CN107133970A - Online multi-object tracking method and device based on movable information - Google Patents
Online multi-object tracking method and device based on movable information Download PDFInfo
- Publication number
- CN107133970A CN107133970A CN201710203635.5A CN201710203635A CN107133970A CN 107133970 A CN107133970 A CN 107133970A CN 201710203635 A CN201710203635 A CN 201710203635A CN 107133970 A CN107133970 A CN 107133970A
- Authority
- CN
- China
- Prior art keywords
- path segment
- target
- tracking
- confidence level
- characteristic information
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of online multi-object tracking method and device based on movable information, belongs to field of computer technology.This method includes：N number of tracking target in current t two field pictures is recognized, t is the integer more than or equal to 3, and N is the integer more than or equal to 2；N number of tracking clarification of objective information in t two field pictures is obtained, characteristic information includes positional information and dimension information；Obtain N number of path segment, each one tracking target movement locus in the preceding two field pictures of t 1 of path segment correspondence；According to the confidence level of each path segment, one in one in N number of path segment and N number of characteristic information is associated, the path segment of N number of tracking target is updated, and the confidence level of path segment meets the uncertain MGU theories of many Gausses.The online multi-object tracking method and device based on movable information that the present invention is provided, improve the degree of accuracy of multiple target tracking.
Description
Technical field
The present invention relates to field of computer technology, more particularly to a kind of online multi-object tracking method based on movable information
And device.
Background technology
Multiple target tracking problem is an important topic of computer vision field, and it is in robot visual guidance, traffic
There is important application value in the fields such as monitoring, intelligent transportation system, medical diagnosis.
Online multiple target tracking, refers to utilize video current image frame and the before target information in picture frame, to many
Individual target carries out real-time tracking.With developing rapidly for detection technique, existing multi-object tracking method is mostly using based on detection
Tracking (tracking-by-detection, TBD) method, the target signature information obtained using online object detector is real
Now track, but during tracking, the appearance information of target is only through as similarity evaluation, if comparing phase for outward appearance
As multiple targets, then can cause track mistake.
Therefore, using existing tracking so that the degree of accuracy of multiple target tracking is not high.
The content of the invention
The present invention provides a kind of online multi-object tracking method and device based on movable information, to improve multiple target tracking
The degree of accuracy.
The embodiment of the present invention provides a kind of online multi-object tracking method based on movable information, including：
N number of tracking target in current t two field pictures is recognized, t is the integer more than or equal to 3, and N is whole more than or equal to 2
Number；
N number of tracking clarification of objective information in the t two field pictures is obtained, the characteristic information includes positional information
And dimension information；
N number of path segment is obtained, each described one tracking target of path segment correspondence moves rail in preceding t-1 two field pictures
Mark；
According to the confidence level of each path segment, by one in N number of path segment and N number of feature
One in information is associated, and updates the path segment of N number of tracking target, and the confidence level of the path segment is met
It is theoretical that many Gausses do not know MGU.
In an embodiment of the present invention, the confidence level of each path segment described in the basis, by N number of track piece
Before one in section is associated with one in N number of characteristic information, in addition to：
It is determined that the confidence level of each path segment.
In an embodiment of the present invention, the confidence level of each path segment described in the determination, including：
According toIt is determined that
The confidence level of each path segment；
Wherein,Represent target i relative positionStandard deviation,Represent that target i's is relatively fast
DegreeStandard deviation,For the target i theoretical upper bounds of MGU, L=| Ti| it is TiLength, λ be and detection
The performance-relevant coefficient of device,For the similarity between the path segment and the characteristic information,
TiFor the track segment,For the characteristic information, DtFor the set of the characteristic information of t two field pictures.
In an embodiment of the present invention, the confidence level of each path segment described in the basis, by N number of track piece
After one in section is associated with one in N number of characteristic information, in addition to：
Re-association is carried out to the path segment for updating obtained N number of tracking target.
In an embodiment of the present invention, it is described that the path segment for updating obtained N number of tracking target is carried out
Before re-association, in addition to：
Re-association judgement is carried out to the path segment for updating obtained N number of tracking target；
Determine that the path segment of the N number of tracking target for updating and obtaining meets the condition of re-association.
In an embodiment of the present invention, it is described that the path segment for updating obtained N number of tracking target is carried out
Re-association judgement, including：
According toDescribed in being obtained to the renewal
The path segment of N number of tracking target carries out re-association judgement；
Wherein, Fre-assDecision function is represented, ε is jump function,For preliminary associated confidence,For
Confidence level transient change situation since apart from current Δ f frames,For in the accumulated change of Δ f frame ins
Situation, α, beta, gamma is respectively corresponding threshold value.
In an embodiment of the present invention, after the path segment for updating N number of tracking target, in addition to：
Update the confidence level of each path segment.
The embodiment of the present invention also provides a kind of online multiple target tracking device based on movable information, including：
Identification module, for recognizing N number of tracking target in current t two field pictures, t is the integer more than or equal to 3, and N is
Integer more than or equal to 2；
Acquisition module, for obtaining N number of tracking clarification of objective information, the feature letter in the t two field pictures
Breath includes positional information and dimension information；
The acquisition module, is additionally operable to obtain N number of path segment, and each described one tracking target of path segment correspondence exists
Movement locus in preceding t-1 two field pictures；
Relating module, for the confidence level according to each path segment, by one in N number of path segment
It is associated with one in N number of characteristic information, updates the path segment of N number of tracking target, and the track piece
The confidence level of section meets many Gausses and does not know MGU theories.
In an embodiment of the present invention, in addition to：
Determining module, the confidence level for determining each path segment.
In an embodiment of the present invention, the determining module, specifically for basis
It is determined that the confidence level of each path segment；
Wherein,Represent target i relative positionStandard deviation,Represent that target i's is relatively fast
DegreeStandard deviation,For the target i theoretical upper bounds of MGU, L=| Ti| it is TiLength, λ is and detector
Performance-relevant coefficient,For the similarity between the path segment and the characteristic information, Ti
For the track segment,For the characteristic information, DtFor the set of the characteristic information of t two field pictures.
Online multi-object tracking method and device provided in an embodiment of the present invention based on movable information, are realizing multiple target
During online tracking, by recognizing N number of tracking target in current t two field pictures；Obtain N number of tracking in t two field pictures
Clarification of objective information；N number of path segment is obtained, each one tracking target of path segment correspondence is transported in preceding t-1 two field pictures
Dynamic rail mark；According to the confidence level of each path segment, by one in one in N number of path segment and N number of characteristic information
It is associated, updates the path segment of N number of tracking target, and the confidence level of path segment meets the uncertain MGU theories of many Gausses.
As can be seen here, the online multi-object tracking method provided in an embodiment of the present invention based on movable information, by the way that MGU theories are introduced
Multiple target online tracking in, with using the target movable information in certain period of time to associating into row constraint, so as to improve many
The degree of accuracy of target following.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the accompanying drawing used required in technology description to do one simply to introduce, it should be apparent that, drawings in the following description are this hairs
Some bright embodiments, for those of ordinary skill in the art, without having to pay creative labor, can be with
Other accompanying drawings are obtained according to these accompanying drawings.
Fig. 1 illustrates for a kind of flow of online multi-object tracking method based on movable information provided in an embodiment of the present invention
Figure；
Fig. 2 shows for the flow of another online multi-object tracking method based on movable information provided in an embodiment of the present invention
It is intended to；
Fig. 3 is a kind of structural representation of the online multiple target tracking device based on movable information provided in an embodiment of the present invention
Figure；
Fig. 4 shows for the structure of another online multiple target tracking device based on movable information provided in an embodiment of the present invention
It is intended to.
Embodiment
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described embodiment is
A part of embodiment of the present invention, rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art
The every other embodiment obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.
Term " first ", " second ", " the 3rd " in description and claims of this specification and above-mentioned accompanying drawing, "
The (if present)s such as four " are for distinguishing similar object, without for describing specific order or precedence.It should manage
The data that solution is so used can be exchanged in the appropriate case, so as to embodiments of the invention described herein, for example can be with
Order in addition to those for illustrating or describing herein is implemented.In addition, term " comprising " and " having " and their times
What is deformed, it is intended that covering is non-exclusive to be included, for example, contain the process of series of steps or unit, method, system,
Product or equipment are not necessarily limited to those steps clearly listed or unit, but may include not list clearly or for
The intrinsic other steps of these processes, method, product or equipment or unit.
It should be noted that these specific embodiments can be combined with each other below, for same or analogous concept
Or process may be repeated no more in certain embodiments.
Fig. 1 illustrates for a kind of flow of online multi-object tracking method based on movable information provided in an embodiment of the present invention
Figure, being somebody's turn to do the online multi-object tracking method based on movable information can be held by the online multiple target tracking device based on movable information
OK, example, should online multiple target tracking based on movable information can be independently arranged, can also it is integrated within a processor.Tool
Body is shown in Figure 1, and being somebody's turn to do the online multi-object tracking method based on movable information can include：
N number of tracking target in S101, the current t two field pictures of identification.
Wherein, t is the integer more than or equal to 3, and N is the integer more than or equal to 2.
, it is necessary to predefine the plurality of tracking target before realizing that multiple target is tracked online.Example, of the invention real
Apply in example, after obtaining current t two field pictures online, it is possible to recognize N number of tracking target in the t two field pictures in advance, with
Just N number of tracking target is tracked online.
N number of tracking clarification of objective information in S102, acquisition t two field pictures.
Wherein, characteristic information includes positional information and dimension information.
After N number of tracking target is recognized, it is possible to using the object detector of training in advance, tracking target is examined
Survey with positioning, so as to obtain each tracking clarification of objective information.Example, the object detector can use traditional SVM
Detector, Adaboost detectors or detector based on deep learning etc..
S103, the N number of path segment of acquisition, each one tracking target of path segment correspondence are moved in preceding t-1 two field pictures
Track.
, can be straight if there is the path segment of tracking target before t two field pictures when obtaining each path segment
Obtain and take tracking target movement locus in preceding t-1 two field pictures.If the track piece of tracking target is not present before t two field pictures
Section, then need to the not associated characteristic information in each two field picture before t two field pictures, calculate its with previous image frame
Duplication between not associated characteristic information, so as to carry out the initialization of path segment according to the Duplication.
Example, it is assumed that the corresponding area of characteristic information of current frame image is Scur, the corresponding feature letter of previous frame image
The area of breath is Spre, its overlapping region area is Sinter, then Duplication can be expressed as：
roverlap=Sinter/(Scur+Spre-Sinter)
If Duplication roverlap>rth(rthFor threshold value set in advance, it can be set according to different application scenarios different
Value, for example, when tracking target is pedestrian, generally setting rth=0.4), then it is assumed that two characteristic informations belong to same target, and will
It is marked., can be as the follow-up track for tracking process when continuous 3 two field picture has same clarification of objective information
Fragment is inputted, and to complete the initialization of path segment, and its track segment confidence level is initialized as 0.75, so as to obtain each
Track target movement locus in preceding t-1 two field pictures.
After N number of path segment is got, N number of path segment can just constitute a path segment set.
It should be noted that in embodiments of the present invention, sequencing is had no between S102 and S103, it can first carry out
S102, then perform S103, can also first carry out S103, then perform S102, it is of course also possible to S102 and S103 is performed simultaneously,
This, the embodiment of the present invention is simply to first carry out S102, then performs and illustrate exemplified by S103, but does not represent and of the invention only limit to
In this.
S104, the confidence level according to each path segment, by one in N number of path segment and N number of characteristic information
One be associated, update it is N number of tracking target path segment, and path segment confidence level meet many Gausses do not know
MGU is theoretical.
Wherein, MGU theories give obey the variables (position) of many Gaussian Profiles and its derivative (speed) this to antithesis amount
Between relation, give its uncertain theory upper bound from uncertainty principle angle：
Wherein, k=mass*v is target momentum, σz,σkThe respectively standard deviation of position and momentum,For planck constant,
ξ be tolerance factor and 1≤ξ≤4,(M is Gauss number),In respectively each Gaussian component with
The related parameter maxima and minima of variance.
In embodiments of the present invention,, can be straight if the corresponding confidence level of path segment is high confidence level when being associated
Connect and be associated the corresponding path segment of the high confidence level with characteristic information, to update the path segment of tracking target；If rail
The corresponding confidence level of mark fragment then may be used to there is the corresponding path segment of high confidence level in low confidence, and path segment set
So that the corresponding path segment of the low confidence, the corresponding path segment of high confidence level and characteristic information to be associated；If track
The corresponding confidence level of fragment then may be used for the corresponding path segment of high confidence level is not present in low confidence, and path segment set
So that directly the corresponding path segment of the low confidence and characteristic information to be associated, to update the path segment of tracking target,
So as to realize the online tracking to multiple target.
Example, for the path segment of high confidence level, when being associated, it can be assumed that there is h bars path segment and n
Individual characteristic information, the association cost matrix between them is calculated as follows：
Wherein,Calculated by similarity defined formula, recycle Hungary Algorithm come realize path segment and
Bipartite Matching between characteristic information.When the association cost between a pair of path segments and characteristic information is less than what is pre-defined
During threshold value-log θ,With regard to energy and Ti(hi)Associate.After association is set up, it is possible to according toState, utilize confidence
Spend defined formula and calculate the T after associationi(hi)Confidence level, so that according to the confidence level of each path segment by path segment and spy
Reference breath is associated, and obtains the movement locus of each tracking target.
For the path segment of low confidence, when being associated, blocked because target is present, motion state becomes
Change, low confidence path segment is relative to high confidence level path segment more fragmentation, at this point it is possible to assume to exist, l bars are low to be put
Reliability Ti(lo)With h bar high confidence levels Tj(hi)Path segment, m not associated characteristic informations in first layer associationAt this
Stage, it is necessary to consider following three kinds of situations, the first：Ti(lo)With Tj(hi)It is associated；Second：Ti(lo)Terminate, such as target
Leave the visual field；The third：Ti(lo)WithIt is associated.The association cost matrix of three cases above is defined as follows：
Wherein, A=[aij],aij=-log (Λ (Ti(lo),Tj(hi))) calculate Ti(lo)With Tj(hi)Between association cost；B
=diag [b1,...,bl],Calculate and terminate Ti(lo)Cost；C=[cij],Calculate Ti(lo)WithBetween association cost.It is identical with high confidence level path segment processing,
Whether effectively threshold θ is used to determine association to, obtaining after cost matrix G, recycling Hungary Algorithm to obtain Optimum Matching pair,
Afterwards, the T after association is calculated further according to confidence level formulai(lo)Confidence level, so that by one in N number of path segment and N number of spy
One in reference breath is associated, and updates the path segment of N number of tracking target.
Online multi-object tracking method provided in an embodiment of the present invention based on movable information, realize multiple target it is online with
During track, by recognizing N number of tracking target in current t two field pictures；Obtain N number of tracking target in t two field pictures
Characteristic information；Obtain N number of path segment, each one tracking target movement locus in preceding t-1 two field pictures of path segment correspondence；
According to the confidence level of each path segment, one in one in N number of path segment and N number of characteristic information is closed
Connection, updates the path segment of N number of tracking target, and the confidence level of path segment meets the uncertain MGU theories of many Gausses.Thus may be used
See, the online multi-object tracking method provided in an embodiment of the present invention based on movable information, by the way that MGU theories are introduced into multiple target
In online tracking, with using the target movable information in certain period of time to associating into row constraint so that improve multiple target with
The degree of accuracy of track.
It is further, shown in Figure 2 on the basis of the corresponding embodiments of Fig. 1 based on the corresponding embodiments of Fig. 1,
Fig. 2 is the schematic flow sheet of another online multi-object tracking method based on movable information provided in an embodiment of the present invention, should
Online multi-object tracking method based on movable information can also include：
N number of tracking target in S201, the current t two field pictures of identification.
Wherein, t is the integer more than or equal to 3, and N is the integer more than or equal to 2.
N number of tracking clarification of objective information in S202, acquisition t two field pictures.
Wherein, characteristic information includes positional information and dimension information.
S203, the N number of path segment of acquisition, each one tracking target of path segment correspondence are moved in preceding t-1 two field pictures
Track.
It should be noted that S201-S203 description can be found in above-mentioned S101-S103 record herein, here, of the invention
No longer repeated.
S204, the confidence level for determining each path segment.
Optionally, in embodiments of the present invention, can basis
Determine the confidence level of each path segment.
Wherein,Represent target i relative positionStandard deviation,Represent that target i's is relatively fast
DegreeStandard deviation,For the target i theoretical upper bounds of MGU, L=| Ti| it is TiLength, λ is and detector
Performance-relevant coefficient,For the similarity between path segment and characteristic information, TiFor the rail
Mark segment,For the characteristic information, DtFor the set of the characteristic information of t two field pictures.Example,Can root
According to Λ (X, Y)=ΛA(X,Y)ΛS(X, Y) is calculated, X, and Y can be path segment or characteristic information, ΛA(X, Y) is by both colors
Histogrammic Pasteur's distance is calculated,Counted by the wide w and high h of target
Calculate, X is T hereini, Y isDue toThen whenWhen for high confidence level path segment, it is on the contrary then
For low confidence path segment.
S205, the confidence level according to each path segment, by one in N number of path segment and N number of characteristic information
One be associated, update it is N number of tracking target path segment, and path segment confidence level meet many Gausses do not know
MGU is theoretical.
As can be seen here, the online multi-object tracking method provided in an embodiment of the present invention based on movable information, by by MGU
Theory is introduced into multiple target online track in, with using the target movable information in certain period of time to associating into row constraint so that
Improve the degree of accuracy of multiple target tracking.
Further, in embodiments of the present invention, S205 is according to the confidence level of each path segment, by N number of track piece
After one in section is associated with one in N number of characteristic information, it can also include：
S206, the path segment of the N number of tracking target obtained to renewal carry out re-association judgement.
Optionally, in embodiments of the present invention, the path segment of the N number of tracking target obtained to renewal carries out re-association and sentenced
It is disconnected, it can include：
According toThe N number of tracking obtained to renewal
The path segment of target carries out re-association judgement.
Wherein, Fre-assDecision function is represented, ε is jump function,For preliminary associated confidence,
Threshold value.
The path segment for N number of tracking target that S207, determination renewal are obtained meets the condition of re-association.
Example, in embodiments of the present invention, work as Fre-assWhen value is equal to 1, then it represents that tentatively associate obtained N number of tracking
The path segment of target meets the condition of re-association, that is, needs to carry out two secondary associations, work as Fre-assWhen value is not equal to 1, then it represents that just
The path segment for N number of tracking target that step association is obtained is unsatisfactory for the condition of re-association, i.e., need not carry out two secondary associations.
S208, the path segment of the N number of tracking target obtained to renewal carry out re-association.
When carrying out re-association to association results, it is necessary to interrupt the preliminary association of movement locus, to these path segments with
Characteristic information carries out re-association, so as to export the association results for meeting global objective function, can so correct in preliminary association
Mistake that may be present, to improve the accuracy and robustness of tracker, so as to further increase the accurate of multiple target tracking
Property.
Example, the angle of the overall track confidence level of optimization is considered from global angle, the global mesh of re-association is defined
Scalar functions.The path segment that making moment t needs to carry out re-association is Ti, i ∈ { 1 ..., n }, characteristic information is dj,j∈
{ 1 ..., m }, the track confidence level after re-association isRe-association can then be obtained compared to the lifting effect tentatively associated
ForAnd all path segment set got a promotion
Wherein, global objective function is defined as follows：
Wherein, l=| Tre| it is set TreThe number of middle element, a, b is balance parameter.
Optionally, after the path segment for updating N number of tracking target, in addition to：Update the confidence of each path segment
Degree.
Example, (Gaussian mixture can be assumed to many Gausses containing M component by association results
Model, GMM) model parameter carry out online updating, to carry out next frame tracking.To m (∈ { 1,2 ..., M }) individual Gauss
Component has：WithIts
In, ρ=0.1.
Further, can be respectively to x-axis and the relative position of y-axis both direction in order to simplify calculating
The weight of each Gaussian component of tByCalculated.Wherein α=0.1,For
The interrelated decision functional value of each obtained Gaussian component is calculated using the MGU upper bounds, the value is 1 if the MGU upper bounds are met, no
It is then 0, so that the confidence level of each path segment is updated, so as in association process next time, according to putting after renewal
Path segment and characteristic information are associated by reliability, so as to realize that multiple target is tracked online.
Fig. 3 is that a kind of structure of the online multiple target tracking device 30 based on movable information provided in an embodiment of the present invention is shown
It is intended to, shown in Figure 3, certainly, the embodiment of the present invention is simply illustrated by taking Fig. 3 as an example, but does not represent the present invention only
It is confined to this.The online multiple target tracking device 30 based on movable information can include：
Identification module 301, for recognizing N number of tracking target in current t two field pictures, t is the integer more than or equal to 3, N
For the integer more than or equal to 2.
Acquisition module 302, for obtaining N number of tracking clarification of objective information in t two field pictures, characteristic information includes position
Information and dimension information.
Acquisition module 302, is additionally operable to obtain N number of path segment, each one tracking target of path segment correspondence is in preceding t-1
Movement locus in two field picture.
Relating module 303, for the confidence level according to each path segment, by one in N number of path segment and N number of
One in characteristic information is associated, and updates the path segment of N number of tracking target, and how high the confidence level of path segment meet
This uncertain MGU is theoretical.
Optionally, shown in Figure 4, Fig. 4 is provided in an embodiment of the present invention another based on the online of movable information
The structural representation of multiple target tracking device 30, being somebody's turn to do the online multiple target tracking device 30 based on movable information also includes：
Determining module 304, the confidence level for determining each path segment.
Optionally, determining module 304, specifically for basisIt is determined that
The confidence level of each path segment.
Wherein,Represent target i relative positionStandard deviation,Represent that target i's is relatively fast
DegreeStandard deviation,For the target i theoretical upper bounds of MGU, L=| Ti| it is TiLength, λ is and detector
Performance-relevant coefficient,For the similarity between path segment and characteristic information, TiFor the rail
Mark segment,For the characteristic information, DtFor the set of the characteristic information of t two field pictures.
Optionally, relating module 303, are additionally operable to the path segment progress re-association of N number of tracking target obtained to renewal.
Optionally, determining module 304, are additionally operable to the path segment progress re-association of N number of tracking target obtained to renewal
Judge；It is determined that the path segment for updating obtained N number of tracking target meets the condition of re-association.
Optionally, determining module 304, specifically for basisIt is right
The path segment for updating obtained N number of tracking target carries out re-association judgement.
Wherein, Fre-assDecision function is represented, ε is jump function,For preliminary associated confidence,
Threshold value.
Optionally, being somebody's turn to do the online multiple target tracking device 30 based on movable information also includes：
Update module 305, the confidence level for updating each path segment.
The online multiple target tracking device 30 based on movable information shown in the embodiment of the present invention, can perform the above method
Technical scheme shown in embodiment, its realization principle and beneficial effect are similar, are no longer repeated herein.
One of ordinary skill in the art will appreciate that：Realizing all or part of step of above-mentioned each method embodiment can lead to
The related hardware of programmed instruction is crossed to complete.Foregoing program can be stored in a computer read/write memory medium.The journey
Sequence upon execution, performs the step of including above-mentioned each method embodiment；And foregoing storage medium includes：ROM, RAM, magnetic disc or
Person's CD etc. is various can be with the medium of store program codes.
Finally it should be noted that：Various embodiments above is merely illustrative of the technical solution of the present invention, rather than its limitations；To the greatest extent
The present invention is described in detail with reference to foregoing embodiments for pipe, it will be understood by those within the art that：Its according to
The technical scheme described in foregoing embodiments can so be modified, or which part or all technical characteristic are entered
Row equivalent substitution；And these modifications or replacement, the essence of appropriate technical solution is departed from various embodiments of the present invention technology
The scope of scheme.
Claims (10)
1. a kind of online multi-object tracking method based on movable information, it is characterised in that including：
N number of tracking target in current t two field pictures is recognized, t is the integer more than or equal to 3, and N is the integer more than or equal to 2；
N number of tracking clarification of objective information in the t two field pictures is obtained, the characteristic information includes positional information and chi
Very little information；
N number of path segment is obtained, each described one tracking target movement locus in preceding t-1 two field pictures of path segment correspondence；
According to the confidence level of each path segment, by one in N number of path segment and N number of characteristic information
In one be associated, update it is described it is N number of tracking target path segment, and the path segment confidence level meet it is how high
This uncertain MGU is theoretical.
2. according to the method described in claim 1, it is characterised in that the confidence level of each path segment described in the basis,
Before one in one in N number of path segment and N number of characteristic information is associated, in addition to：
It is determined that the confidence level of each path segment.
3. method according to claim 2, it is characterised in that the confidence level of each path segment described in the determination,
Including：
According to
It is determined that the confidence level of each path segment；
Wherein,Represent target i relative positionStandard deviation,Represent target i relative velocityStandard deviation,For the target i theoretical upper bounds of MGU, L=| Ti| it is TiLength, λ be and detector
The related coefficient of energy,For the similarity between the path segment and the characteristic information, TiFor
The track segment,For the characteristic information, DtFor the set of the characteristic information of t two field pictures.
4. according to the method described in claim 1, it is characterised in that the confidence level of each path segment described in the basis,
After one in one in N number of path segment and N number of characteristic information is associated, in addition to：
Re-association is carried out to the path segment for updating obtained N number of tracking target.
5. method according to claim 4, it is characterised in that described to update obtained N number of tracking target to described
Path segment carry out re-association before, in addition to：
Re-association judgement is carried out to the path segment for updating obtained N number of tracking target；
Determine that the path segment of the N number of tracking target for updating and obtaining meets the condition of re-association.
6. method according to claim 5, it is characterised in that described to update obtained N number of tracking target to described
Path segment carry out re-association judgement, including：
According toTo it is described update obtain it is described it is N number of with
The path segment of track target carries out re-association judgement；
Wherein, Fre-assDecision function is represented, ε is jump function,For preliminary associated confidence,For
Confidence level transient change situation since apart from current Δ f frames,For in the accumulated change of Δ f frame ins
Situation, α, beta, gamma is respectively corresponding threshold value.
7. according to the method described in claim 1, it is characterised in that the path segment for updating N number of tracking target it
Afterwards, in addition to：
Update the confidence level of each path segment.
8. a kind of online multiple target tracking device based on movable information, it is characterised in that including：
Identification module, for recognizing N number of tracking target in current t two field pictures, t is the integer more than or equal to 3, N be more than
Integer equal to 2；
Acquisition module, for obtaining N number of tracking clarification of objective information, the characteristic information bag in the t two field pictures
Include positional information and dimension information；
The acquisition module, is additionally operable to obtain N number of path segment, and each described one tracking target of path segment correspondence is in preceding t-
Movement locus in 1 two field picture；
Relating module, for the confidence level according to each path segment, by one in N number of path segment and institute
State one in N number of characteristic information to be associated, the path segment of renewal N number of tracking target, and the path segment
Confidence level meets many Gausses and does not know MGU theories.
9. device according to claim 8, it is characterised in that also include：
Determining module, the confidence level for determining each path segment.
10. device according to claim 9, it is characterised in that
The determining module, specifically for basisDetermine institute
State the confidence level of each path segment；
Wherein,Represent target i relative positionStandard deviation,Represent target i relative velocityStandard deviation,For the target i theoretical upper bounds of MGU, L=| Ti| it is TiLength, λ be and detector
The related coefficient of energy,For the similarity between the path segment and the characteristic information, TiFor
The track segment,For the characteristic information, DtFor the set of the characteristic information of t two field pictures.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710203635.5A CN107133970B (en) | 2017-03-30 | 2017-03-30 | Online multi-target tracking method and device based on motion information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710203635.5A CN107133970B (en) | 2017-03-30 | 2017-03-30 | Online multi-target tracking method and device based on motion information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107133970A true CN107133970A (en) | 2017-09-05 |
| CN107133970B CN107133970B (en) | 2020-02-14 |
Family
ID=59715865
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710203635.5A Active CN107133970B (en) | 2017-03-30 | 2017-03-30 | Online multi-target tracking method and device based on motion information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107133970B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108875666A (en) * | 2018-06-27 | 2018-11-23 | 腾讯科技（深圳）有限公司 | Acquisition methods, device, computer equipment and the storage medium of motion profile |
| CN110349184A (en) * | 2019-06-06 | 2019-10-18 | 南京工程学院 | The more pedestrian tracting methods differentiated based on iterative filtering and observation |
| CN110855935A (en) * | 2018-08-02 | 2020-02-28 | 高新兴科技集团股份有限公司 | Personnel track generation system and method based on multiple cameras |
| CN112102357A (en) * | 2020-09-08 | 2020-12-18 | 杭州海康威视数字技术股份有限公司 | Track adjusting method, device and equipment and storage medium |
| CN112507949A (en) * | 2020-12-18 | 2021-03-16 | 北京百度网讯科技有限公司 | Target tracking method and device, road side equipment and cloud control platform |
| CN112651994A (en) * | 2020-12-18 | 2021-04-13 | 零八一电子集团有限公司 | Ground multi-target tracking method |
| CN113032498A (en) * | 2019-12-24 | 2021-06-25 | 深圳云天励飞技术有限公司 | Method and device for judging track similarity, electronic equipment and storage medium |
| CN113256686A (en) * | 2021-06-28 | 2021-08-13 | 上海齐感电子信息科技有限公司 | System and method for tracking accurate visual target |
| CN113421288A (en) * | 2021-08-23 | 2021-09-21 | 杭州云栖智慧视通科技有限公司 | Static track fragment improvement method in multi-target real-time track tracking |
| CN114299112A (en) * | 2021-12-24 | 2022-04-08 | 萱闱(北京)生物科技有限公司 | Multi-target-based track identification method, device, medium and computing equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105809714A (en) * | 2016-03-07 | 2016-07-27 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | Track confidence coefficient based multi-object tracking method |
| CN106373145A (en) * | 2016-08-30 | 2017-02-01 | 上海交通大学 | Multi-target tracking method based on tracking fragment confidence and discrimination appearance learning |
-
2017
- 2017-03-30 CN CN201710203635.5A patent/CN107133970B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105809714A (en) * | 2016-03-07 | 2016-07-27 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | Track confidence coefficient based multi-object tracking method |
| CN106373145A (en) * | 2016-08-30 | 2017-02-01 | 上海交通大学 | Multi-target tracking method based on tracking fragment confidence and discrimination appearance learning |
Non-Patent Citations (1)
| Title |
|---|
| BAOCHANG ZHANG等: "Bounding Multiple Gaussians Uncertainty with Application", 《INTERNATIONAL JOURNAL OF COMPUTER VISION》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108875666A (en) * | 2018-06-27 | 2018-11-23 | 腾讯科技（深圳）有限公司 | Acquisition methods, device, computer equipment and the storage medium of motion profile |
| CN108875666B (en) * | 2018-06-27 | 2023-04-18 | 腾讯科技（深圳）有限公司 | Method and device for acquiring motion trail, computer equipment and storage medium |
| CN110855935A (en) * | 2018-08-02 | 2020-02-28 | 高新兴科技集团股份有限公司 | Personnel track generation system and method based on multiple cameras |
| CN110855935B (en) * | 2018-08-02 | 2023-08-25 | 高新兴科技集团股份有限公司 | Personnel track generation system and method based on multiple cameras |
| CN110349184B (en) * | 2019-06-06 | 2022-08-09 | 南京工程学院 | Multi-pedestrian tracking method based on iterative filtering and observation discrimination |
| CN110349184A (en) * | 2019-06-06 | 2019-10-18 | 南京工程学院 | The more pedestrian tracting methods differentiated based on iterative filtering and observation |
| CN113032498B (en) * | 2019-12-24 | 2024-05-07 | 深圳云天励飞技术有限公司 | Method and device for judging track similarity, electronic equipment and storage medium |
| CN113032498A (en) * | 2019-12-24 | 2021-06-25 | 深圳云天励飞技术有限公司 | Method and device for judging track similarity, electronic equipment and storage medium |
| CN112102357A (en) * | 2020-09-08 | 2020-12-18 | 杭州海康威视数字技术股份有限公司 | Track adjusting method, device and equipment and storage medium |
| CN112102357B (en) * | 2020-09-08 | 2023-07-25 | 杭州海康威视数字技术股份有限公司 | Track adjustment method, track adjustment device, track adjustment equipment and storage medium |
| CN112507949A (en) * | 2020-12-18 | 2021-03-16 | 北京百度网讯科技有限公司 | Target tracking method and device, road side equipment and cloud control platform |
| CN112651994A (en) * | 2020-12-18 | 2021-04-13 | 零八一电子集团有限公司 | Ground multi-target tracking method |
| CN113256686B (en) * | 2021-06-28 | 2021-10-08 | 上海齐感电子信息科技有限公司 | System and method for tracking accurate visual target |
| CN113256686A (en) * | 2021-06-28 | 2021-08-13 | 上海齐感电子信息科技有限公司 | System and method for tracking accurate visual target |
| CN113421288A (en) * | 2021-08-23 | 2021-09-21 | 杭州云栖智慧视通科技有限公司 | Static track fragment improvement method in multi-target real-time track tracking |
| CN114299112A (en) * | 2021-12-24 | 2022-04-08 | 萱闱(北京)生物科技有限公司 | Multi-target-based track identification method, device, medium and computing equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107133970B (en) | 2020-02-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107133970A (en) | Online multi-object tracking method and device based on movable information | |
| Huang et al. | Multiple target tracking by learning-based hierarchical association of detection responses | |
| Song et al. | DynaVINS: a visual-inertial SLAM for dynamic environments | |
| CN111127513A (en) | Multi-target tracking method | |
| Meuter et al. | A decision fusion and reasoning module for a traffic sign recognition system | |
| CN111161315B (en) | Multi-target tracking method and system based on graph neural network | |
| CN113269073A (en) | Ship multi-target tracking method based on YOLO V5 algorithm | |
| CN106934817B (en) | Multi-attribute-based multi-target tracking method and device | |
| CN108416258A (en) | A kind of multi-human body tracking method based on human body model | |
| Liu et al. | RGB-D inertial odometry for a resource-restricted robot in dynamic environments | |
| CN110765906A (en) | Pedestrian detection algorithm based on key points | |
| CN104778465B (en) | A kind of matched method for tracking target of distinguished point based | |
| JP2021026644A (en) | Article detection apparatus, article detection method, and article-detecting computer program | |
| Zhang et al. | Lane detection model based on spatio-temporal network with double convolutional gated recurrent units | |
| CN105374049A (en) | Multi-angle-point tracking method based on sparse optical flow method and apparatus thereof | |
| Prokaj et al. | Using 3d scene structure to improve tracking | |
| CN116592897A (en) | Improved ORB-SLAM2 positioning method based on pose uncertainty | |
| Xie et al. | Hierarchical forest based fast online loop closure for low-latency consistent visual-inertial SLAM | |
| Yu et al. | Accurate and robust visual localization system in large-scale appearance-changing environments | |
| Swedish et al. | Deep visual teach and repeat on path networks | |
| Samadzadeh et al. | Srvio: Super robust visual inertial odometry for dynamic environments and challenging loop-closure conditions | |
| Ai et al. | Deep visual navigation under partial observability | |
| Yuan et al. | A novel approach to image-sequence-based mobile robot place recognition | |
| Mittal et al. | Pedestrian detection and tracking using deformable part models and Kalman filtering | |
| Gong et al. | An accurate, robust visual odometry and detail-preserving reconstruction system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |