CN108320300A - A kind of space-time context visual tracking method of fusion particle filter - Google Patents
A kind of space-time context visual tracking method of fusion particle filter Download PDFInfo
- Publication number
- CN108320300A CN108320300A CN201810002288.4A CN201810002288A CN108320300A CN 108320300 A CN108320300 A CN 108320300A CN 201810002288 A CN201810002288 A CN 201810002288A CN 108320300 A CN108320300 A CN 108320300A
- Authority
- CN
- China
- Prior art keywords
- target
- particle
- indicate
- weights
- particle filter
- 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
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
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30241—Trajectory
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
A kind of space-time context visual tracking method of fusion particle filter is claimed in the present invention.First, the first frame image information is initialized, and by the way that experiment parameter is arranged, automatically selects the rectangular area where first frame target；Secondly, using Bhattacharyya coefficients as the foundation for judging whether target blocks；Finally, when target, which is in, blocks, target is estimated and is predicted in the position of subsequent image frames and movement locus by introducing particle filter algorithm, realizes the accurate tracking of target.The present invention can be applied not only to the visual target tracking under the complex backgrounds such as illumination variation, target rotation, background area interference, and blocking with robustness to target, meet requirement of real-time.
Description
Technical field
The invention belongs to the space-time context vision tracking of technical field of image processing, more particularly to fusion particle filter
Method.
Background technology
Vision tracking is one of hot spots of area researches such as computer vision, image procossing and pattern-recognition, is commonly used to
The fields such as video monitoring, man-machine interface and imaging of medical.In recent years, Visual Tracking causes the wide of academia and industrial quarters
General concern, domestic and international researcher pass through unremitting effort, it is proposed that including tracking-study-detection (Tracking-Learning-
Detection, TLD), compression tracking (Compressive Tracking, CT), core correlation filtering (Kernelized
Correlation Filter, KCF) and space-time context (Spatio-Temporal Context, STC) etc. it is a variety of outstanding
Vision Tracking.However, during actual tracking, one side target is there may be deformation or is rotated, separately
On the one hand due to the influence of the external environments such as illumination or shelter, cause above-mentioned algorithm that cannot accurately realize target following.Cause
This, it is most important to design effective Vision Tracking.
STC Vision Trackings establish the time-space relationship model of target and its context area, by calculating confidence map
The target location of a new frame is obtained, since the study of space-time model and the detection of target are all to pass through Fast Fourier Transform (FFT)
(Fast Fourier Transform, FFT) is realized, therefore accelerates arithmetic speed.On the basis of STC algorithms, current
Improvement has：(1) dynamic zoning method processing target context area is used, while original algorithm is improved to reach according to weights are assigned
To the purpose of more preferable tracking effect；(2) picture relatively rough in STC algorithms is replaced using the middle level feature of weighting super-pixel
It the low-level image feature information of plain grade gray scale and is fixed with the parameter for solving space-time context model using auto-adaptive parameter update method
Problem.Although the above method can effectively track target to a certain extent, in a practical situation, when target is in complex background
It is middle when blocking, target following can be made to fail.
For the target occlusion problem in processing STC algorithms, there are some researchers to consider and utilize Kalman filtering algorithm
It solves, but seldom research particle filter algorithm is in solving STC algorithms target the video frequency object tracking during and blocks
When the tracking drifting problem that generates.
Invention content
It is in and blocks and complex background holds in target mainly for STC algorithms the object of the present invention is to provide one kind
A kind of the problem of being also easy to produce tracking drift, it is proposed that space-time context visual tracking method of fusion particle filter.This method is not
The visual target tracking that can be only suitable under complex background, and blocking with robustness to target, meet real-time and want
It asks.Technical scheme is as follows：
A kind of space-time context visual tracking method of fusion particle filter comprising following steps：
Step 1：Target following video is obtained, initialized target tracks the first frame image information of video, and passes through setting
Parameter alpha, β in experiment parameter, including target confidence map function, β indicate form parameter, α indicate scale parameter, learning rate because
The variance parameter σ of sub- ρ and the particle weights in particle filter algorithm automatically select the square where target in the first frame image information
Shape region；
Step 2：Rectangular area where in the first frame target image information obtained to step 1 utilizes
Bhattacharyya Pasteur's coefficients judge whether target blocks；
Step 3：When target, which is in, blocks, by introducing improved particle filter algorithm to target in subsequent image frames
Position and movement locus are estimated and are predicted, realize that target is in accurate tracking when blocking, the improvement particle filter is calculated
Method is mainly improved：Importance sampling density function pair t+1 frame particle weights are introduced in the weight update of particle to carry out more
Newly.
Further, parameter alpha=2.25, β=1 in the step 1 in target confidence map function, learning rate factor ρ
Variance parameter σ=2 of particle weights in as the parameter value in STC algorithms also 0.075, particle filter algorithm.
Further, using Bhattacharyya coefficients as the foundation for judging whether target blocks in the step 2,
It specifically includes：
The color histogram of known target and context area is defined as：
Wherein, q is color histogram, and c is normalization coefficient, and y indicates the center pixel of target, xiIndicate ith pixel,
I indicates that a variable, n indicate that the quantity of pixel, ω () are gaussian weighing function, and δ () is Dirac function, b (xi) be
Pixel xiColor histogram mapping function, μ be colored pixels probability characteristics, miIndicate the probability of i-th of target area pixel,
At this point, the context area pixel value around target is defined as：
Ii=miOi+(1-mi)Li
Wherein, OiFor target image, LiFor local context area image；
Bhattacharyya coefficients are defined as：
Wherein, qi(μ) is goal histogram, qi *(μ) is tracing area histogram.
Further, the step 3 by introduce improved particle filter algorithm to target in the position of subsequent image frames
And movement locus is estimated and is predicted, realizes target in accurate tracking when blocking, specifically includes：
(1) it initializes, if the particle assembly in t moment stochastical sampling tracking target and surrounding context region is：Then corresponding weights areM indicates the quantity of particle,Indicate i-th of particle in t
The weights at moment,It indicates i-th of particle of stochastical sampling, and meets
(2) status predication of target；Appoint at interval of time t and take two neighboring point in m sampled point, calculates between them
Vector be：yiIndicate the ordinate of i-th of particle, xiIndicate the abscissa of i-th of particle.
Ri=(dxi,dyi)=(xi-xi-1,yi-yi-1)
According to above formula, the moving direction of t moment is predicted, and t+1 moment status predication vectors are：
Above formula is substituted into following formula status predication equation, calculates state-noise covariance matrix Qt；
Wherein, Xt+1For status predication vector, A is state-transition matrix, WtFor system white Gaussian noise vector and Wt∈N
(0,Qt)；
(3) the weight update of particle；The similitude expression formula of particle model and object module is：
Wherein, ρi *Indicate Bhattacharyya coefficients,For the characteristic information of particle, qμFor target signature information；Pass through
Distance between the two indicates that its similar degree, expression formula are：
Then the calculation formula of t frames particle weights is：
Wherein, σ indicates the variance of particle weights；
It introduces importance sampling density function pair t+1 frame particle weights at this time to be updated, expression formula is：
Wherein,Indicate likelihood function,Indicate probability transition density function, Zt+1Table
Show the target observation variable at t+1 moment,Indicate state variable of i-th of particle at the t+1 moment,
Indicate importance sampling density function；
(4) particle weights are normalized to：
Wherein, i=1,2 ... M；Indicate j-th of particle the t+1 moment weighted value；
(5) resampling of particle；Before the resampling of particle starts, thresholding sample points are first set as Nth, effective grain
Subnumber is Neff, the expression formula of number of effective particles is：
Work as Neff＜ NthWhen, carry out the resampling of particle, the particle that right of retention is greater than the set value again is given up weight and is less than and sets
The particle of definite value, and obtained all particle weights are set as equal；
(6) state estimation of target；The expression formula such as following formula of t+1 moment observation vectors, calculating observation noise covariance square
Battle array Rt+1；
Yt+1=BXt+1+Vt+1
Wherein, Yt+1For observation vector, B is observed differential matrix, Vt+1For process white Gaussian noise vector and Vt+1∈N(0,
Rt+1)；
By above step, the center expression formula for obtaining final goal estimation is：
Above formula is iterated into space-time context model, realizes target accurately tracking when complex background blocks.
It advantages of the present invention and has the beneficial effect that：
What the present invention was studied in the case where target is in complex background and circumstance of occlusion, target in STC algorithms can be solved
A kind of drifting problem in video tracking, it is proposed that space-time context visual tracking method of fusion particle filter.This method
Innovation be Bhattacharyya coefficients are added to carry out target in the target local context region of STC algorithms blocking
Differentiate, when judging that target is not blocked, target can effectively be tracked by directly executing STC algorithms.When judging target
When blocking, if still directly executing STC algorithms, tracking target can drift about, but be moved to target by introducing particle filter
Track carries out predictive estimation to reach real-time, accurate, robust tracking target, can solve STC algorithms when target is in and blocks
Tracking drifting problem.
Description of the drawings
Fig. 1 is the flow diagram that the present invention provides preferred embodiment.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, detailed
Carefully describe.Described embodiment is only a part of the embodiment of the present invention.
The present invention solve above-mentioned technical problem technical solution be：
The invention discloses a kind of space-time context visual tracking methods of fusion particle filter, as shown in Figure 1, specifically
Steps are as follows：
Step 1：In video frequency object tracking, the first frame image information is initialized, and by the way that experiment parameter is arranged, automatically
The rectangular area where first frame target is selected, is specifically included：
Parameter alpha=2.25, β=1 in target confidence map function, the parameter value one in learning rate factor ρ and STC algorithm
Sample also 0.075.Variance parameter σ=2 of particle weights in particle filter algorithm.
Step 2：Using Bhattacharyya coefficients as the foundation for judging whether target blocks, specifically include：
The color histogram of known target and context area is defined as：
Wherein, q is color histogram, and c is normalization coefficient, and ω () is gaussian weighing function, and δ () is dirac letter
Number, b (xi) it is pixel xiColor histogram mapping function, μ be colored pixels probability characteristics, miIndicate i-th of target area picture
The probability of element.At this point, the context area pixel value around target is defined as：
Ii=miOi+(1-mi)Li
Wherein, OiFor target image, LiFor local context area image.
Bhattacharyya coefficients are defined as：
Wherein, qi(μ) is goal histogram, qi *(μ) is tracing area histogram.
Step 3：Vision Tracking：When target, which is in, blocks, by introducing particle filter algorithm to target follow-up
The position of picture frame and movement locus are estimated and are predicted, realize the accurate tracking of target, specifically include：
(1) it initializes.If t moment stochastical sampling tracks target and the particle assembly in surrounding context region is：Then corresponding weights areAnd meet
(2) status predication of target.Appoint at interval of time t and take two neighboring point in m sampled point, calculates between them
Vector be：
Ri=(dxi,dyi)=(xi-xi-1,yi-yi-1)
According to above formula, the moving direction of t moment can be predicted, and t+1 moment status predication vectors are：
Above formula is substituted into following formula status predication equation, calculates state-noise covariance matrix Qt。
Wherein, Xt+1For status predication vector, A is state-transition matrix, WtFor system white Gaussian noise vector and Wt∈N
(0,Qt)。
(3) the weight update of particle.The similitude expression formula of particle model and object module is：
Wherein, ρi *Indicate Bhattacharyya coefficients,For the characteristic information of particle, qμFor target signature information.It can be with
Indicate that its similar degree, expression formula are by distance between the two：
Then the calculation formula of t frames particle weights is：
Wherein, σ indicates the variance of particle weights.
It introduces importance sampling density function pair t+1 frame particle weights at this time to be updated, expression formula is：
Wherein,Indicate likelihood function,Indicate probability transition density function,Indicate importance sampling density function.
(4) normalization of particle weights.At the moment, it is to particle weights normalization result：
Wherein, i=1,2 ... M.
(5) resampling of particle.Before the resampling of particle starts, thresholding sample points are first set as Nth, effective grain
Subnumber is Neff, the expression formula of number of effective particles is：
Work as Neff＜ NthWhen, the resampling of particle is carried out, retains the larger particle of weight, gives up the smaller particle of weight, and
Obtained all particle weights are set as equal.
(6) state estimation of target.The expression formula such as following formula of t+1 moment observation vectors, calculating observation noise covariance square
Battle array Rt+1。
Yt+1=BXt+1+Vt+1
Wherein, Yt+1For observation vector, B is observed differential matrix, Vt+1For process white Gaussian noise vector and Vt+1∈N(0,
Rt+1)。
By above step, the center expression formula for obtaining final goal estimation is：
Above formula is iterated into space-time context model, realizes target accurately tracking when complex background blocks.
The above embodiment is interpreted as being merely to illustrate the present invention rather than limit the scope of the invention.
After the content for having read the record of the present invention, technical staff can make various changes or modifications the present invention, these equivalent changes
Change and modification equally falls into the scope of the claims in the present invention.
Claims (4)
1. a kind of space-time context visual tracking method of fusion particle filter, which is characterized in that include the following steps：
Step 1：Target following video is obtained, initialized target tracks the first frame image information of video, and is tested by being arranged
Parameter alpha, β in parameter, including target confidence map function, β indicate that form parameter, α indicate scale parameter, learning rate factor ρ
With the variance parameter σ of the particle weights in particle filter algorithm, the rectangle where target in the first frame image information is automatically selected
Region；
Step 2：Rectangular area where in the first frame target image information obtained to step 1, utilizes Bhattacharyya
Pasteur's coefficient judges whether target blocks；
Step 3：When target, which is in, blocks, by introducing improved particle filter algorithm to target in the position of subsequent image frames
And movement locus is estimated and is predicted, realizes that target is in accurate tracking when blocking, the improved particle filter algorithm master
It improves：Importance sampling density function pair t+1 frame particle weights are introduced in the weight update of particle to be updated.
2. the space-time context visual tracking method of fusion particle filter according to claim 1, which is characterized in that described
Parameter alpha=2.25, β=1 in step 1 in target confidence map function, the parameter value one in learning rate factor ρ and STC algorithm
Sample also 0.075, variance parameter σ=2 of the particle weights in particle filter algorithm.
3. the space-time context visual tracking method of fusion particle filter according to claim 1, which is characterized in that described
Using Bhattacharyya coefficients as the foundation for judging whether target blocks in step 2, specifically include：
The color histogram of known target and context area is defined as：
Wherein, q is color histogram, and c is normalization coefficient, and y indicates the center pixel of target, xiIndicate that ith pixel, i indicate
One variable, n indicate that the quantity of pixel, ω () are gaussian weighing function, and δ () is Dirac function, b (xi) it is pixel xi
Color histogram mapping function, μ be colored pixels probability characteristics, miIndicate the probability of i-th of target area pixel, at this point,
Context area pixel value around target is defined as：
Ii=miOi+(1-mi)Li
Wherein, OiFor target image, LiFor local context area image；
Bhattacharyya coefficients are defined as：
Wherein, qi(μ) is goal histogram, qi *(μ) is tracing area histogram.
4. the space-time context visual tracking method of fusion particle filter according to claim 3, which is characterized in that described
Step 3 by introduce improved particle filter algorithm to target the position of subsequent image frames and movement locus carry out estimation and
Prediction realizes target in accurate tracking when blocking, specifically includes：
(1) it initializes, if the particle assembly in t moment stochastical sampling tracking target and surrounding context region is：
Then corresponding weights areM indicates the quantity of particle,Indicate i-th of particle t moment weights,It indicates i-th of particle of stochastical sampling, and meets
(2) status predication of target；Appoint at interval of time t and take two neighboring point in m sampled point, calculate between them to
Amount is：yiIndicate the ordinate of i-th of particle, xiIndicate the abscissa of i-th of particle；
Ri=(dxi,dyi)=(xi-xi-1,yi-yi-1)
According to above formula, the moving direction of t moment is predicted, and t+1 moment status predication vectors are：
Above formula is substituted into following formula status predication equation, calculates state-noise covariance matrix Qt；
Wherein, Xt+1For status predication vector, A is state-transition matrix, WtFor system white Gaussian noise vector and Wt∈N(0,
Qt)；
(3) the weight update of particle；The similitude expression formula of particle model and object module is：
Wherein, ρi *Indicate Bhattacharyya coefficients,For the characteristic information of particle, qμFor target signature information；Both pass through
The distance between indicate its similar degree, expression formula is：
Then the calculation formula of t frames particle weights is：
Wherein, σ indicates the variance of particle weights；
It introduces importance sampling density function pair t+1 frame particle weights at this time to be updated, expression formula is：
Wherein,Indicate likelihood function,Indicate probability transition density function, Zt+1Indicate t+1
The target observation variable at moment,Indicate state variable of i-th of particle at the t+1 moment,It indicates
Importance sampling density function；
(4) particle weights are normalized to：
Wherein, i=1,2 ... M；Indicate j-th of particle the t+1 moment weighted value；
(5) resampling of particle；Before the resampling of particle starts, thresholding sample points are first set as Nth, number of effective particles
For Neff, the expression formula of number of effective particles is：
Work as Neff＜ NthWhen, the resampling of particle is carried out, the particle that right of retention is greater than the set value again gives up weight and is less than setting value
Particle, and obtained all particle weights are set as equal；
(6) state estimation of target；The expression formula such as following formula of t+1 moment observation vectors, calculating observation noise covariance matrix
Rt+1；
Yt+1=BXt+1+Vt+1
Wherein, Yt+1For observation vector, B is observed differential matrix, Vt+1For process white Gaussian noise vector and Vt+1∈N(0,
Rt+1)；
By above step, the center expression formula for obtaining final goal estimation is：
Above formula is iterated into space-time context model, realizes target accurately tracking when complex background blocks.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810002288.4A CN108320300A (en) | 2018-01-02 | 2018-01-02 | A kind of space-time context visual tracking method of fusion particle filter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810002288.4A CN108320300A (en) | 2018-01-02 | 2018-01-02 | A kind of space-time context visual tracking method of fusion particle filter |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108320300A true CN108320300A (en) | 2018-07-24 |
Family
ID=62892945
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810002288.4A Pending CN108320300A (en) | 2018-01-02 | 2018-01-02 | A kind of space-time context visual tracking method of fusion particle filter |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108320300A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109308457A (en) * | 2018-08-31 | 2019-02-05 | 华南理工大学 | Multiparticle three-dimensional method for tracing under a kind of high concentration |
| CN110135314A (en) * | 2019-05-07 | 2019-08-16 | 电子科技大学 | A kind of multi-object tracking method based on depth Trajectory prediction |
| CN110244322A (en) * | 2019-06-28 | 2019-09-17 | 东南大学 | Pavement construction robot environment sensory perceptual system and method based on Multiple Source Sensor |
| CN110363113A (en) * | 2019-06-28 | 2019-10-22 | 华南理工大学 | VLC-ITS information source based on particle filter tracks extracting method |
| CN110580711A (en) * | 2019-08-23 | 2019-12-17 | 天津大学 | video tracking method adopting particle filtering |
| CN110910416A (en) * | 2019-11-20 | 2020-03-24 | 河北科技大学 | Moving obstacle tracking method and device and terminal equipment |
| CN111862157A (en) * | 2020-07-20 | 2020-10-30 | 重庆大学 | Multi-vehicle target tracking method integrating machine vision and millimeter wave radar |
| CN112053385A (en) * | 2020-08-28 | 2020-12-08 | 西安电子科技大学 | Remote sensing video shielding target tracking method based on deep reinforcement learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105631895A (en) * | 2015-12-18 | 2016-06-01 | 重庆大学 | Temporal-spatial context video target tracking method combining particle filtering |
| CN107038431A (en) * | 2017-05-09 | 2017-08-11 | 西北工业大学 | Video target tracking method of taking photo by plane based on local sparse and spatio-temporal context information |
-
2018
- 2018-01-02 CN CN201810002288.4A patent/CN108320300A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105631895A (en) * | 2015-12-18 | 2016-06-01 | 重庆大学 | Temporal-spatial context video target tracking method combining particle filtering |
| CN107038431A (en) * | 2017-05-09 | 2017-08-11 | 西北工业大学 | Video target tracking method of taking photo by plane based on local sparse and spatio-temporal context information |
Non-Patent Citations (7)
| Title |
|---|
| KAIHUA ZHANG等: "Fast Tracking via Spatio-Temporal Context Learning", 《ARXIV:1311.1939V1 [CS.CV]》 * |
| TING LIU等: "Real-time part-based visual tracking via adaptive correlation filters", 《2015 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION(CVPR)》 * |
| WU WEN等: "Tracking Algorithm of Improved Spatio-Temporal Context with Particle Filter", 《2017 3RD IEEE INTERNATIONAL CONFERENCE ON COMPUTER AND COMMUNICATIONS》 * |
| ZIYANG WANG等: "Particle Filter Based on Context Tracking Algorithm for Real-world Hazy Scenes", 《2017 IEEE 2ND ADVANCED INFORMATION TECHNOLOGY, ELECTRONIC AND AUTOMATION CONTROL CONFERENCE》 * |
| 刘维亭等: "基于重要性重采样粒子滤波器的机动目标跟踪方法", 《江苏科技大学学报(自然科学版)》 * |
| 夏瑜等: "一种抗遮挡自适应粒子滤波目标跟踪方法", 《光电子·激光》 * |
| 贺永峰: "粒子滤波算法在视频跟踪中的应用研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109308457A (en) * | 2018-08-31 | 2019-02-05 | 华南理工大学 | Multiparticle three-dimensional method for tracing under a kind of high concentration |
| CN109308457B (en) * | 2018-08-31 | 2022-03-29 | 华南理工大学 | Multi-particle three-dimensional tracking method under high concentration |
| CN110135314A (en) * | 2019-05-07 | 2019-08-16 | 电子科技大学 | A kind of multi-object tracking method based on depth Trajectory prediction |
| CN110244322A (en) * | 2019-06-28 | 2019-09-17 | 东南大学 | Pavement construction robot environment sensory perceptual system and method based on Multiple Source Sensor |
| CN110363113A (en) * | 2019-06-28 | 2019-10-22 | 华南理工大学 | VLC-ITS information source based on particle filter tracks extracting method |
| CN110363113B (en) * | 2019-06-28 | 2021-09-21 | 华南理工大学 | VLC-ITS information source tracking extraction method based on particle filtering |
| CN110580711A (en) * | 2019-08-23 | 2019-12-17 | 天津大学 | video tracking method adopting particle filtering |
| CN110910416A (en) * | 2019-11-20 | 2020-03-24 | 河北科技大学 | Moving obstacle tracking method and device and terminal equipment |
| CN111862157A (en) * | 2020-07-20 | 2020-10-30 | 重庆大学 | Multi-vehicle target tracking method integrating machine vision and millimeter wave radar |
| CN111862157B (en) * | 2020-07-20 | 2023-10-10 | 重庆大学 | Multi-vehicle target tracking method integrating machine vision and millimeter wave radar |
| CN112053385A (en) * | 2020-08-28 | 2020-12-08 | 西安电子科技大学 | Remote sensing video shielding target tracking method based on deep reinforcement learning |
| CN112053385B (en) * | 2020-08-28 | 2023-06-02 | 西安电子科技大学 | Remote sensing video shielding target tracking method based on deep reinforcement learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108320300A (en) | A kind of space-time context visual tracking method of fusion particle filter | |
| CN109344725B (en) | Multi-pedestrian online tracking method based on space-time attention mechanism | |
| CN110009665B (en) | Target detection tracking method in shielding environment | |
| CN107452015B (en) | Target tracking system with re-detection mechanism | |
| CN103886325B (en) | Cyclic matrix video tracking method with partition | |
| CN105469054B (en) | The model building method of normal behaviour and the detection method of abnormal behaviour | |
| CN108521554A (en) | Large scene multi-target cooperative tracking method, intelligent monitor system, traffic system | |
| CN109359549A (en) | A kind of pedestrian detection method based on mixed Gaussian and HOG_LBP | |
| CN104036526A (en) | Gray target tracking method based on self-adaptive window | |
| CN110717934A (en) | Anti-occlusion target tracking method based on STRCF | |
| Qi et al. | Small infrared target detection utilizing local region similarity difference map | |
| CN102509414B (en) | Smog detection method based on computer vision | |
| CN117011381A (en) | Real-time surgical instrument pose estimation method and system based on deep learning and stereoscopic vision | |
| CN109102520A (en) | The moving target detecting method combined based on fuzzy means clustering with Kalman filter tracking | |
| CN106485283B (en) | A kind of particle filter pedestrian target tracking based on Online Boosting | |
| Wang et al. | A real-time active pedestrian tracking system inspired by the human visual system | |
| CN112733770A (en) | Regional intrusion monitoring method and device | |
| CN117252908A (en) | Anti-occlusion multi-target tracking method based on attention | |
| CN108985216B (en) | Pedestrian head detection method based on multivariate logistic regression feature fusion | |
| Ren et al. | A multiCell visual tracking algorithm using multi-task particle swarm optimization for low-contrast image sequences | |
| CN116193103A (en) | Video picture jitter level assessment method | |
| CN114429489A (en) | Mobile robot target occlusion tracking method based on multi-core correlation filtering fusion | |
| CN110322474B (en) | Image moving target real-time detection method based on unmanned aerial vehicle platform | |
| Zhou et al. | Real-time detection and spatial segmentation of difference image motion changes | |
| JPH07320068A (en) | Method and processor for image processing |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180724 |