CN106909935A - A kind of method for tracking target and device - Google Patents
A kind of method for tracking target and device Download PDFInfo
- Publication number
- CN106909935A CN106909935A CN201710039664.2A CN201710039664A CN106909935A CN 106909935 A CN106909935 A CN 106909935A CN 201710039664 A CN201710039664 A CN 201710039664A CN 106909935 A CN106909935 A CN 106909935A
- Authority
- CN
- China
- Prior art keywords
- target
- target object
- tracking
- similarity
- size
- 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
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The method for tracking target that the present invention is provided, the home position of target object is obtained first, target object is tracked according to the position, obtain tracing positional, then the similarity of the tracing positional and the home position of target object is calculated, if the similarity is more than predetermined threshold value, the size of the target object in tracing positional is obtained；Size according to target object updates original object object.In the program, tracking result is proofreaded by cascade detectors, if the difference with home position is little, then explanation is tracked successfully, original image is further updated by the size of target image, so when exist block or target object change in size when, newest target image can be obtained, so as to go further tracking according to the target image, tracking of the prior art can be overcome causes tracking to fail when target object is blocked or size changes, and improves the performance of existing algorithm.
Description
Technical field
The present invention relates to computer vision field, and in particular to a kind of method for tracking target and device.
Background technology
In computer vision field, target following is always one of hot research field.So-called target following, is one
In individual continuous image sequence, the process of lasting positioning is carried out to target interested.Target following be widely used in it is military,
The multiple fields such as traffic, monitoring.Due to illumination variation, target deformation, the factor such as target occlusion and real-time influences, accurately
Target tracking algorism is also difficult to.
It is the preferable track algorithm of a kind of effect developed in recent years that nuclear phase closes filter tracking algorithm, is moved using circulation
The thought of position, constructs substantial amounts of sample to train grader, while using discrete Fourier transform reduction classifier training and inspection
Operand during survey.
But nuclear phase closes filter tracking algorithm has that some are intrinsic：First, when target object is continuously blocked,
Shelter can shelter from target, so as to cause tracking result to deviate.Secondly, in the larger change of yardstick or form generation of target object
In the case of change, tracking performance is very restricted.Therefore how to be preferably tracked when tracking target changes
Referred to as problem demanding prompt solution.
The content of the invention
Therefore, the technical problem to be solved in the present invention be tracking of the prior art when target object is blocked or
Tracking is caused to fail when size changes.
Therefore, the present invention provides a kind of method for tracking target, comprise the following steps：Obtain the home position of target object；
Target object is tracked according to the home position, obtains tracing positional；The tracing positional is calculated with target object
The similarity in home position；Judge the similarity whether more than predetermined threshold value；If the similarity is more than predetermined threshold value, obtain
The size of the target object in tracing positional；Size according to target object updates original object object.
Preferably, also target object is tracked according to the target size after renewal including next two field picture.
Preferably, if being also not more than predetermined threshold value including the similarity, the tracking result is incorrect, and tracking is lost
Lose.
Preferably, in the step of being tracked to target object, filter tracking algorithm is closed using nuclear phase and is tracked.
Preferably, by tracing positional described in classifier calculated and the similarity in the home position of target object.
Preferably, the grader and trained by positive negative sample.
Additionally, the present invention also provides a kind of target tracker, including：Home position extraction unit, obtains target object
Home position；Tracking cell, is tracked according to the home position to target object, obtains tracing positional；Similarity meter
Unit is calculated, the similarity of the tracing positional and the home position of target object is calculated；Judging unit, it is described similar for judging
Whether degree is more than predetermined threshold value；Size acquiring unit, if being more than predetermined threshold value for the similarity, in acquisition tracing positional
The size of target object；Target update unit, for updating original object object according to the size of target object.
Preferably, also including follow-up tracking cell, for next two field picture according to the target size after renewal to object
Body is tracked.
Preferably, also including tracking failure unit, if being not more than predetermined threshold value, the tracking knot for the similarity
It is really incorrect, tracking failure.
Preferably, filter tracking algorithm is closed using nuclear phase to be tracked.
Technical solution of the present invention, has the following advantages that：
The method for tracking target that the present invention is provided, obtains the home position of target object, according to the position to target first
Object is tracked, and obtains tracing positional, then calculates the similarity of the tracing positional and the home position of target object, if
The similarity is more than predetermined threshold value, obtains the size of the target object in tracing positional；Size according to target object updates
Original object object.In the program, tracking result is proofreaded by cascade detectors, if with the difference in home position not
Greatly, then explanation is tracked successfully, and original image is further updated by the size of target image, so when presence is blocked or object
When body change in size, newest target image can be obtained, so as to go further tracking according to the target image, can
Overcome tracking of the prior art causes tracking to fail when target object is blocked or size changes, and improves existing
The performance of algorithm.
Brief description of the drawings
In order to illustrate more clearly of the specific embodiment of the invention or technical scheme of the prior art, below will be to specific
The accompanying drawing to be used needed for implementation method or description of the prior art is briefly described, it should be apparent that, in describing below
Accompanying drawing is some embodiments of the present invention, for those of ordinary skill in the art, before creative work is not paid
Put, other accompanying drawings can also be obtained according to these accompanying drawings.
Fig. 1 is an a kind of flow chart for specific example of method for tracking target in the embodiment of the present invention 1.
Fig. 2 is an a kind of structured flowchart for specific example of target tracker in the embodiment of the present invention 2；
Specific embodiment
Technical scheme is clearly and completely described below in conjunction with accompanying drawing, it is clear that described implementation
Example is a part of embodiment of the invention, rather than whole embodiments.Based on the embodiment in the present invention, ordinary skill
The every other embodiment that personnel are obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.
In the description of the invention, it is necessary to explanation, term " " center ", " on ", D score, "left", "right", " vertical ",
The orientation or position relationship of the instruction such as " level ", " interior ", " outward " be based on orientation shown in the drawings or position relationship, merely to
Be easy to the description present invention and simplify describe, rather than indicate imply signified device or element must have specific orientation,
With specific azimuth configuration and operation, therefore it is not considered as limiting the invention.Additionally, term " first ", " second ",
" the 3rd " is only used for describing purpose, and it is not intended that indicating or implying relative importance.
In the description of the invention, it is necessary to illustrate, unless otherwise clearly defined and limited, term " installation ", " phase
Company ", " connection " should be interpreted broadly, for example, it may be being fixedly connected, or being detachably connected, or be integrally connected；Can
Being to mechanically connect, or electrically connect；Can be joined directly together, it is also possible to be indirectly connected to by intermediary, can be with
It is two connections of element internal, can is wireless connection, or wired connection.For one of ordinary skill in the art
For, above-mentioned term concrete meaning in the present invention can be understood with concrete condition.
As long as additionally, technical characteristic involved in invention described below different embodiments non-structure each other
Can just be combined with each other into conflict.
Embodiment 1
A kind of method for tracking target is provided in the present embodiment, for tracking target object, particularly with partial occlusion or chi
The very little target object for changing can preferably be tracked.During the method can be with smart machines such as user computer, servers,
Target object in the video of input is tracked.
Method for tracking target in the present embodiment comprises the following steps：
S1, the home position for obtaining target object.In video initial frame, to the initial position that sets the goal.
S2, target object is tracked according to the home position, obtains tracing positional, tracing positional is a rectangle
Frame.In the step of being tracked to target object, filter tracking algorithm is closed using nuclear phase and is tracked.
The tracking process of core correlation filtering can be decomposed into following several steps：
The first step, in initial two field picture, target initial position is located at position p (t).In I (t) two field pictures, in place
Put p (t) nearby to sample, train a recurrence device, this returns device and can calculate the response of each sampling wicket.
Second step, in I (t+1) frame, is sampled near previous frame position p (t), and the recurrence device trained with previous frame is calculated
The response of each sampling window, responds target location p (t+1) of the maximum window as this frame, and the target location is tracking
As a result.
Core correlation filtering has advantages below：First, gathered just using the circular matrix of target area peripheral region
Negative sample, using ridge regression training return device, and successfully using circular matrix Fourier space diagonalizable property, greatly
It is big to reduce operand, improve algorithm arithmetic speed.Secondly, linear ridge regression is mapped to non-linear space by kernel function, is made
The sample of linearly inseparable can divide in non-linear space.
In core correlation filtering, all of training sample is obtained by target sample cyclic shift, and cyclic shift can be by
Permutation matrix is obtained, for two dimensional image, can loopy moving realizes the movement of diverse location respectively by x-axis and y-axis.
The two dimensional image training sample constructed by cyclic shift is as follows：
So all of training sample is just obtained, and then trains classification by the method for ridge regression and Fourier transformation
Device.
If training sample set (xi, yi), then its linear regression function f (xi)=wTxi, w is that column vector represents weight system
Number, can be solved by least square method,
By seeking partial derivative, abbreviation is carried out, obtain the form of last analytic solutions：
W=(XHX+λ)-1XHy
Then by introducing Fourier transformation, the computation complexity of matrix inversion is reduced.
The similarity in the home position of S3, the calculating tracing positional and target object.
The similarity of tracing positional and the home position of target object can be calculated by nearest neighbor classifier, nearest
The home position of target object is prestored in adjacent grader.Tracking result return target rectangle frame, calculate the position with it is nearest
Related similarity S (r) of the real target location preserved in adjacent grader.
S4, judge the similarity whether less than predetermined threshold value.S (r) then thinks to track successfully more than or equal to threshold value, holds
Row S5-S6；If S (r) is less than the threshold value of setting, current goal is considered as because quick movement causes tracking to fail, perform S7-
S9。
If S5, the similarity are more than predetermined threshold value, the size of the target object in tracing positional is obtained.
S6, according to the size of target object update original object object, next two field picture is according to the target size after renewal
Target object is tracked.In this case think to track successfully, but size modes due to object etc. change, this
When need size further to target object to be adjusted so that its with block or change in size after image be adapted.Energy
Enough overcome tracking of the prior art causes tracking to fail when target object is blocked or size changes, and improves existing
There is the performance of algorithm.
If S7, the similarity are less than predetermined threshold value, multiple candidate's tracing areas are obtained by cascade detectors.
Because track algorithm is when a new frame is tracked, sampled near the target location of previous frame, calculate to return and ring
Should, this results in algorithm and cannot solve the target quickly mobile situation with dimensional variation, therefore introduces cascade detectors part pair
Both of these case is modified.
Cascade detectors scan input picture by sliding window, then judge in each window either with or without target, level
Connection detector mainly includes three modules：Variance grader, integrated classifier and nearest neighbor classifier.
Initial phase：Being obtained according to certain step-length and scaling can be comprising all possible size and conversion
Preliminary sweep grid, positive sample is obtained to the initial position synthesis that sets the goal.In the scanning grid nearest apart from initial rectangular frame
In, 10 rectangle frames are selected, for each rectangle frame, we are by geometric transformation (± 1% skew, the contracting of ± 1% ratio
Put, ± 10 ° of Plane Rotation) the different rectangle frames of generation 20, and it is additional with the normal Gaussian noise that variance is 5, finally obtain
200 positive samples.Negative sample searches what choosing was obtained from around initial rectangular frame, it is not necessary to carry out affine transformation.
The positive negative sample obtained by initialization, trains integrated classifier and nearest neighbor classifier.
Related similarity S (r) excursion is from 0 to 1 in nearest neighbor classifier.Value is bigger to represent that image block is more possible to
It is target area.
Wherein S (+) is positive sample similarity：
Wherein S (-) is negative sample similarity：
In nearest neighbor classifier calculating process, related similarity S (r) be used to point out an arbitrary image block and mesh
Part in mark model has great similar.If Sr(p, M) ＞ θNN, then image p is classified as positive sample, is otherwise just divided
Class is negative sample.
Scan image sorting phase：In variance classifier modules, image block variance is calculated, if the side of this image block
Difference just refuses this image block less than the half of target image block variance.Integrated classifier is by a series of basic classification device groups
Into judging whether image block exports by calculating the LBP features of image.By first two steps, image block has obtained significantly subtracting
It is few, using nearest neighbor classifier, final target area is exported, target area herein is candidate's tracing area.
S8, it is tracking result by candidate's tracing area cluster confidence level highest region.For cascade detectors meeting
Return to several candidate's tracing areas, one related similarity of each candidate's tracing area correspondence, by cluster, obtain it is several most
Believable target area, using weighted average, exports a final testing result.
S9, the home position that the target object is updated with the tracking result, the position in next two field picture in the updated
Put and target object is tracked.So as to improve the degree of accuracy of subsequent frame image trace.
Compared with prior art, a large amount of samples are effectively constructed using the method for circular matrix based on core correlation filtering
This, can obtain preferable grader, and using Fourier transformation reduction operation time, while the present invention utilizes cascade detectors
Testing result, quickly mobile and target is multiple dimensioned effectively to solve the problems, such as target.
Embodiment 2：
A kind of target tracker is provided in the present embodiment, for being tracked to the target object in video image, especially
It is applied to target object produce block or target object size situation about changing, the apparatus structure is as shown in Fig. 2 bag
Include：
Home position extraction unit 21, the home position for obtaining target object；
Tracking cell 22, for being tracked to target object according to the home position, obtains tracing positional；Using core
Correlation filtering track algorithm is tracked.
Similarity calculated 23, the similarity for calculating the tracing positional and the home position of target object；
Judging unit 24, for judging the similarity whether more than predetermined threshold value；
Size acquiring unit 25, if being more than predetermined threshold value for the similarity, obtains the target object in tracing positional
Size；
Target update unit 26, for updating original object object according to the size of target object.Next two field picture according to
Target size after renewal is tracked to target object.
It should be understood by those skilled in the art that, embodiments of the invention can be provided as method, system or computer program
Product.Therefore, the present invention can be using the reality in terms of complete hardware embodiment, complete software embodiment or combination software and hardware
Apply the form of example.And, the present invention can be used and wherein include the computer of computer usable program code at one or more
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) is produced
The form of product.
The present invention is the flow with reference to method according to embodiments of the present invention, equipment (system) and computer program product
Figure and/or block diagram are described.It should be understood that every first-class during flow chart and/or block diagram can be realized by computer program instructions
The combination of flow and/or square frame in journey and/or square frame and flow chart and/or block diagram.These computer programs can be provided
The processor of all-purpose computer, special-purpose computer, Embedded Processor or other programmable data processing devices is instructed to produce
A raw machine so that produced for reality by the instruction of computer or the computing device of other programmable data processing devices
The device of the function of being specified in present one flow of flow chart or multiple one square frame of flow and/or block diagram or multiple square frames.
These computer program instructions may be alternatively stored in can guide computer or other programmable data processing devices with spy
In determining the computer-readable memory that mode works so that instruction of the storage in the computer-readable memory is produced and include finger
Make the manufacture of device, the command device realize in one flow of flow chart or multiple one square frame of flow and/or block diagram or
The function of being specified in multiple square frames.
These computer program instructions can be also loaded into computer or other programmable data processing devices so that in meter
Series of operation steps is performed on calculation machine or other programmable devices to produce computer implemented treatment, so as in computer or
The instruction performed on other programmable devices is provided for realizing in one flow of flow chart or multiple flows and/or block diagram one
The step of function of being specified in individual square frame or multiple square frames.
Obviously, above-described embodiment is only intended to clearly illustrate example, and not to the restriction of implementation method.It is right
For those of ordinary skill in the art, can also make on the basis of the above description other multi-forms change or
Change.There is no need and unable to be exhaustive to all of implementation method.And the obvious change thus extended out or
Among changing still in the protection domain of the invention.
Claims (10)
1. a kind of method for tracking target, it is characterised in that comprise the following steps：
Obtain the home position of target object；
Target object is tracked according to the home position, obtains tracing positional；
Calculate the similarity of the tracing positional and the home position of target object；
Judge the similarity whether more than predetermined threshold value；
If the similarity is more than predetermined threshold value, the size of the target object in tracing positional is obtained；
Size according to target object updates original object object.
2. method according to claim 1, it is characterised in that also including next two field picture according to the target size after renewal
Target object is tracked.
3. method according to claim 1 and 2, it is characterised in that if being also not more than predetermined threshold value including the similarity,
Then the tracking result is incorrect, tracking failure.
4. method according to claim 3, it is characterised in that in the step of being tracked to target object, using nuclear phase
Filter tracking algorithm is closed to be tracked.
5. method according to claim 4, it is characterised in that by tracing positional described in classifier calculated and target object
Home position similarity.
6. method according to claim 5, it is characterised in that the grader and trained by positive negative sample.
7. a kind of target tracker, it is characterised in that including：
Home position extraction unit, obtains the home position of target object；
Tracking cell, is tracked according to the home position to target object, obtains tracing positional；
Similarity calculated, calculates the similarity of the tracing positional and the home position of target object；
Judging unit, for judging the similarity whether more than predetermined threshold value；
Size acquiring unit, if being more than predetermined threshold value for the similarity, obtains the size of the target object in tracing positional；
Target update unit, for updating original object object according to the size of target object.
8. target tracker according to claim 7, it is characterised in that also including follow-up tracking cell, for next
Two field picture is tracked according to the target size after renewal to target object.
9. the target tracker according to claim 7 or 8, it is characterised in that also including tracking failure unit, if for
The similarity is not more than predetermined threshold value, then the tracking result is incorrect, tracking failure.
10. target tracker according to claim 9, it is characterised in that closing filter tracking algorithm using nuclear phase is carried out
Tracking.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710039664.2A CN106909935B (en) | 2017-01-19 | 2017-01-19 | Target tracking method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710039664.2A CN106909935B (en) | 2017-01-19 | 2017-01-19 | Target tracking method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106909935A true CN106909935A (en) | 2017-06-30 |
| CN106909935B CN106909935B (en) | 2021-02-05 |
Family
ID=59207156
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710039664.2A Active CN106909935B (en) | 2017-01-19 | 2017-01-19 | Target tracking method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106909935B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107480704A (en) * | 2017-07-24 | 2017-12-15 | 南开大学 | It is a kind of that there is the real-time vision method for tracking target for blocking perception mechanism |
| CN109740441A (en) * | 2018-12-10 | 2019-05-10 | 平安科技（深圳）有限公司 | Object detection method, device and terminal device based on image recognition |
| CN111242981A (en) * | 2020-01-21 | 2020-06-05 | 北京捷通华声科技股份有限公司 | Tracking method and device for fixed object and security equipment |
| CN114565638A (en) * | 2022-01-25 | 2022-05-31 | 上海安维尔信息科技股份有限公司 | Multi-target tracking method and system based on tracking chain |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102945554A (en) * | 2012-10-25 | 2013-02-27 | 西安电子科技大学 | Target tracking method based on learning and speeded-up robust features (SURFs) |
| CN102982340A (en) * | 2012-10-31 | 2013-03-20 | 中国科学院长春光学精密机械与物理研究所 | Target tracking method based on semi-supervised learning and random fern classifier |
| CN103325126A (en) * | 2013-07-09 | 2013-09-25 | 中国石油大学(华东) | Video target tracking method under circumstance of scale change and shielding |
| CN103440669A (en) * | 2013-09-10 | 2013-12-11 | 武汉大学 | Dynamic Mean shift kernel bandwidth updating method based on compressed domain fusion |
| CN103927762A (en) * | 2013-01-11 | 2014-07-16 | 浙江大华技术股份有限公司 | Target vehicle automatic tracking method and device |
| CN104240266A (en) * | 2014-09-04 | 2014-12-24 | 成都理想境界科技有限公司 | Target object tracking method based on color-structure features |
-
2017
- 2017-01-19 CN CN201710039664.2A patent/CN106909935B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102945554A (en) * | 2012-10-25 | 2013-02-27 | 西安电子科技大学 | Target tracking method based on learning and speeded-up robust features (SURFs) |
| CN102982340A (en) * | 2012-10-31 | 2013-03-20 | 中国科学院长春光学精密机械与物理研究所 | Target tracking method based on semi-supervised learning and random fern classifier |
| CN103927762A (en) * | 2013-01-11 | 2014-07-16 | 浙江大华技术股份有限公司 | Target vehicle automatic tracking method and device |
| CN103325126A (en) * | 2013-07-09 | 2013-09-25 | 中国石油大学(华东) | Video target tracking method under circumstance of scale change and shielding |
| CN103440669A (en) * | 2013-09-10 | 2013-12-11 | 武汉大学 | Dynamic Mean shift kernel bandwidth updating method based on compressed domain fusion |
| CN104240266A (en) * | 2014-09-04 | 2014-12-24 | 成都理想境界科技有限公司 | Target object tracking method based on color-structure features |
Non-Patent Citations (1)
| Title |
|---|
| 雷帮军 等: "《视频目标跟踪系统分步详解》", 31 December 2015, 北京:国防工业出版社 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107480704A (en) * | 2017-07-24 | 2017-12-15 | 南开大学 | It is a kind of that there is the real-time vision method for tracking target for blocking perception mechanism |
| CN107480704B (en) * | 2017-07-24 | 2021-06-29 | 无际智控(天津)智能科技有限公司 | Real-time visual target tracking method with shielding perception mechanism |
| CN109740441A (en) * | 2018-12-10 | 2019-05-10 | 平安科技（深圳）有限公司 | Object detection method, device and terminal device based on image recognition |
| CN111242981A (en) * | 2020-01-21 | 2020-06-05 | 北京捷通华声科技股份有限公司 | Tracking method and device for fixed object and security equipment |
| CN114565638A (en) * | 2022-01-25 | 2022-05-31 | 上海安维尔信息科技股份有限公司 | Multi-target tracking method and system based on tracking chain |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106909935B (en) | 2021-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106952288B (en) | Based on convolution feature and global search detect it is long when block robust tracking method | |
| Yang et al. | SiamAtt: Siamese attention network for visual tracking | |
| CN103208008B (en) | Based on the quick adaptive method of traffic video monitoring target detection of machine vision | |
| CN106909935A (en) | A kind of method for tracking target and device | |
| CN109087337B (en) | Long-time target tracking method and system based on hierarchical convolution characteristics | |
| CN111091101B (en) | High-precision pedestrian detection method, system and device based on one-step method | |
| CN106909885A (en) | A kind of method for tracking target and device based on target candidate | |
| Zhu et al. | Tiny object tracking: A large-scale dataset and a baseline | |
| Xiao et al. | MeMu: Metric correlation Siamese network and multi-class negative sampling for visual tracking | |
| Tao et al. | Gap detection of switch machines in complex environment based on object detection and image processing | |
| Lu et al. | An improved target detection method based on multiscale features fusion | |
| CN104537686A (en) | Tracing method and device based on target space and time consistency and local sparse representation | |
| CN110348492A (en) | A kind of correlation filtering method for tracking target based on contextual information and multiple features fusion | |
| CN106897735A (en) | The tracking and device of a kind of Fast Moving Object | |
| Zhang et al. | Spatio-temporal matching for siamese visual tracking | |
| Tang et al. | Foreign object detection for transmission lines based on Swin Transformer V2 and YOLOX | |
| Li et al. | Parallel dual networks for visual object tracking | |
| Jiang et al. | MultiBSP: multi-branch and multi-scale perception object tracking framework based on siamese CNN | |
| CN117152206A (en) | Multi-target long-term tracking method for unmanned aerial vehicle | |
| CN106909934A (en) | A kind of method for tracking target and device based on adaptable search | |
| Feng et al. | Sparse representation combined with context information for visual tracking | |
| CN104182993B (en) | Target tracking method | |
| Yang et al. | An instance segmentation algorithm based on improved mask R-CNN | |
| CN106033550A (en) | Target tracking method and apparatus thereof | |
| CN106372650A (en) | Motion prediction-based compression tracking method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |