CN106570471A - Scale adaptive multi-attitude face tracking method based on compressive tracking algorithm - Google Patents
Scale adaptive multi-attitude face tracking method based on compressive tracking algorithm Download PDFInfo
- Publication number
- CN106570471A CN106570471A CN201610947868.1A CN201610947868A CN106570471A CN 106570471 A CN106570471 A CN 106570471A CN 201610947868 A CN201610947868 A CN 201610947868A CN 106570471 A CN106570471 A CN 106570471A
- Authority
- CN
- China
- Prior art keywords
- face
- tracking
- target face
- algorithm
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
Abstract
The invention discloses a scale adaptive multi-attitude face tracking method based on a compressive tracking algorithm. A target face is roughly positioned by using the compressive tracking algorithm so that the search range of a face detection algorithm can be reduced and the accuracy and the real-time performance of the target face detection algorithm can be enhanced; accurate positioning of the target face is realized by using the face detection algorithm, and scale adaptive tracking of the target face is also realized; the problem of tracking failure caused by the fact that the target face leaves the lens and then enters the lens again can be solved by using the target face detection algorithm; and the problem of tracking continuity under the condition of detection failure of the target face detection algorithm can be realized by using the time continuity of the target face movement process. With application of the method, accurate and effective multi-attitude scale adaptive tracking of the camera for the target face can be guaranteed so that the method can be widely applied to video monitoring, human-computer interaction, virtual reality and various security systems like an ATM monitoring entrance guard system.
Description
Technical field
The present invention relates to image steganalysis field, more particularly to a kind of dimension self-adaption people based on compression track algorithm
Face track algorithm.
Background technology
In recent years, scientific research personnel achieves huge progress in face tracking technology.In the Web-based instruction, video conference, prison
Depending on being required for tracking target face in real time with the specific occasion such as monitoring, data transfer and analysis.Remote teaching, regard
The aspects such as frequency communication, videophone, identity validation, man-machine interaction are all closely bound up with face tracking.
At present, having had, and such as Camshift track algorithms, is covered based on sequence
The particle filter method of special Caro, Mean shift algorithms etc., although these algorithms can be carried out accurately to target face
Tracking, but, when the attitude of target face changes, tracking is easily failed.
To solve the problems, such as to carry out multi-pose Face effectively tracking in real time, Kaihua Zhang propose one kind and are based on
Theoretical compression tracking (Compressive Tracking, the CT) algorithm of compressed sensing (compressive sensing, CS).
CT algorithms solve the problems, such as to carry out real-time tracking to multi-pose Face well, while having operand little, tracking velocity is fast,
Real-time the advantages of, but, there are problems that yardstick can not be adaptive during being tracked to target.Meanwhile, when
After target leaves camera lens, it is again introduced into, tracking failure.
The content of the invention
The technical problem to be solved in the present invention is：For above-mentioned technical problem, propose a kind of based on compression track algorithm
Dimension self-adaption multi-pose Face tracking, effectively realizes that the dimension self-adaption to multi-pose target face is tracked.
The present invention is adopted the following technical scheme that to solve above-mentioned technical problem：
A kind of dimension self-adaption multi-pose Face tracking based on compression track algorithm, it is characterised in that combine face
Detection algorithm and CT algorithms carry out detecting and tracking to target face；First on video display window, target face is examined
Survey, frame selects target face, then start CT algorithms and positioning is tracked to the target face selected by frame；On the basis of CT algorithms
On, then by the Face datection algorithm based on Adaboost and Haar features, target face is accurately positioned.
In above-mentioned technical proposal, including following key step：
Step S1：Photographic head is opened, video data stream is read in, meanwhile, start target person face detection algorithm and realize to video
The detection of target face in stream；
Step S2：On the basis of the target face yardstick for detecting, generation window of the size more than target face is generated
Mouthful；
Step S3：With the window initialization CT algorithms for generating, the Bayes classifier of the positive negative sample of CT algorithms is built, together
When, start CT algorithms, rough tracking is carried out to target face；
Step S4：Using the tracking box of CT algorithms as target face detection window；
Step S5：Judge whether detection window touches the border of video display window, if the judgment is Yes, return to step
S1；Conversely, start the Face datection algorithm based on Adaboost and Haar features in detection window, inside detection window
Target face is detected；
Step S6：Face whether is detected in judging step S5, if the judgment is Yes, is recorded in present frame, target person
The size and location of face, and as display box；Conversely, being bold little and position using target person in the former frame recorded
Put, generate the display box of present frame；
Wherein, each parameter is defined as follows：Detection window：The detection zone of target face, while and CT algorithms tracking
Frame；Tracking box：The tracking box of CT algorithms；Display box：The tracking effect of algorithm final output.
In above-mentioned technical proposal, the detection of step S1 target face refers to specific face is identified from complex background.
In above-mentioned technical proposal, step S2 generates window on the basis of the target face center for detecting, and length is a width of right
Answer target face length and width yardstick 1.2～1.8 times, to ensure accuracy of the CT algorithms to target face tracking.
In above-mentioned technical proposal, start CT algorithms in step S3 carries out rough tracking to target face, including generate with
Machine calculation matrix, the structure of compression tracking Bayes classifier and renewal, are specifically carried out as follows：
Step S31：T-th moment, when t frames picture reads in, is entered by the background to target face and its surrounding
Row sampling, so as to get several target face positive samples and the background negative sample around target face；Then pass through one
Individual sparse calculation matrix aligns negative sample and carries out feature extraction, then with the features training Bayes classifier for extracting, quite
In the Bayes classifier for initializing positive negative sample, it is that the startup of next step CT algorithm is laid the groundwork；
Step S32：When t+1 frames picture reads in, using the position of target face in t frame pictures and size as
Benchmark, is sampled around which, generates n detection block, feature extraction, feature extraction institute are then carried out to this n detection block
Using sparseness measuring matrix and step S1 it is identical to sparseness measuring matrix involved in the detection of target face in video flowing；
Reuse these features that the initialized Bayes classifier of t frames extracts n detection block to classify, what classification was obtained
The window of maximum ratio is tracking box；New target window has been got thus.
In above-mentioned technical proposal, the generating process of n described detection block is：With target face place rectangular area position
The upper left corner be the center of circle, with 4 pixels as radius, choose 45 positive samples；With 8 pixels as inside radius, 25 pixels are outer
Radius chooses 50 negative samples.
In above-mentioned technical proposal, following tool is included based on the Face datection algorithm of Adaboost and Haar features in step S5
Body step：
Step S51：By face with Haar-Like features being described, using the method for integrogram come the spy to face
The eigenvalue levied is calculated；
Step S52：Classified using Adaboost algorithm, so as to select the feature that can most represent face, namely
Then these Weak Classifiers are combined by Haar-Like rectangular characteristic blocks, and then construct a strong classifier；
Step S53：The strong classifier that training is obtained is connected, so as to the stacking for constituting a cascade structure is classified
Device.
In above-mentioned technical proposal, the above-mentioned dimension self-adaption multi-pose Face tracking based on compression track algorithm, base
Visual Studio 2010 under 7 operating systems of windows, and the OpenCV storehouses of increasing income that version is 2.4.4.
In above-mentioned technical proposal, the opening of photographic head, the reading of video and the formation of video frame are all based on
The built-in function in OpenCV storehouses.
During in prior art to target face tracking, although Compression alone track algorithm can realize it is colourful
The continuous tracking of state face, but yardstick is but unable to self adaptation；And simple Face datection algorithm, in detection process, although
The dimension self-adaption to target face tracking can be accomplished, but, due to Face datection algorithm it cannot be guaranteed that each frame can be examined
Target face is measured, tracking picture occurs non-continuous event, and when occurring multiple faces in picture, non-targeted face pair
The tracking of target face be present.It is proposed by the present invention based on compression tracking dimension self-adaption multi-pose Face with
Track algorithm, combining compression track algorithm can carry out continuous, quick, effective tracking and based on Adaboost to multi-pose Face
The Face datection algorithm of learning algorithm can carry out the advantage of quick effective detection to face, and realizing carries out chi to multi-pose Face
Degree adaptive tracing.Coarse localization is carried out to target face using compression track algorithm, so as to reduce Face datection algorithm
Hunting zone, and then improve target person face detection algorithm accuracy and real-time；Using Face datection algorithm, target face is realized
Be accurately positioned, while realize target face dimension self-adaption tracking；Using target person face detection algorithm, target face is solved
When leaving camera lens and being again introduced into, the problem of failure is tracked；Using target face motor process seriality in time, mesh is realized
Tracking continuity problem under mark Face datection algorithm detection failure scenarios.The present invention ensure that photographic head by above method
Accurately and effectively multi-pose is carried out to target face, dimension self-adaption tracking can be widely used in video monitoring, man-machine friendship
Mutually, among virtual reality and various safety-protection systems such as：ATM monitoring gate control system etc..
Description of the drawings
Fig. 1 is the dimension self-adaption multi-pose Face tracking flow chart based on compression track algorithm of the present invention.
Specific embodiment
In order to further illustrate technical scheme, this programme is described in detail below in conjunction with accompanying drawing 1.
The dimension self-adaption multi-pose Face tracking based on compression track algorithm of the present invention as shown in Figure 1, in reality
During existing, first on the display window of video, target face is detected, frame selects target face, then start pressure
Contracting track algorithm is tracked positioning to the target face selected by frame.On the basis of compression track algorithm, then known by face
Other algorithm, is accurately positioned to target face.For convenience of the explanation of problem, now it is defined as follows：
Detection window：The detection zone of target face；
Tracking box：The tracking box of track algorithm；
Display box：The tracking effect of algorithm final output.
In such scheme, the technology platform of needs is the Visual Studio 2010 under 7 operating systems of windows, with
And version is the OpenCV storehouses of increasing income of 2.4.4.
In such scheme, the opening of photographic head, the reading of video, and the formation of video frame, it is all based on OpenCV storehouses
Built-in function.
Face recognition algorithms in such scheme, in the OpenCV storehouses that face recognition algorithms are.
In such scheme, following steps are specifically included：
Step S1：Photographic head is opened, video data stream is read in, meanwhile, start target person face detection algorithm and realize to video
The detection of target face in stream.
Target Face datection refers to specific face is identified from complex background.
Step S2：On the basis of the target face for detecting, a window for being slightly larger in dimension than target face is generated.
To further illustrate the concrete methods of realizing for generating window, following steps are supplemented：
Step S21：Jing tests find that, during being tracked to target, target face is bigger, and tracking is got over for CT algorithms
Accurately.Therefore it is during window is generated, of the invention on the basis of the target face center for detecting, a width of correspondence mesh of length
1.2～1.8 times of mark face length and width yardstick, to ensure accuracy of the CT algorithms to target face tracking.
Step S3：With initialization CT algorithms are generated, the Bayes classifier of the positive negative sample of CT algorithms is built, meanwhile, start
CT algorithms, carry out rough tracking to target face.Compression track algorithm is as follows：
Step S31：When t frames picture reads in, sampled by the background to target face and its surrounding, from
And get several positive samples (target face) and negative sample (background around target face).Then it is sparse by one
Calculation matrix align negative sample and carry out feature extraction (realizing dimensionality reduction), then with the features training Bayes's classification for extracting
Device, equivalent to the Bayes classifier for initializing positive negative sample, is that the startup of next step compression track algorithm is laid the groundwork.
Step S32：When t+1 frames picture reads in, using the position of target face in t frame pictures and size as
Benchmark, is sampled around which, disposably generates n detection block, feature extraction is then carried out to this n detection block and (is adopted
Sparseness measuring matrix with it is identical in step S1).Reusing the initialized Bayes classifier of t frames these features is carried out point
Class, the window of the maximum ratio for obtaining of classifying are target window.New target window has been got thus.
To further illustrate the concrete methods of realizing of compression track algorithm, following steps are supplemented：
Step S321：Generate random measurement matrix
One very typical random measurement matrix is random Gaussian matrix, matrix element meet N (0,1) be distributed.But,
When needing the image space higher to dimension to carry out dimensionality reduction, this matrix can not meet actual demand in sparse degree,
A kind of very sparse mxn (m rows, n row) random measurement matrix R, element r in matrix are employed in CT algorithmsij(represent random to survey
I-th row of moment matrix, jth row)
As s=2,1-1/s=1/2, the element for having 1/2 in that is to say random measurement matrix are zero, and amount of calculation is changed into former
1/2 for coming；
As s=3,1-1/s=2/3, the element for having 2/3 in that is to say random measurement matrix are zero, and amount of calculation is changed into former
2/3 for coming；
By this matrix, amount of calculation just greatly reduces.S=m/4 is have chosen in CT algorithms, matrix R's is each
Row only need to calculate c element value (c represents the number of the sampling block for generating at random, less than 4, generally 2 or 3).So
Its computation complexity is O (cn).In addition, we only need to the nonzero element for storing R, so required memory space is also very
It is few.
Step S322：The structure of compression tracking Bayes classifier and renewal
Grader employed in compression track algorithm is Bayes classifier, at the dimensionality reduction through random measurement matrix
After reason, corresponding eigenvalue is acquired, it is assumed that each element is all independently distributed.The then criteria for classification of grader H (v)
For：
Wherein, y ∈ (0, be 1) classification samples label, the value 0,1 of y represents positive sample and negative sample, v respectivelyiRepresent i-th
Individual given sample, P (vi| y=1) and P (vi| y=0) random sample sheet is represented to respectively for positive sample and the probability of negative sample.It is assumed that two
The priori conditions of class sample are identical, i.e. P (y=0)=0.5=P (y=1).It is random that Diaconis and Freedman demonstrate higher-dimension
The accidental projection of vector is nearly all Gauss distribution.Therefore it is presumed that the conditional probability P (v in grader H (v)i| y=1)
With P (vi| y=0) Gauss distribution is fallen within, and can be with four parametersTo describe：
In formula,Positive negative sample is represented respectively, is above designated as 1 expression positive sample grader, is above designated as 0 expression
Negative sample grader,iRepresent theiIt is individual,WithRepresent that average is respectivelyVariance isGauss distribution and
Average isVariance isGauss distribution.Correspondence is updated for the parameter of Bayes classifier, is had：
In formula：λ is Studying factors, and λ>0.
Step S4：Using the tracking box of CT algorithms as target face detection window.
As the tracking box of CT algorithms is slightly larger in dimension than target face, therefore, when target face is near photographic head, according to
Inside of the target face in detection window is ensure that so, so as to ensure yardstick of the target face during photographic head certainly
Adapt to.
Step S5：Judge whether detection window touches the border of video window, if the judgment is Yes, return to step S1；
Conversely, start the Face datection algorithm based on Adaboost algorithm and Haar-Like features in detection window, to detection window
Internal target face is detected.
By the judgement on border, target face tracking and detection process are reasonably switched, solve CT algorithms with
The tracking Problem of Failure occurred during track target face (during being tracked to target item using CT algorithms, works as mesh
After mark leaves camera lens, when being again introduced into, it may appear that the problem of tracking failure).
To further illustrate the Face datection algorithm based on Adaboost algorithm and Haar-Like features, following walking is supplemented
Suddenly：
Step S51：By face with Haar-Like features being described, using the method for integrogram come the spy to face
The eigenvalue levied is calculated.
Step S52：Classified using Adaboost algorithm (this is equivalent to a weak typing), most can generation so as to select
Then these Weak Classifiers are combined by the feature (Haar-Like rectangular characteristic blocks) of table face, and then construct one by force
Grader.
Step S53：The strong classifier that training is obtained is connected, so as to the stacking for constituting a cascade structure is classified
Device, by way of cascade, can effectively improve the speed of detection.
Step S6：Whether face is detected in judging step S5, if the judgment is Yes, in record present frame, target face
Size and location, and as display box；Conversely, using target person face size and location in the former frame recorded,
Generate the display box of present frame.
The dimension self-adaption multi-pose Face tracking based on compression track algorithm of the present invention, it is characterised in that：Two
Individual algorithm --- the combination of compression tracking (Compressive Tracking) algorithm and Face datection algorithm, advantage are obtained
It is complementary.During algorithm is realized, the Visual Studio 2010 and version under 7 platforms of windows is mainly make use of to be
2.4.4 increase income OpenCV storehouses to realize algorithm.
In due to OpenCV storehouses, based on Adaboost algorithm and the front face of Haar-Like eigenface detection algorithms
Detection angles scope substantially [- 20., 20.], when the deflection angle of face exceedes this scope, detection can be failed, and then make
Process must be tracked to occur discontinuously.
It is that face deflection angle is excessive in view of the main cause of detection failure, and in deflection, face is relative to mirror
The distance of head does not occur more significantly to change, and therefore, it can the size of target face in former frame as in present frame
The size of target face.And the change that CT algorithms have preferable robustness, i.e. attitude for the attitudes vibration of target face is not
Tracking effect can be significantly had influence on more, thus it can be assumed that in adjacent two field pictures, target item and tracking box central point
Between relative position be constant, and then by former frame, the relative position between the central point of target item and tracking box, with
And in present frame tracking box central point position determining the position of target item in present frame.
To sum up, the dimension self-adaption multi-pose Face tracking based on compression track algorithm proposed by the invention, profit
Can be real-time with compression track algorithm, quickly target face is tracked, then on the basis of compression tracking, people is added
Face detection algorithm, detects to target face, so as to realize tracking.The algorithm is substantially overcomed compression track algorithm application
During face tracking, yardstick can not adaptive problem and single Face datection algorithm can not exclude non-targeted people in picture
Interference of the face to target face tracking, and during single Face datection realizes tracking, due to being not detected by target face
So that the discontinuous problem of tracking.The proposition of the present invention so that the tracking of target face is simpler quick in photographic head, has
Good expansion and practicality.
Claims (9)
1. it is a kind of based on the dimension self-adaption multi-pose Face tracking for compressing track algorithm, it is characterised in that to examine with reference to face
Method of determining and calculating and CT algorithms carry out detecting and tracking to target face；First on video display window, target face is detected,
Frame selects target face, then starts CT algorithms and is tracked positioning to the target face selected by frame；On the basis of CT algorithms,
Again by the Face datection algorithm based on Adaboost and Haar features, target face is accurately positioned.
2. according to claim 1 based on the dimension self-adaption multi-pose Face tracking for compressing track algorithm, which is special
Levy is to include following key step：
Step S1：Photographic head is opened, video data stream is read in, meanwhile, start target person face detection algorithm and realize in video flowing
The detection of target face；
Step S2：On the basis of the target face yardstick for detecting, generation window of the size more than target face is generated；
Step S3：With the window initialization CT algorithms for generating, the Bayes classifier of the positive negative sample of CT algorithms is built, meanwhile, open
Dynamic CT algorithms, carry out rough tracking to target face；
Step S4：Using the tracking box of CT algorithms as target face detection window；
Step S5：Judge whether detection window touches the border of video display window, if the judgment is Yes, return to step S1；
Conversely, start the Face datection algorithm based on Adaboost and Haar features in detection window, to the mesh inside detection window
Mark face is detected；
Step S6：Face whether is detected in judging step S5, if the judgment is Yes, is recorded in present frame, target face
Size and location, and as display box；Conversely, using target person face size and location in the former frame recorded, it is raw
Into the display box of present frame；
Wherein, each parameter is defined as follows：Detection window：The detection zone of target face, while and CT algorithms tracking box；With
Track frame：The tracking box of CT algorithms；Display box：The tracking effect of algorithm final output.
3. according to claim 2 based on the dimension self-adaption multi-pose Face tracking for compressing track algorithm, which is special
Levy is that the detection of step S1 target face refers to specific face is identified from complex background.
4. according to claim 2 based on the dimension self-adaption multi-pose Face tracking for compressing track algorithm, which is special
It is that step S2 generates window on the basis of the target face center for detecting to levy, and grows a width of correspondence target face length and width yardstick
1.2～1.8 times, to ensure accuracy of the CT algorithms to target face tracking.
5. according to claim 2 based on the dimension self-adaption multi-pose Face tracking for compressing track algorithm, which is special
It is to start CT algorithms in step S3 to carry out target face rough tracking to levy, including generate random measurement matrix, compression with
The structure of track Bayes classifier and renewal, are specifically carried out as follows：
Step S31：T-th moment, when t frames picture reads in, is adopted by the background to target face and its surrounding
Sample, so as to get several target face positive samples and the background negative sample around target face；Then it is dilute by one
Thin calculation matrix aligns negative sample and carries out feature extraction, then with the features training Bayes classifier for extracting, equivalent to first
The Bayes classifier of the positive negative sample of beginningization, is that the startup of next step CT algorithm is laid the groundwork；
Step S32：When t+1 frames picture reads in, using the position of target face in t frame pictures and size as base
Standard, is sampled around which, is generated n detection block, is then carried out feature extraction to this n detection block, and feature extraction is adopted
Sparseness measuring matrix is identical to sparseness measuring matrix involved in the detection of target face in video flowing with step S1；Again
These features extracted to n detection block using the initialized Bayes classifier of t frames are classified, and classification is obtained most
The window of vast scale is tracking box；New target window has been got thus.
6. according to claim 5 based on the dimension self-adaption multi-pose Face tracking for compressing track algorithm, which is special
Levy is that the generating process of n described detection block is：The upper left corner with target face place rectangular area position as the center of circle, with
4 pixels are radius, choose 45 positive samples；With 8 pixels as inside radius, 25 pixels are that outer radius chooses 50 negative samples
This.
7. according to claim 2 based on the dimension self-adaption multi-pose Face tracking for compressing track algorithm, which is special
Levy is to be comprised the following specific steps that based on the Face datection algorithm of Adaboost and Haar features in step S5：
Step S51：By face with Haar-Like features being described, using the method for integrogram come the feature to face
Eigenvalue is calculated；
Step S52：Classified using Adaboost algorithm, so as to select the feature that can most represent face, namely Haar-
Then these Weak Classifiers are combined by Like rectangular characteristic blocks, and then construct a strong classifier；
Step S53：The strong classifier that training is obtained is connected, so as to constitute the cascade filtering of a cascade structure.
8. according to claim 2 based on the dimension self-adaption multi-pose Face tracking for compressing track algorithm, which is special
It is the above-mentioned dimension self-adaption multi-pose Face tracking based on compression track algorithm to levy, and based on the operations of windows 7 is
Visual Studio 2010 under system, and the OpenCV storehouses of increasing income that version is 2.4.4.
9. according to claim 2 based on the dimension self-adaption multi-pose Face tracking for compressing track algorithm, which is special
It is the opening of photographic head, the reading of video and the formation of video frame to levy, and is all based on the built-in function in OpenCV storehouses.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610947868.1A CN106570471B (en) | 2016-10-26 | 2016-10-26 | Dimension self-adaption multi-pose Face tracking based on compression track algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610947868.1A CN106570471B (en) | 2016-10-26 | 2016-10-26 | Dimension self-adaption multi-pose Face tracking based on compression track algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106570471A true CN106570471A (en) | 2017-04-19 |
| CN106570471B CN106570471B (en) | 2019-04-26 |
Family
ID=58535097
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610947868.1A Active CN106570471B (en) | 2016-10-26 | 2016-10-26 | Dimension self-adaption multi-pose Face tracking based on compression track algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106570471B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108805094A (en) * | 2018-06-19 | 2018-11-13 | 合肥工业大学 | Data enhancement methods based on artificial face |
| CN109271883A (en) * | 2018-08-28 | 2019-01-25 | 武汉科技大学 | A kind of method for tracking target merging study mechanism |
| CN109990893A (en) * | 2017-12-29 | 2019-07-09 | 浙江宇视科技有限公司 | Photometric parameter calculation method, device and terminal device |
| CN110705478A (en) * | 2019-09-30 | 2020-01-17 | 腾讯科技（深圳）有限公司 | Face tracking method, device, equipment and storage medium |
| CN111259907A (en) * | 2020-03-12 | 2020-06-09 | Oppo广东移动通信有限公司 | Content identification method and device and electronic equipment |
| CN113591607A (en) * | 2021-07-12 | 2021-11-02 | 辽宁科技大学 | Station intelligent epidemic prevention and control system and method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103905733A (en) * | 2014-04-02 | 2014-07-02 | 哈尔滨工业大学深圳研究生院 | Method and system for conducting real-time tracking on faces by monocular camera |
-
2016
- 2016-10-26 CN CN201610947868.1A patent/CN106570471B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103905733A (en) * | 2014-04-02 | 2014-07-02 | 哈尔滨工业大学深圳研究生院 | Method and system for conducting real-time tracking on faces by monocular camera |
Non-Patent Citations (3)
| Title |
|---|
| 刘晴 等: "尺度自适应的多模型压缩跟踪算法", 《系统工程与电子技术》 * |
| 熊池亮: "基于Adaboost算法的驾驶员疲劳驾驶检测", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
| 钟权 等: "一种改进的实时压缩跟踪算法", 《光电工程》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109990893A (en) * | 2017-12-29 | 2019-07-09 | 浙江宇视科技有限公司 | Photometric parameter calculation method, device and terminal device |
| CN109990893B (en) * | 2017-12-29 | 2021-03-09 | 浙江宇视科技有限公司 | Photometric parameter calculation method and device and terminal equipment |
| CN108805094A (en) * | 2018-06-19 | 2018-11-13 | 合肥工业大学 | Data enhancement methods based on artificial face |
| CN109271883A (en) * | 2018-08-28 | 2019-01-25 | 武汉科技大学 | A kind of method for tracking target merging study mechanism |
| CN110705478A (en) * | 2019-09-30 | 2020-01-17 | 腾讯科技（深圳）有限公司 | Face tracking method, device, equipment and storage medium |
| CN111259907A (en) * | 2020-03-12 | 2020-06-09 | Oppo广东移动通信有限公司 | Content identification method and device and electronic equipment |
| WO2021179856A1 (en) * | 2020-03-12 | 2021-09-16 | Oppo广东移动通信有限公司 | Content recognition method and apparatus, electronic device, and storage medium |
| CN111259907B (en) * | 2020-03-12 | 2024-03-12 | Oppo广东移动通信有限公司 | Content identification method and device and electronic equipment |
| CN113591607A (en) * | 2021-07-12 | 2021-11-02 | 辽宁科技大学 | Station intelligent epidemic prevention and control system and method |
| CN113591607B (en) * | 2021-07-12 | 2023-07-04 | 辽宁科技大学 | Station intelligent epidemic situation prevention and control system and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106570471B (en) | 2019-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109819208B (en) | Intensive population security monitoring management method based on artificial intelligence dynamic monitoring | |
| CN106570471A (en) | Scale adaptive multi-attitude face tracking method based on compressive tracking algorithm | |
| Mahmood et al. | WHITE STAG model: Wise human interaction tracking and estimation (WHITE) using spatio-temporal and angular-geometric (STAG) descriptors | |
| CN109934176B (en) | Pedestrian recognition system, recognition method, and computer-readable storage medium | |
| Yogameena et al. | Computer vision based crowd disaster avoidance system: A survey | |
| Thome et al. | A real-time, multiview fall detection system: A LHMM-based approach | |
| Lavi et al. | Survey on deep learning techniques for person re-identification task | |
| Harville et al. | Fast, integrated person tracking and activity recognition with plan-view templates from a single stereo camera | |
| US9569531B2 (en) | System and method for multi-agent event detection and recognition | |
| Lu et al. | Face detection and recognition algorithm in digital image based on computer vision sensor | |
| Sikandar et al. | ATM crime detection using image processing integrated video surveillance: a systematic review | |
| CN109711344A (en) | A kind of intelligentized specific exceptions behavioral value method in front end | |
| Gualdi et al. | Contextual information and covariance descriptors for people surveillance: an application for safety of construction workers | |
| CN106326851B (en) | A kind of method of number of people detection | |
| Jiang et al. | A tree-based approach to integrated action localization, recognition and segmentation | |
| Badave et al. | Head pose estimation based robust multicamera face recognition | |
| CN116994209A (en) | Image data processing system and method based on artificial intelligence | |
| Vo et al. | An effective approach for human actions recognition based on optical flow and edge features | |
| Izo et al. | Unsupervised modeling of object tracks for fast anomaly detection | |
| CN112541403A (en) | Indoor personnel falling detection method utilizing infrared camera | |
| Gualdi et al. | Covariance descriptors on moving regions for human detection in very complex outdoor scenes | |
| Archana et al. | Face recognition using Deep Neural Network and Support Based Anchor Boxes Methods in video frames | |
| Zou et al. | Occupancy measurement by object tracking at building entrances | |
| Vinay et al. | Dominant feature based convolutional neural network for faces in videos | |
| Ji et al. | Visual-based view-invariant human motion analysis: A review |
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 |