CN106372650A - Motion prediction-based compression tracking method - Google Patents
Motion prediction-based compression tracking method Download PDFInfo
- Publication number
- CN106372650A CN106372650A CN201610701199.XA CN201610701199A CN106372650A CN 106372650 A CN106372650 A CN 106372650A CN 201610701199 A CN201610701199 A CN 201610701199A CN 106372650 A CN106372650 A CN 106372650A
- Authority
- CN
- China
- Prior art keywords
- target
- motion
- tracking
- sample
- motion prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
Abstract
The invention discloses a motion prediction-based compression tracking method. The method comprises the steps of performing motion prediction on a video target according to a current frame to obtain a motion direction of the target; calculating a distance of target motion of first two frames according to a motion vector, automatically adjusting a search range according to the distance, and reducing acquisition of candidate samples; and performing optimization by using an adaptive tracking window, acquiring positive and negative sample sets again, extracting features of the two sample sets, performing update of a Naive Bayes classifier, recording a target position of the current frame, tracking the target position, and updating parameters. The method has the beneficial effects that the search time is greatly shortened, and the acquisition of the candidate samples is reduced, so that the real-time property and robustness of the compression tracking method in certain complex scenes are improved.
Description
Technical field
The invention belongs to technical field of computer vision, more particularly, to a kind of compression tracking based on motion prediction.
Background technology
Developing rapidly with electronic computer technology, computer vision becomes popular research topic.Intelligent video is supervised
Control has gradually infiltrated into daily life, automatically analyzes using video sequence image to detect, follow the tracks of and to identify monitoring field
Target in scape, so analyze judge target and make countermeasure.And video frequency object tracking is the crucial portion in intelligent monitor system
Point, merge multi-field, the multidisciplinary problem such as image procossing, pattern recognition, signal processing and control.Because tracking is subject to
The impact of several factors, is particularly due to the type change of target, the change of illumination, the problems such as blocking of object, therefore, sets up one
Robust, adaptive method for tracking target remain a challenging problem.
In recent years, finding a kind of efficient and robust tracking enjoys research worker to pay close attention to.2012, zhang et al.
[document 1] (zhang k h, zhang l, yang m h.real-time compressive tracking [c] .european
Conference on computer vision, 2012:864-877) the compression track algorithm (compressive that proposes
Tracking, ct), algorithm passes through experiment and employs optimal experimental configuration, and the method to next frame image procossing is according to front
Around one frame target upper left angle point, the rectangle frame of 20 Euclidean distance radiuses is all as candidate region, to each extracted region 50
Haar-like eigenvalue.Finally using Naive Bayes Classifier, the eigenvalue of these candidate regions is screened, select
The region of macrotaxonomy device number of responses, the as target area of present frame.When determining target area, using this extensive
Search strategy is to compare the waste calculating time.Therefore, motion prediction is incorporated in track algorithm, with prediction direction, to this
Chosen on a large scale on individual direction, to other direction minimizing candidate region numbers.
Conventional motion forecast method has a lot, such as luo et al. [document 2] (luo h l, zhong b k, kong f
s.object tracking algorithm by combining the predicted target position with
Compressive tracking [j] .journal of image and graphics, 2014,19 (6): 875-885.) carry
Go out carries out target motion prediction using mean shift, and affects tracing positional by the position of prediction, improves tracking accuracy.
Yang Dongdong et al. [document 3] (Yang Dongdong, Chang Danhua, Han Xia, etc. the improvement of moving object detection and tracking algorithm
With realize [j]. laser with infrared, 2010,40 (2): 205-209) prediction moved using motion history image, raising
Tracking accuracy.
2016, zhang et al. [document 4] (zhang k h, liu q s, wu y, et al.robust visual
tracking via convolutional networks without training[j].ieee transactions on
Image processing, 2016,25 (4): 1779-1792) the convolutional neural networks algorithm (convolutional proposing
Networks without training, cnt), tracking performance is obviously improved.Using these algorithms during tracking, can
To improve the performance of target following.But, these method comparison are complicated, and computational complexity is higher it is impossible to preferably satisfaction is real
Border demand.
Content of the invention
The present invention is to overcome the deficiencies in the prior art, there is provided a kind of complexity is low, robustness is high based on motion prediction
Compression tracking, specifically realized by technical scheme below:
The described compression tracking based on motion prediction, comprising:
Initialization, chooses in the first two field picture and follows the tracks of target area, in the first two field picture close-proximity target zone sampled images
Block, carries out feature extraction to described image block and dimensionality reduction obtains the characteristic vector of each image block, sets up grader；
Follow the tracks of target and target is carried out with motion prediction: for t+1 two field picture, t is the integer more than or equal to 2, use
Target location in front cross frame image obtains target motion vectors and prediction direction, and according to described target motion vectors and prediction
Direction draws the predicted position of target, sampled images block near the predicted position that t frame traces into, and then obtains each figure
As the characteristic vector of block, the target that corresponding for maximum grader image block is traced into by described grader as present frame；
Determine the final tracking result of present frame: adaptive optimization is carried out to tracking window, in described target proximity, again
Sampled, and selected by grader, corresponding for maximum grader image block is followed the tracks of target as final；And record
Present frame target location, traces into target location, updates classifier parameters.
The design further of the described compression tracking based on motion prediction is, described grader is divided using Bayes
Class device h (v), such as formula (1),
Wherein, p (y=1)=p (y=0), y ∈ { 0,1 } is positive negative sample mark, and y=0 represents negative sample, and y=1 generation
Table positive sample, low dimensional space v=(v1,....,vn)t,viFor i-th element in v.
The design further of the described compression tracking based on motion prediction is, sets the condition distribution p in grader h (v)
(vi| y=1) and p (vi| y=0) it is Gauss distribution, and meet
WhereinWithIt is respectively expectation and the standard deviation of positive sample probability, andWithIt is respectively expectation and the standard of negative sample probability
Difference,For being desired forWith standard deviation it isGauss distribution,For being desired forWith standard deviation it is's
Gauss distribution, n represents the symbol of Gauss distribution, and Bayes classifier parameter updates as formula (2), formula (3):
Wherein, μ1、σ1Represent the expectation of positive sample and standard deviation after updating, λ ＞ 0 is Studying factors, n representative sample number, viK () represents kth sample and represents in lower dimensional space, k
Represent k-th sample.
The design further of the described compression tracking based on motion prediction is, the sampling of described image block is by excellent
The positive and negative samples selection changed is realized gathering, positive sample selection gist success rate formula, success rate formula such as formula (4):
Wherein, ci is the target frame region that algorithm calculates, and ri is realistic objective region；When the value of rt in a frame is more than
0.5 then it is assumed that tracking result is correct.
The design further of the described compression tracking based on motion prediction is, the optimization of positive and negative samples selection, according to
According to described success rate formula, the optimization formula such as formula (5) of positive and negative samples selection:
Wherein, w, h are width and the height of target window, and l is side-play amount, n and m is horizontal and vertical offset component,
Positive sample selection region is region more than 0.9 for the success rate, and negative sample is success rate in 0.5 area below.
Described based on motion prediction compression tracking design further be, described motion prediction specifically include as
Lower step:
Step 1) set former frame target location as (x1,y1), present frame target location is (x2,y2), then motion vector β
For:
β=(x2,y2)-(x1,y1)；
Step 2) according to target direction of motion, definition of search region, define a matrix a, described matrix a size and search
Range size is identical, and it is divided into four quadrants according to rectangular coordinate system, and the quadrant that β is located is set as candidate region, mesh
Mark candidate region and matrix a carry out with computing after, the just only candidate target in the remaining direction of motion relying on former frame prediction.
Step 3) define another matrix b, the tracking in the case of the unexpected break-in of target in reply present frame:
It is used matrix a and matrix b to carry out or computing is as sampling matrix；
Step 4) according to motion vector, calculate the motion of front cross frame target apart from d, be adaptively adjusted according to described distance
Hunting zone, arranges weights by the distance and target window size of motion, then weights ω is:
Wherein, w, h are width and the height of target window；Using weights ω self-adaptative adjustment initial search frequency range.
The design further of the described compression tracking based on motion prediction is, the adaptive optimization of tracking window,
Particularly as follows: being adjusted to window size with the distance of two pixels every time, carrying out feature extraction in these windows, and using pattra leaves
This grader is classified, and finds best matching result, that is, obtain the adaptive optimization of tracking window.
Advantages of the present invention:
The present invention gives target initial shape based on the compression tracking of motion prediction in the first frame of target video sequence
In the case of state, motion prediction is carried out to video object according to present frame, the direction of motion obtaining target is according to motion vector, meter
Calculate the distance of front cross frame target motion, according to this apart from adjust automatically hunting zone, greatly reduce search time, reduce
The collection of candidate samples；Using adaptive tracing window optimization, thus improve compression tracking under some complex scenes
Real-time and robustness.
Brief description
The FB(flow block) of Fig. 1 the inventive method.
Fig. 2 the inventive method positive sample selects to optimize schematic diagram.
Fig. 3 compressed sensing algorithm keeps track schematic diagram and the inventive method adaptive tracing window optimization follow the tracks of schematic diagram.
Fig. 4 the inventive method, ct and cnt tracking result schematic diagram.
Fig. 5 the inventive method, ct and cnt follow the tracks of another result schematic diagram.
Another tracking result schematic diagram of Fig. 6 the inventive method, ct and cnt.
Specific embodiment
In order that the objects, technical solutions and advantages of the present invention become more apparent, below in conjunction with the accompanying drawings the application is entered
One step describes in detail.
As Fig. 1, the compression tracking based on motion prediction of the present embodiment, comprising: initialization, choose the first two field picture
Middle tracking target area, in the first two field picture close-proximity target zone sampled images block, carries out feature extraction and dimensionality reduction to image block
Obtain the characteristic vector of each image block, set up grader.Follow the tracks of target and motion prediction is carried out to target: for t+1 frame
Image, using front cross frame, the value of t is greater than the integer equal to 2.Target location in image obtain target motion vectors with pre-
Survey direction, and the predicted position of target, the predicted position tracing in t frame is drawn according to target motion vectors and prediction direction
Neighbouring sampled images block, and then obtain the characteristic vector of each image block, grader is by corresponding for maximum grader image block
The target tracing into as present frame.Determine the final tracking result of present frame: adaptive optimization is carried out to tracking window, in mesh
Near mark, sampled again, and selected by grader, corresponding for maximum grader image block is followed the tracks of as final
Target；And record present frame target location, trace into target location, update classifier parameters, such as Fig. 3.
In the present embodiment, grader adopt Bayes classifier h (v), such as formula (1),
Wherein, p (y=1)=p (y=0), y ∈ { 0,1 } is positive negative sample mark, and y=0 represents negative sample, and y=1 generation
Table positive sample, low dimensional space v=(v1,....,vn)t,viFor i-th element in v.
Further, set the condition distribution p (v in grader h (v)i| y=1) and p (vi| y=0) it is Gauss distribution, and
MeetWhereinWithIt is respectively the phase of positive sample probability
Hope and standard deviation, andWithIt is respectively expectation and the standard deviation of negative sample probability,For being desired forAnd standard deviation
ForGauss distribution,For being desired forWith standard deviation it isGauss distribution, n represents the symbol of Gauss distribution,
Bayes classifier parameter updates as formula (2), formula (3):
Wherein, μ1、σ1Represent the expectation of positive sample and standard deviation after updating, λ ＞ 0 is Studying factors, n representative sample number, vi(k) represent kth sample represent in lower dimensional space, k
Represent k-th sample.
In the present embodiment, the sampling of image block realizes collection, positive sample selection gist by the positive and negative samples selection optimizing
Success rate formula, success rate formula such as formula (4):
Wherein, ci is the target frame region that algorithm calculates, and ri is realistic objective region；When the value of rt in a frame is more than
0.5 then it is assumed that tracking result is correct.
Further, as Fig. 2, the optimization of positive and negative samples selection, according to success rate formula, the optimization of positive and negative samples selection is public
Formula such as formula (5):
Wherein, w, h are width and the height of target window, and l is side-play amount, n and m is horizontal and vertical offset component,
Positive sample selection region is region more than 0.9 for the success rate, and negative sample is success rate in 0.5 area below.
In the present embodiment, motion prediction specifically includes following steps:
Step 1) set former frame target location as (x1,y1), present frame target location is (x2,y2), then motion vector β
For:
β=(x2,y2)-(x1,y1)；
Step 2) according to target direction of motion, definition of search region, define a matrix a, this matrix size and search model
Enclose size identical, it is divided into four quadrants according to rectangular coordinate system, the quadrant that β is located is set as candidate region, target
Candidate region and matrix a carry out with computing after, the just only candidate target in the remaining direction of motion relying on former frame prediction.
Step 3) define another matrix b, the tracking in the case of the unexpected break-in of target in reply present frame:
It is used matrix a and matrix b to carry out or computing is as sampling matrix；
Step 4) according to motion vector, calculate the motion of front cross frame target apart from d, be adaptively adjusted according to described distance
Hunting zone, arranges weights by the distance and target window size of motion, then weights ω is:
Wherein, w, h are width and the height of target window；Using weights ω self-adaptative adjustment initial search frequency range.
In order to improve the robustness of tracking, the adaptive optimization of tracking window, particularly as follows: every time with the distance of two pixels
Window size is adjusted, carries out feature extraction in these windows, and classified using Bayes classifier, find best
Join result, that is, obtain the adaptive optimization of tracking window.
The effect to the present embodiment method of the application has carried out experimental verification, using ct [document 1] and cnt [document 4],
In order to be compared with the effect of the inventive method, in conjunction with the tracking result figure shown in Fig. 4, Fig. 5 and Fig. 6.The tracking of ct algorithm
Result No. 1 frame solid line of redness marks, and cnt arithmetic result No. 2 frame solid lines of green mark, and result of the present invention is using blue No. 3
Dotted box marks.As seen from the figure, the accuracy of the tracking of the present embodiment either target following is still in tracking window
Target location be all better than other two methods.Experiment is verified in 15 kinds of challenging sequences, altogether tests
7531 two field pictures.This 15 kinds of videos are public video library [document 5] (http://cvlab.hanyang.ac.kr/tracker_
Benchmark/ randomly select in).Experimental facilitiess are configured to, 2.5ghz dominant frequency four core core i5cpu, internal memory 4gb,
Windows7 32-bit operating system, and run in matlab 2014a development platform.
For evaluation algorithms performance of target tracking, inventor is come using tracking success rate rt and center offset two indices
Weigh the accuracy rate followed the tracks of.Wherein, the definition of success rate rt is:
Wherein, ci is the target frame region that algorithm calculates, and ri is realistic objective region.So, when the value of rt in a frame
More than 0.5 then it is assumed that tracking result is correct.
The definition of center offset is the Euclidean distance with actual frames central point for the central point of tracking result frame.Result is as follows
Shown in table.
Table 1 time and success rate
Table 2 center offset
As seen from the above table, the inventive method improves robustness and the real-time of tracking.Compared to cnt algorithm, the present invention
Method has very big advantage in terms of real-time, and the time of use is more than one the percent of cnt algorithm；Come following the tracks of success rate
See, cnt algorithm doing well in individual video, but overall the inventive method to be slightly worse than.Meanwhile, the inventive method
Mean center side-play amount is minimum, the center offset highest of ct algorithm.Therefore, in terms of the degree of off-centring, side of the present invention
Method there has also been larger improvement.
The compression tracking based on motion prediction of the present embodiment is at the beginning of the given target of the first frame of target video sequence
In the case of beginning state, motion prediction is carried out to video object according to present frame, the direction of motion obtaining target is according to motion arrow
Amount, calculates the distance of front cross frame target motion, according to this apart from adjust automatically hunting zone, greatly reduces search time,
Reduce the collection of candidate samples；Using adaptive tracing window optimization, thus improve compression tracking in some complicated fields
Real-time under scape and robustness.
The above, the only present invention preferably specific embodiment, but protection scope of the present invention is not limited to this, appoints
What those familiar with the art the invention discloses technical scope in, technology according to the present invention scheme and its this
Inventive concept equivalent or change in addition, is all included within the scope of the present invention.
Claims (7)
1. a kind of compression tracking based on motion prediction is it is characterised in that include:
Initialization, chooses in the first two field picture and follows the tracks of target area, in the first two field picture close-proximity target zone sampled images block, right
Described image block carries out feature extraction and dimensionality reduction obtains the characteristic vector of each image block, sets up grader；
Follow the tracks of target and target is carried out with motion prediction: for t+1 two field picture, t is the integer more than or equal to 2, using front two
Target location in two field picture obtains target motion vectors and prediction direction, and according to described target motion vectors and prediction direction
Draw the predicted position of target, sampled images block near the predicted position that t frame traces into, and then obtain each image block
Characteristic vector, the target that corresponding for maximum grader image block is traced into by described grader as present frame；
Determine the final tracking result of present frame: adaptive optimization is carried out to tracking window, in described target proximity, carries out again
Sampling, and selected by grader, corresponding for maximum grader image block is followed the tracks of target as final；And record current
Frame target location, traces into target location, updates classifier parameters.
2. the compression tracking based on motion prediction according to claim 1 is it is characterised in that described grader adopts
Bayes classifier h (v), such as formula (1),
Wherein, p (y=1)=p (y=0), y ∈ { 0,1 } are positive negative sample marks, and y=0 represents negative sample, and y=1 just represents
Sample, low dimensional space v=(v1,....,vn)t,viFor i-th element in v.
3. the compression tracking based on motion prediction according to claim 2 is it is characterised in that set the bar in grader h (v)
Part distribution p (vi| y=1) and p (vi| y=0) it is Gauss distribution, and meet
WhereinWithIt is respectively expectation and the standard deviation of positive sample probability, andWithIt is respectively expectation and the mark of negative sample probability
It is accurate poor,For being desired forWith standard deviation it isGauss distribution,For being desired forWith standard deviation it is
Gauss distribution, n represents the symbol of Gauss distribution, and Bayes classifier parameter updates as formula (2), formula (3):
Wherein, μ1、σ1Represent the expectation of positive sample and standard deviation after updating,
λ ＞ 0 is Studying factors, n representative sample number, viK () represents that kth sample represents in lower dimensional space, k represents k-th sample.
4. the compression tracking based on motion prediction according to claim 1 is it is characterised in that the adopting of described image block
The positive and negative samples selection that sample passes through to optimize is realized gathering, positive sample selection gist success rate formula, success rate formula such as formula (4):
Wherein, ci is the target frame region that algorithm calculates, and ri is realistic objective region；When the value of rt in a frame is more than 0.5, then
Think that tracking result is correct.
5. the compression tracking based on motion prediction according to claim 4 is it is characterised in that positive and negative samples selection
Optimize, according to described success rate formula, the optimization formula such as formula (5) of positive and negative samples selection:
Wherein, w, h are width and the height of target window, and l is side-play amount, n and m is horizontal and vertical offset component, positive sample
This selection region is region more than 0.9 for the success rate, and negative sample is success rate in 0.5 area below.
6. the compression tracking based on motion prediction according to claim 1 is it is characterised in that described motion prediction has
Body comprises the steps:
Step 1) set former frame target location as (x1,y1), present frame target location is (x2,y2), then motion vector β is:
β=(x2,y2)-(x1,y1)；
Step 2) according to target direction of motion, definition of search region, define a matrix a, described matrix a size and hunting zone
Size is identical, and it is divided into four quadrants according to rectangular coordinate system, and the quadrant that β is located is set as candidate region, and target is waited
Favored area and matrix a carry out with computing after, the just only candidate target in the remaining direction of motion relying on former frame prediction.
Step 3) define another matrix b, the tracking in the case of the unexpected break-in of target in reply present frame:
It is used matrix a and matrix b to carry out or computing is as sampling matrix；
Step 4) according to motion vector, calculate the motion of front cross frame target apart from d, search is adaptively adjusted according to described distance
Scope, arranges weights by the distance and target window size of motion, then weights ω is:
Wherein, w, h are width and the height of target window；Using weights ω self-adaptative adjustment initial search frequency range.
7. according to claim 2 based on motion prediction compression tracking it is characterised in that tracking window adaptive
Should optimizing, particularly as follows: adjusting to window size with the distance of two pixels every time, in these windows, carrying out feature extraction, and
Classified using Bayes classifier, found best matching result, that is, obtained the adaptive optimization of tracking window.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610701199.XA CN106372650B (en) | 2016-08-19 | 2016-08-19 | A kind of compression tracking based on motion prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610701199.XA CN106372650B (en) | 2016-08-19 | 2016-08-19 | A kind of compression tracking based on motion prediction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106372650A true CN106372650A (en) | 2017-02-01 |
| CN106372650B CN106372650B (en) | 2019-03-19 |
Family
ID=57879528
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610701199.XA Active CN106372650B (en) | 2016-08-19 | 2016-08-19 | A kind of compression tracking based on motion prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106372650B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108304808A (en) * | 2018-02-06 | 2018-07-20 | 广东顺德西安交通大学研究院 | A kind of monitor video method for checking object based on space time information Yu depth network |
| CN109859242A (en) * | 2019-01-16 | 2019-06-07 | 重庆邮电大学 | A kind of method for tracking target for predicting adaptive learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050018879A1 (en) * | 2003-07-22 | 2005-01-27 | Wataru Ito | Object tracking method and object tracking apparatus |
| CN101867798A (en) * | 2010-05-18 | 2010-10-20 | 武汉大学 | Mean shift moving object tracking method based on compressed domain analysis |
| CN102881024A (en) * | 2012-08-24 | 2013-01-16 | 南京航空航天大学 | Tracking-learning-detection (TLD)-based video object tracking method |
| CN104683802A (en) * | 2015-03-24 | 2015-06-03 | 江南大学 | H.264/AVC compressed domain based moving target tracking method |
-
2016
- 2016-08-19 CN CN201610701199.XA patent/CN106372650B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050018879A1 (en) * | 2003-07-22 | 2005-01-27 | Wataru Ito | Object tracking method and object tracking apparatus |
| CN101867798A (en) * | 2010-05-18 | 2010-10-20 | 武汉大学 | Mean shift moving object tracking method based on compressed domain analysis |
| CN102881024A (en) * | 2012-08-24 | 2013-01-16 | 南京航空航天大学 | Tracking-learning-detection (TLD)-based video object tracking method |
| CN104683802A (en) * | 2015-03-24 | 2015-06-03 | 江南大学 | H.264/AVC compressed domain based moving target tracking method |
Non-Patent Citations (1)
| Title |
|---|
| 王哲: ""基于压缩感知的运动目标跟踪算法的研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108304808A (en) * | 2018-02-06 | 2018-07-20 | 广东顺德西安交通大学研究院 | A kind of monitor video method for checking object based on space time information Yu depth network |
| CN108304808B (en) * | 2018-02-06 | 2021-08-17 | 广东顺德西安交通大学研究院 | Monitoring video object detection method based on temporal-spatial information and deep network |
| CN109859242A (en) * | 2019-01-16 | 2019-06-07 | 重庆邮电大学 | A kind of method for tracking target for predicting adaptive learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106372650B (en) | 2019-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jiao et al. | New generation deep learning for video object detection: A survey | |
| CN106570486B (en) | Filtered target tracking is closed based on the nuclear phase of Fusion Features and Bayes's classification | |
| CN109993775B (en) | Single target tracking method based on characteristic compensation | |
| CN107633226B (en) | Human body motion tracking feature processing method | |
| CN108416266A (en) | A kind of video behavior method for quickly identifying extracting moving target using light stream | |
| CN107403175A (en) | Visual tracking method and Visual Tracking System under a kind of movement background | |
| CN108961308B (en) | Residual error depth characteristic target tracking method for drift detection | |
| CN107590427B (en) | Method for detecting abnormal events of surveillance video based on space-time interest point noise reduction | |
| CN110084165A (en) | The intelligent recognition and method for early warning of anomalous event under the open scene of power domain based on edge calculations | |
| CN108647694A (en) | Correlation filtering method for tracking target based on context-aware and automated response | |
| Zhang et al. | A swarm intelligence based searching strategy for articulated 3D human body tracking | |
| CN106529441B (en) | Depth motion figure Human bodys' response method based on smeared out boundary fragment | |
| CN112801019B (en) | Method and system for eliminating re-identification deviation of unsupervised vehicle based on synthetic data | |
| CN107230219A (en) | A kind of target person in monocular robot is found and follower method | |
| CN112616023A (en) | Multi-camera video target tracking method in complex environment | |
| CN112329784A (en) | Correlation filtering tracking method based on space-time perception and multimodal response | |
| CN110569706A (en) | Deep integration target tracking algorithm based on time and space network | |
| Yang et al. | Visual tracking with long-short term based correlation filter | |
| Zhang et al. | A survey on instance segmentation: Recent advances and challenges | |
| Hui | RETRACTED ARTICLE: Motion video tracking technology in sports training based on Mean-Shift algorithm | |
| CN109241932B (en) | Thermal infrared human body action identification method based on motion variance map phase characteristics | |
| Zhang et al. | Residual memory inference network for regression tracking with weighted gradient harmonized loss | |
| CN106372650B (en) | A kind of compression tracking based on motion prediction | |
| CN106934395B (en) | Rigid body target tracking method adopting combination of SURF (speeded Up robust features) and color features | |
| CN109102520A (en) | The moving target detecting method combined based on fuzzy means clustering with Kalman filter tracking |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |