CN103281476A - Television image moving target-based automatic tracking method - Google Patents
Television image moving target-based automatic tracking method Download PDFInfo
- Publication number
- CN103281476A CN103281476A CN2013101417296A CN201310141729A CN103281476A CN 103281476 A CN103281476 A CN 103281476A CN 2013101417296 A CN2013101417296 A CN 2013101417296A CN 201310141729 A CN201310141729 A CN 201310141729A CN 103281476 A CN103281476 A CN 103281476A
- Authority
- CN
- China
- Prior art keywords
- template
- target
- subtemplate
- equation
- moving target
- 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
- 238000000034 method Methods 0.000 title claims abstract description 38
- 239000011159 matrix material Substances 0.000 claims description 15
- 238000010168 coupling process Methods 0.000 claims description 14
- 238000005859 coupling reaction Methods 0.000 claims description 14
- 230000008878 coupling Effects 0.000 claims description 13
- 238000006467 substitution reaction Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 abstract description 8
- 239000013589 supplement Substances 0.000 abstract 1
- 230000000875 corresponding effect Effects 0.000 description 9
- 230000008901 benefit Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000013011 mating Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000007789 sealing Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a television image moving target-based automatic tracking method. The television image moving target-based automatic tracking method comprises the following steps of: 1, matching images by utilizing a plurality of sub-templates, and when the peaks of all the sub-templates are almost the same, matching the positions corresponding to the peaks with the initial positions of all the sub-templates; 2, updating and formulating a new template according to the matching degree at the optimal matching position between each template and the current image; and 3, predicting the region where the next frame of a moving target appears by utilizing a Kalman filter, and performing matching operation in the predicted region to find an optimal related matching point. By the method, the moving target in a non-stationary background is tracked accurately by matching the images by utilizing the plurality of sub-templates and the supplement of a suitable template updating strategy; and then by fully utilizing the predicting function of the Kalman filter, the operating region of a correlation matching algorithm is greatly reduced, and the real-time performance of the system is improved, so that the correlation tracking of the target becomes more active.
Description
Technical field
The present invention relates to image processing field, be specifically related to a kind of automatic tracking method based on the television image moving target.
Background technology
TV tracker system converts the aerial sports target to image sequence by video camera, and adopts the realization of a series of images Processing Algorithm to extract real-time and the tracking of target.TV tracker system is an integrated system that relates to numerous areas such as Digital Image Processing, pattern recognition, artificial intelligence, adaptive control, and its tracking accuracy, real-time, reliability, anti-complex background ability to work, intelligent degree have very high requirement.The tracking of target is unusual critical step, should accomplish accurate tracking target in the practical application scene, improves the real-time of system again.
A kind of tracking scheme of prior art is based on the tracking of active contour, be to utilize the curved profile of sealing to come the expressive movement target, and this profile can upgrade automatically and continuously.Profile is expressed the advantage that reduces computation complexity, if beginning can reasonably separate each moving target and realize the initialized words of profile, both made under the situation that has partial occlusion to exist and also can follow the tracks of continuously, yet initialization is normally very difficult.
Another kind of scheme is based on the tracking of model, utilizes point, line, zone that tracked target is fitted to a geometrical model, and the tracking of moving target has become the target identification problem.This method contains high-rise semantic description and knowledge, therefore, compares with other tracking, and this method has a lot of advantages, and this advantage seems particularly outstanding under complex environment.Its shortcoming is that amount of calculation is bigger, and need know a large amount of prioris about want tracking target.
The difficulty that aspect moving object detection and tracking, exists be to a great extent since in the actual environment complexity of target travel and particularity, the complexity that video data has cause.Such as because the variation of on-the-spot light luminance makes background image also change thereupon, thus be difficult to these change with image in since the variation that the introducing of foreground target causes distinguished; The dash area of foreground target of motion may cause that partial picture brightness changes in the background, in addition between Yun Dong the target, and the target of moving and the overlapping covering between the background, all may change shape and the further feature of the moving target that detects; When the foreground target of motion is similar aspect external appearance characteristics such as CF to the scenery in the background, the difficulty of telling foreground target from background will be increased; Background is not exclusively static etc.
Therefore, be necessary to provide a kind of automatic tracking method of new television image moving target to solve above-mentioned defective.
Summary of the invention
The purpose of this invention is to provide a kind of tracking and matching accurately based on the automatic tracking method of television image moving target, and described method has forecast function, make target following have more real-time and initiative.
The invention provides a kind of automatic tracking method based on the television image moving target, may further comprise the steps: step 1: utilize the multi-tool plate that image is mated, when the basically identical of the top of each subtemplate, the location matches of corresponding position, top each subtemplate when initial then; Step 2: upgrade according to the matching degree at the best match position place of template and present image and to formulate new template; Step 3: utilize the next frame of Kalman filter predicted motion target the zone to occur, in the estimation range, carry out matching operation, find best relevant matches point.
Preferably, described step 2 further comprises: step 21: the 1st frame in tracing process, mate with template, obtain a best match position, and a plurality of subtemplate mates all in the tram, calculate the best correlation of each subtemplate respectively, at this moment, each subtemplate is tightly adjacent; Step 22: according to the adjacent situation of each subtemplate position new template more, only adjacent subtemplate is handled, according to confidence level old template and new template are weighted, the template after obtaining to upgrade, method of weighting is as follows: M
+=α M
-+ (1-α) M
n, wherein, α is weight coefficient 0≤α≤1, M
+Be the template after upgrading, M
-Be the old template of using before refreshing, and M
nSubimage for match point place corresponding templates in the current frame image; Step 23: the new template after utilization is upgraded is carried out the coupling of next frame image, obtains new current best match position, and repeating step 22 then, and the circulation coupling is revised template.
Preferably, described Kalman filter is specially: state equation and the observational equation of establishing linear system are respectively: state equation: x
k=Ax
K-1+ ω
K-1, observational equation: z
k=H
kx
k+ υ
k, wherein, x
kBe that k maintains the system state vector in n * 1 constantly; z
kIt is k m * 1 dimension observation vector constantly; A is that n * n maintains the system state-transition matrix; H
kBe that m * n maintains overall view survey matrix; ω
kIt is k n * 1 dimension random disturbances noise vector constantly; υ
kBe that k maintains overall view survey noise vector, ω in m * 1 constantly
k, υ
kUsually be assumed to be mutually independently zero-mean white Gaussian noise vector, make Q
kAnd R
kBe respectively their covariance matrixes:
Q
k=E{ω
kω
k T}
R
k=E{υ
kυ
k T}
Because system is definite, then A and H
kKnown, and ω
K-1And υ
kSatisfy certain hypothesis, also known, establish P
kBe x
kCovariance matrix, P
k' be x
kWith
The error covariance matrix, the Kalman filter reduces to minimum to the error covariance of the posterior estimate of the system mode of each moment point k, the Kalman filter equation is as follows:
The status predication equation:
Error covariance predictive equation: P
k'=AP
K-1A
TQ
Kalman gain coefficient equation: K
k=P
k' H
k T(H
kP
k' H
k T+ R)
-1
The state update equation:
Covariance update equation: P
k=(I-K
kH
k) P
k'.
Preferably, described step 3 further comprises: step 31: initialization, when using the Kalman filter for the first time, to carry out initialization to filter, with x
0Initialize is initial position and the speed of target, under the speed condition of unknown, can be made as 0, and the record present image constantly, establishes initial error covariance P simultaneously
0=0; Step 32: prediction, in every two field picture of new input, carry out match search before, the time interval Δ t of record and previous frame image, in the substitution formula, the motion state of prediction current goal
, the error of prediction is designated as Δ p
k=ω
k-s
k, being used for the calculating of next frame region of search, substitution error covariance predictive equation is predicted new error covariance; Step 33: coupling, set with
In (xs
k, ys
k) centered by the zone be the region of search, in this zone, seek best match position, find optimal moving target, target area image is copied to T
K+1And first pixel coordinate of the upper left corner, target area is two-dimensional observation vector (x ω
k, y ω
k), substitution state update equation obtains (xs
K+1, ys
K+1), calculate the measuring speed υ of target simultaneously
K+1=(s
K+1-s
k)/Δ t; Step 34: revise, obtain Kalman filter gain coefficient, with z
k=(x ω
k, y ω
k)
TIn the substitution state update equation, obtain by the revised state vector of current actual observation, simultaneously the round-off error covariance matrix.
Compared with prior art, the automatic tracking method based on the television image moving target of the present invention, the present invention adopted multi-tool version matching process to come tracking target, the tracking problem when having solved target by partial occlusion.The template renewal and the correction strategy that propose have overcome that variations such as size, shape, attitude may appear in target in tracing process, the long-time stable tracking of the system that kept.The forecast function that takes full advantage of the Kalman filter is predicted the zone that the next frame target may occur, in less estimation range, carry out the relevant matches computing then, find best relevant matches point, significantly reduced amount of calculation, antijamming capability also significantly strengthens, and makes the target correlation tracking have more initiative.
Description of drawings
In order to be illustrated more clearly in the embodiment of the invention or technical scheme of the prior art, to do to introduce simply to the accompanying drawing of required use in embodiment or the description of the Prior Art below, apparently, accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills, under the prerequisite of not paying creative work, can also obtain other accompanying drawing according to these accompanying drawings.
Fig. 1 is the automatic tracking method flow chart based on the television image moving target of the present invention;
Fig. 2 is that the embodiment of the invention adopts 4 subtemplates to carry out the schematic diagram of template matches;
Fig. 3 a-3d is the schematic diagram of 4 subtemplates shown in Figure 2 different situations of mating.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the invention, the technical scheme in the embodiment of the invention is clearly and completely described, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, those of ordinary skills belong to the scope of protection of the invention not making all other embodiment that obtain under the creative work prerequisite.
As mentioned above, the invention provides a kind of automatic tracking method based on the television image moving target, utilize multi-tool plate correlation matching algorithm, be aided with suitable masterplate update strategy again, accomplished the accurate tracking to moving target under the nonstatic background; Take full advantage of the forecast function of Kalman filter then, dwindle the computing zone of correlation matching algorithm greatly, improved the real-time of system, make target correlation tracking more initiatively.
With reference to figure 1, the automatic tracking method based on the television image moving target of the embodiment of the invention may further comprise the steps:
Step S001: utilize the multi-tool plate that image is mated, when the basically identical of the top of each subtemplate, the location matches of corresponding position, top each subtemplate when initial then;
Step S002: upgrade according to the matching degree at the best match position place of template and present image and to formulate new template;
Step S003: utilize the next frame of Kalman filter predicted motion target the zone to occur, in the estimation range, carry out matching operation, find best relevant matches point.
Suppose that template is N * N dimension, and the realtime graphic frame is M * M dimension (M〉N).At first, reference map being divided into n size is
Subtemplate, respectively each subtemplate and real-time frame are carried out correlation ratio then, because the pixel of subtemplate is less, so when partial pixel changes, mainly influence is corresponding subtemplate, and less to the influence of other subtemplate, set up comprehensive matching result with the matching result of these subtemplates, the problem that occurs in following the tracks of is made preferably being judged accordingly.The algorithm of multi-tool plate is as follows:
M(u,v)=M
1(u
1,v
1)∪M
2(U
2,V
2)∪···∪M
n(u
n,v
n)
Wherein, M
n(u
n, v
n) representing each subtemplate, n represents the number of subtemplate.Because multi-tool plate matching algorithm has adopted the integrated conduct method that is similar to k-out-of-n system, therefore need 2 subtemplates just can draw the update information of physical location at least.
When the basically identical of the top of each subtemplate, then corresponding position, top row of each subtemplate when initial is similar to the position; Pseudo-peak occurs as one or several subtemplate, they have formed unreasonable relative position, then can come out by the position resolution of certain criterion with these pseudo-peak correspondences.So multi-tool plate related algorithm has the ability that certain identification false appearance is closed peak value.
Because positioning accuracy is directly proportional with relevant area, the group template size is reduced into by N * N dimension
During dimension, its positioning accuracy descends nearly n doubly; But, when n subtemplate acting in conjunction, positioning accuracy can be improved n again doubly, so multi-tool plate related algorithm has possessed the coupling positioning accuracy of large form under the situation that does not change single mode plate size, significantly reduce the influence of geometric distortion simultaneously again.
With reference to figure 2, select N=4, namely utilize 4 subtemplates to carry out correlation computations, the division of subtemplate is as shown in the figure.V and U are the sizes of whole template, and the weight coefficient of each subtemplate is 1, their size and in image corresponding hunting zone as follows:
First subtemplate: size:
The field of search: 0≤x＜X-U+1,0≤y＜Y-V+1;
Second subtemplate: size:
The field of search: 0≤x＜X-U+1,
The 3rd subtemplate: size:
The field of search:
0≤y＜Y-V+1;
The 4th subtemplate: size:
The field of search:
Carry out correlation computations in the corresponding field of search respectively with these 4 subtemplates, and record is maximum or the position of minimum correlation peak, in corresponding array as a result, is superimposed with the probability distribution stencil value that target exists then.
Subtemplate carries out correlation computations respectively and gets correlation R in the field of search
i(x, y), establishing its peak point is P
Ip(x
Ip, y
Ip), target is at this point and obey probability distribution on every side, thus, can obtain target and get probability (comprehensive matching result) at every bit:
P(x,y)=∑P
i(x,y)
Can determine the possibility that target exists according to the size of value, as P
p(x
p, y
p)=1 o'clock represents that the equal indicating target of each subtemplate matching result in a certain position, works as P
Max(x
p, y
p)=max{P
i(x, y) } time, represent that target has been lost.
Target under the correlation tracking method energy tenacious tracking complex background of multi-tool plate, and can judge the subproblem that occurs in the tracking.But, owing to all need to comprise certain object pixel in each subtemplate, so this method is not suitable for the tracking of Weak target.For making each subtemplate correlated results reliable, the size of each subtemplate is more preferably greater than 4 * 4.
With reference to figure 3a to 3d, the relative position between each subtemplate is very important, and we can do further more careful judgement to coupling tracking situation the observation by location information.Relative position between each subtemplate divides four kinds of situations in simple terms, respectively as shown in Fig. 3 a to 3d:
When the subtemplate situation was shown in Fig. 3 a, four subtemplates all mated fully, and the match is successful for the expression target; When the subtemplate situation is shown in Fig. 3 b (it is adjacent, right adjacent, upward adjacent, adjacent down to contain a left side), expression has only two sub-template matches successes, loses for all the other two, and can determine the tram of target this moment according to these two templates that the match is successful; When the subtemplate situation was shown in Fig. 3 c, representing had one to lose in four subtemplates, and can determine the tram of target this moment according to three remaining subtemplates; When the subtemplate situation is shown in Fig. 3 d, can't determine the tram of target according to the relative position between the subtemplate, must assist and determine which subtemplate with out of Memory the match is successful or four subtemplates are all lost.When each subtemplate is adjacent, the phase contact is always arranged, establishing this phase contact is P, and under the normal condition, this P point should be the central point of template.When tracking target is the expansion target, adjacent half that can be defined as ultimate range between two subtemplate borders and be no more than the subtemplate length of side, and after the coupling, subtemplate one, subtemplate three centers are always on the left side at subtemplate two, subtemplate four centers, always in the top at subtemplate three, subtemplate four centers, phase contact P also will redefine for subtemplate one, subtemplate two centers.
Target is being carried out in the correlation tracking process, To Template has maintained the dynamic process of whole tracking, in sequence image, variations such as size, shape, attitude may appear in target, add the various interference in the battlefield surroundings, and image handles the precision problem of minimum unit of measurement, real image certainly existing distortion, noise, block etc. and to change, slight error on the single frames can accumulate gradually along with motion, finally can cause the drift of target following point.It is that long-time stable key is kept in correlation tracking that template is reasonably upgraded, and selects effective template renewal and correction strategy can overcome these variations to a certain extent to the influence of tracking effect.Therefore, described step S002 further comprises:
Step S021: the 1st frame in tracing process, mate with template, obtain a best match position, and a plurality of subtemplate mates all in the tram, calculate the best correlation of each subtemplate respectively, at this moment, each subtemplate is tightly adjacent;
Step S022: according to the adjacent situation of each subtemplate position new template more, only adjacent subtemplate is handled, according to confidence level old template and new template are weighted, the template after obtaining to upgrade, method of weighting is as follows: M
+=α M
-+ (1-α) M
n, wherein, M
+Be the template after upgrading, M
-Be the old template of using before refreshing, and M
nSubimage for match point place corresponding templates in the current frame image; Here α is weight coefficient 0≤α≤1, and the value of α is important, if value is excessive, the ratio regular meeting that old template accounts for is excessive, when big as if object variations in the present frame, can cause losing of target; If too small, the ratio regular meeting that current frame image accounts for is excessive, when the target part temporarily is blocked, carries out To Template and refreshes the characteristic that then makes template can not reflect target, can lose target equally.
Step S023: the new template after utilization is upgraded is carried out the coupling of next frame image, obtain new current best match position, repeating step S022 then, enter in the circulation of " coupling---obtain current best match position---calculate matching degree---revise template---coupling ", the circulation coupling is revised template.
In multi-tool plate coupling, the confidence calculations here is only at the adjacent situation of first three seed pattern in the subtemplate relative position, and coefficient correlation is the mean value of adjacent subtemplate.When α=0 upgrades subtemplate, if the complete non-conterminous situation of each subtemplate, new template and continue to use old template more not then, if the adjacent situation of first three seed pattern, then whole large form is upgraded in the position of ordering according to P.
Preferably, among the described step S003, the Kalman filter is specially: state equation and the observational equation of establishing linear system are respectively: state equation: x
k=Ax
K-1+ ω
K-1, observational equation: z
k=H
kx
k+ υ
k, wherein, x
kBe that k maintains the system state vector in n * 1 constantly; z
kIt is k m * 1 dimension observation vector constantly; A is that n * n maintains the system state-transition matrix; H
kBe that m * n maintains overall view survey matrix; ω
kIt is k n * 1 dimension random disturbances noise vector constantly; υ
kBe that k maintains overall view survey noise vector, ω in m * 1 constantly
k, υ
kUsually be assumed to be mutually independently zero-mean white Gaussian noise vector, make Q
kAnd R
kBe respectively their covariance matrixes:
Q
k=E{ω
kω
k T}
R
k=E{υ
kυ
k T}
Because system is definite, then A and H
kKnown, and ω
K-1And υ
kSatisfy certain hypothesis, also known, establish P
kBe x
kCovariance matrix, P
k' be x
kWith
The error covariance matrix, the Kalman filter reduces to minimum to the error covariance of the posterior estimate of the system mode of each moment point k, the Kalman filter equation is as follows:
The status predication equation:
Error covariance predictive equation: P
k'=AP
K-1A
TQ
Kalman gain coefficient equation: K
k=P
k' H
k T(H
kP
k' H
k T+ R)
-1
The state update equation:
Covariance update equation: P
k=(I-K
kH
k) P
k'.
Preferably, described step 3 further comprises: step 31: initialization, when using the Kalman filter for the first time, to carry out initialization to filter, with x
0Initialize is initial position and the speed of target, under the speed condition of unknown, can be made as 0, and the record present image constantly, establishes initial error covariance P simultaneously
0=0;
Step 32: prediction, in every two field picture of new input, carry out match search before, the time interval Δ t of record and previous frame image, in the substitution formula, the motion state of prediction current goal
, the error of prediction is designated as Δ p
k=ω
k-s
k, being used for the calculating of next frame region of search, substitution error covariance predictive equation is predicted new error covariance;
Step 33: coupling, set with x
In (xs
k, ys
k) centered by the zone be the region of search, in this zone, seek best match position, find optimal moving target, target area image is copied to T
K+1And first pixel coordinate of the upper left corner, target area is two-dimensional observation vector (x ω
k, y ω
k), substitution state update equation obtains (xs
K+1, ys
K+1), calculate the measuring speed υ of target simultaneously
K+1=(s
K+1-s
k)/Δ t;
Step 34: revise, obtain Kalman filter gain coefficient, with z
k=(x ω
k, y ω
k)
TIn the substitution state update equation, obtain by the revised state vector of current actual observation, simultaneously the round-off error covariance matrix.
Compared with prior art, the automatic tracking method based on the television image moving target of the present invention has adopted multi-tool version matching process to come tracking target, the tracking problem when having solved target by partial occlusion.The template renewal and the correction strategy that propose have overcome that variations such as size, shape, attitude may appear in target in tracing process, the long-time stable tracking of the system that kept.The forecast function that takes full advantage of the Kalman filter is predicted the zone that the next frame target may occur, in less estimation range, carry out the relevant matches computing then, find best relevant matches point, significantly reduced amount of calculation, antijamming capability also significantly strengthens, and makes the target correlation tracking have more initiative.
More than the automatic tracking method based on the television image moving target that the embodiment of the invention is provided, be described in detail, used specific case among the present invention principle of the present invention and execution mode are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, the part that all can change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.
Claims (4)
1. the automatic tracking method based on the television image moving target is characterized in that, may further comprise the steps:
Step 1: utilize the multi-tool plate that image is mated, when the basically identical of the top of each subtemplate, the location matches of corresponding position, top each subtemplate when initial then;
Step 2: upgrade according to the matching degree at the best match position place of template and present image and to formulate new template;
Step 3: utilize the next frame of Kalman filter predicted motion target the zone to occur, in the estimation range, carry out matching operation, find best relevant matches point.
2. the automatic tracking method based on the television image moving target as claimed in claim 1 is characterized in that, described step 2 further comprises:
Step 21: the 1st frame in tracing process, mate with template, obtain a best match position, and a plurality of subtemplate mates all in the tram, calculate the best correlation of each subtemplate respectively, at this moment, each subtemplate is tightly adjacent;
Step 22: according to the adjacent situation of each subtemplate position new template more, only adjacent subtemplate is handled, according to confidence level old template and new template are weighted, the template after obtaining to upgrade, method of weighting is as follows: M
+=α M
-+ (1-α) M
n, wherein, α is weight coefficient 0≤α≤1, M
+Be the template after upgrading, M
-Be the old template of using before refreshing, and M
nSubimage for match point place corresponding templates in the current frame image;
Step 23: the new template after utilization is upgraded is carried out the coupling of next frame image, obtains new current best match position, and repeating step 22 then, and the circulation coupling is revised template.
3. the automatic tracking method based on the television image moving target as claimed in claim 1 is characterized in that, described Kalman filter is specially:
If the state equation of linear system and observational equation are respectively: state equation: x
k=Ax
K-1+ ω
K-1, observational equation: z
k=H
kx
k+ υ
k, wherein, x
kBe that k maintains the system state vector in n * 1 constantly; z
kIt is k m * 1 dimension observation vector constantly; A is that n * n maintains the system state-transition matrix; H
kBe that m * n maintains overall view survey matrix; ω
kIt is k n * 1 dimension random disturbances noise vector constantly; υ
kBe that k maintains overall view survey noise vector, ω in m * 1 constantly
k, υ
kUsually be assumed to be mutually independently zero-mean white Gaussian noise vector, make Q
kAnd R
kBe respectively their covariance matrixes:
Q
k=E{ω
kω
k T}
R
k=E{υ
kυ
k T}
Because system is definite, then A and H
kKnown, and ω
K-1And υ
kSatisfy certain hypothesis, also known, establish P
kBe x
kCovariance matrix, P
k' be x
kWith
The error covariance matrix, the Kalman filter reduces to minimum to the error covariance of the posterior estimate of the system mode of each moment point k, the Kalman filter equation is as follows:
The status predication equation:
Error covariance predictive equation: P
k'=AP
K-1A
TQ
Kalman gain coefficient equation: K
k=P
k' H
k T(H
kP
k' H
k T+ R)
-1
The state update equation:
Covariance update equation: P
k=(I-K
kH
k) P
k'.
4. the automatic tracking method based on the television image moving target as claimed in claim 3 is characterized in that, described step 3 further comprises:
Step 31: initialization, when using the Kalman filter for the first time, to carry out initialization to filter, with x
0Initialize is initial position and the speed of target, under the speed condition of unknown, can be made as 0, and the record present image constantly, establishes initial error covariance P simultaneously
0=0;
Step 32: prediction, in every two field picture of new input, carry out match search before, the time interval Δ t of record and previous frame image, in the substitution formula, the motion state of prediction current goal
, the error of prediction is designated as Δ p
k=ω
k-s
k, being used for the calculating of next frame region of search, substitution error covariance predictive equation is predicted new error covariance;
Step 33: coupling, set with
In (xs
k, ys
k) centered by the zone be the region of search, in this zone, seek best match position, find optimal moving target, target area image is copied to T
K+1And first pixel coordinate of the upper left corner, target area is two-dimensional observation vector (x ω
k, y ω
k), substitution state update equation obtains (xs
K+1, ys
K+1), calculate the measuring speed υ of target simultaneously
K+1=(s
K+1-s
k)/Δ t;
Step 34: revise, obtain Kalman filter gain coefficient, with z
k=(x ω
k, y ω
k)
TIn the substitution state update equation, obtain by the revised state vector of current actual observation, simultaneously the round-off error covariance matrix.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013101417296A CN103281476A (en) | 2013-04-22 | 2013-04-22 | Television image moving target-based automatic tracking method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013101417296A CN103281476A (en) | 2013-04-22 | 2013-04-22 | Television image moving target-based automatic tracking method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103281476A true CN103281476A (en) | 2013-09-04 |
Family
ID=49063910
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013101417296A Pending CN103281476A (en) | 2013-04-22 | 2013-04-22 | Television image moving target-based automatic tracking method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103281476A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105022889A (en) * | 2015-08-04 | 2015-11-04 | 深圳市哈工大交通电子技术有限公司 | Video-based fast template matching GPU implementation method |
| CN105260740A (en) * | 2015-09-23 | 2016-01-20 | 广州视源电子科技股份有限公司 | Element identification method and device |
| CN105894535A (en) * | 2016-03-30 | 2016-08-24 | 中国科学院地理科学与资源研究所 | Bayes-based vortex automatic tracking method |
| CN108038868A (en) * | 2017-10-17 | 2018-05-15 | 国网河南省电力公司郑州供电公司 | Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model |
| CN110244771A (en) * | 2019-05-22 | 2019-09-17 | 安徽翔翼智能科技有限公司 | A kind of unmanned plane mission payload real training adaptive tracking control method |
| WO2020078140A1 (en) * | 2018-10-16 | 2020-04-23 | 北京理工大学 | Optimization system and method for motion model |
| CN111640140A (en) * | 2020-05-22 | 2020-09-08 | 北京百度网讯科技有限公司 | Target tracking method and device, electronic equipment and computer readable storage medium |
| CN117644516A (en) * | 2024-01-09 | 2024-03-05 | 中核四川环保工程有限责任公司 | Nuclear waste treatment monitoring robot based on image processing |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100224752B1 (en) * | 1997-06-11 | 1999-10-15 | 윤종용 | Target tracking method and apparatus |
| CN102646279A (en) * | 2012-02-29 | 2012-08-22 | 北京航空航天大学 | Anti-shielding tracking method based on moving prediction and multi-sub-block template matching combination |
| CN102737386A (en) * | 2012-05-15 | 2012-10-17 | 北京硅盾安全技术有限公司 | Moving target anti-fusion shielding tracking algorithm |
-
2013
- 2013-04-22 CN CN2013101417296A patent/CN103281476A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100224752B1 (en) * | 1997-06-11 | 1999-10-15 | 윤종용 | Target tracking method and apparatus |
| CN102646279A (en) * | 2012-02-29 | 2012-08-22 | 北京航空航天大学 | Anti-shielding tracking method based on moving prediction and multi-sub-block template matching combination |
| CN102737386A (en) * | 2012-05-15 | 2012-10-17 | 北京硅盾安全技术有限公司 | Moving target anti-fusion shielding tracking algorithm |
Non-Patent Citations (1)
| Title |
|---|
| 魏小鹏，刘瑞，张强，肖伯祥: "基于模板匹配的人体运动捕获数据处理方法", 《系统仿真学报》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105022889A (en) * | 2015-08-04 | 2015-11-04 | 深圳市哈工大交通电子技术有限公司 | Video-based fast template matching GPU implementation method |
| CN105260740A (en) * | 2015-09-23 | 2016-01-20 | 广州视源电子科技股份有限公司 | Element identification method and device |
| CN105894535A (en) * | 2016-03-30 | 2016-08-24 | 中国科学院地理科学与资源研究所 | Bayes-based vortex automatic tracking method |
| CN105894535B (en) * | 2016-03-30 | 2019-05-31 | 中国科学院地理科学与资源研究所 | A kind of vortex method for automatic tracking based on Bayes |
| CN108038868A (en) * | 2017-10-17 | 2018-05-15 | 国网河南省电力公司郑州供电公司 | Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model |
| WO2020078140A1 (en) * | 2018-10-16 | 2020-04-23 | 北京理工大学 | Optimization system and method for motion model |
| CN110244771A (en) * | 2019-05-22 | 2019-09-17 | 安徽翔翼智能科技有限公司 | A kind of unmanned plane mission payload real training adaptive tracking control method |
| CN111640140A (en) * | 2020-05-22 | 2020-09-08 | 北京百度网讯科技有限公司 | Target tracking method and device, electronic equipment and computer readable storage medium |
| WO2021232652A1 (en) * | 2020-05-22 | 2021-11-25 | 北京百度网讯科技有限公司 | Target tracking method and apparatus, electronic device, and computer-readable storage medium |
| JP2023500969A (en) * | 2020-05-22 | 2023-01-11 | ベイジン バイドゥ ネットコム サイエンス テクノロジー カンパニー リミテッド | Target Tracking Method, Apparatus, Electronics, Computer Readable Storage Medium and Computer Program Product |
| JP7375192B2 (en) | 2020-05-22 | 2023-11-07 | ベイジン バイドゥ ネットコム サイエンス テクノロジー カンパニー リミテッド | Target tracking methods, devices, electronic devices, computer readable storage media and computer program products |
| CN117644516A (en) * | 2024-01-09 | 2024-03-05 | 中核四川环保工程有限责任公司 | Nuclear waste treatment monitoring robot based on image processing |
| CN117644516B (en) * | 2024-01-09 | 2024-06-11 | 中核四川环保工程有限责任公司 | Nuclear waste treatment monitoring robot based on image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103281476A (en) | Television image moving target-based automatic tracking method | |
| CN107292911B (en) | Multi-target tracking method based on multi-model fusion and data association | |
| CN111190981B (en) | Method and device for constructing three-dimensional semantic map, electronic equipment and storage medium | |
| CN109800689B (en) | Target tracking method based on space-time feature fusion learning | |
| CN108242079B (en) | VSLAM method based on multi-feature visual odometer and graph optimization model | |
| CN110533687B (en) | Multi-target three-dimensional track tracking method and device | |
| CN109102547A (en) | Robot based on object identification deep learning model grabs position and orientation estimation method | |
| Liu et al. | Detection and pose estimation for short-range vision-based underwater docking | |
| CN113674416B (en) | Three-dimensional map construction method and device, electronic equipment and storage medium | |
| CN108694724A (en) | A kind of long-time method for tracking target | |
| CN110717927A (en) | Indoor robot motion estimation method based on deep learning and visual inertial fusion | |
| CN111489394B (en) | Object posture estimation model training method, system, device and medium | |
| CN104517275A (en) | Object detection method and system | |
| JP7263216B2 (en) | Object Shape Regression Using Wasserstein Distance | |
| KR20220043847A (en) | Method, apparatus, electronic device and storage medium for estimating object pose | |
| CN112947419B (en) | Obstacle avoidance method, device and equipment | |
| CN105488811A (en) | Depth gradient-based target tracking method and system | |
| CN110260866A (en) | A kind of robot localization and barrier-avoiding method of view-based access control model sensor | |
| CN109543694A (en) | A kind of visual synchronization positioning and map constructing method based on the sparse strategy of point feature | |
| CN112967340A (en) | Simultaneous positioning and map construction method and device, electronic equipment and storage medium | |
| CN115063447A (en) | Target animal motion tracking method based on video sequence and related equipment | |
| CN111998862A (en) | Dense binocular SLAM method based on BNN | |
| CN111354022A (en) | Target tracking method and system based on kernel correlation filtering | |
| CN115457086A (en) | Multi-target tracking algorithm based on binocular vision and Kalman filtering | |
| CN112233149A (en) | Scene flow determination method and device, storage medium and electronic device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20130904 |