CN103679753A - Track identifying method of probability hypothesis density filter and track identifying system - Google Patents
Track identifying method of probability hypothesis density filter and track identifying system Download PDFInfo
- Publication number
- CN103679753A CN103679753A CN201310690685.2A CN201310690685A CN103679753A CN 103679753 A CN103679753 A CN 103679753A CN 201310690685 A CN201310690685 A CN 201310690685A CN 103679753 A CN103679753 A CN 103679753A
- Authority
- CN
- China
- Prior art keywords
- gauss
- item
- identify label
- constantly
- weight
- 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
Images
Abstract
The invention provides a track identifying method of a probability hypothesis density filter suitable for the technical field of multi-sensors. The track identifying method includes the first step of determining a predicted gauss item and an identity identification of the gauss item according to a gauss item and an identity identification of the gauss item at the previous moment and adding a non-exclusive identity identification to each newly produced gauss item, the second step of determining updated gauss items and identity identifications of the updated gauss items according to the predicted gauss item, the newly produced gauss items and the corresponding identity identifications, the third step of cutting and combining the updated gauss items and the identity identifications, the fourth step of regulating the identity identifications whose weight is larger than a preset weight threshold value according to the cut and combined gauss items and the corresponding identity identifications, and the fifth step of extracting the gauss items whose weight is larger than the preset weight threshold value to serve as output of the filter and outputting the corresponding identity identifications. According to the method, the identity identifications are added to the gauss items, target states of different moments are correlated, and accordingly motion trails of targets are obtained.
Description
Technical field
The present invention relates to multiple sensor information amalgamation method, relate in particular to a kind of track identification method and track tag system of probability hypothesis density wave filter.
Background technology
In the situation that existing false-alarm, undetected and number of targets unknown, the probability hypothesis density wave filter that Mahler proposes is the new method that solves target detection and tracking.Probability hypothesis density wave filter has been avoided the direct correlation between observed reading and state value, and its great advantage is from posteriority square, to estimate number of targets.For solving the reluctant problem of integral operation in the prediction of probability hypothesis density wave filter and renewal equation, Vo etc. have proposed particle probability hypothesis density wave filter and Gaussian Mixture probability hypothesis density wave filter.At present, probability hypothesis density wave filter has been obtained application more widely in the field such as the target following in passive positioning, passive radar target following, video tracking, sonar image and group target following.
Yet, although existing probability hypothesis density filtering algorithm can effectively estimate state and the target number of target, but owing to having avoided data correlation, so cannot determine not the corresponding relation of dbjective state in the same time, can not obtain the movement locus of each target.The movement locus problem of establishing target is a key technical problem that needs exploration in probability hypothesis density wave filter and solve.
Summary of the invention
Technical matters to be solved by this invention is to provide a kind of track identification method and track tag system of probability hypothesis density wave filter, is intended to solve while using probability hypothesis density wave filter to multiple target tracking, and target trajectory is difficult to definite problem.
The present invention is achieved in that a kind of track identification method of probability hypothesis density wave filter, comprises the following steps:
Step 1: determine Gauss's item of prediction and the identify label of prediction term according to the identify label of Gauss's item of previous moment and previous moment, and add non-proprietary identify label to Gauss's item of each current time new life;
Step 2: determine the identify label of Gauss's item and the renewal item of renewal according to the identify label of Gauss's item of the identify label of Gauss's item of described prediction, prediction term, new life and new life's item;
Step 3, the identify label of Gauss's item of described renewal and described renewal item is carried out to cutting and merging;
Step 4, Gauss's item and identify label thereof according to cutting and after merging, the identify label that weight is greater than to Gauss's item of default weight threshold is adjusted, and makes identify label have uniqueness;
Step 5, extract weight and be greater than Gauss's item of default weight threshold as the output of wave filter, and identify label after adjusting, that weight is greater than Gauss's item of default weight threshold is exported together.
Further, in described step 1, with k-1, represent previous moment, with k, represent current time, described non-proprietary identify label 0 mark, k-1 Gauss's item set expression is constantly
i=1,2 ..., n
k-1, k-1 identify label set expression is constantly
i=1,2 ..., n
k-1,
not being to show that Gauss's item identity specifies at 0 o'clock, is that 0 o'clock indicate identification is not yet specified, and wherein, i is call number, and value is from 1 to n
k-1, n
k-1represent the k-1 number of Gauss's item constantly,
represent the k-1 weight of Gauss's item constantly,
represent the k-1 average of Gauss's item constantly,
represent the k-1 variance of Gauss's item constantly;
By k-1 Gauss's item constantly, gathered
i=1,2 ..., n
k-1with k-1 identify label set constantly
i=1,2 ..., n
k-1obtain Gauss's item set that k predicts constantly
i=1,2 ..., n
k-1, and k predicts the identify label set of Gauss's item constantly
i=1,2 ..., n
k-1, wherein
represent that k predicts the weight of Gauss's item constantly,
represent that k predicts the average of Gauss's item constantly,
represent that k predicts the variance of Gauss's item constantly,
Here, p
sfor the probability that target is survived, F
k-1for state-transition matrix, Q
k-1for process noise covariance matrix; K constantly Gauss's item set of new life is
i=1,2 ..., n
γ, k, wherein
be given parameter, represent respectively weight, average and the variance of newborn Gauss's item, n
γ,
knumber for newborn Gauss's item; The k constantly identify label set of newborn Gauss's item is
i=1,2 ..., n
γ, k, the described k identify label unification of newborn Gauss's item is constantly set to non-proprietary sign 0,
i=1,2 ..., n
γ, k.
Further, in described step 2, Gauss's item set of constantly predicting according to described k
i=1,2 ..., n
k-1with k Gauss's item set of new life constantly
i=1,2 ..., n
γ, kand the identify label set of prediction Gauss item
i=1,2 ..., n
k-1identify label set with newborn Gauss's item
i=1,2 ..., n
γ, k, obtain Gauss's item set that k upgrades constantly
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k) and upgrade the identify label set of Gauss's item
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k), wherein, m
kfor k moment measuring set Z
kthe number of middle measurement,
represent that k upgrades the weight of Gauss's item constantly,
represent that k upgrades the average of Gauss's item constantly,
represent that k upgrades the variance of Gauss's item constantly;
As i≤n
k-1+ n
γ, ktime,
Work as i>n
k-1+ n
γ, ktime, make q=int[(i-1)/(n
k+ n
γ, k)], p=i-q * (n
k+ n
γ, k),
?
Wherein, int represents to round, p
dfor acquisition probability, λ
cfor clutter density, H
kfor observing matrix, R
kfor the covariance of observation noise, K
pfor filter gain,
be the average of p prediction or newborn Gauss's item,
be the variance of p prediction or newborn Gauss's item, I is unit matrix,
be the weight of p prediction or newborn Gauss's item, N represents Gaussian distribution, the state that ξ is target,
be the average of e prediction or newborn Gauss's item,
be the variance of e prediction or newborn Gauss's item, z
qfor k moment measuring set Z
kin q measurement, the number of the Gauss's item after renewal is n
k|k=(n
k-1+ n
γ, k) (m
k+ 1).
Further, in described step 3, Gauss's item that k upgrades is constantly expressed as
i=1,2 ..., n
k|k, the identify label set that k upgrades Gauss's item is constantly
i=1,2 ..., n
k|k, n wherein
k|k=(n
k-1+ n
γ, k) (m
k+ 1); Delete weight fully little,
gauss's item, wherein, τ is cutting thresholding;
Will be apart from d in Gauss's item of remainder from cutting
ijgauss's item of <U is merged into one, and wherein U is for merging thresholding, and combined distance is defined as
after merging, the identify label of Gauss's item determines that principle is: if the identify label of Gauss's item is 0 entirely before merging, after merging, the identify label of Gauss's item is also 0; Otherwise identify label is not in Gauss's item of 0 before merging, the heavy maximum Gauss's item identify label of weighting is as the identify label of Gauss's item after merging; When Gauss's item is merged, weight is greater than other weight of Gauss Xiang Buyu of 0.5 and is greater than Gauss's item of 0.5 and merges.
Further, in described step 4, according to cutting with merge after Gauss's item, the identify label that weight is greater than to Gauss's item of 0.5 is judged: if the identify label of Gauss's item is greater than 0, show that its identity is definite; If 0, show that its identity not yet specifies, now need to specify a unique non-zero sign as its proprietary identify label to it.
Further, for making proprietary identify label there is uniqueness, after a proprietary identify label is used, proprietary identify label adds 1 automatically, if a plurality of Gauss's items have identical proprietary identify label, the identify label of Gauss's item that weight coefficient is the highest remains unchanged, and the identify label of other Gauss's item is reassigned.
The present invention also provides a kind of track tag system of probability hypothesis density wave filter, comprising:
Prediction module, determines Gauss's item of prediction and the identify label of prediction according to the identify label of Gauss's item of previous moment and previous moment, and adds non-proprietary identify label to each newborn Gauss's item;
Update module, determines the identify label of Gauss's item and the renewal item of renewal according to the identify label of Gauss's item of the identify label of Gauss's item of described prediction, prediction term, new life and new life's item;
Cutting and merging module, carry out cutting and merging to the identify label of Gauss's item of described renewal and described renewal item;
Adjusting module, Gauss's item and identify label thereof according to cutting and after merging, the identify label that weight is greater than to Gauss's item of default weight threshold is adjusted, and makes identify label have uniqueness;
Dbjective state extraction module, extracts weight and is greater than Gauss's item of default weight threshold as the output of wave filter, and identify label after adjusting, that weight is greater than Gauss's item of default weight threshold is exported together.
Further, in described prediction module, with k-1, represent previous moment, with k, represent current time, described non-proprietary identify label 0 mark, k-1 Gauss's item set expression is constantly
i=1,2 ..., n
k-1, k-1 identify label set expression is constantly
i=1,2 ..., n
k-1,
not being to show that Gauss's item identity specifies at 0 o'clock, is that 0 o'clock indicate identification is not yet specified, and wherein, i is call number, and value is from 1 to n
k-1, n
k-1represent the k-1 number of Gauss's item constantly,
represent the k-1 weight of Gauss's item constantly,
represent the k-1 average of Gauss's item constantly,
represent the k-1 variance of Gauss's item constantly;
By k-1 Gauss's item constantly, gathered
i=1,2 ..., n
k-1with k-1 identify label set constantly
i=1,2 ..., n
k-1obtain Gauss's item set that k predicts constantly
i=1,2 ..., n
k-1, and the identify label set of k prediction Gauss item constantly
i=1,2 ..., n
k-1, wherein
represent that k predicts the weight of Gauss's item constantly,
represent that k predicts the average of Gauss's item constantly,
represent that k predicts the variance of Gauss's item constantly,
Here, p
sfor the probability that target is survived, F
k-1for state-transition matrix, Q
k-1for process noise covariance matrix; K constantly Gauss's item set of new life is
i=1,2 ..., n
γ, k, wherein
be given parameter, represent respectively weight, average and the variance of newborn Gauss's item, n
γ, knumber for newborn Gauss's item; The k constantly identify label set of newborn Gauss's item is
i=1,2 ..., n
γ, k, and the described k identify label unification of newborn Gauss's item is constantly set to non-proprietary sign 0,
i=1,2 ..., n
γ, k.
Further, in described update module, Gauss's item set of constantly predicting according to described k
i=1,2 ..., n
k-1with k Gauss's item set of new life constantly
i=1,2 ..., n
γ, kand the identify label set of prediction Gauss item
i=1,2 ..., n
k-1identify label set with newborn Gauss's item
i=1,2 ..., n
γ, k, obtain Gauss's item set that k upgrades constantly
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k) and upgrade the identify label set of Gauss's item
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k), wherein, m
kfor k moment measuring set Z
kthe number of middle measurement,
represent that k upgrades the weight of Gauss's item constantly,
represent that k upgrades the average of Gauss's item constantly,
represent that k upgrades the variance of Gauss's item constantly;
As i≤n
k-1+ n
γ, ktime,
Work as i>n
k-1+ n
γ, ktime, make q=int[(i-1)/(n
k+ n
γ, k)], p=i-q * (n
k+ n
γ, k),
?
Wherein, int represents to round, p
dfor acquisition probability, λ
cfor clutter density, H
kfor observing matrix, R
kfor the covariance of observation noise, K
pfor filter gain,
be the average of p prediction or newborn Gauss's item,
be the variance of p prediction or newborn Gauss's item, I is unit matrix,
be the weight of p prediction or newborn Gauss's item, N represents Gaussian distribution, the state that ξ is target,
be the average of e prediction or newborn Gauss's item,
be the variance of e prediction or newborn Gauss's item, z
qfor k moment measuring set Z
kin q measurement, the number of the Gauss's item after renewal is n
k|k=(n
k-1+ n
γ, k) (m
k+ 1).
Further, in cutting and merging module, Gauss's item that k upgrades is constantly expressed as
i=1,2 ..., n
k|k, the identify label set that k upgrades Gauss's item is constantly
i=1,2 ..., n
k|k, n wherein
k|k=(n
k-1+ n
γ, k) (m
k+ 1); Delete weight fully little,
gauss's item, wherein, τ is cutting thresholding;
Will be apart from d in Gauss's item of remainder from cutting
ijgauss's item of <U is merged into one, and wherein U is for merging thresholding, and combined distance is defined as
after merging, the identify label of Gauss's item determines that principle is: if the identify label of Gauss's item is 0 entirely before merging, after merging, the identify label of Gauss's item is also 0; Otherwise identify label is not in Gauss's item of 0 before merging, the heavy maximum Gauss's item identify label of weighting is as the identify label of Gauss's item after merging; When Gauss's item is merged, weight is greater than other weight of Gauss Xiang Buyu of 0.5 and is greater than Gauss's item of 0.5 and merges;
In described adjusting module, according to cutting with merge after Gauss's item, the identify label that weight is greater than to Gauss's item of 0.5 is judged: if the identify label of Gauss's item is greater than 0, show that its identity is definite; If 0, show that its identity not yet specifies, now need to specify a unique non-zero sign as its proprietary identify label to it.
Compare with existing probability hypothesis density wave filter, the track identification method of probability hypothesis density wave filter of the present invention utilizes to Gauss's item and adds identify label, will be not in the same time dbjective state associate, thereby obtain the movement locus of each target, and each different target can be made a distinction.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of track identification method in the probability hypothesis density wave filter that provides of the embodiment of the present invention;
Fig. 2 is the block diagram of the track tag system of the probability hypothesis density wave filter that provides of the embodiment of the present invention;
Fig. 3 is the emulated data that the track identification method of probability hypothesis density wave filter of the present invention adopts;
Fig. 4 is that verification and measurement ratio is 0.99 o'clock definite target trajectory of the inventive method.
Embodiment
In order to make object of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not intended to limit the present invention.
The track identification method of probability hypothesis density wave filter of the present invention adds proprietary identify label can to the Gauss's item as wave filter output, when wave filter prediction, renewal and cutting merge, utilize inheritance to carry out respective handling to the identify label of Gauss's item, during wave filter output, proprietary identify label is exported together in company with dbjective state, can determine the movement locus of each target.
As shown in Figure 1, be this bright preferred embodiment, a kind of track identification method of probability hypothesis density wave filter, comprises the following steps:
Step 1: determine Gauss's item of prediction and the identify label of prediction term according to the identify label of Gauss's item of previous moment and previous moment, and add non-proprietary identify label to each newborn Gauss's item.
With k-1, represent previous moment, with k, represent current time, non-proprietary identify label 0 mark, k-1 Gauss's item set expression is constantly
i=1,2 ..., n
k-1, k-1 identify label set expression is constantly
i=1,2 ..., n
k-1,
not being to show that Gauss's item identity specifies at 0 o'clock, is that 0 o'clock indicate identification is not yet specified, and wherein, i is call number, and value is from 1 to n
k-1, n
k-1represent the k-1 number of Gauss's item constantly,
represent the k-1 weight of Gauss's item constantly,
represent the k-1 average of Gauss's item constantly,
represent the k-1 variance of Gauss's item constantly.
By k-1 Gauss's item constantly, gathered
i=1,2 ..., n
k-1with k-1 identify label set constantly
i=1,2 ..., n
k-1obtain Gauss's item set that k predicts constantly
i=1,2 ..., n
k-1, and k predicts the identify label set of Gauss's item constantly
i=1,2 ..., n
k-1, wherein
represent that k predicts the weight of Gauss's item constantly,
represent that k predicts the variance of Gauss's item constantly,
Here, p
sfor the probability that target is survived, F
k-1for state-transition matrix, Q
k-1for process noise covariance matrix; K constantly Gauss's item set of new life is
=1,2 ..., n
γ, k, wherein
be given parameter, represent respectively weight, average and the variance of newborn Gauss's item, n
γ, knumber for newborn Gauss's item.The k constantly identify label set of newborn Gauss's item is
i=1,2 ..., n
γ, k, the described k identify label unification of newborn Gauss's item is constantly set to non-proprietary sign 0,
i=1,2 ..., n
γ, k.
Step 2: determine the identify label of Gauss's item and the renewal item of renewal according to the identify label of Gauss's item of the identify label of Gauss's item of described prediction, prediction term, new life and new life's item.
Gauss's item set of constantly predicting according to described k
i=1,2 ..., n
k-1with k Gauss's item set of new life constantly
i=1,2 ..., n
γ, kand the identify label set of prediction Gauss item
i=1,2 ..., n
k-1identify label set with newborn Gauss's item
i=1,2 ..., n
γ, k, obtain Gauss's item set that k upgrades constantly
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k) and upgrade the identify label set of Gauss's item
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k), wherein, m
kfor k moment measuring set Z
kthe number of middle measurement,
represent that k upgrades the weight of Gauss's item constantly,
represent that k upgrades the average of Gauss's item constantly,
represent that k upgrades the variance of Gauss's item constantly.
As i≤n
k-1+ n
γ, ktime,
Work as i>n
k-1+ n
γ, ktime, make q=int[(i-1)/(n
k+ n
γ, k)], p=i-q * (n
k+ n
γ, k),
?
Wherein, int represents to round, p
dfor acquisition probability, λ
cfor clutter density (being the clutter number in unit volume), H
kfor observing matrix, R
kfor the covariance of observation noise, K
pfor Kalman filter gain,
be the average of p prediction or newborn Gauss's item,
be the variance of p prediction or newborn Gauss's item, I is unit matrix,
be the weight of p prediction or newborn Gauss's item, N represents Gaussian distribution, the state that ξ is target,
be the average of e prediction or newborn Gauss's item,
be the variance of e prediction or newborn Gauss's item, z
qfor k moment measuring set Z
kin q measurement, the number of the Gauss's item after renewal is n
k|kindividual, n
k|k=(n
k-1+ n
γ, k) (m
k+ 1).
Step 3, the identify label of Gauss's item of described renewal and described renewal is carried out to cutting and merging.
Gauss's item that k upgrades is constantly expressed as
i=1,2 ..., n
k|k, the identify label set that k upgrades Gauss's item is constantly
i=1,2 ..., n
k|k, n wherein
k|k=(n
k-1+ n
γ, k) (m
k+ 1); Delete weight fully little,
gauss's item, wherein, τ is cutting thresholding.
Will be apart from d in Gauss's item of remainder from cutting
ijgauss's item of <U is merged into one, and wherein U is for merging thresholding, and combined distance is defined as
after merging, the identify label of Gauss's item determines that principle is: if the identify label of Gauss's item is 0 entirely before merging, after merging, the identify label of Gauss's item is also 0; Otherwise identify label is not in Gauss's item of 0 before merging, the heavy maximum Gauss's item identify label of weighting is as the identify label of Gauss's item after merging.When Gauss's item is merged, weight is greater than other weight of Gauss Xiang Buyu of 0.5 and is greater than Gauss's item of 0.5 and merges.
Step 4, Gauss's item and identify label thereof according to cutting and after merging, the identify label that weight is greater than to Gauss's item of default weight threshold is adjusted, and makes identify label have uniqueness.
Described default weight threshold is 0.5, according to cutting with merge after Gauss's item, the identify label that weight is greater than to Gauss's item of 0.5 is judged: if the identify label of Gauss's item is greater than 0, show that its identity is definite; If 0, show that its identity not yet specifies, now need to specify a unique non-zero sign as its proprietary identification number to it.
In order to guarantee that proprietary mark identity is known, do not repeat, after a proprietary identify label (target lot number) is used, proprietary identify label adds 1 automatically; If a plurality of Gauss's items have identical proprietary identify label (being designated non-zero), for guaranteeing the uniqueness of identify label, the identify label of Gauss's item that weight coefficient is the highest remains unchanged, and the identify label of other Gauss's item is reassigned.
Step 5, extract weight and be greater than Gauss's item of default weight threshold as the output of wave filter, and identify label after adjusting, that weight is greater than Gauss's item of default weight threshold is exported together.The identify label that utilizes each moment wave filter to export, associates each dbjective state that wave filter is exported in the same time, thereby obtains the movement locus of each target.
Track identification method based on above-mentioned probability hypothesis density wave filter, the present invention also provides a kind of track tag system of probability hypothesis density wave filter, as shown in Figure 2, this track tag system comprises: prediction module 201, update module 202, cutting and merging module 203, adjusting module 204, dbjective state extraction module 205.Prediction module 201 is determined Gauss's item of prediction and the identify label of prediction term according to the identify label of Gauss's item of previous moment and previous moment, and adds non-proprietary identify label 0 to each newborn Gauss's item.Update module 202 is determined the identify label of Gauss's item and the renewal item of renewal according to the identify label of Gauss's item of the identify label of Gauss's item of described prediction, prediction term, new life and new life's item.Cutting and merging are carried out in Gauss's item of 203 pairs of described renewals of cutting and merging module and the identify label of described renewal item.Gauss item and the identify label thereof of adjusting module 204 according to cutting and after merging, the identify label that weight is greater than to Gauss's item of default weight threshold is adjusted, and makes identify label have uniqueness.Dbjective state extraction module 205 extracts weights and is greater than Gauss's item of default weight threshold as the output of wave filter, and identify label after adjusting, that weight is greater than Gauss's item of default weight threshold is exported together.
In described prediction module 201, with k, represent current time, k-1 Gauss's item set expression is constantly
i=1,2 ..., n
k-1, k-1 identify label set expression is constantly
i=1,2 ..., n
k-1,
not being to show that Gauss's item identity specifies at 0 o'clock, is that 0 o'clock indicate identification is not yet specified, and wherein, i is call number, and value is from 1 to n
k-1, n
k-1represent the k-1 number of Gauss's item constantly,
represent the k-1 weight of Gauss's item constantly,
represent the k-1 average of Gauss's item constantly,
represent the k-1 variance of Gauss's item constantly.
By k-1 Gauss's item constantly, gathered
i=1,2 ..., n
k-1with k-1 identify label set constantly
i=1,2 ..., n
k-1obtain Gauss's item set that k predicts constantly
i=1,2 ..., n
k-1, and the identify label set of k prediction Gauss item constantly
i=1,2 ..., n
k-1, wherein
represent that k predicts the weight of Gauss's item constantly,
represent that k predicts the average of Gauss's item constantly,
represent that k predicts the variance of Gauss's item constantly,
Here, p
sfor the probability that target is survived, F
k-1for state-transition matrix, Q
k-1for process noise covariance matrix.K constantly Gauss's item set of new life is
i=1,2 ..., n
γ, k, wherein
be given parameter, represent respectively weight, average and the variance of newborn Gauss's item, n
γ, knumber for newborn Gauss's item; The k constantly identify label set of newborn Gauss's item is
i=1,2 ..., n
γ, k, and the described k identify label unification of newborn Gauss's item is constantly set to non-proprietary sign 0,
i=1,2 ..., n
γ, k.
In update module 202, Gauss's item set of constantly predicting according to described k
i=1,2 ..., n
k-1with k Gauss's item set of new life constantly
i=1,2 ..., n
γ, kand the identify label set of prediction Gauss item
i=1,2 ..., n
k-1identify label set with newborn Gauss's item
i=1,2 ..., n
γ, k, obtain Gauss's item set that k upgrades constantly
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k) and upgrade the identify label set of Gauss's item
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k), wherein, m
kfor k moment measuring set Z
kthe number of middle measurement,
represent that k upgrades the weight of Gauss's item constantly,
represent that k upgrades the average of Gauss's item constantly,
represent that k upgrades the variance of Gauss's item constantly.
As i≤n
k-1+ n
γ, ktime,
Work as i>n
k-1+ n
γ, ktime, make q=int[(i-1)/(n
k+ n
γ, k)], p=i-q * (n
k+ n
γ, k),
?
Wherein, int represents to round, p
dfor acquisition probability, λ
cfor clutter density, H
kfor observing matrix, R
kfor the covariance of observation noise, K
pfor Kalman filter gain,
be the average of p prediction or newborn Gauss's item,
be the variance of p prediction or newborn Gauss's item, I is unit matrix,
be the weight of p prediction or newborn Gauss's item, N represents Gaussian distribution, the state that ξ is target,
be the average of e prediction or newborn Gauss's item,
be the variance of e prediction or newborn Gauss's item, z
qfor k moment measuring set Z
kin q measurement, the number of the Gauss's item after renewal is n
k|k=(n
k-1+ n
γ, k) (m
k+ 1).
In cutting and merging module 203, Gauss's item that k upgrades is constantly expressed as
i=1,2 ..., n
k|k, the identify label set of Gauss's item that k upgrades is constantly
i=1,2 ..., n
k|k, n wherein
k|k=(n
k-1+ n
γ, k) (m
k+ 1); Delete weight fully little,
gauss's item, wherein, τ is cutting thresholding.
Will be apart from d in Gauss's item of remainder from cutting
ijgauss's item of <U is merged into one, and wherein U is for merging thresholding, and combined distance is defined as
after merging, the identify label of Gauss's item determines that principle is: if the identify label of Gauss's item is 0 entirely before merging, after merging, the identify label of Gauss's item is also 0; Otherwise identify label is not in Gauss's item of 0 before merging, the heavy maximum Gauss's item identify label of weighting is as the identify label of Gauss's item after merging; When Gauss's item is merged, weight is greater than other weight of Gauss Xiang Buyu of 0.5 and is greater than Gauss's item of 0.5 and merges.
In described adjusting module 204, default weight threshold is 0.5, according to cutting with merge after Gauss's item, the identify label that weight is greater than to Gauss's item of 0.5 is judged: if the identify label of Gauss's item is greater than 0, show that its identity is definite; If 0, show that its identity not yet specifies, now need to specify a unique non-zero sign as its proprietary identify label to it.
The present invention adds proprietary identify label can to the Gauss's item as wave filter output, when wave filter prediction, renewal and cutting merge, utilize inheritance to carry out respective handling to the identify label of Gauss's item, during wave filter output, proprietary identify label is exported together in company with dbjective state, by identify label by wave filter not in the same time the dbjective state of output associate, thereby can determine the movement locus of each target.At clutter density λ
c=5 * 10
-6, target detection probability is 0.99, in situation that target disappears, the emulated data shown in Fig. 3 is processed appears and have in existing fresh target, the tracking results that the present invention obtains is as shown in Figure 4.As can be seen from Figure 4 the wave filter that, we propose can detect whole 10 batches of targets of observation space and can effectively follow the tracks of.Because wave filter output tape has proprietary identify label, and proprietary identify label provided and do not exported in the same time the relevance between data, thereby can determine the movement locus of each target.Although have decoy in wave filter output data, between decoy and follow-up output data, there is not relevance, there is different identify labels, on Fig. 4, decoy shows as some isolated points, and this isolated point is the point representing with circle.
The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, all any modifications of doing within the spirit and principles in the present invention, be equal to and replace and improvement etc., within all should being included in protection scope of the present invention.
Claims (10)
1. a track identification method for probability hypothesis density wave filter, is characterized in that, comprises the following steps:
Step 1: determine Gauss's item of prediction and the identify label of prediction term according to the identify label of Gauss's item of previous moment and previous moment, and add non-proprietary identify label to Gauss's item of each current time new life;
Step 2: determine the identify label of Gauss's item and the renewal item of renewal according to the identify label of Gauss's item of the identify label of Gauss's item of described prediction, prediction term, new life and new life's item;
Step 3, the identify label of Gauss's item of described renewal and described renewal item is carried out to cutting and merging;
Step 4, Gauss's item and identify label thereof according to cutting and after merging, the identify label that weight is greater than to Gauss's item of default weight threshold is adjusted, and makes identify label have uniqueness;
Step 5, extract weight and be greater than Gauss's item of default weight threshold as the output of wave filter, and identify label after adjusting, that weight is greater than Gauss's item of default weight threshold is exported together.
2. track identification method according to claim 1, is characterized in that, in described step 1, with k-1, represents previous moment, with k, represents current time, described non-proprietary identify label 0 mark, and k-1 Gauss's item set expression is constantly
i=1,2 ..., n
k-1, k-1 identify label set expression is constantly
i=1,2 ..., n
k-1,
not being to show that Gauss's item identity specifies at 0 o'clock, is that 0 o'clock indicate identification is not yet specified, and wherein, i is call number, and value is from 1 to n
k-1, n
k-1represent the k-1 number of Gauss's item constantly,
represent the k-1 weight of Gauss's item constantly,
represent the k-1 average of Gauss's item constantly,
represent the k-1 variance of Gauss's item constantly;
By k-1 Gauss's item constantly, gathered
i=1,2 ..., n
k-1with k-1 identify label set constantly
i=1,2 ..., n
k-1obtain Gauss's item set that k predicts constantly
i=1,2 ..., n
k-1, and k predicts the identify label set of Gauss's item constantly
i=1,2 ..., n
k-1, wherein
represent that k predicts the weight of Gauss's item constantly,
represent that k predicts the average of Gauss's item constantly,
represent that k predicts the variance of Gauss's item constantly,
Here, p
sfor the probability that target is survived, F
k-1for state-transition matrix, Q
k-1for process noise covariance matrix; K constantly Gauss's item set of new life is
i=1,2 ..., n
γ, k, wherein
be given parameter, represent respectively weight, average and the variance of newborn Gauss's item, n
γ,
knumber for newborn Gauss's item; The k constantly identify label set of newborn Gauss's item is
i=1,2 ..., n
γ, k, the described k identify label unification of newborn Gauss's item is constantly set to non-proprietary sign 0,
i=1,2 ..., n
γ, k.
3. track identification method according to claim 2, is characterized in that, in described step 2, and Gauss's item set of constantly predicting according to described k
i=1,2 ..., n
k-1with k Gauss's item set of new life constantly
i=1,2 ..., n
γ, kand the identify label set of prediction Gauss item
i=1,2 ..., n
k-1identify label set with newborn Gauss's item
i=1,2 ..., n
γ, k, obtain Gauss's item set that k upgrades constantly
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k) and upgrade the identify label set of Gauss's item
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k), wherein, m
kfor k moment measuring set Z
kthe number of middle measurement,
represent that k upgrades the weight of Gauss's item constantly,
represent that k upgrades the average of Gauss's item constantly,
represent that k upgrades the variance of Gauss's item constantly;
As i≤n
k-1+ n
γ, ktime,
Work as i>n
k-1+ n
γ, ktime, make q=int[(i-1)/(n
k+ n
γ, k)], p=i-q * (n
k+ n
γ, k),
?
Wherein, int represents to round, p
dfor acquisition probability, λ
cfor clutter density, H
kfor observing matrix, R
kfor the covariance of observation noise, K
pfor filter gain,
be the average of p prediction or newborn Gauss's item,
be the variance of p prediction or newborn Gauss's item, I is unit matrix,
be the weight of p prediction or newborn Gauss's item, N represents Gaussian distribution, the state that ξ is target,
be the average of e prediction or newborn Gauss's item,
be the variance of e prediction or newborn Gauss's item, z
qfor k moment measuring set Z
kin q measurement, the number of the Gauss's item after renewal is n
k|k=(n
k-1+ n
γ, k) (m
k+ 1).
4. track identification method according to claim 3, is characterized in that, in described step 3, Gauss's item that k upgrades is constantly expressed as
i=1,2 ..., n
k|k, the identify label set that k upgrades Gauss's item is constantly
i=1,2 ..., n
k|k, n wherein
k|k=(n
k-1+ n
γ, k) (m
k+ 1); Delete weight fully little,
gauss's item, wherein, τ is cutting thresholding;
Will be apart from d in Gauss's item of remainder from cutting
ijgauss's item of <U is merged into one, and wherein U is for merging thresholding, and combined distance is defined as
after merging, the identify label of Gauss's item determines that principle is: if the identify label of Gauss's item is 0 entirely before merging, after merging, the identify label of Gauss's item is also 0; Otherwise identify label is not in Gauss's item of 0 before merging, the heavy maximum Gauss's item identify label of weighting is as the identify label of Gauss's item after merging; When Gauss's item is merged, weight is greater than other weight of Gauss Xiang Buyu of 0.5 and is greater than Gauss's item of 0.5 and merges.
5. track identification method according to claim 1, it is characterized in that, in described step 4, default weight threshold is 0.5, Gauss's item according to cutting and after merging, the identify label that weight is greater than to Gauss's item of 0.5 is judged: if the identify label of Gauss's item is greater than 0, show that its identity is definite; If 0, show that its identity not yet specifies, now need to specify a unique non-zero sign as its proprietary identify label to it.
6. track identification method according to claim 5, it is characterized in that, for making proprietary identify label there is uniqueness, after a proprietary identify label is used, proprietary identify label adds 1 automatically, if a plurality of Gauss's items have identical proprietary identify label, the identify label of Gauss's item that weight coefficient is the highest remains unchanged, and the identify label of other Gauss's item is reassigned.
7. a track tag system for probability hypothesis density wave filter, is characterized in that, comprising:
Prediction module, determines Gauss's item of prediction and the identify label of prediction according to the identify label of Gauss's item of previous moment and previous moment, and adds non-proprietary identify label to each newborn Gauss's item;
Update module, determines the identify label of Gauss's item and the renewal item of renewal according to the identify label of Gauss's item of the identify label of Gauss's item of described prediction, prediction term, new life and new life's item;
Cutting and merging module, carry out cutting and merging to the identify label of Gauss's item of described renewal and described renewal item;
Adjusting module, Gauss's item and identify label thereof according to cutting and after merging, the identify label that weight is greater than to Gauss's item of default weight threshold is adjusted, and makes identify label have uniqueness;
Dbjective state extraction module, extracts weight and is greater than Gauss's item of default weight threshold as the output of wave filter, and identify label after adjusting, that weight is greater than Gauss's item of default weight threshold is exported together.
8. track tag system according to claim 7, is characterized in that, in described prediction module, with k-1, represents previous moment, with k, represents current time, described non-proprietary identify label 0 mark, and k-1 Gauss's item set expression is constantly
i=1,2 ..., n
k-1, k-1 identify label set expression is constantly
i=1,2 ..., n
k-1,
not being to show that Gauss's item identity specifies at 0 o'clock, is that 0 o'clock indicate identification is not yet specified, and wherein, i is call number, and value is from 1 to n
k-1, n
k-1represent the k-1 number of Gauss's item constantly,
represent the k-1 weight of Gauss's item constantly,
represent the k-1 average of Gauss's item constantly,
represent the k-1 variance of Gauss's item constantly;
By k-1 Gauss's item constantly, gathered
i=1,2 ..., n
k-1with k-1 identify label set constantly
i=1,2 ..., n
k-1obtain Gauss's item set that k predicts constantly
i=1,2 ..., n
k-1, and k predicts the identify label set of Gauss's item constantly
i=1,2 ..., n
k-1, wherein
represent that k predicts the weight of Gauss's item constantly,
represent that k predicts the average of Gauss's item constantly,
represent that k predicts the variance of Gauss's item constantly,
Here, p
sfor the probability that target is survived, F
k-1for state-transition matrix, Q
k-1for process noise covariance matrix; K constantly Gauss's item set of new life is
i=1,2 ..., n
γ, k, wherein
be given parameter, represent respectively weight, average and the variance of newborn Gauss's item, n
γ, knumber for newborn Gauss's item; The k constantly identify label set of newborn Gauss's item is
i=1,2 ..., n
γ, k, the described k identify label unification of newborn Gauss's item is constantly set to non-proprietary sign 0,
i=1,2 ..., n
γ, k.
9. track tag system according to claim 8, is characterized in that, in described update module, and Gauss's item set of constantly predicting according to described k
i=1,2 ..., n
k-1with k Gauss's item set of new life constantly
i=1,2 ..., n
γ, kand the identify label set of prediction Gauss item
i=1,2 ..., n
k-1identify label set with newborn Gauss's item
i=1,2 ..., n
γ, k, obtain Gauss's item set that k upgrades constantly
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k) and the identify label set of Gauss's item of upgrading
i=1 ..., (m
k+ 1) (n
k-1+ n
γ, k), wherein, m
kfor k moment measuring set Z
kthe number of middle measurement,
represent that k upgrades the weight of Gauss's item constantly,
represent that k upgrades the average of Gauss's item constantly,
represent that k upgrades the variance of Gauss's item constantly;
As i≤n
k-1+ n
γ, ktime,
Work as i>n
k-1+ n
γ, ktime, make q=int[(i-1)/(n
k+ n
γ, k)], p=i-q * (n
k+ n
γ, k),
?
Wherein, int represents to round, p
dfor acquisition probability, λ
cfor clutter density, H
kfor observing matrix, R
kfor the covariance of observation noise, K
pfor filter gain,
be the average of p prediction or newborn Gauss's item,
be the variance of p prediction or newborn Gauss's item, I is unit matrix,
be the weight of p prediction or newborn Gauss's item, N represents Gaussian distribution, the state that ξ is target,
be the average of e prediction or newborn Gauss's item,
be the variance of e prediction or newborn Gauss's item, z
qfor k moment measuring set Z
kin q measurement, the number of the Gauss's item after renewal is n
k|k=(n
k-1+ n
γ, k) (m
k+ 1).
10. track tag system according to claim 9, is characterized in that, in cutting and merging module, Gauss's item that k upgrades is constantly expressed as
i=1,2 ..., n
k|k, the identify label set that k upgrades Gauss's item is constantly
i=1,2 ..., n
k|k, n wherein
k|k=(n
k-1+ n
γ, k) (m
k+ 1); Delete weight fully little,
gauss's item, wherein, τ is cutting thresholding;
Will be apart from d in Gauss's item of remainder from cutting
ijgauss's item of <U is merged into one, and wherein U is for merging thresholding, and combined distance is defined as
after merging, the identify label of Gauss's item determines that principle is: if the identify label of Gauss's item is 0 entirely before merging, after merging, the identify label of Gauss's item is also 0; Otherwise identify label is not in Gauss's item of 0 before merging, the heavy maximum Gauss's item identify label of weighting is as the identify label of Gauss's item after merging; When Gauss's item is merged, weight is greater than other weight of Gauss Xiang Buyu of 0.5 and is greater than Gauss's item of 0.5 and merges;
In described adjusting module, according to cutting with merge after Gauss's item, the identify label that weight is greater than to Gauss's item of 0.5 is judged: if the identify label of Gauss's item is greater than 0, show that its identity is definite; If 0, show that its identity not yet specifies, now need to specify a unique non-zero sign as its proprietary identify label to it.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310690685.2A CN103679753A (en) | 2013-12-16 | 2013-12-16 | Track identifying method of probability hypothesis density filter and track identifying system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310690685.2A CN103679753A (en) | 2013-12-16 | 2013-12-16 | Track identifying method of probability hypothesis density filter and track identifying system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103679753A true CN103679753A (en) | 2014-03-26 |
Family
ID=50317203
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310690685.2A Pending CN103679753A (en) | 2013-12-16 | 2013-12-16 | Track identifying method of probability hypothesis density filter and track identifying system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103679753A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103902829A (en) * | 2014-04-11 | 2014-07-02 | 深圳大学 | Target tracking method and system transmitting edge distribution and existence probability |
CN104063615A (en) * | 2014-07-03 | 2014-09-24 | 深圳大学 | Target tracking method and tracking system based on variable coefficient alpha-beta filter |
WO2015196377A1 (en) * | 2014-06-25 | 2015-12-30 | 华为技术有限公司 | Method and device for determining user identity category |
CN105320143A (en) * | 2014-07-31 | 2016-02-10 | 霍尼韦尔国际公司 | Two step pruning in a phd filter |
WO2016187870A1 (en) * | 2015-05-28 | 2016-12-01 | 深圳大学 | Target tracking method and tracking system measuring and driving by propagating marginal distribution |
WO2017185688A1 (en) * | 2016-04-26 | 2017-11-02 | 深圳大学 | Method and apparatus for tracking on-line target |
CN108333569A (en) * | 2018-01-19 | 2018-07-27 | 杭州电子科技大学 | A kind of asynchronous multiple sensors fusion multi-object tracking method based on PHD filtering |
CN112906743A (en) * | 2021-01-19 | 2021-06-04 | 中国人民解放军国防科技大学 | Rapid multi-sensor set potential probability hypothesis density filtering method |
US11175142B2 (en) | 2014-07-31 | 2021-11-16 | Honeywell International Inc. | Updating intensities in a PHD filter based on a sensor track ID |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110097012A1 (en) * | 2009-10-22 | 2011-04-28 | Canon Kabushiki Kaisha | Image processing apparatus and method of controlling the same |
CN102307041A (en) * | 2011-04-29 | 2012-01-04 | 浙江大学 | Designing of current-statistical-model-based probability hypothesis density particle filter and filter |
CN103324835A (en) * | 2013-05-30 | 2013-09-25 | 深圳大学 | Probability hypothesis density filter target information maintaining method and information maintaining system |
CN103390107A (en) * | 2013-07-24 | 2013-11-13 | 深圳大学 | Target tracking method based on Dirac weighted sum and target tracking system |
-
2013
- 2013-12-16 CN CN201310690685.2A patent/CN103679753A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110097012A1 (en) * | 2009-10-22 | 2011-04-28 | Canon Kabushiki Kaisha | Image processing apparatus and method of controlling the same |
CN102307041A (en) * | 2011-04-29 | 2012-01-04 | 浙江大学 | Designing of current-statistical-model-based probability hypothesis density particle filter and filter |
CN103324835A (en) * | 2013-05-30 | 2013-09-25 | 深圳大学 | Probability hypothesis density filter target information maintaining method and information maintaining system |
CN103390107A (en) * | 2013-07-24 | 2013-11-13 | 深圳大学 | Target tracking method based on Dirac weighted sum and target tracking system |
Non-Patent Citations (1)
Title |
---|
刘宗香等: "一种有轨迹标识的利用测量生成新目标密度的GM-PHD滤波器", 《信号处理》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103902829A (en) * | 2014-04-11 | 2014-07-02 | 深圳大学 | Target tracking method and system transmitting edge distribution and existence probability |
CN103902829B (en) * | 2014-04-11 | 2017-02-15 | 深圳大学 | Target tracking method and system transmitting edge distribution and existence probability |
WO2015196377A1 (en) * | 2014-06-25 | 2015-12-30 | 华为技术有限公司 | Method and device for determining user identity category |
WO2016000487A1 (en) * | 2014-07-03 | 2016-01-07 | 深圳大学 | Target tracking method and tracking system based on variable coefficient α-β filter |
CN104063615A (en) * | 2014-07-03 | 2014-09-24 | 深圳大学 | Target tracking method and tracking system based on variable coefficient alpha-beta filter |
CN104063615B (en) * | 2014-07-03 | 2017-02-15 | 深圳大学 | Target tracking method and tracking system based on variable coefficient alpha-beta filter |
CN105320143A (en) * | 2014-07-31 | 2016-02-10 | 霍尼韦尔国际公司 | Two step pruning in a phd filter |
US10605607B2 (en) | 2014-07-31 | 2020-03-31 | Honeywell International Inc. | Two step pruning in a PHD filter |
US11175142B2 (en) | 2014-07-31 | 2021-11-16 | Honeywell International Inc. | Updating intensities in a PHD filter based on a sensor track ID |
WO2016187870A1 (en) * | 2015-05-28 | 2016-12-01 | 深圳大学 | Target tracking method and tracking system measuring and driving by propagating marginal distribution |
WO2017185688A1 (en) * | 2016-04-26 | 2017-11-02 | 深圳大学 | Method and apparatus for tracking on-line target |
CN108333569A (en) * | 2018-01-19 | 2018-07-27 | 杭州电子科技大学 | A kind of asynchronous multiple sensors fusion multi-object tracking method based on PHD filtering |
CN112906743A (en) * | 2021-01-19 | 2021-06-04 | 中国人民解放军国防科技大学 | Rapid multi-sensor set potential probability hypothesis density filtering method |
CN112906743B (en) * | 2021-01-19 | 2021-11-19 | 中国人民解放军国防科技大学 | Rapid multi-sensor set potential probability hypothesis density filtering method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103679753A (en) | Track identifying method of probability hypothesis density filter and track identifying system | |
Ruano et al. | Seismic detection using support vector machines | |
Laxhammar et al. | Inductive conformal anomaly detection for sequential detection of anomalous sub-trajectories | |
CN104714225B (en) | Dynamic programming tracking-before-detection method based on generalized likelihood ratios | |
CN103324835A (en) | Probability hypothesis density filter target information maintaining method and information maintaining system | |
CN110365647A (en) | A kind of false data detection method for injection attack based on PCA and BP neural network | |
CN106546975A (en) | A kind of small-sized unmanned plane based on radar data and flying bird classifying identification method | |
CN116778292B (en) | Method, device, equipment and storage medium for fusing space-time trajectories of multi-mode vehicles | |
CN104268657A (en) | Drainage basin water ecological risk early warning and distinguishing method based on remote sensing | |
CN113850838A (en) | Ship voyage intention acquisition method and device, computer equipment and storage medium | |
CN112907632A (en) | Single-towing ship target identification method and device | |
Moraffah et al. | Use of hierarchical Dirichlet processes to integrate dependent observations from multiple disparate sensors for tracking | |
Yildirim et al. | Ship detection in optical remote sensing images using YOLOv4 and Tiny YOLOv4 | |
Tedeschini et al. | Cooperative lidar sensing for pedestrian detection: Data association based on message passing neural networks | |
CN103076595B (en) | Abnormal type identifying method for multivariate discrete radar emitter signal | |
CN109614926B (en) | Distributed optical fiber sensing signal mode identification method and system based on prediction model | |
Safaei et al. | Open-set automatic target recognition | |
CN116662904A (en) | Method, device, computer equipment and medium for detecting variation of data type | |
CN114898222B (en) | Ship target track identification method and device | |
CN115294172A (en) | Target detection method and device, electronic equipment and storage medium | |
Hanusa et al. | Posterior distribution preprocessing with the JPDA algorithm: PACsim data set | |
CN112016539A (en) | Signal identification method and device, electronic equipment and storage medium | |
Darian et al. | Detecting GNSS spoofing using deep learning | |
Boily | Performance evaluation of a multisensor data fusion implementation: a progress report | |
CN115795353B (en) | Underground metal target classification method and system based on unbalanced data set |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140326 |