CN109993771A - Modulation domain infrared object tracking method based on state vector increment - Google Patents
Modulation domain infrared object tracking method based on state vector increment Download PDFInfo
- Publication number
- CN109993771A CN109993771A CN201711480518.XA CN201711480518A CN109993771A CN 109993771 A CN109993771 A CN 109993771A CN 201711480518 A CN201711480518 A CN 201711480518A CN 109993771 A CN109993771 A CN 109993771A
- Authority
- CN
- China
- Prior art keywords
- target
- frame
- infrared
- template
- particle
- 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
- G06T7/251—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Studio Devices (AREA)
- Aiming, Guidance, Guns With A Light Source, Armor, Camouflage, And Targets (AREA)
- Closed-Circuit Television Systems (AREA)
Abstract
The invention discloses a kind of modulation domain infrared object tracking methods based on state vector increment.This method changes the problems such as complicated for low contrast present in infrared motion target tracking, low signal-to-noise ratio, background, and adoption status vector increment is extended moving target state vector, and stable tracking is carried out to infrared target in modulation domain.Firstly, the Gabor filter group by 18 channels analyzes the infrared video of input, the AM-FM characteristic model of moving target is established；Then, the AM significant characteristics of moving target are extracted using conspicuousness constituent analysis (DCA) algorithm；Then, tenacious tracking is carried out to the moving target in infrared video under particle filter frame using the AM significant characteristics of moving target, during tracking, excludes interference of the complex background to target using the state vector increment of target；Secondly, judging whether currently used target template needs to update, if meeting template renewal condition, current goal template is updated；Finally, the tracking result of output present frame, realizes the tenacious tracking of infrared target.
Description
Technical field
The invention belongs to infrared motion target tracking technique fields in computer vision, and in particular to one kind based on state to
Measure the modulation domain infrared object tracking method of increment.
Background technique
The features such as infrared object tracking technology is because of its low contrast, low signal-to-noise ratio, and background variation is complicated, targeted mutagenesis, with
Visible light motion target tracking, which is compared, has bigger challenge.During actual tracking, These characteristics frequently can lead to
Phenomena such as track result offset, target is lost, stable tracking (1.Han, Tony X., Ming cannot be carried out to infrared target
Liu,and Thomas S.Huang."A drifting-proof framework for tracking and online
appearance learning."Applications of Computer Vision,2007.WACV'07.IEEE
Workshop on.IEEE,2007；2.Pan,Jiyan,and Bo Hu."Robust object tracking against
template drift."Image Processing,2007.ICIP 2007.IEEE International Conference
on.Vol.3.IEEE,2007.).It therefore, is still computer vision field suitable for the Moving Target Tracking Algorithm of infrared video
A big research hotspot.
Since the modulation field characteristics of infrared target can distinguish moving target and complex background, can will adjust
Width-frequency modulation (AM-FM) target signature model theory is applied to infrared object tracking field.Conspicuousness constituent analysis (DCA) algorithm
AM significant characteristics can be extracted from the AM-FM characteristic model of target, therefore also be widely used in image segmentation, referred to
Line identification, the fields such as target following (1.Havlicek, J.P., P.C.Tay, and A.C.Bovik. " AM-FM image
models:Fundamental techniques and emerging trends."Handbook of Image and
Video Processing 2(2005)；2.Prakash,R.Senthil,and Rangarajan Aravind."
Modulation-domain particle filter for template tracking."Pattern Recognition,
2008.ICPR 2008.19th International Conference on.IEEE,2008.)。
Chuong et al. points out in infrared object tracking that infrared motion target and complex background can be by modulation domains
Target property is effectively distinguished (Nguyen, Chuong T., and Joseph P.Havlicek. " Modulation
domain features for discriminating infrared targets and backgrounds."Image
Processing,2006IEEE International Conference on.IEEE,2006.).It is mentioned according to the theoretical thought
The modulation domain infrared object tracking method based on normalized crosscorrelation template out can track infrared target
(Nguyen,Chuong T.,Joseph P.Havlicek,and Mark Yeary."Modulation domain
template tracking."Computer Vision and Pattern Recognition,2007.CVPR'07.IEEE
Conference on.IEEE,2007.).But target morphology becomes suddenly during this method does not account for target actual tracking
The influence of the factors such as change, haves the defects that certain.Jesyca et al. proposes a kind of target following based on state vector increment
Method is tracked (Bello, Jesyca to target during Target state estimator is added in the velocity information of target
C.Fuenmayor,and Joseph P.Havlicek."A state vector augmentation technique for
incorporating indirect velocity information into the likelihood function of
the sir video target tracking filter."Image Analysis and Interpretation
(SSIAI),2016IEEE Southwest Symposium on.IEEE,2016.)。
Summary of the invention
It is an object of the invention to propose a kind of modulation domain infrared object tracking method based on state vector increment.Using
Conspicuousness constituent analysis (DCA) algorithm is extracted by the AM significant characteristics in AM-FM characteristic model, sharp under particle filter frame
Interference of the complex background to target is excluded with the state vector increment of target, realizes the tenacious tracking of infrared target.With it is traditional
Pixel domain infrared object tracking algorithm is compared, and this method can overcome infrared object tracking in conventional pixel domain is easy to be lost to lack
It falls into, tenacious tracking is carried out to the infrared target under complex background, there is feasibility.
In order to solve the above-mentioned technical problem, the present invention is based on the modulation domain infrared object tracking method of state vector increment,
The moving target in infrared video sequence is effectively tracked using following steps:
Step 1: the infrared video sequence I containing moving target is shot using infrared camerak(x, y), k=1,2 ...,
Object as target following is carried out subsequent processing by n, the moving target in image；
Step 2: by the Gabor filter group in 18 channels to the infrared video sequence I of inputk(x, y), k=1,
2 ..., n is analyzed, and the AM-FM characteristic model s of moving target is establishedk(x, y), k=1,2 ..., n；The model such as formula
(1) shown in:
In formula (1), sk(x, y) indicates input picture IkThe plural extension image of (x, y), M=18 indicate Gabor filtering
The port number of filter, a in device groupk,m(x, y) indicates to be obtained at point (x, y) in kth frame image by m-th of Gabor filter
AM function, i.e. AM feature,It indicates to obtain the FM letter at point (x, y) in kth frame image by m-th of Gabor filter
Number, i.e. FM feature, exp indicate that exponent arithmetic, j indicate imaginary number.
Step 3, using conspicuousness constituent analysis (DCA) algorithm from the AM-FM characteristic model s of moving targetk(x,y),k
=1,2 ..., the AM significant characteristics a of moving target is extracted in nD,k(x, y), k=1,2 ..., n, i.e. Gabor filter
In 18 channels of group, AM characteristic value corresponding to the maximum channel of channel response；Define AM significant characteristics aD(x, y) is
In 18 channels of Gabor filter group, the corresponding AM characteristic value in the maximum channel of channel response.The channel selection standard definition
Are as follows:
In formula (2), Γ (x, y) indicates channel selection standard, ym(x, y), (m=1,2 ..., 18) indicate Gabor filter
The channel response of wave device group, Gm() indicates the frequency response of m-th of Gabor filter, and max indicates maximizing, | | table
Show absolute value.
Step 4: the first frame I in input infrared video sequence image1(x, y), according to target's center coordinate (x1,y1) and
The wide w of target1, high h1True value initialized target template T1(m,n)；
Step 5: the AM significant characteristics a of moving target is utilizedD,k(x, y), k=1,2 ..., n is in particle filter frame
Under to infrared video sequence Ik(x, y), the moving target in k=2,3 ..., n are tracked, and target's center's coordinate (x is exportedk,
yk), k=2,3 ..., n and the wide w of targetk, k=2,3 ..., n, high hk, k=2,3 ..., n is as tracking result；
Moving target posterior probability needed for infrared object tracking process is calculated using the method as shown in formula (3),
To realize the tenacious tracking of infrared target,
In formula (3),It is the weight of i-th of particle in kth frame, ∝ expression is proportional to；It is tracking process
The moving target posterior probability of middle acquisition,It is the delta state vector of i-th of particle in kth frame,It is moving target observation
Value；Exp indicates exponent arithmetic, and λ is the proportion adjustable factor,It is i-th of particle journey similar to the frame target template in kth frame
Degree.
The similarity degree of i-th particle and the frame target template in the kth frameIt calculates and obtains by the following method:
Present frame target AM significant characteristics and current are compared by normalized crosscorrelation algorithm (NCC) using formula (4)
The AM significant characteristics of template, solve the similarity degree of i-th particle and the frame target template in kth frame
In formula (4), (m, n) is pixel position coordinates, and M, N are the width of present frame, high, Tk(m, n) is in kth frame
Target template,It is the target template in -1 frame of kth,It is the observation of particle in kth frame,
It is the observation of particle in -1 frame of kth, Σ indicates that summation, ∈ expression belong to.
Step 6: judging whether currently used target template needs to update, if meeting template renewal condition, goes to step
Seven；Otherwise, step 5 is gone to；
Template renewal condition are as follows: continuous nDFrameThe threshold value th both greater than set.
Step 7: target template is updated, step 5 is gone to.Specific target template update method are as follows:
Input picture Ik(x,y)；Calculate the similarity degree of i-th particle and the frame target template in kth frameExist simultaneously
Particle assemblyMiddle find has maximum similarity degreeParticle?
In frame image, optimal particle is usedCalculate similarity degreeMaximum similarity degree coefficient is selected to correspond to frame
k*, new target template is obtained using formula (5)
Compared with prior art, the present invention its remarkable advantage is:
1) the AM significant characteristics in modulation domain can be completely separable by infrared motion target and complex background；
2) the state vector increment of target is increased by being extended to the original state vector of target in historical frames
The judgement of moving target state can be excluded effectively during particle filter due to particle multiplicity caused by particle resampling
Property missing problem；
3) update method of target template can reduce the drift of the tracking result as caused by accumulated error in renewal process；
The method of the present invention can overcome the defect that infrared object tracking is easy to be lost in conventional pixel domain, under complex background
Infrared target carries out tenacious tracking, has feasibility.
Detailed description of the invention
Fig. 1 is the flow chart of the modulation domain infrared object tracking method the present invention is based on state vector increment.
Fig. 2 is that infrared target AM significant characteristics extract schematic diagram.
Fig. 3 is the modulation domain infrared object tracking schematic diagram based on state increment.
Fig. 4 is that target template updates flow chart.
Specific embodiment
One, the concept of infrared target particle filter tracking
Define the state vector x of a hexa-atomic elementkIndicate each frame image IkMoving target state in (x, y):
In formula (1), xk,ykIndicate target's center's position coordinates,Indicate the speed on target level and vertical direction
Spend information, δk,γkIndicate the size of target morphology model.
The linear goal state renewal equation of standard can indicate are as follows:
xk+1=Axk+vk (2)
In formula (2),
In formula (3),
In formula (4), | δ | < 1 indicates the dimensional variation of target morphology model in each frame image.
In formula (2),
The target observation model of standard can indicate are as follows:
zk=h (xk,nk) (6)
In formula (6), nkIndicate observation noise, zkIndicate observation, h () indicates observation function.
For the set of N number of particleThe weight of each particle isThen the posterior probability of standard can indicate
Are as follows:
In formula (7), the likelihood function (weight) of each particleIt can indicate are as follows:
In formula (8), ∝ expression is proportional to, and exp indicates exponent arithmetic,It is i-th of particle and frame target in kth frame
The similarity degree of template can be solved by normalized crosscorrelation algorithm (NCC):
In formula (9), (m, n) is pixel position coordinates, and M, N are the width of present frame, high, Tk(m, n) is in kth frame
Target template,It is the observation of particle in kth frame, ∑ indicates that summation, ∈ expression belong to.
Finally, the center of moving target can indicate are as follows:
Two, the concept that infrared target AM significant characteristics extract
The Gabor filter group that the present invention passes through 18 channels first analyzes the infrared video of input, establishes movement
The AM-FM characteristic model of target；Then, the AM conspicuousness of moving target is extracted using conspicuousness constituent analysis (DCA) algorithm
Feature.Fig. 2 is the schematic diagram that infrared target AM significant characteristics extract, the specific steps are as follows:
Step 1: the AM-FM characteristic model based on Gabor filter group is established.
Each frame image I is obtained using part Hilbert transformationkThe plural extension image s of (x, y)k(x, y):
In formula (11), j indicates imaginary number,Indicate part Hilbert transformation.
Plural extension image s in formula (11)k(x, y) can be rented with AM-FM function and be indicated, it may be assumed that
In formula (11), M=18 indicates the port number of filter in Gabor filter group, ak,m(x, y) indicates kth frame figure
The AM function at point (x, y), i.e. AM feature are obtained by m-th of Gabor filter as in,It indicates in kth frame image
FM function at point (x, y), i.e. FM feature are obtained by m-th of Gabor filter, exp indicates that exponent arithmetic, j indicate imaginary number.
AM-FM feature can pass through the channel response y of Gabor filter groupm(x, y), (m=1,2 ..., 18) it calculates
It arrives:
In formula (13), the real part of Re [] representative function,Indicate gradient.
In formula (14), Gm() indicates the frequency response of m-th of Gabor filter.
The part Hilbert transformation and the foundation of AM-FM characteristic model are detailed in document (Havlicek, Joseph P., et
al."Skewed 2D Hilbert transforms and computed AM-FM models."Image Processing,
1998.ICIP 98.Proceedings.1998International Conference on.Vol.1.IEEE,1998.)。
Formula (13-14) is the AM-FM characteristic model that the present invention establishes.
Step 2: AM significant characteristics extract.
Define a channel selection standard Γ (x, y):
In formula (15), max indicates maximizing.
Define AM significant characteristics aD(x, y) is the maximum channel pair of channel response in 18 channels of Gabor filter group
The AM characteristic value answered.
Three, based on the modulation domain infrared object tracking of state vector increment
The present invention is using the AM significant characteristics of moving target to the movement mesh in infrared video under particle filter frame
Mark carries out tenacious tracking, during tracking, excludes interference of the complex background to target using the state vector increment of target；Sentence
Whether the currently used target template that breaks needs to update, if meeting template renewal condition, updates current goal template, thus real
The tenacious tracking of existing infrared target.Fig. 3 is the modulation domain infrared object tracking schematic diagram based on state increment, and detailed process is such as
Under:
1, the statement of state vector increment
The original state vector of target is extended, definition status vector incremental vector are as follows:
In formula (16), subscript k indicates that present frame, k-1 indicate former frame, xk,yk,xk-1,yk-1Indicate target's center position
Coordinate,Indicate the velocity information on target level and vertical direction, δk,γk,δk-1,γk-1Indicate target shape
The size of states model.
Then the dbjective state equation of transfer based on state vector increment can indicate are as follows:
In formula (17), A is identical as formula (3), vkIt is identical as formula (5).
Target observation model based on state vector increment can indicate are as follows:
2, based on the modulation domain infrared object tracking of state vector increment
For input video, the location information of target in first frame image is extracted by the way of choosing manually, simultaneously
Extract target template.Since the second frame image, using the modulation domain infrared object tracking method based on state vector increment
Target is tracked.Specific tracking process is as follows:
For each frame image Ik(x, y), during the calculating infrared object tracking of the method as shown in formula (19)
Required moving target posterior probability, thus realize the tenacious tracking of infrared target,
In formula (19),It is the weight of i-th of particle in kth frame, ∝ expression is proportional to,It is tracking process
The moving target posterior probability of middle acquisition,It is the delta state vector of i-th of particle in kth frame,It is corresponding observation,
Exp indicates exponent arithmetic, and λ is the proportion adjustable factor,It is the similarity degree of i-th particle and the frame target template in kth frame.
In formula (19), the similarity degree of i-th particle and the frame target template in kth frameIt calculates by the following method
It obtains:
Present frame target AM significant characteristics and current are compared by normalized crosscorrelation algorithm (NCC) using formula (20)
The AM significant characteristics of template, solve the similarity degree of i-th particle and the frame target template in kth frame
In formula (2), (m, n) is pixel position coordinates, and M, N are the width of present frame, high, Tk(m, n) is in kth frame
Target template,It is the target template in -1 frame of kth,It is the observation of particle in kth frame,
It is the observation of particle in -1 frame of kth, ∑ indicates that summation, ∈ expression belong to.
3, target template updates
Fig. 4 is that target template updates flow chart.Steps are as follows for specific target template update:
Step 1: input picture Ik(x,y)；
Step 2: the similarity degree of i-th particle and the frame target template in kth frame is calculatedSimultaneously in particle assemblyMiddle find has maximum similarity degreeParticle
Step 3: compare in present frameThreshold value th size is updated with the target template of setting, if continuous nDFrameAll
Greater than the threshold value of setting, then needs to carry out target template update, go to step 4；Otherwise, into next frame, step 5 is gone to；
Step 4:In frame image, optimal particle is usedCalculate phase
Like degreeMaximum similarity degree coefficient is selected to correspond to frame k*, new target template is obtained using formula (21)
Step 5: completing the tracking of present frame, into next frame.
Four, a process of the method for the present invention is executed
Step 1: the infrared video sequence I containing moving target is shot using infrared camerak(x, y), k=1,2 ...,
Object as target following is carried out subsequent processing by n, the moving target in image；
Step 2: by the Gabor filter group in 18 channels to the infrared video sequence I of inputk(x, y), k=1,
2 ..., n is analyzed, and the AM-FM characteristic model s of moving target is establishedk(x, y), k=1,2 ..., n；
Step 3, using conspicuousness constituent analysis (DCA) algorithm from the AM-FM characteristic model s of moving targetk(x,y),k
=1,2 ..., the AM significant characteristics a of moving target is extracted in nD,k(x, y), k=1,2 ..., n, i.e. Gabor filter
In 18 channels of group, AM characteristic value corresponding to the maximum channel of channel response；
Step 4: the first frame I in input infrared video sequence image1(x, y), according to target's center coordinate (x1,y1) and
The wide w of target1, high h1True value initialized target template T1(m,n)；
Step 5: the AM significant characteristics a of moving target is utilizedD,k(x, y), k=1,2 ..., n is in particle filter frame
Under to infrared video sequence Ik(x, y), the moving target in k=2,3 ..., n are tracked, and target's center's coordinate (x is exportedk,
yk), k=2,3 ..., n and the wide w of targetk, k=2,3 ..., n, high hk, k=2,3 ..., n is as tracking result；
Step 6: judging whether currently used target template needs to update, if meeting template renewal condition, goes to step
Seven；Otherwise, step 5 is gone to；
Step 7: target template is updated, step 5 is gone to.
Beneficial effects of the present invention can be further illustrated by following experiment:
Conventional pixel domain particle filter tracking algorithm (PDT), conventional modulated domain particle filter tracking algorithm are used in experiment
(MDT), the modulation domain infrared object tracking method based on state vector increment (no target template updates, MDVT), and be based on
Four kinds of algorithms of modulation domain infrared object tracking method (having target template update, MDVT_DU) of state vector increment are compared.
As can be seen that target loss, MDVT finally has occurred as target is separate in PDT and MDT algorithm from experimental result
With MDVT_DU algorithm due to joined dbjective state vector increment, the judgement to moving target state in historical frames, energy are increased
It is enough effectively exclude during particle filter due to caused by particle resampling particle diversity lack problem, even if target is remote
From visual field, size changes, and is also able to achieve the tenacious tracking of infrared target.
Claims (7)
1. a kind of modulation domain infrared object tracking method based on state vector increment, it is characterised in that steps are as follows:
Step 1: the infrared video sequence I containing moving target is shot using infrared camerak(x, y), k=1,2 ..., n, image
In moving target by the object as target following carry out subsequent processing；
Step 2: by the Gabor filter group in 18 channels to the infrared video sequence I of inputk(x, y), k=1,2 ..., n into
Row analysis, establishes the AM-FM characteristic model s of moving targetk(x, y), k=1,2 ..., n；
Step 3, using conspicuousness constituent analysis DCA algorithm from the AM-FM characteristic model s of moving targetk(x, y), k=1,
2 ..., the AM significant characteristics a of moving target is extracted in nD,k(x, y), k=1,2 ..., n, i.e. Gabor filter group 18
In a channel, AM characteristic value corresponding to the maximum channel of channel response；
Step 4: the first frame I in input infrared video sequence image1(x, y), according to target's center coordinate (x1,y1) and target
Wide w1, high h1True value initialized target template T1(m,n)；
Step 5: the AM significant characteristics a of moving target is utilizedD,k(x, y), k=1,2 ..., n is right under particle filter frame
Infrared video sequence Ik(x, y), the moving target in k=2,3 ..., n are tracked, and target's center's coordinate (x is exportedk,yk),k
=2,3 ..., n and the wide w of targetk, k=2,3 ..., n, high hk, k=2,3 ..., n is as tracking result；
Step 6: judging whether currently used target template needs to update, if meeting template renewal condition, executes step 7；
Otherwise, step 5 is gone to；
Step 7: target template is updated, step 5 is gone to.
2. modulation domain infrared object tracking method as described in claim 1, it is characterised in that: in step 2: using 18 channels
Gabor filter group the infrared video of input is analyzed, establish the AM-FM characteristic model of moving target, the model is such as
Shown in formula (1):
In formula (1), sk(x, y) indicates input picture IkThe plural extension image of (x, y), M=18 indicate Gabor filter group
The port number of middle filter, ak,m(x, y) indicates to obtain the AM letter at point (x, y) in kth frame image by m-th of Gabor filter
Number, i.e. AM feature,It indicates to obtain the FM function at point (x, y) by m-th of Gabor filter in kth frame image, i.e.,
FM feature, exp indicate that exponent arithmetic, j indicate imaginary number.
3. modulation domain infrared object tracking method as described in claim 1, it is characterised in that: in step 3: using conspicuousness
Constituent analysis DCA algorithm extracts the AM significant characteristics of moving target, defines AM significant characteristics aD(x, y) is Gabor filter
In 18 channels of wave device group, the corresponding AM characteristic value in the maximum channel of channel response, channel selection standard is defined as:
In formula (2), Г (x, y) indicates channel selection standard, ym(x, y), (m=1,2 ..., 18) indicate Gabor filter group
Channel response, Gm() indicates the frequency response of m-th of Gabor filter, and max indicates maximizing, | | indicate absolute
Value.
4. modulation domain infrared object tracking method as described in claim 1, it is characterised in that: in step 5, use such as formula
(3) method shown in calculates moving target posterior probability needed for infrared object tracking process, to realize infrared target
Tenacious tracking,
In formula (3),It is the weight of i-th of particle in kth frame, ∝ expression is proportional to；It is that tracking obtains in the process
The moving target posterior probability obtained,It is the delta state vector of i-th of particle in kth frame,It is moving target observation；
Exp indicates exponent arithmetic, and λ is the proportion adjustable factor,It is the similarity degree of i-th particle and the frame target template in kth frame.
5. modulation domain infrared object tracking method as claimed in claim 4, it is characterised in that: i-th of particle in the kth frame
With the similarity degree of the frame target templateIt calculates and obtains by the following method:
Present frame target AM significant characteristics and current template are compared by normalized crosscorrelation algorithm (NCC) using formula (4)
AM significant characteristics, solve the similarity degree of i-th particle and the frame target template in kth frame
In formula (4), (m, n) is pixel position coordinates, and M, N are the width of present frame, high, Tk(m, n) is the target in kth frame
Template,It is the target template in -1 frame of kth,It is the observation of particle in kth frame,It is
The observation of particle in k-1 frame.
6. modulation domain infrared object tracking method as described in claim 1, it is characterised in that: target template described in step 6
Update condition are as follows: continuous nDFrameThe threshold value th both greater than set.
7. modulation domain infrared object tracking method as described in claim 1, it is characterised in that: target template described in step 7
Update method is as follows:
Step 1: input picture Ik(x,y)；
Step 2: the similarity degree of i-th particle and the frame target template in kth frame is calculatedSimultaneously in particle assembly
Middle find has maximum similarity degreeParticle
Step 3:In frame image, optimal particle is usedCalculate similar journey
DegreeMaximum similarity degree coefficient is selected to correspond to frame k*, new target template is obtained using formula (5)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711480518.XA CN109993771B (en) | 2017-12-29 | 2017-12-29 | Modulation domain infrared target tracking method based on state vector increment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711480518.XA CN109993771B (en) | 2017-12-29 | 2017-12-29 | Modulation domain infrared target tracking method based on state vector increment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109993771A true CN109993771A (en) | 2019-07-09 |
| CN109993771B CN109993771B (en) | 2022-09-13 |
Family
ID=67109111
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711480518.XA Active CN109993771B (en) | 2017-12-29 | 2017-12-29 | Modulation domain infrared target tracking method based on state vector increment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109993771B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115205510A (en) * | 2022-07-12 | 2022-10-18 | 中国科学院长春光学精密机械与物理研究所 | Complex scene infrared point target identification method based on saliency characteristic information discrimination |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105389807A (en) * | 2015-10-26 | 2016-03-09 | 南京理工大学 | Particle filter infrared tracking method with fusion of gradient feature and adaptive template |
| CN107146240A (en) * | 2017-05-05 | 2017-09-08 | 西北工业大学 | The video target tracking method of taking photo by plane detected based on correlation filtering and conspicuousness |
-
2017
- 2017-12-29 CN CN201711480518.XA patent/CN109993771B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105389807A (en) * | 2015-10-26 | 2016-03-09 | 南京理工大学 | Particle filter infrared tracking method with fusion of gradient feature and adaptive template |
| CN107146240A (en) * | 2017-05-05 | 2017-09-08 | 西北工业大学 | The video target tracking method of taking photo by plane detected based on correlation filtering and conspicuousness |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115205510A (en) * | 2022-07-12 | 2022-10-18 | 中国科学院长春光学精密机械与物理研究所 | Complex scene infrared point target identification method based on saliency characteristic information discrimination |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109993771B (en) | 2022-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107274419B (en) | Deep learning significance detection method based on global prior and local context | |
| CN103077521B (en) | A kind of area-of-interest exacting method for video monitoring | |
| CN108346159A (en) | A kind of visual target tracking method based on tracking-study-detection | |
| CN106709472A (en) | Video target detecting and tracking method based on optical flow features | |
| CN103886325B (en) | Cyclic matrix video tracking method with partition | |
| CN105513053B (en) | One kind is used for background modeling method in video analysis | |
| CN111340824B (en) | Image feature segmentation method based on data mining | |
| CN102542289A (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| CN110321937B (en) | Motion human body tracking method combining fast-RCNN with Kalman filtering | |
| Vosters et al. | Background subtraction under sudden illumination changes | |
| CN107680116A (en) | A kind of method for monitoring moving object in video sequences | |
| CN101216943B (en) | A method for video moving object subdivision | |
| Kothiya et al. | A review on real time object tracking in video sequences | |
| CN101945257A (en) | Synthesis method for extracting chassis image of vehicle based on monitoring video content | |
| CN109859246B (en) | Low-altitude slow unmanned aerial vehicle tracking method combining correlation filtering and visual saliency | |
| CN107563397A (en) | Cloud cluster method for calculation motion vector in a kind of satellite cloud picture | |
| CN106570885A (en) | Background modeling method based on brightness and texture fusion threshold value | |
| CN103886324B (en) | Scale adaptive target tracking method based on log likelihood image | |
| CN108038856B (en) | Infrared small target detection method based on improved multi-scale fractal enhancement | |
| CN104200455B (en) | A kind of key poses extracting method based on movement statistics signature analysis | |
| CN106446832B (en) | Video-based pedestrian real-time detection method | |
| CN109993771A (en) | Modulation domain infrared object tracking method based on state vector increment | |
| CN107564029B (en) | Moving target detection method based on Gaussian extreme value filtering and group sparse RPCA | |
| CN113066077B (en) | Flame detection method and device | |
| CN114283157A (en) | Ellipse fitting-based ellipse object segmentation method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |