CN104574442B - Adaptive particle swarm optimization particle filter motion target tracking method - Google Patents
Adaptive particle swarm optimization particle filter motion target tracking method Download PDFInfo
- Publication number
- CN104574442B CN104574442B CN201510019086.7A CN201510019086A CN104574442B CN 104574442 B CN104574442 B CN 104574442B CN 201510019086 A CN201510019086 A CN 201510019086A CN 104574442 B CN104574442 B CN 104574442B
- Authority
- CN
- China
- Prior art keywords
- particle
- adaptive
- swarm optimization
- particle filter
- tracking
- 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.)
- Active
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
- G06T7/277—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
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
Abstract
The invention discloses a kind of adaptive particle swarm optimization particle filter motion target tracking method, belong to image procossing and intelligent Video Surveillance Technology field.The inventive method is optimized during the tracking of moving target is carried out with particle filter method, using particle group optimizing method to the position of particle;When being optimized using particle group optimizing method to the position of particle, the quantity of particle is adaptively adjusted according to the change in location situation of global optimum's particle.Compared with prior art, the present invention not only can effectively mitigate the local optimum phenomenon in particle group optimizing, improve the degree of accuracy of target following, while have the advantages of algorithm complex is low, and real-time is good again.
Description
Technical field
The present invention relates to a kind of motion target tracking method, more particularly to a kind of adaptive particle swarm optimization particle filter to transport
Tracking of maneuvering target method, belong to image procossing and intelligent Video Surveillance Technology field.
Background technology
Motion target tracking is the focus of image procossing and intelligent Video Surveillance Technology area research, and robot is led
The key technology in the fields such as boat, precise guidance, there is extensive use.Such as:The track and posture that target is moved in capture experiment,
Detection to traffic highway flow, the video monitoring of important events and analysis etc..Therefore, the tracking of moving target has very heavy
The Research Significance wanted.The tracking of moving target includes several portion of techniques such as the detection, feature extraction and matched jamming of target
Composition.Wherein, the detection of target, clarification of objective extraction need certain priori, then further according to certain algorithm, profit
With Given information prediction subsequent time target movable information (position, speed etc.) before, motion target tracking is realized.One good
Good track algorithm should have the characteristics such as reliability is high, real-time is good, accuracy is accurate.
Many effective algorithms have been proposed in tracking on moving target, wherein, establish in kalman filtering theory
On the basis of target following technology receive very big concern.But Kalman filtering is only applicable to linear system, therefore,
Kalman filtering is further applicable to non-linear domain variability and proposes EKF (EKF) by Sunahara, Buey et al.,
But this method computational complexity considerably increases, therefore it is not used widely in reality.Another kind of conventional method
It is mean shift algorithm --- Mean-shift algorithms [D.Comaniciu, V.Ramesh and P.Meer, " Real-time
tracking of non-rigid objects using Mean Shift",Proceedings of IEEE Computer
Society Conference on Computer Vision and Pattern Recognition,2:142-149,
2000], it is a kind of printenv algorithm for estimating based on kernel function, it is not necessary to priori, and also convergence rate ratio is very fast, but
It is for having limitation under quick motion and non-Gaussian noise environment.Particle filter (PF:Particle Filtering) algorithm
Increasingly it is taken seriously in target tracking domain in recent years, it is both not only restricted to linear system, does not also require that noise obeys Gauss
Distribution, and reliable tracking can also be realized when target is blocked.But there is particle dilution and amount of calculation in particle filter
The problem of big is the significant obstacle of practical application.
In order to solve the problems, such as the particle dilution present in particle filter motion target tracking, many scholars are selected particle
Colony optimization algorithm (Particle Swarm optimization, abbreviation PSO) is applied among particle filter, forms population
Optimized particle filter algorithm (abbreviation PSOPF), so that the diversity of particle is ensured, but in existing PSOPF algorithms
Particle swarm optimization algorithm easily causes particle to be absorbed in local best points, so as to cause the positional information positioning to target not accurate enough
Really.
For the problem, some adaptive particle swarm optimization particle filter algorithms are suggested, for example, there is researcher to propose one
The dynamic particles group optimized particle filter algorithm algorithms that kind new neighborhood adaptively adjusts consider the neighborhood information of particle,
The neighborhood particle quantity of particle is adaptively adjusted jointly using Diversity factor, the neighborhood extending factor and neighborhood restriction factor
It is whole, influence of the control particle to neighborhood, mitigate local optimum phenomenon, reach the optimum balance of convergence rate and optimizing ability;
There is researcher to propose to assign each particle different weights, and particle weights are adaptively adjusted in an iterative process.This
Although a little methods can solve the problem that particle swarm optimization algorithm is easily so that particle is absorbed in local optimum in PSOPF algorithms to some extent
The problem of point, but the deficiency that algorithm is complicated, computationally intensive be present, the real-time of motion target tracking is unsatisfactory.
The content of the invention
The technical problems to be solved by the invention are to overcome existing particle group optimizing particle filter motion target tracking skill
The deficiency not accurate enough, amount of calculation is larger is positioned present in art, there is provided a kind of adaptive particle swarm optimization particle filter motion
Method for tracking target, while the motion target tracking degree of accuracy is accurately improved, its computation complexity is lower, and target following is real-time
Property is more preferable.
The present invention specifically uses following technical scheme:
Adaptive particle swarm optimization particle filter motion target tracking method, moving target is being carried out with particle filter method
Tracking during, the position of particle is optimized using particle group optimizing method;Particle group optimizing method is being utilized to grain
When the position of son optimizes, the quantity of particle is adaptively adjusted according to the change in location situation of global optimum's particle,
It is specific as follows:Position such as global optimum's particle changes all the time in 1 iteration of continuous N, then reduces number of particles;Such as the overall situation most
The position of excellent particle is constant all the time in 2 iteration of continuous N, then increases number of particles;M1, M2 are default to be more than or equal to 3
Integer.
M1 be able to can not also be waited with M2 with equal.
Compared with prior art, the invention has the advantages that:
Adaptive particle swarm optimization (APSO) technology is combined by the present invention with particle filter (PF), and moving target is carried out
Tracking, can effectively overcome particle dilution phenomenon, and further the position of particle is being optimized using particle group optimizing method
When, the quantity of particle is adaptively adjusted according to the change in location situation of global optimum's particle, so as to effectively mitigate grain
Local optimum phenomenon in the optimization of subgroup, the degree of accuracy of target following is improved, at the same it is again low with algorithm complex, and real-time is good
The advantages of.
Brief description of the drawings
Fig. 1 is the basic procedure schematic diagram of particle filter algorithm;
Fig. 2 is the basic procedure schematic diagram of particle swarm optimization algorithm;
Fig. 3 is the schematic flow sheet of adaptive particle swarm optimization particle filter motion target tracking method of the present invention;
Is the inventive method is respectively adopted and conventional particle filtering method carries out motion mesh to Browse1 video sequences in Fig. 4
Mark the Contrast on effect of tracking.
Embodiment
Technical scheme is described in detail below in conjunction with the accompanying drawings:
The thinking of the present invention is the positioning present in existing particle group optimizing particle filter motion target tracking technology
Deficiency not accurate enough, amount of calculation is larger, makes improvements, i.e., the position of particle is being carried out using particle group optimizing method
During optimization, the quantity of particle is adaptively adjusted according to the change in location situation of global optimum's particle:In particle iteration more
In new process, if globe optimum continuous several times all update, illustrate that current particle group is in and continually develop new state
Process, now suitably reduce number of particles;, whereas if globe optimum continuous several times all do not update, now population
In a convergent state, it is possible to be absorbed in local best points and can not jump out, i.e., target location is possible to tracking not
Accurately, at this time need to increase number of particles, so as to help population to jump out this point, the scope of expanded search.
For the ease of public understanding technical scheme, first below to particle filter and grain involved in the present invention
Subgroup optimisation technique is simply introduced.
Fig. 1 shows the basic procedure of particle filter algorithm.Particle filter (PF:Particle Filtering) thought
It is to be based on DSMC, can be with any type of spatial model the problem of using particle collection to represent probability.It
Core concept be to express its distribution by extracting the particle of stochastic regime from posterior probability, be a kind of order importance
The method of sampling.In simple terms, it is exactly that probability density function is carried out by finding one group of random sample propagated in state space
Approximation, the computing of integration is replaced with sample average, so as to obtain the process for making state realize minimum variance distribution.
Generally, the state-space model of particle filter can be described as:
xk=f (xk-1)+uk-1
yk=h (xk)+wk
xkIt is system in the state value at k moment, ykFor system mode xkMeasuring value, uk-1,wkRespectively nonlinear system
Process noise and measure noise figure.
The basic step of particle filter target following is as follows:
1. initialization:Initial tracking frame k, then according to prior distribution p (xk), sample primary collection
2.For k=1,2 ...
A) importance sampling:From suggest profile samples particle collection,
B) importance weights:
Importance weighted formula is:
It is normalized i.e. followed by the weight to particle:
C) resampling:If Neff< NthThen carry out resampling forSo as to obtain
New particle collection:Resampling steps need not be carried out if condition is unsatisfactory for.
3. the estimation of state:There are position and the weight of particle, it is possible to estimate the position of target.
4. it is end frame next to judge, if it is tracking terminates, if not then into next state
Tracking process, goes to step 2.
Fig. 2 illustrates the basic framework of particle swarm optimization algorithm.Particle group optimizing (Particle Swarm
Optimization, PSO) in all particles have the adaptive value that an optimised function determines, each particle has a speed
Degree determines the direction and position that they circle in the air.Then it is close to be all intended to current optimal particle for all particles.PSO passes through initial
Change a group random particles, then update oneself by tracking current particle extreme value and global extremum point come continuous iteration.Specific step
Suddenly it is:Population scale N selected first, i-th of particle is expressed as to the vector x of a N-dimensionali=(xi1,xi2,...,xiN), i=
1,2 ..., m, i.e. position of i-th of the particle in N-dimensional search space is xiCarrying it into object function can, to calculate it suitable
It should be worth, so as to weigh out xiQuality.The speed of i-th of particle is also the vector of a N-dimensional, is designated as vi=(vi1,vi2,...,
viN), speed determines self-care convergent speed in search space.The optimal location that i-th of particle searched so far arrives is pi=
(pi1,pi2,...,piN), the optimum point that whole population searches more to so far is pg=(pg1,pg2,...pgN)。
Come speed and the position of more new particle then according to equation:
vi=w*vi+c1*rand1*(pi-xi)+c2*rand2*(pg-xi)
xi+1=xi+vi
After the position of more new particle, constantly the optimum point of current particle and globe optimum are updated.Iteration again
Need to judge whether the most adaptive value of current goal state meets condition according to newest target state estimator state after terminating, such as
Fruit meets then to jump out iterative process, terminates to find optimal point process, otherwise continue iteration.
Although particle filter algorithm can solve the problems, such as nonlinear and non-Gaussian, the algorithm still has some and asked
Topic.Wherein most importantly needing could be well close to the posterior probability density of system with substantial amounts of sample size, therefore counts
Calculation amount is also just very big, and the resampling stage can cause sample availability and multifarious loss, cause the dilution phenomenon of sample.
It is frequently encountered for particle group optimizing and is absorbed in the problem of local best points can not jump out for multi-peak problem particle,
Therefore globe optimum can not be obtained, and then influences tracking effect.Therefore the present invention combines both, while in population
The change in location situation of global optimum's particle is adaptively adjusted to the quantity of particle in optimization process, so as on the one hand overcome
The problem of conventional particle filtering method is computationally intensive, particle dilution, on the other hand overcomes and is absorbed in caused by local optimum
Tracking result inaccuracy problem.Meanwhile compared to existing various adaptive particle swarm optimization particle filter algorithms, for example, existing
Dynamic particles group's optimized particle filter algorithm that neighborhood adaptively adjusts needs to calculate Diversity factor, the neighborhood extending factor and neighbour
Domain restriction factor is adaptively adjusted to the neighborhood particle quantity of particle, and the inventive method is only needed according in iterative process
The continuous update status of global optimum is the adaptive adjustment that population can be achieved, and its computation complexity is lower, and real-time is more
It is good.
Fig. 3 is the basic procedure schematic diagram of adaptive particle swarm optimization particle filter motion target tracking method of the present invention,
Its specific algorithm flow is as follows:
Video sequence initial frame k frames are read in first, select the target to be tracked, and color is based on to the region foundation of selection
The color histogram feature p (x of informationk), the equation of initialization system tracking, number of particles N, noise R during trackingk, and
The initial position of particle is to obey the Gaussian function distribution centered on the geometric center of the target area initialized.
State shifts:When video sequence moves to the next frame moment, according to the parameter of setting, Particles Moving to a certain position
Put, the colouring information histogram that current particle is then calculated according to the position of i-th of particle isBy calculate its with it is first
Pasteur's distance of the color histogram feature of beginning frame obtains similarity relation between the two, and then obtains the power of current particle
Weight, to needing to carry out weights normalization after N number of particle statistics, and according to weight and position relationshipObtain pre-
The target location of survey.
In the step of advancing to adaptive particle swarm optimization:
The iterations threshold value T of particle group optimizing is jumped out in initializationm.Afterwards according to the speed of particle group optimizing and position side
Journey constantly updates the speed of particle and position, and whether extreme value p is updated here, be according to the function result of Pasteur's distance come
Judge, if similarity is high, corresponding extreme value is just updated.piIt is current particle i individual extreme value, pgFor whole grain
The global extremum of subgroup.Each particle needs to judge that the target location for seeing prediction is similar to initial target again after all updating
Whether degree reaches given threshold, particle group optimizing process can be jumped out if threshold value is reached, if iteration TmIt is secondary all not reach
Arrive, also jump out the process of particle group optimizing.
Two integer M1, M2 more than or equal to 3 are preset, if during continuous iteration updates, global optimum
Point pg1 iteration of continuous N all updates, then illustrates that current particle group is in the process of new state continually developed, then appropriate to subtract
Few number of particles, such as number of particles is subtracted 1;, whereas if 2 iteration of globe optimum continuous N all do not update, now grain
Subgroup is in a convergent state, it is possible to is absorbed in local best points and can not jump out, i.e., target location is possible to track
Inaccuracy, at this time need to increase number of particles, such as number of particles added 1, so as to help population to jump out this point, expand
The scope of search is opened up, after particle estimated state threshold value is met or meets to jump out particle after maximum iteration threshold value
Group's optimization process.Both M1 and M2 value can be set according to actual conditions, and both can be the same or different, in the present embodiment
Value it is identical, be T.
Jump out after adaptive particle swarm optimization:
Need to recalculate particle weight, and the processing that weight is normalized.
The estimation of state:According to the position of particle and weight, the position of target is estimated.And estimation target location is shown
Come, to identify.Next judge to be end frame, if it is tracking terminates, if not then entering next state
Tracking process, go to state transfer step.
As described above as can be seen that in the adaptive particle swarm optimization particle filter algorithm of the present invention, resampling
It is unwanted, is not in that weight concentrates on because particle is all intended to the position of optimum point after particle group optimizing
Above certain several particle.
In order to verify the effect of the present invention, following confirmatory experiment has been carried out:Choose three sections of different video sequences
(Browse1, Fight_RunAway1 derive from http://homepages.inf.ed.ac.uk/rbf/CAVIARDATA1,
Aircraft is conventional aircraft tracking sequence), the particle filter algorithm and 30 particle present invention of 200 particles is respectively adopted
Method carries out target following to it, and is tracked the time statistics (time that i.e. frame of average treatment one is spent) of processing.Institute
Obtained experimental result is as shown in table 1, and wherein PF represents particle filter algorithm, and APSOPF represents the inventive method.
The results contrast of the confirmatory experiment of table 1
Video sequence | PF (200 particles) | APSOPF (30 particles) | Time is saved |
Browse1 | 0.3284s | 0.2120s | 35.44% |
Fight_RunAway1 | 0.3946s | 0.2429s | 38.44% |
Aircraft | 0.3735s | 0.2264s | 39.38% |
Fig. 4 shows position coordinates of the estimated tracking target of two methods in Browse1 video sequences (with picture
Element is unit) contrast between the curve and virtual condition curve that change over time.Figure 4, it is seen that the inventive method
Target following accuracy be substantially better than traditional particle filter method for tracking target.
Claims (5)
1. adaptive particle swarm optimization particle filter motion target tracking method, moving target is being carried out with particle filter method
During tracking, the position of particle is optimized using particle group optimizing method;Characterized in that, utilizing particle group optimizing
When method optimizes to the position of particle, the quantity of particle is carried out according to the change in location situation of global optimum's particle adaptive
It should adjust, it is specific as follows:Position such as global optimum's particle changes all the time in 1 iteration of continuous N, then reduces number of particles;
As global optimum's particle position in 2 iteration of continuous N it is constant all the time, then increase number of particles;M1, M2 are more than to be default
Integer equal to 3.
2. adaptive particle swarm optimization particle filter motion target tracking method as claimed in claim 1, it is characterised in that M1 etc.
In M2.
3. adaptive particle swarm optimization particle filter motion target tracking method as claimed in claim 1, it is characterised in that every time
The number of particles increased or decreased is 1.
4. adaptive particle swarm optimization particle filter motion target tracking method as claimed in claim 1, it is characterised in that with
It is used to be characterized as color histogram during particle filter method carries out the tracking of moving target.
5. adaptive particle swarm optimization particle filter motion target tracking method as claimed in claim 4, it is characterised in that use
The similarity of Pasteur's distance metric color histogram.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510019086.7A CN104574442B (en) | 2015-01-14 | 2015-01-14 | Adaptive particle swarm optimization particle filter motion target tracking method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510019086.7A CN104574442B (en) | 2015-01-14 | 2015-01-14 | Adaptive particle swarm optimization particle filter motion target tracking method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104574442A CN104574442A (en) | 2015-04-29 |
CN104574442B true CN104574442B (en) | 2018-02-09 |
Family
ID=53090410
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510019086.7A Active CN104574442B (en) | 2015-01-14 | 2015-01-14 | Adaptive particle swarm optimization particle filter motion target tracking method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104574442B (en) |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105136469A (en) * | 2015-07-23 | 2015-12-09 | 江苏大学 | Unmanned vehicle speed control method based on PSO and RBF neutral network |
CN105303585A (en) * | 2015-09-29 | 2016-02-03 | 燕山大学 | Rapid target tracking method and device |
CN105427348A (en) * | 2015-12-03 | 2016-03-23 | 山东理工大学 | Video object tracking method based on bat algorithm |
CN106250979B (en) * | 2016-08-16 | 2019-06-14 | 安阳师范学院 | A kind of particle group optimizing method based on mutual information similarity Strategy |
CN106600624B (en) * | 2016-12-06 | 2020-05-08 | 昆山鲲鹏无人机科技有限公司 | Particle filter video target tracking method based on particle swarm |
CN106558064A (en) * | 2016-12-07 | 2017-04-05 | 江南大学 | A kind of method for tracking target based on cellular quantum-behaved particle swarm optimization |
CN107869993A (en) * | 2017-11-07 | 2018-04-03 | 西北工业大学 | Small satellite attitude method of estimation based on adaptive iteration particle filter |
CN109961129B (en) * | 2017-12-25 | 2021-02-09 | 中国科学院沈阳自动化研究所 | Improved particle swarm-based marine static target searching scheme generation method |
CN110320472B (en) * | 2019-05-17 | 2021-06-01 | 枣庄学院 | Self-correction SOC estimation method for mining lithium battery |
JP2022032776A (en) * | 2020-08-14 | 2022-02-25 | 富士通株式会社 | Image processing device and screen processing program |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101576880A (en) * | 2008-05-06 | 2009-11-11 | 山东省标准化研究院 | Database query optimization method based on extremum optimization |
CN103942612A (en) * | 2014-04-08 | 2014-07-23 | 河海大学 | Cascade reservoir optimal operation method based on adaptive particle swarm optimization algorithm |
CN104023394A (en) * | 2014-06-17 | 2014-09-03 | 河海大学 | WSN positioning method based on self-adaptation inertia weight |
CN104036525A (en) * | 2014-06-18 | 2014-09-10 | 上海交通大学 | Self-adaptive particle swarm optimization (PSO)-based particle filter tracking method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8112369B2 (en) * | 2009-02-11 | 2012-02-07 | The United States Of America As Represented By The Secretary Of The Navy | Methods and systems of adaptive coalition of cognitive agents |
-
2015
- 2015-01-14 CN CN201510019086.7A patent/CN104574442B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101576880A (en) * | 2008-05-06 | 2009-11-11 | 山东省标准化研究院 | Database query optimization method based on extremum optimization |
CN103942612A (en) * | 2014-04-08 | 2014-07-23 | 河海大学 | Cascade reservoir optimal operation method based on adaptive particle swarm optimization algorithm |
CN104023394A (en) * | 2014-06-17 | 2014-09-03 | 河海大学 | WSN positioning method based on self-adaptation inertia weight |
CN104036525A (en) * | 2014-06-18 | 2014-09-10 | 上海交通大学 | Self-adaptive particle swarm optimization (PSO)-based particle filter tracking method |
Non-Patent Citations (3)
Title |
---|
Dynamic Population Size in PSO-based Multiobjective Optimization;Wen-Fung Leong 等;《IEEE Congress on Evolutionary Computation》;20060911;第1718-1725页 * |
基于自适应粒子群优化的新型粒子滤波在目标跟踪中的应用;陈志敏 等;《控制与决策》;20130228;第28卷(第2期);第193-200页 * |
整车物流网络规划问题的混合粒子群算法研究;秦绪伟 等;《系统工程理论与实践》;20060730;正文第4节 * |
Also Published As
Publication number | Publication date |
---|---|
CN104574442A (en) | 2015-04-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104574442B (en) | Adaptive particle swarm optimization particle filter motion target tracking method | |
CN109816689B (en) | Moving target tracking method based on adaptive fusion of multilayer convolution characteristics | |
CN105335986B (en) | Method for tracking target based on characteristic matching and MeanShift algorithm | |
CN107424177B (en) | Positioning correction long-range tracking method based on continuous correlation filter | |
CN110033473B (en) | Moving target tracking method based on template matching and depth classification network | |
CN106683118B (en) | Unmanned aerial vehicle target tracking method based on hierarchical model | |
CN107633226B (en) | Human body motion tracking feature processing method | |
CN109086663B (en) | Natural scene text detection method based on scale self-adaption of convolutional neural network | |
CN111739053B (en) | Online multi-pedestrian detection tracking method under complex scene | |
CN108154159A (en) | A kind of method for tracking target with automatic recovery ability based on Multistage Detector | |
CN109448023A (en) | A kind of satellite video Small object method for real time tracking of combination space confidence map and track estimation | |
CN110717934A (en) | Anti-occlusion target tracking method based on STRCF | |
Guo | Moving target localization in sports image sequence based on optimized particle filter hybrid tracking algorithm | |
CN112184762A (en) | Gray wolf optimization particle filter target tracking algorithm based on feature fusion | |
CN112164093A (en) | Automatic person tracking method based on edge features and related filtering | |
CN112991394B (en) | KCF target tracking method based on cubic spline interpolation and Markov chain | |
CN108280845B (en) | Scale self-adaptive target tracking method for complex background | |
CN112884799A (en) | Target tracking method in complex scene based on twin neural network | |
CN109448024A (en) | Visual tracking method, the system of constraint correlation filter are constructed using depth data | |
CN111242972B (en) | On-line cross-scale multi-fluid target matching tracking method | |
CN110717930B (en) | Mutation moving target tracking method based on extended Wang-Landau Monte Carlo and KCF | |
CN116630376A (en) | Unmanned aerial vehicle multi-target tracking method based on ByteTrack | |
CN111241965A (en) | Target tracking method for occlusion detection based on multi-example learning | |
Yu | Moving target tracking based on improved Meanshift and Kalman filter algorithm | |
CN108009272B (en) | Low-altitude moving target searching method based on directed weighted graph |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |