimation needs to solve one group of nonlinearity equation；Second is that obeying exponential distribution in target strength
Under the conditions of, the algorithm for estimating is entirely ineffective.
Summary of the invention
The technical problem to be solved by the present invention is under target prior information unknown condition weak extension target joint-detection with
Tracking.To overcome the shortcomings of in existing method and disadvantage, the present invention provides track side before a kind of detection towards weak extension target
Method, this method accumulate multiframe sensor intensity measurement data, using the average value of ionization meter as source power estimation,
The sensing data that can be effectively adapted under the conditions of different model profiles guarantees the adaptive ability of detection with tracking, realizes weak
Extend quick detection and the tenacious tracking of target.
The present invention is achieved by the following technical solutions:
Tracking before a kind of detection towards weak extension target, this method includes the following steps:
Step 1 initializes filter: model and parameter including determining system, and the prior information of setting system, base
In following steps:
(1.1) target dynamics model x is definedk=f (xk-1)+wkWith sensor ionization meter model zk=h (xk)+vk, wherein
xkIt is the state vector of k moment target, f () is dbjective state transfer function, wkIt is to obey pwThe process noise of distribution, it is assumed that
It is independent identically distributed；zkIt is the ionization meter set of k moment sensor, h () describes dbjective state vector xkWith ionization meter
zkBetween relationship, vkIt is to obey pvIt is distributed and independently of wkMeasurement noise, also assume that be independent identically distributed.
(1.2) system prior information is set.Since filter is Recursion process data, iteration requires one each time
The particle state and survival probability at moment.Therefore, the survival probability of initial time target and for generating new particle state
Distribution function all must priori setting.The setting of prior information including the following three aspects:
A. the survival probability of initial time target is set as p0|0；
B. by the existence m of targetkIt is modeled as two-state Markov Chain, and sets pbAnd psFor constant constant, wherein
pb=prob { mk=1 | mk-1=0 } indicate that k-1 moment target is not present and in the probability of k moment " new life ", ps=prob { mk
=1 | mk-1=1 } indicate that target exists at the k-1 moment and in the probability of k moment " survival "；
C. two are set for generating the distribution function of new particle state, that is, corresponds to pbSuggestion probability distribution bk|k-1
(x) and corresponding to psSuggestion probability distributionUnder conditions of wherein target prior information is unknown, bk|k-1(x)
It can be assumed that be uniformly distributed；It is then derived by by target dynamics model, is pw(xk-f(xk-1))。
(1.3) determine that the population N+B of particle filter and target maximum occupy resolution cell number Lmax, that is, complete filter
Initialization, wherein N indicates each moment " survivals " population, each moment " new life " population of B expression.
Step 2 predicts target survival probability and particle state: since filter is Recursion process in time, thus by k-
The target survival probability p that 1 moment was estimatedk-1|k-1And particle stateIt is transmitted to the k moment,Indicate k-1
The motion state of n-th of particle of moment.It can be determined according to particle filter:
(2.1) pass through target " new life " probability set in step 1, " survival " probability and the survival probability at k-1 moment,
Predict the survival probability of k moment target are as follows:
pk|k-1=pb(1-pk-1|k-1)+pspk-1|k-1
Wherein, pk|k-1For the predicted value of k moment target survival probability.
(2.2) by the two suggestion probability distribution set in step 1, the state of N+B particle of k moment is predictedAre as follows:
Wherein,The predicted state for indicating n-th of particle, with " pseudo- weight " directly proportional to particle weights
Indicate the prediction weight of n-th of particle.
It thereby determines that comprising the particle prediction state set including " new life " particle and " survival " particleWith target k moment survival probability predicted value pk|k-1。
The intensity measurement data that sensor generates is input in filter by step 3: under Bayesian filter frame, being completed
After the predicted operation that step 2 refers to, filter by obtain the measurement data of k moment sensor to the state of particle collection into
Row correction updates.Each resolution cell intensity measurement data that k moment sensor is provided collects input filter as measurement, at this time
The numerical value that measurement collects each unit corresponds to the echo power of the corresponding resolution cell in monitoring region, is denoted as
Wherein I is sensor resolution cell sum,Indicate the echo power of i-th of resolution cell of k moment.
Step 4 target power parameter Estimation: it for intensity measurement data, is contained simultaneously by the resolution cell of object effects
The power information of noise and target.Assuming that the parameter of likelihood function, needs in order to obtain under conditions of noise power a priori known
Target power information is estimated from the measurement data that step 3 inputs.It is different from conventional method, the present invention is existed using particle state
Gradually to the convergent characteristic of target time of day after successive ignition, frame number length W is accumulated in settinglAfterwards, each institute of particle is accumulated
The intensity measurements for accounting for resolution cell take expectation after subtracting noise power, as target after accumulation frame number reaches cumulative length
The estimated value of power parameter.At the k moment, since step 3 inputs the strength measurement message of all resolution cells, each particle will be each
It is run up in corresponding particle measurement set from the ionization meter for occupying resolution cell.When the measuring assembly length of particle reaches Wl
Afterwards, the power parameter P of each resolution cell occupied by corresponding particlen(r) can estimate to obtain according to the following formula:
Wherein t indicates that frame number locating for measurement data, r are r-th of Range resolution unit that particle n is occupied,Indicate t
The echo power of r-th of Range resolution unit in frame measurement data,Indicate the Range resolution unit set that particle n is occupied,
σ2It is measurement noise power, LmaxResolution cell number is occupied for target maximum, i.e.,In order to guarantee this
Estimated value has specific physical significance, i.e., permanent on magnitude of power to be greater than zero, this method, which joined, takes maximum between estimated value and zero
Value Operations.
Step 5 corrects target survival probability and particle state: in Bayesian filter frame, the mesh that is obtained according to step 2
Survival probability predicted value and particle prediction state are marked, on the basis of the measurement data of step 3 input filter, is survived to target general
Rate and particle state are corrected update.Collect z in measurement firstly, for each particle statekAll resolution cells on count
Calculate corresponding likelihood ratioAnd the weight after being corrected using the ratio multiplied by corresponding particle weights as particle,
Target power parameter is provided by step 4 when middle calculating likelihood ratio value；Secondly, combining the weight and step of all particles after correction
2.1 obtained target prediction survival probability pklk-1, to k moment target survival probability pk|kIt is updated；Finally, to the power of particle
It reforms normalization and estimates the state at target k momentDetailed process are as follows:
(5.1) likelihood ratio of each particle is calculated:
Wherein, g1() is the likelihood function in the presence of target, g0() is likelihood function when only having noise；It is to be for stateLikelihood ratio of the particle on entire measurement collection, PnIt (r) is to estimate in step 4
What is obtained corresponds to particleTarget power.
(5.2) particle weights are updated using " pseudo- weight " correction in likelihood ratio and step 2 obtained in step 5.1:
Meanwhile it enabling
(5.3) correction of target survival probability: the power after prediction survival probability and step 5.2 correction of step 2.1 is utilized
Weight more fresh target survival probability:
pk|k=s/ (1-pk|k-1+s)
Wherein,
(5.4) normalized weight and estimate dbjective state:
Wherein,Particle weights after indicating normalization, according to the following formula to target-like under minimum mean square error criterion
State is estimated:
So far, the target survival probability p after available k time correctionk|kWith the state of all particlesThe estimated state at target k moment is obtained simultaneously
Step 6 particle resampling: in order to avoid particle filter leads to the problem of particle exhaustion after successive ignition, system is introduced
Systemization resampling steps.By weighted value size, resampling obtains N number of particle from the particle after correction, and particle is transmitted to
In iterative filtering next time, to complete the Recursion process of filter.
Continuous extension of the step 7 with moment k, iteration step 2~step 6.
Compared with prior art, technical effect of the invention is as follows:
First, the method for the present invention can play the advantage of joint-detection and tracking using TBD technology, in sensor output
It is filtered on each resolution cell intensity measurement data, realization is quickly detected to weak extension target and tenacious tracking.
Second, the method for the present invention, come approximate evaluation target power parameter, is avoided by accumulation that particle measures multiframe
The Solving Nonlinear Systems of Equations difficult problem that MLE estimation method faces；Meanwhile the distribution that the algorithm for estimating meets measurement data
Type does not require specifically, has versatility.
Detailed description of the invention
Fig. 1 is processing method the general frame of the present invention.
Specific embodiment
Technical solution in order to better illustrate the present invention is made embodiments of the present invention below in conjunction with attached drawing further
Description.Fig. 1 is tracking before a kind of detection towards weak extension target of the invention, wherein described is a kind of particle of method
Filter approximate implementation, comprising the following steps:
Step 1 initializes filter: initialization step is related to determining system model and parameter, and the priori of setting system
Information.
(1.1) by the data recipient of the kinematics character of target (such as uniform motion or uniformly accelerated motion) and sensor
Formula defines target dynamics model xk=f (xk-1)+wkWith sensor ionization meter model zk=h (xk)+vk, wherein xkIt is the k moment
The state vector of target, f () are dbjective state transfer function, wkIt is to obey pwThe process noise of distribution, it is assumed that be independent same point
Cloth；zkIt is the ionization meter set of k moment sensor, h () describes dbjective state xkWith ionization meter zkBetween relationship,
vkIt is to obey pvIt is distributed and independently of wkMeasurement noise, also assume that be independent identically distributed.
(1.2) system prior information is set.Since filter is Recursion process data, iteration requires one each time
The particle state and survival probability at moment.Therefore, the survival probability of initial time target and for generating new particle state
Distribution function all must priori setting.The setting of prior information is divided into following three aspects:
A. the survival probability of initial time target is set as p0|0, the usual probability takes lesser numerical value, and preferably value is
0.01；
B. by the existence m of targetkIt is modeled as two-state Markov Chain, and sets pbAnd psFor constant constant, wherein
pb=prob { mk=1 | mk-1=0 } indicate that k-1 moment target is not present and in the probability of k moment " new life ", ps=prob { mk
=1 | mk-1=1 } indicate that target exists at the k-1 moment and in the probability of k moment " survival ".Due to the probability of target " survival "
The significantly larger than probability of target " new life ", therefore, pbUsual value is smaller and psValue is larger, and preferably value is respectively 0.01 He
0.99；
C. two are set for generating the distribution function of new particle state, that is, corresponds to pbSuggestion probability distribution bk|k-1
(x) and corresponding to psSuggestion probability distributionUnder conditions of wherein target prior information is unknown, bk|k-1(x)
It can be assumed that be uniformly distributed；It is then derived by by target dynamics model, is pw(xk-f(xk-1))。
(1.3) determine that the population N+B of particle filter and target maximum occupy resolution cell number Lmax, that is, complete filter
Initialization, wherein N indicates each moment " survivals " population, each moment " new life " population of B expression.
Step 2 predicts target survival probability and particle state: since filter is Recursion process in time, thus by k-
The target survival probability p that 1 moment was estimatedk-1|k-1And particle stateIt is transmitted to the k moment, whereinTable
Show the motion state of n-th of particle.It can be determined according to particle filter:
(2.1) pass through target " new life " probability set in step 1, " survival " probability and the survival probability at k-1 moment,
Predict the survival probability of k moment target are as follows:
pk|k-1=pb(1-pk-1|k-1)+pspk-1|k-1
Wherein, pk|k-1For the predicted value of k moment survival probability.
(2.2) by the two suggestion probability distribution set in step 1, the state of N+B particle of k moment is predictedAre as follows:
Wherein,Indicate the prediction weight of n-th of particle,Indicate the predicted state of n-th of particle；It is counting
" pseudo- weight " is used in calculationMethod simplify and reduce computation complexity, it may be assumed that
It is hereby achieved that including the particle prediction state set including " new life " particle and " survival " particleWith target k moment survival probability predicted value pk|k-1。
The intensity measurement data that sensor generates is input in filter by step 3: under Bayesian filter frame, being completed
After the predicted operation that step 2 refers to, filter need to obtain k moment sensor measurement data could state to particle collection into
Row correction updates.Each resolution cell intensity measurement data that k moment sensor is provided collects input filter as measurement, at this time
The numerical value that measurement collects each unit corresponds to the echo power of the corresponding resolution cell in monitoring region, is denoted as
Wherein I is sensor resolution cell sum,Indicate the echo power of i-th of resolution cell of k moment.
Step 4 target power parameter Estimation: it for intensity measurement data, is contained simultaneously by the resolution cell of object effects
The power information of noise and target.Assuming that the parameter of likelihood function, needs in order to obtain under conditions of noise power a priori known
Target power information is estimated from the measurement data that step 3 inputs.It is different from conventional method, the present invention is existed using particle state
Gradually to the convergent characteristic of target time of day after successive ignition, frame number length W is accumulated in settinglAfterwards, each institute of particle is accumulated
The ionization meter for accounting for resolution cell takes expectation after subtracting noise power, as target function after accumulation frame number reaches cumulative length
The estimated value of rate parameter.At the k moment, since step 3 inputs the strength measurement message of all resolution cells, each particle will be respective
The ionization meter for occupying resolution cell is run up in corresponding particle measurement accumulation set.When the measurement accumulation set length of particle
Reach WlAfterwards, the target power parameter P of each resolution cell occupied by corresponding particlen(r) can estimate to obtain by following formula:
Wherein, r is r-th of Range resolution unit that particle n is occupied,Indicate r-th of distance point in t frame measurement data
Distinguish the echo power of unit,Indicate the Range resolution unit set that particle n is occupied, σ2It is measurement noise power, LmaxFor mesh
Mark maximum occupies resolution cell number, i.e.,It is to guarantee that the estimated value has specific physical significance, i.e., full
Permanent on sufficient magnitude of power to be greater than zero, this method, which joined, is maximized operation between estimated value and zero.
Step 5 corrects target survival probability and particle state: in Bayesian filter frame, the mesh that is obtained according to step 2
Survival probability predicted value and particle prediction state are marked, on the basis of the measurement data of step 3 input filter, is survived to target general
Rate and particle state are corrected update.Collect z in measurement firstly, for each particle statekAll resolution cells on count
Calculate corresponding likelihood ratioAnd the weight after being corrected using the ratio multiplied by corresponding particle weights as particle,
Target power parameter is provided by step 4 when middle calculating likelihood ratio value；Secondly, combining the weight and step of all particles after correction
2.1 obtained target prediction survival probability pk|k-1, to k moment target survival probability pk|kIt is updated；Finally, to the power of particle
It reforms normalization and estimates the state at target k momentDetailed process are as follows:
(5.1) likelihood ratio of each particle is calculated:
Wherein, g1() is the likelihood function in the presence of target, g0() is likelihood function when only having noise；It is to be for stateLikelihood ratio of the particle on entire measurement collection, PnIt (r) is to estimate in step 4
What is obtained corresponds to particleTarget power.
(5.2) particle weights are corrected using " the pseudo- weight " in likelihood ratio and step 2 obtained in step 5.1:
Meanwhile it enabling
(5.3) correction of target survival probability: the power after prediction survival probability and step 5.2 correction of step 2.1 is utilized
Weight more fresh target survival probability:
pk|k=s/ (1-pk|k-1+s)
Wherein,
(5.4) normalized weight and estimate dbjective state:
Wherein,Particle weights after indicating normalization, according to the following formula to target-like under minimum mean square error criterion
State is estimated:
So far, the target survival probability p after available k time correctionk|kWith the state of all particlesThe estimated state at target k moment is obtained simultaneously
Step 6 particle resampling: in order to avoid particle filter leads to the problem of particle exhaustion after successive ignition, system is introduced
Systemization resampling steps.Using the method for systematization resampling, by weighted value size, resampling is obtained from the particle after correction
It is transmitted in the iterative filtering of subsequent time to N number of particle, and by particle, to complete the Recursion process of filter.
Continuous extension of the step 7 with moment k, iteration step 2~step 6.
The above is only one of embodiments of the present invention, protection scope of the present invention is not limited in examples detailed above, all categories
Technical solution under thinking of the present invention all belongs to the scope of protection of the present invention.It should be pointed out that for the common of the art
For technical staff, several improvements and modifications without departing from the principles of the present invention should be considered as falling into guarantor of the invention
Protect range.
Claims (2)
1. tracking before a kind of detection towards weak extension target, which is characterized in that this method includes the following steps:
Step 1 initialize filter: including determine system model and parameter, and setting system prior information, based on
Lower step:
(1.1) target dynamics model x is definedk=f (xk-1)+wkWith sensor ionization meter model zk=h (xk)+vk, wherein xkIt is
The state vector of k moment target, f () are dbjective state transfer function, wkIt is to obey pwThe process noise of distribution, it is assumed that be only
Vertical same distribution；zkIt is the ionization meter set of k moment sensor, h () describes dbjective state vector xkWith ionization meter zkIt
Between relationship, vkIt is to obey pvIt is distributed and independently of wkMeasurement noise, also assume that be independent identically distributed；
(1.2) system prior information is set, following three aspects are divided into:
A. the survival probability of initial time target is set as p0|0；
B. by the existence m of targetkIt is modeled as two-state Markov Chain, and sets pbAnd psFor constant constant, wherein pb=
prob{mk=1 | mk-1=0 } indicate that k-1 moment target is not present and in the probability of k moment " new life ", ps=prob { mk=1 |
mk-1=1 } indicate that target exists at the k-1 moment and in the probability of k moment " survival "；
C. two are set for generating the distribution function of new particle state, that is, corresponds to pbSuggestion probability distribution bk|k-1(x) and
Corresponding to psSuggestion probability distributionUnder conditions of wherein target prior information is unknown, bk|k-1It (x) can be false
It is set to and is uniformly distributed；It is then derived by by target dynamics model, is pw(xk-f(xk-1))；
(1.3) determine that the population N+B of particle filter and target maximum occupy resolution cell number Lmax, i.e., completion filter is first
Beginningization, wherein N indicates each moment " survival " population, and B indicates each moment " new life " population；
Step 2 predicts target survival probability and particle state: the target survival probability p that the k-1 moment is estimatedk-1|k-1And grain
Sub- stateIt is transmitted to the k moment,Indicate the motion state of n-th of particle of k-1 moment；According to particle filter
Device can determine:
(2.1) pass through target " new life " probability set in step 1, " survival " probability and the survival probability at k-1 moment, prediction
The survival probability of k moment target are as follows:
pk|k-1=pb(1-pk-1|k-1)+pspk-1|k-1
Wherein, pk|k-1For the predicted value of k moment target survival probability；
(2.2) by the two suggestion probability distribution set in step 1, the state of N+B particle of k moment is predictedAre as follows:
Wherein,The predicted state for indicating n-th of particle, with " pseudo- weight " directly proportional to particle weightsIt indicates
The prediction weight of n-th of particle；
It thereby determines that comprising the particle prediction state set including " new life " particle and " survival " particleWith target k moment survival probability predicted value pk|k-1；
The intensity measurement data that sensor generates is input in filter by step 3: each resolution that k moment sensor is provided is single
For first intensity measurement data as measurement collection input filter, the numerical value that measurement at this time collects each unit corresponds to monitoring region phase
The echo power for answering resolution cell, is denoted asWherein I is sensor resolution cell sum,Indicate the k moment
The echo power of i-th of resolution cell；
Step 4 target power parameter Estimation: using particle state gradually to the convergent spy of target time of day after successive ignition
Property, frame number length W is accumulated in settinglAfterwards, the intensity measurements of accumulation particle shared resolution cell every time, reach in accumulation frame number
After cumulative length, expectation is taken after subtracting noise power, the estimated value as target power parameter；At the k moment, since step 3 is defeated
Enter the strength measurement message of all resolution cells, each particle runs up to the ionization meter for respectively occupying resolution cell corresponding
In particle measurement set；When the measuring assembly length of particle reaches WlAfterwards, the power ginseng of each resolution cell occupied by corresponding particle
Number Pn(r) can estimate to obtain according to the following formula:
Wherein t indicates that frame number locating for measurement data, r are r-th of Range resolution unit that particle n is occupied,Indicate that t frame is surveyed
The echo power of r-th of Range resolution unit in data is measured,Indicate the Range resolution unit set that particle n is occupied, σ2It is
Measure noise power, LmaxResolution cell number is occupied for target maximum, i.e.,Pass through estimated value and zero
Between be maximized operation and guarantee permanent on magnitude of power to be greater than zero；
Step 5 corrects target survival probability and particle state: collecting z in measurement firstly, for each particle statekAll points
It distinguishes and calculates corresponding likelihood ratio on unitAnd it is corrected using the ratio multiplied by corresponding particle weights as particle
Weight afterwards, wherein target power parameter is provided by step 4 when calculating likelihood ratio；Secondly, all particles after joint correction
The target prediction survival probability p that weight and step 2.1 obtaink|k-1, to k moment target survival probability pk|kIt is updated；Most
Afterwards, normalization is done to the weight of particle and estimates the state at target k momentDetailed process are as follows:
(5.1) likelihood ratio of each particle is calculated:
Wherein, g1() is the likelihood function in the presence of target, g0() is likelihood function when only having noise；It is to be for stateLikelihood ratio of the particle on entire measurement collection, PnIt (r) is to estimate in step 4
What is obtained corresponds to particleTarget power；
(5.2) particle weights are updated using " pseudo- weight " correction in likelihood ratio and step 2 obtained in step 5.1:
Meanwhile it enabling
(5.3) correction of target survival probability: more using the weight after prediction survival probability and step 5.2 correction of step 2.1
Fresh target survival probability:
pk|k=s/ (1-pk|k-1+s)
Wherein,
(5.4) normalized weight and estimate dbjective state:
Wherein,Indicate normalization after particle weights, under minimum mean square error criterion according to the following formula to dbjective state into
Row estimation:
So far, the target survival probability p after available k time correctionk|kWith the state of all particlesThe estimated state at target k moment is obtained simultaneously
Step 6 particle resampling: by weighted value size, resampling obtains N number of particle from the particle after correction, and by particle
It is transmitted in iterative filtering next time, to complete the Recursion process of filter；
Continuous extension of the step 7 with moment k, iteration step 2~step 6.
2. tracking before a kind of detection towards weak extension target according to claim 1, which is characterized in that the step
Suddenly p in (1.2)0|0Preferred value be 0.01, pbAnd psPreferred value be 0.01 and 0.99.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910273270.2A CN109991597A (en) | 2019-04-04 | 2019-04-04 | Weak-expansion-target-oriented tracking-before-detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910273270.2A CN109991597A (en) | 2019-04-04 | 2019-04-04 | Weak-expansion-target-oriented tracking-before-detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109991597A true CN109991597A (en) | 2019-07-09 |
Family
ID=67130933
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910273270.2A Pending CN109991597A (en) | 2019-04-04 | 2019-04-04 | Weak-expansion-target-oriented tracking-before-detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109991597A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111580084A (en) * | 2020-05-13 | 2020-08-25 | 中国人民解放军国防科技大学 | Multi-distance extended target-oriented multi-Bernoulli pre-detection tracking method |
| CN113408422A (en) * | 2021-06-21 | 2021-09-17 | 电子科技大学 | Multi-frame joint detection tracking and classification method suitable for weak targets |
| WO2022116375A1 (en) * | 2020-12-01 | 2022-06-09 | 中国人民解放军海军航空大学 | Method for performing tracking-before-detecting on multiple weak targets by high-resolution sensor |
| CN115097437A (en) * | 2022-06-06 | 2022-09-23 | 哈尔滨工程大学 | Underwater target tracking track approaching and crossing solution method based on label multi-Bernoulli pre-detection tracking algorithm |
| CN115407299A (en) * | 2022-09-15 | 2022-11-29 | 中国人民解放军国防科技大学 | Weak and small target detection method and device based on Bernoulli filter algorithm |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014106133A (en) * | 2012-11-28 | 2014-06-09 | Toshiba Corp | Target detection device, and target detection method |
| CN106569193A (en) * | 2016-11-01 | 2017-04-19 | 西安电子科技大学 | Sea-surface small target detection method based on front-back revenue reference particle filter |
| CN107340514A (en) * | 2017-07-10 | 2017-11-10 | 中国人民解放军海军航空工程学院 | Hypersonic weak signal target RAE HT TBD integration detection methods in three dimensions |
| CN108919225A (en) * | 2018-07-26 | 2018-11-30 | 中国人民解放军海军航空大学 | Range extension target Multichannel fusion detection method under the uniform environment in part |
| CN109471081A (en) * | 2018-11-07 | 2019-03-15 | 中国人民解放军国防科技大学 | Single pulse radar weak and small target combined detection and state estimation method |
-
2019
- 2019-04-04 CN CN201910273270.2A patent/CN109991597A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014106133A (en) * | 2012-11-28 | 2014-06-09 | Toshiba Corp | Target detection device, and target detection method |
| CN106569193A (en) * | 2016-11-01 | 2017-04-19 | 西安电子科技大学 | Sea-surface small target detection method based on front-back revenue reference particle filter |
| CN107340514A (en) * | 2017-07-10 | 2017-11-10 | 中国人民解放军海军航空工程学院 | Hypersonic weak signal target RAE HT TBD integration detection methods in three dimensions |
| CN108919225A (en) * | 2018-07-26 | 2018-11-30 | 中国人民解放军海军航空大学 | Range extension target Multichannel fusion detection method under the uniform environment in part |
| CN109471081A (en) * | 2018-11-07 | 2019-03-15 | 中国人民解放军国防科技大学 | Single pulse radar weak and small target combined detection and state estimation method |
Non-Patent Citations (1)
| Title |
|---|
| 蔡飞: ""雷达弱小目标检测与跟踪技术研究"", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111580084A (en) * | 2020-05-13 | 2020-08-25 | 中国人民解放军国防科技大学 | Multi-distance extended target-oriented multi-Bernoulli pre-detection tracking method |
| WO2022116375A1 (en) * | 2020-12-01 | 2022-06-09 | 中国人民解放军海军航空大学 | Method for performing tracking-before-detecting on multiple weak targets by high-resolution sensor |
| CN113408422A (en) * | 2021-06-21 | 2021-09-17 | 电子科技大学 | Multi-frame joint detection tracking and classification method suitable for weak targets |
| CN115097437A (en) * | 2022-06-06 | 2022-09-23 | 哈尔滨工程大学 | Underwater target tracking track approaching and crossing solution method based on label multi-Bernoulli pre-detection tracking algorithm |
| CN115407299A (en) * | 2022-09-15 | 2022-11-29 | 中国人民解放军国防科技大学 | Weak and small target detection method and device based on Bernoulli filter algorithm |
| CN115407299B (en) * | 2022-09-15 | 2024-06-18 | 中国人民解放军国防科技大学 | Weak and small target detection method and device based on Bernoulli filter algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109991597A (en) | Weak-expansion-target-oriented tracking-before-detection method | |
| CN105717505B (en) | The data correlation method of multiple target tracking is carried out using Sensor Network | |
| CN104766320B (en) | Many Bernoulli Jacob under thresholding is measured filter Faint target detection and tracking | |
| CN107703496B (en) | Interactive multimode Bernoulli filtering maneuvering weak target tracking-before-detection method | |
| CN106526585B (en) | Tracking before target detection based on the filtering of Gaussian particle gesture probability hypothesis density | |
| CN108802722A (en) | It is a kind of based on tracking before the Faint target detection virtually composed | |
| CN105740203A (en) | Multi-sensor passive synergic direction finding and positioning method | |
| CN108333569A (en) | A kind of asynchronous multiple sensors fusion multi-object tracking method based on PHD filtering | |
| CN110058222B (en) | Double-layer particle filter tracking-before-detection method based on sensor selection | |
| CN110187335A (en) | Tracking before being detected for the particle filter with discontinuous characteristic target | |
| CN106291498A (en) | A kind of detecting and tracking combined optimization method based on particle filter | |
| CN107202989A (en) | A kind of complicated Faint target detection and tracking suitable for passive starboard ambiguity of towed linear array sonar | |
| CN110007299A (en) | A kind of dim target detection tracking based on hybrid coordinate puppet spectral technology | |
| CN109239702B (en) | Airport low-altitude flying bird number statistical method based on target state set | |
| CN106295674B (en) | A kind of image object matching detection method and system based on multiparticle group's algorithm | |
| CN107730537A (en) | The Faint target detection and tracking of density filtering are assumed based on case particle probabilities | |
| CN107271973A (en) | CFAR detection method based on degree of skewness and average ratio under Weibull clutter environment | |
| CN109471091A (en) | Method for simultaneously resolving ambiguity, detecting and tracking intermediate repetition frequency PD radar | |
| CN103616680B (en) | Mobile dim target tracking-before-detecting method based on discrete variable rate sampling | |
| CN115755015A (en) | Method, device, equipment and medium for detecting living body in cabin | |
| CN111259332A (en) | Fuzzy data association method and multi-target tracking method in clutter environment | |
| CN110244289A (en) | A kind of adaptive particle filter ground wave radar target integrative detection method | |
| CN116541668A (en) | Swimming stroke number determining method, device, equipment and storage medium | |
| CN113740843B (en) | Motion state estimation method and system for tracking target and electronic device | |
| CN112748735B (en) | Extended target tracking method introducing color features |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190709 |
|
| WD01 | Invention patent application deemed withdrawn after publication |