CN109064489A - Method, apparatus, equipment and medium for face tracking - Google Patents
Method, apparatus, equipment and medium for face tracking Download PDFInfo
- Publication number
- CN109064489A CN109064489A CN201810786024.2A CN201810786024A CN109064489A CN 109064489 A CN109064489 A CN 109064489A CN 201810786024 A CN201810786024 A CN 201810786024A CN 109064489 A CN109064489 A CN 109064489A
- Authority
- CN
- China
- Prior art keywords
- trace regions
- camera lens
- lens frame
- frame
- region
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000001514 detection method Methods 0.000 claims abstract description 70
- 230000004044 response Effects 0.000 claims abstract description 26
- 238000007689 inspection Methods 0.000 claims description 23
- 238000012545 processing Methods 0.000 claims description 19
- 238000009499 grossing Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 description 38
- 238000010586 diagram Methods 0.000 description 13
- 230000008569 process Effects 0.000 description 12
- 238000013528 artificial neural network Methods 0.000 description 7
- 238000012360 testing method Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000009825 accumulation Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000002035 prolonged effect Effects 0.000 description 3
- 230000009885 systemic effect Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 230000001815 facial effect Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G06T5/70—
-
- 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
-
- 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
- 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
- G06V40/168—Feature extraction; Face representation
-
- 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
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Abstract
According to this disclosure relates to a kind of method, apparatus, equipment and medium for face tracking.According to one embodiment, the method for face tracking is provided.In the method, target face is detected in the first video frame of video sequence to determine the first human face region；Based on the first human face region, trace regions are calculated in the second video frame in the video sequence, trace regions instruction includes the candidate region of target face；Determine that detection zone, detection zone include at least trace regions in the second video frame based on trace regions；Detection target face is in the detection area to determine the second human face region；And trace regions are matched in response to the second human face region, determination is tracked successfully.
Description
Technical field
This disclosure relates to field of image processing, and in particular to method, apparatus, the equipment being tracked to the face in video
And medium.
Background technique
In traditional live teaching broadcast classroom, teacher's personalization difficult to realize to student in direct broadcasting room is paid close attention to.Teacher
Teacher is typically limited to the interaction forms of student and proposes problem, is then answered jointly by whole students.This live teaching broadcast class
Hall is difficult to realize call the roll and answer a question, and the process answered a question of being more difficult to will to call the roll embodies in live streaming.
Summary of the invention
Based on the above issues, it according to the example embodiment of present disclosure, provides a kind of for real-time tracing face
Scheme.
In the first aspect of present disclosure, a kind of method for tracking face is provided.Specifically, this method packet
It includes: detecting target face in the first video frame of video sequence to determine the first human face region；Based on the first human face region,
Trace regions are calculated in the second video frame in video sequence, trace regions instruction includes the candidate region of target face；It is based on
Trace regions determine that detection zone, detection zone include at least trace regions in the second video frame；It detects in the detection area
Target face is to determine the second human face region；And trace regions are matched in response to the second human face region, determination is tracked successfully.
By carrying out real-time inspection to trace regions, the accumulation of error in tracing algorithm can be effectively eliminated, to avoid
Tracking target is lost, tracking precision is improved.
In the certain embodiments of the disclosure, determine that tracking successfully includes: to compare the second human face region and trace regions
To determine the first offset；And in response to the first offset not higher than threshold value is examined, determination is tracked successfully.
In the certain embodiments of the disclosure, this method further comprises: tracking successfully in response to determining, is based on first
Human face region trace regions calculate in the third video frame of video sequence and determine that new trace regions, new trace regions refer to
Show the candidate region including target face.
In the certain embodiments of the disclosure, this method further include: be higher than in response to the first offset and examine threshold value, determine
Tracking failure；And in response to determining tracking failure, it is based on the second human face region, is calculated in the third video frame of video sequence
Determine new trace regions, new trace regions instruction includes the candidate region of target face.
In the certain embodiments of the disclosure, this method further include: camera lens frame is set based on trace regions, camera lens frame is used
The image-region including target face in display video sequence.
In the certain embodiments of the disclosure, trace regions are defined as B 'track=(x, y, w, h), wherein x, y are to chase after
The lower-left angular coordinate in track region, w are the width of trace regions, and h is the height of trace regions；Be provided with camera lens frame include: by
Camera lens frame B 'camIt is set as B'cam=(x-x0,y-y0, W, H), wherein x0、y0For predefined deviant, W is predefined mirror
The width of head frame, H are the height of predefined camera lens frame, wherein W > w and H > h.
In the certain embodiments of the disclosure, this method further include: compare camera lens frame and for the elder generation in video sequence
The first front lens frame of preceding video frame is to determine the second offset；And it is not higher than smooth threshold value in response to the second offset, it is based on camera lens
Frame shows video sequence；And it is higher than smooth threshold value in response to the second offset, camera lens frame is smoothed.
In the certain embodiments of the disclosure, smoothing processing includes: to obtain inserting between camera lens frame and first front lens frame
Value；And camera lens frame is arranged based on interpolation.
In the certain embodiments of the disclosure, trace regions are defined as B 'track=(x, y, w, h), x, y are tracing Area
The coordinate in the lower left corner in domain, w are the width of trace regions, and h is the height of trace regions；Wherein based on trace regions in the second view
Determine that detection zone includes: that will test region R ' in frequency framedetectIt is set as R 'detect=(x-w, y-h, 3*w, 3*h).
In the certain embodiments of the disclosure, the coordinate at the center of the second human face region is (X, Y), the center of trace regions
Coordinate is (X ', Y '), wherein determining that the first offset includes: to set the first offset d to
In the certain embodiments of the disclosure, the lower-left angular coordinate of camera lens frame is (X ', Y '), and the lower left corner of first front lens frame is sat
It is designated as (X, Y), wherein determining that the second offset includes: to set the second offset s to
In the certain embodiments of the disclosure, obtaining interpolation includes: by interpolation B "camIt is set asWherein W is the width of predefined camera lens frame
Degree, H are the height of predefined camera lens frame.
In in the second aspect of the present disclosure, provide a kind of for tracking the device of face.Specifically, the device packet
Include: detection module, the detection module are configured as detecting target face in the first video frame of video sequence to determine first
Human face region；Computing module, the computing module are configured as based on the first human face region, the second video frame in the video sequence
Middle calculating trace regions, trace regions instruction include the candidate region of target face；Inspection module, the inspection module are configured
Are as follows: determine that detection zone, detection zone include at least trace regions in the second video frame based on trace regions；And it is detecting
Target face is detected in region to determine the second human face region；Determining module, determining module are configured to respond to the second face
Region Matching is tracked successfully in trace regions, determination.
In the certain embodiments of the disclosure, determine that tracking successfully includes: to compare the second human face region and trace regions
To determine the first offset；And in response to the first offset not higher than threshold value is examined, determination is tracked successfully.
In the certain embodiments of the disclosure, which further comprises: tracking successfully in response to determining, based on tracking
Region calculates new trace regions in the third video frame of video sequence, and new trace regions instruction includes target face
Candidate region.
In the certain embodiments of the disclosure, the device further include: be higher than in response to the first offset and examine threshold value, determine
Tracking failure；And in response to determining tracking failure, it is based on the second human face region, is calculated in the third video frame of video sequence
New trace regions, new trace regions instruction include the candidate region of target face.
In the certain embodiments of the disclosure, the device further include: display module, display module are configured as being based on chasing after
Camera lens frame is arranged in track region, and camera lens frame is used to show the image-region including target face in video sequence.
In the certain embodiments of the disclosure, trace regions are defined as B 'track=(x, y, w, h), wherein x, y are to chase after
The lower-left angular coordinate in track region, w are the width of trace regions, and h is the height of trace regions；Be provided with camera lens frame include: by
Camera lens frame B 'camIt is set as B'cam=(x-x0,y-y0, W, H), wherein x0、y0For predefined deviant, W is predefined mirror
The width of head frame, H are the height of predefined camera lens frame, wherein W > w and H > h.
In the certain embodiments of the disclosure, wherein display module further includes smooth submodule, and smooth submodule is matched
It is set to: comparing camera lens frame with the first front lens frame for the preceding video frame in video sequence to determine the second offset；And it rings
It should be not higher than smooth threshold value in the second offset, video sequence is shown based on camera lens frame；And it is higher than in response to the second offset smooth
Threshold value is smoothed camera lens frame.
In the certain embodiments of the disclosure, smoothing processing includes: to obtain inserting between camera lens frame and first front lens frame
Value；And camera lens frame is arranged based on interpolation.
In the certain embodiments of the disclosure, trace regions are defined as B 'track=(x, y, w, h), x, y are tracing Area
The coordinate in the lower left corner in domain, w are the width of trace regions, and h is the height of trace regions；Wherein based on trace regions in the second view
Determine that detection zone includes: that will test region R ' in frequency framedetectIt is set as R 'detect=(x-w, y-h, 3*w, 3*h).
In the certain embodiments of the disclosure, the coordinate at the center of the second human face region is (X, Y), the center of trace regions
Coordinate is (X ', Y '), wherein determining that the first offset includes: to set the first offset d to
In the certain embodiments of the disclosure, the lower-left angular coordinate of camera lens frame is (X ', Y '), and the lower left corner of first front lens frame is sat
It is designated as (X, Y), wherein determining that the second offset includes: to set the second offset s to
In the certain embodiments of the disclosure, wherein obtaining interpolation includes: by interpolation B "camIt is set asWherein W is the width of predefined camera lens frame
Degree, H are the height of predefined camera lens frame.
In the third aspect of present disclosure, a kind of equipment, including one or more processors are provided；And storage
Device, for storing one or more programs, when one or more programs are executed by one or more processors so that one or
The method that multiple processors realize the first aspect according to present disclosure.
In the fourth aspect of present disclosure, a kind of computer-readable medium is provided, is stored thereon with computer journey
Sequence realizes the method for the first aspect according to present disclosure when the program is executed by processor.
The disclosure has carried out quality inspection while tracking target, to tracking each time, and in time to error bigger than normal
It is corrected, thus accumulation of error when preventing tracking, and then target loss is avoided, realize higher for a long time chase after
Track precision.In addition, the operational model that lightweight is used only in the disclosure cooperates algorithm policy, energy in the case where height tracks precision
Enough accomplish that the frame per second of average 15FPS or so carries out video output, has achieved the effect that real-time tracing.
In contrast to the method for other application deep learning, the disclosure also has very big advantage in application aspect.This public affairs
Opening quickly to be disposed on CPU, and most neural network models need GPU to carry out deployment and operation, this just subtracts
The lower deployment cost of model is lacked.In addition, the disclosure can handle the low clear video flowing of high definition simultaneously, thus can also quick portion
It affixes one's name in application scenarios.
It should be appreciated that content described in Summary is not intended to limit the embodiment of present disclosure
Crucial or important feature, it is also non-for limiting the scope of the disclosure.The other feature of present disclosure will be by below
Description is easy to understand.
Detailed description of the invention
It refers to the following detailed description in conjunction with the accompanying drawings, it is the above and other feature of each embodiment of present disclosure, excellent
Point and aspect will be apparent.In the accompanying drawings, the same or similar appended drawing reference indicates the same or similar element,
In:
Fig. 1 schematically shows the treatment process block diagram of the method for face tracking according to disclosure embodiment；
Fig. 2 schematically shows the treatment process frame of the method for face tracking according to another embodiment of the disclosure
Figure；
Fig. 3 schematically shows the procedural block diagram of smoothing processing；
Fig. 4 shows the schematic diagram of camera lens frame offset；
Fig. 5 diagrammatically illustrates the block diagram of the device for face tracking according to disclosure embodiment；
Fig. 6 diagrammatically illustrates the block diagram of the device for face tracking according to another embodiment of the disclosure；
Fig. 7 shows the block diagram that can implement the calculating equipment of multiple embodiments of the disclosure.
Specific embodiment
The embodiment of present disclosure is more fully described below with reference to accompanying drawings.Although showing the disclosure in attached drawing
The certain embodiments of content, it should be understood that, present disclosure can be realized by various forms, and not answered
This is construed as limited to embodiments set forth herein, and providing these embodiments on the contrary is for more thorough and complete geography
Solve present disclosure.It should be understood that the being given for example only property of drawings and the embodiments of present disclosure acts on, it is not intended to
Limit the protection scope of present disclosure.
In the description of the embodiment of present disclosure, term " includes " and its similar term should be understood as opening
Include, i.e., " including but not limited to ".Term "based" should be understood as " being based at least partially on ".Term " embodiment "
Or " embodiment " should be understood as " at least one embodiment ".Term " first ", " second " etc. may refer to difference
Or identical object.Hereafter it is also possible that other specific and implicit definition.
Technology in field of image processing based on face tracking helps to solve the problems, such as above-mentioned live teaching broadcast.Closely
Over a little years, with the development of deep learning, it is based on correlation filtering (Correlation Filter) and deep neural network
The object tracking technology of (Deep Neural Network, DNN) emerges one after another.This kind of technologies pass through convolutional neural networks
(Convolutional Neural Network, CNN) or Recognition with Recurrent Neural Network (Recurrent Neural Network,
RNN the target signature in image) is extracted, is then analyzed by correlation filtering, to obtain the location of target.Pass through
Richer characteristics of image is obtained, this kind of method for tracing achieves very high tracking precision.However, deep neural network needs
A large amount of calculation amount is consumed, and needs the plenty of time to initialize and more new model, thus is difficult to accomplish in application scenarios
Tracking in real time, needs processed offline video.
In addition, all there is a common question in the above-mentioned method based on computer vision technique, i.e., as the tracking time increases
Long, tracking error can be accumulated constantly, to eventually lose tracking target after the accurate tracking for completing a period of time.And
And in these methods, due to lacking itself correction mechanism, target is difficult to restore again once be lost, so that can not
Guarantee prolonged tracking.
In traditional method for tracing, tracing process needs to analyze the correlation of target.With prolonged tracking, error meeting
Constantly accumulation eventually leads to target loss.In view of the deficiency in above-mentioned technical proposal, the present disclosure proposes improved face trackings
Method and device.It is provided with the detection zone with tracking result real-time change in the disclosure, by detection zone
The face that face is detected and be will test is compared with tracking result to realize checking function, to eliminate every time
The error that may be accumulated in tracing computation.On the basis of not increasing excessive calculation amount, the disclosure can effectively avoid tracking mesh
Target is lost, and then realizes prolonged real-time tracking.In addition, the disclosure can also to display tracking result camera lens carry out it is flat
Sliding processing avoids the shake of tracking camera lens during live streaming, improves the effect of real-time tracing.
Embodiment of the present disclosure relates generally to carry out real-time tracing to the face in video.But it is to be understood that this
The open tracking for being not limited to be used only for face, it is all to be needed defined by disclosure appended claims spirit and scope
The target being tracked is considered as belonging to the protection scope of the disclosure.
Fig. 1 diagrammatically illustrates the treatment process frame of the method 100 for face tracking according to disclosure embodiment
Figure.As shown in Figure 1, the treatment process of this method 100 may include:
At frame 101, after face tracking is activated, target face is detected in the current video frame of video sequence with true
Determine the human face region in current video frame.
Specifically, using face recognition algorithms, system can detecte the position of all faces in current video frame, then
The face that can be will test is compared with target face information gathered in advance, to obtain the position of target face.
In this embodiment, Face datection algorithm can be existing algorithm, such as the identification based on human face characteristic point
It algorithm, the recognizer based on whole picture facial image, the recognizer based on template and is identified using neural network
Algorithm.
System can store target face information gathered in advance.The target face information of the storage can be by other
Picture or video obtain, and can also be obtained and carrying out Image Acquisition to live video before tracking starting.Target face
Information can also be obtained from other non-image sources.Target face can be corresponding with ID, as long as so that acquisition ID can
With the corresponding target face information of determination.
The position of the target face obtained by face recognition algorithms is human face region.It can be right according to the human face region
System initializes parameter to be used.For example, the human face region detected according to this, can chase after to what is referred to below
Trace regions in track algorithm are initialized, and can also be initialized to the detection zone and camera lens frame referred to below.
Initialization can be completed in frame 101, can also set up other frame and complete, and can also use ginseng according to the use of parameter
It is completed in several respective blocks.
In an embodiment of the disclosure, human face region, trace regions, Face datection region and camera lens frame are defined
Region can be square frame-shaped, be also possible to other any proper shapes.The size of human face region can be according to detecting
The size of face and corresponding change.The size of trace regions, Face datection region and camera lens frame can be fixed and invariable, can also
According to circumstances and suitably to change.What face frame below and tracking frame respectively referred to generation is human face region and trace regions.
In an embodiment of the disclosure, initialization tracking frame may include: according to face frame IfaceCoordinate and
Size determines that the tracking frame of initialization, such as the tracking frame of initialization have coordinate identical with face frame and size.Tracking
Frame can be defined as Btrack=(x, y, w, h), wherein x, y are box lower left corner apex coordinate, and w is the width of box, and h is box
Height, camera lens frame and detection zone below refer to this definition.
It initializes detection zone and camera lens frame includes: according to the tracking frame B after initializationtrack=(x, y, w, h), to determine
The detection zone R of initializationdetect=(x-w, y-h, 3*w, 3*h), wherein the coordinate of detection zone and size are only exemplary
, it also can have other coordinate positions and width and height, such as the height or the width of 6*w of the width of 4*w, 4*h, 5*h
It is high；And according to the tracking frame B after initializationtrack=(x, y, w, h) determines camera lens frame Bcam=(x-x0,y-y0, W, H),
Middle x0、y0For predefined deviant, W is the width of predefined camera lens frame, and H is the height of predefined camera lens frame, such as x0
=y0=200, W=500, H=600 (digital representation pixel therein), the width and height of usual camera lens frame are all larger than and chase after
The height and width of track frame.
At frame 102, based on the human face region detected in frame 101, second video frame in the video sequence
Middle calculating trace regions, trace regions instruction include the candidate region of target face.
Specifically, it is being based on face frame IfaceCoordinate and size the tracking frame B of initialization has been determinedtrackLater, will under
One video frame and the initialization tracking frame determined based on face frame are input in tracing algorithm, and second video frame is calculated
In tracking result (i.e. trace regions or tracking frame, be used to indicate the candidate region including target face).
Wherein above-mentioned tracing algorithm can be existing face tracking algorithm, such as KCF (Kernelized
Correlation Filter, i.e. coring correlation filter) algorithm.
At frame 103, the calculated trace regions of tracing algorithm are based on, detection is determined in second above-mentioned video frame
Region, the detection zone have included at least calculated trace regions.
For example, being B ' based on the calculated trace regions of tracing algorithmtrack=(x ', y ', w, h), detection zone can be by
It is determined as R 'detect=(x '-w, y '-h, 3*w, 3*h).Wherein the coordinate position and width of detection zone and height are only shown
Example property, it also can have other coordinate positions appropriate and width and height.But identified detection zone is usual
It should include above-mentioned trace regions.
At frame 104, the target face human face region new with determination is detected in identified detection zone.Specifically,
The new human face region about target face is detected in the detection zone determined by previous frame using face recognition algorithms.
It will be understood that face recognition algorithms used in the frame can be calculated with the recognition of face used in frame 101
Method is identical.However, the new face recognition algorithms different from frame 101 also can be used in the frame.Due to needing to carry out in the frame
The image-region of detection is smaller, and the face that may include is less (if detection zone restriction is suitable, can to only have target face
In detection zone), therefore the face recognition algorithms that the frame uses can be more simplified compared to algorithm used in frame 101, from
And it reduces calculation amount and saves holding time.
At frame 105 and frame 107, the trace regions in the new human face region detected in frame 104 and frame 102 are judged
Whether match, to determine whether tracking succeeds；If tracked successfully, based on the trace regions in frame 102, in video sequence
New trace regions are calculated in third video frame, if tracking failure, based on the new human face region in frame 104, in video
New trace regions are calculated in the third video frame of sequence.
Specifically, system obtains the face area detected in the central point between region to be matched, such as frame 104 first
The centre coordinate in domain is (X, Y), and the coordinate at the center of the trace regions in frame 102 is (X ', Y ')；Then, two centers are calculated
Offset between coordinate Then, by the inspection threshold value of offset d and systemic presupposition
It compares, if offset d can be determined and be tracked successfully not higher than threshold value is examined, whereas if offset d, which is higher than, examines threshold value,
It can then determine tracking failure.
If tracked successfully, by calculated trace regions in next video frame, i.e. third video frame and frame 102
It is input to and carries out continuing to track in the tracing algorithm of frame 102, and continuing to execute frame 103, frame 104 and frame 105 later.
If tracking failure, will be determined as what initial detecting arrived in the human face region that detection zone detects in frame 104
Human face region, and according to new human face region initialization tracking frame, camera lens frame and detection zone, and frame is being continued to execute later
102, frame 103, frame 104 and frame 105.
Other than frame 101- frame 105, embodiment of the present disclosure can also include output tracking result to be shown
Frame (not shown in figure 1).Specifically, camera lens frame, tracking frame here can be arranged based on tracking frame calculated in frame 102
The mode that camera lens frame is arranged is identical with the frame setting mode of camera lens frame is tracked in above-mentioned initialization, therefore repeats no more.Then will
Identified camera lens frame corresponds in video frame, and intercepts corresponding image, then shows corresponding image.
The embodiment in tracing process simultaneously monitoring, tracing quality, before tracking out active, to tracing algorithm
The case where being corrected, to guarantee being not in pursuing missing in tracing process.
Fig. 2 shows handle block diagram according to the process of the method 200 for face tracking of another embodiment of the disclosure.
The difference is that, the frame for flow display camera lens is increased during the embodiment with above description embodiment
206。
As shown in Fig. 2, frame 206 is arranged between frame 102 and frame 103.In frame 101, camera lens frame is initialized,
Wherein according to the tracking frame B after initializationtrack=(x, y, w, h) determines camera lens frame Bcam=(x-x0,y-y0,W,H).Camera lens
Frame is used to show the image-region including target face in video sequence.Frame 206 is mainly used for the mobile progress to camera lens frame
Flow display.
The treatment process of frame 206 is shown in Fig. 3:
At frame 2061, it is inclined to determine with the first front lens frame for the preceding video frame in video sequence to compare camera lens frame
It moves.
If the coordinate in the current lens frame lower left corner is (X ', Y '), the lower left corner of the first front lens frame of preceding video frame
Coordinate is (X, Y), then can determine offset are as follows:
At frame 2062 and frame 2064, offset is compared with the smooth threshold value of systemic presupposition.
Fig. 4 shows the schematic diagram of camera lens frame offset 400.As shown in figure 4, B 'camIndicate current camera lens frame, BcamIt indicates
Previous camera lens frame, s are indicated from previous camera lens frame to the offset of current camera lens frame.If deviating s is not higher than smooth threshold value,
Video sequence is then shown based on camera lens frame, whereas if offset s is higher than smooth threshold value, then camera lens frame will be smoothed.
For example, smooth threshold value can be 10 (pixels).
The smoothing processing includes the interpolation obtained between camera lens frame and first front lens frame.For example, the interpolation can beThen, camera lens is arranged based on interpolation
Frame.
Fig. 3 is returned, at frame 2063, output tracking result.Specifically, it is obtained after being handled by frame 2062 and frame 2064
Final camera lens frame B "camIt corresponds in video frame, and intercepts corresponding image.Corresponding image is output to front end, thus aobvious
Show the result of tracking.
There is no the relationships between processing frame and frame for conventional method for tracing, thus tracing process will appear many jumps
Point, i.e. tracking quickly jump to another point from a point.This problem is exactly mirror during being embodied in real-time tracing
The fast jitter of head, can be greatly reduced the effect of real-time tracing.In addition to applying real-time tracing in another embodiment of the disclosure
Except method is tracked, gliding smoothing processing also is carried out to tracking camera lens, so as to be more close to true man true for tracing process
The tracking effect of real camera.
Fig. 5 diagrammatically illustrates the block diagram of the device 500 for face tracking according to disclosure embodiment.Specifically
Ground, the device 500 include detection module 501, computing module 502, inspection module 503, determining module 504 and display module 505.
Modules will be successively described in detail below.
In an embodiment of the disclosure, detection module 501 is configured as: after face tracking is activated, being regarded
Target face is detected in the current video frame of frequency sequence to determine the human face region in current video frame.
In an embodiment of the disclosure, detection module can execute face recognition algorithms, work as so as to detect
The position of all faces in preceding video frame, the face that then can be will test and target face information gathered in advance carry out
Compare, and then obtains the position of target face.
In an embodiment of the disclosure, Face datection algorithm can be existing algorithm, such as based on face spy
It levies the recognizer put, the recognizer based on whole picture facial image, the recognizer based on template and utilizes neural network
The algorithm identified.
In an embodiment of the disclosure, it can store in detection module or outside detection module and adopted in advance
The target face information of collection.The target face information of the storage can be obtained by other pictures or video, can also tracked
It is obtained and carrying out Image Acquisition to live video before starting.Target face information can also be from other non-image sources
Middle acquisition.Target face can be corresponding with ID, as long as so that obtaining ID is assured that corresponding target face information.
In an embodiment of the disclosure, the position for the target face that detection module is obtained by face recognition algorithms
As human face region.Parameter to be used can be initialized device by detection module according to the human face region.For example, root
According to the human face region detected or face frame, detection module can be initialized the trace regions in tracing algorithm,
Detection zone and camera lens frame can be initialized.
In an embodiment of the disclosure, human face region, trace regions, Face datection region and camera lens frame are defined
Region can be square frame-shaped, be also possible to other any proper shapes.The size of human face region can be according to detecting
The size corresponding change of face.The size of trace regions, Face datection region and camera lens frame can be fixed and invariable, can also be with
According to circumstances and suitably change.
In an embodiment of the disclosure, initialization tracking frame may include: according to face frame IfaceCoordinate and
Size determines that the tracking frame of initialization, such as the tracking frame of initialization have coordinate identical with face frame and size.Tracking
Frame can be defined as Btrack=(x, y, w, h), wherein x, y are box lower left corner apex coordinate, and w is the width of box, and h is box
Height, camera lens frame and detection zone below refer to this definition.
In an embodiment of the disclosure, initializes detection zone and camera lens frame includes: according to chasing after after initialization
Track frame Btrack=(x, y, w, h), to determine the detection zone R of initializationdetect=(x-w, y-h, 3*w, 3*h), wherein detection zone
The coordinate and size in domain are merely exemplary, and also can have other coordinate positions and width and height, such as 4*w
The height of the width of the wide, height of 4*h or 6*w, 5*h；And come according to the tracking frame B after initializationtrack=(x, y, w, h) determines mirror
Head frame Bcam=(x-x0,y-y0, W, H), wherein wherein x0、y0For predefined deviant, W is the width of predefined camera lens frame
Degree, H are the height of predefined camera lens frame, such as x0=y0=200, W=500, H=600 (digital representation pixel therein
Point), the width and height of usual camera lens frame are all larger than the height and width of tracking frame.
In an embodiment of the disclosure, computing module 502 is configured as: based on the people detected in detection module
Face region calculates trace regions in second video frame in the video sequence, and trace regions instruction includes target face
Candidate region.
In an embodiment of the disclosure, it is being based on face frame IfaceCoordinate and size chasing after for initialization has been determined
Track frame BtrackLater, next video frame and the initialization tracking frame determined based on face frame are input to tracking and calculated by computing module
In method, be calculated in second video frame tracking result (i.e. trace regions or tracking frame, be used to indicate including target face
Candidate region).
In an embodiment of the disclosure, above-mentioned tracing algorithm can be existing face tracking algorithm, such as KCF
(Kernelized Correlation Filter, i.e. coring correlation filter) algorithm.
In an embodiment of the disclosure, inspection module 503 is configured as: being based on the calculated tracking of tracing algorithm
Region determines that detection zone, the detection zone have included at least calculated trace regions in second above-mentioned video frame；
The target face human face region new with determination is detected in identified detection zone.
It is B ' based on the calculated trace regions of tracing algorithm in an embodiment of the disclosuretrack=(x ', y ',
W, h), inspection module can will test region and be determined as R 'detect=(x '-w, y '-h, 3*w, 3*h).The wherein seat of detection zone
Cursor position and width and height are merely exemplary, and also can have other coordinate positions appropriate and width and height
Degree.But identified detection zone usually should include above-mentioned trace regions.
In an embodiment of the disclosure, inspection module is using face recognition algorithms in identified detection zone
Detect the new human face region about target face.
In an embodiment of the disclosure, performed face recognition algorithms can be with detection mould in inspection module
Face recognition algorithms used in block are identical.However, inspection module can also execute the new face different from detection module
Recognizer.Since the image-region detected in inspection module is smaller, the face that may include is less (if inspection
The suitable of region restriction is surveyed, can only have target face in detection zone), therefore recognition of face used in inspection module is calculated
Method can be more simplified compared to algorithm used in detection module, to reduce calculation amount and save holding time.
In an embodiment of the disclosure, determining module 504 is configured as: judge that inspection module detects is new
Whether human face region matches with trace regions calculated in computing module, to determine whether tracking succeeds；If tracking at
Function is based on the calculated trace regions of computing module, calculates new trace regions in the third video frame of video sequence, such as
Fruit tracking failure is calculated new based on the new human face region that inspection module detects in the third video frame of video sequence
Trace regions.
In an embodiment of the disclosure, determining module obtains the central point between region to be matched, example first
The centre coordinate of the human face region detected such as inspection module is (X, Y), the coordinate at the center of the calculated trace regions of computing module
For (X ', Y ')；It is then determined module calculates the offset between two centre coordinates
Then, it is determined that module compares offset d with preset inspection threshold value, if offset d can be determined not higher than threshold value is examined
It tracks successfully, whereas if offset d, which is higher than, examines threshold value, then can determine tracking failure.
In an embodiment of the disclosure, if tracked successfully, computing module is by next video frame, i.e. third
A video frame and the calculated trace regions of computing module, which are input to, to carry out continuing to track in tracing algorithm.
In an embodiment of the disclosure, if tracking failure, the people that detection module detects inspection module
Face region is determined as the human face region that initial detecting arrives, and initializes tracking frame, camera lens frame and detection zone according to the human face region
Domain.
In an embodiment of the disclosure, in addition, the device 500 of the disclosure can also include display module 505, show
Show that module 505 can be configured as based on the calculated trace regions setting camera lens frame of computing module.Here trace regions are set
The mode for setting camera lens frame is identical with the frame setting mode of camera lens frame is tracked in above-mentioned initialization, therefore repeats no more.Then, it shows
Module 505 corresponds to set camera lens frame in video frame, and intercepts corresponding image, then shows corresponding image
Show.
Fig. 6 shows the block diagram of the device 600 for face tracking according to another embodiment.With dress shown in fig. 5
Set the difference is that, provided with can be to the display module 605 that the movement of camera lens frame is smoothed.Specifically, should
Display module 605 can have smooth submodule 6051, which is configured as: comparing camera lens frame and is used to regard
The first front lens frame of preceding video frame in frequency sequence is to determine offset；Offset and the smooth threshold value of systemic presupposition are compared
Compared with；And output tracking result.
In an embodiment of the disclosure, if the coordinate in the current lens frame lower left corner is (X ', Y '), previous video
The coordinate in the lower left corner of the first front lens frame of frame is (X, Y), then smooth submodule can determine offset are as follows:
In an embodiment of the disclosure, if offset is not higher than smooth threshold value, smooth submodule module is based on
Camera lens frame shows video sequence, whereas if offset is higher than smooth threshold value, then smooth submodule will smoothly locate camera lens frame
Reason.For example, smooth threshold value can be 10 (pixels).
The smoothing processing includes the interpolation obtained between camera lens frame and first front lens frame.For example, the interpolation can beThen, smooth submodule based on interpolation come
Camera lens frame is set.
The final camera lens frame B " that display module will also obtaincamIt corresponds in video frame, and intercepts corresponding image.It will be right
The image answered is output to display module, to show the result of tracking.
According to the example implementations of present disclosure, a kind of equipment, including one or more processors are provided；With
And storage device, for storing one or more programs.When one or more programs are executed by one or more processors, make
One or more processors realization is obtained according to the method for present disclosure.
According to the example implementations of present disclosure, a kind of computer-readable medium is provided, is stored thereon with meter
Calculation machine program realizes the method according to present disclosure when the program is executed by processor.
Fig. 7 shows the block diagram that can implement the calculating equipment 700 of multiple embodiments of present disclosure.As schemed
Show, equipment 700 includes central processing unit (CPU) 701, can be according to the calculating being stored in read-only memory (ROM) 702
Machine program instruction is loaded into the computer program instructions in random access storage device (RAM) 703 from storage unit 708, comes
Execute various movements appropriate and processing.In RAM 703, it can also store equipment 700 and operate required various programs and data.
CPU 701, ROM 702 and RAM 703 are connected with each other by bus 704.Input/output (I/O) interface 705 is also connected to always
Line 704.
Multiple components in equipment 700 are connected to I/O interface 705, comprising: input unit 706, such as keyboard, mouse etc.；
Output unit 707, such as various types of displays, loudspeaker etc.；Storage unit 708, such as disk, CD etc.；And it is logical
Believe unit 709, such as network interface card, modem, wireless communication transceiver etc..Communication unit 709 allows equipment 700 by such as
The computer network of internet and/or various telecommunication networks exchange information/data with other equipment.
Processing unit 701 executes each method as described above and processing, such as method 100 and/or method 200.Example
Such as, in some embodiments, method 100 and/or method 200 can be implemented as computer software programs, visibly be wrapped
Contained in machine readable media, such as storage unit 708.In some embodiments, some or all of of computer program can
To be loaded into and/or be installed in equipment 700 via ROM 702 and/or communication unit 709.When computer program loads arrive
RAM 703 and by CPU 701 execute when, one or more frames of method as described above 100 and/or method 200 can be executed.
Alternatively, in other embodiments, CPU 701 can pass through other any modes (for example, by means of firmware) appropriate
It is configured as execution method 100 and/or method 200.
Function described herein can be executed at least partly by one or more hardware logic components.Example
Such as, without limitation, the hardware logic component for the exemplary type that can be used includes: field programmable gate array (FPGA), dedicated
Integrated circuit (ASIC), Application Specific Standard Product (ASSP), the system (SOC) of system on chip, load programmable logic device
(CPLD) etc..
Program code for implementing the method for present disclosure can be using any group of one or more programming languages
It closes to write.These program codes can be supplied to general purpose computer, special purpose computer or other programmable data processing units
Processor or controller so that program code when by processor or controller execution when make to be advised in flowchart and or block diagram
Fixed function/operation is carried out.Program code can be executed completely on machine, partly be executed on machine, as independence
Software package partly executes on machine and partly executes or hold on remote machine or server on the remote machine completely
Row.
In the context of present disclosure, machine readable media can be tangible medium, may include or stores
The program for using or being used in combination with instruction execution system, device or equipment for instruction execution system, device or equipment.Machine
Device readable medium can be machine-readable signal medium or machine-readable storage medium.Machine readable media may include but unlimited
In times of electronics, magnetic, optical, electromagnetism, infrared or semiconductor system, device or equipment or above content
What appropriate combination.The more specific example of machine readable storage medium will include the electrical connection of line based on one or more, portable
Formula computer disks, hard disk, random access memory (RAM), read-only memory (ROM), Erasable Programmable Read Only Memory EPROM
(EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage are set
Standby or above content any appropriate combination.
Although this should be understood as requiring operating in this way with shown in addition, depicting each operation using certain order
Certain order out executes in sequential order, or requires the operation of all diagrams that should be performed to obtain desired result.
Under certain environment, multitask and parallel processing be may be advantageous.Similarly, although containing several tools in being discussed above
Body realize details, but these be not construed as to scope of the present disclosure limitation.In individual embodiment
Certain features described in context can also be realized in combination in single realize.On the contrary, in the context individually realized
Described in various features can also realize individually or in any suitable subcombination in multiple realizations.
Although having used specific to this theme of the language description of structure feature and/or method logical action, answer
When understanding that theme defined in the appended claims is not necessarily limited to special characteristic described above or movement.On on the contrary,
Special characteristic described in face and movement are only to realize the exemplary forms of claims.
Claims (26)
1. a kind of method for face tracking, comprising:
Target face is detected in the first video frame of video sequence to determine the first human face region；
Based on first human face region, trace regions, the tracking are calculated in the second video frame in the video sequence
Region instruction includes the candidate region of the target face；
Determine that detection zone, the detection zone include at least described chase after in second video frame based on the trace regions
Track region；
The target face is detected in the detection zone to determine the second human face region；And
The trace regions are matched in response to second human face region, determination is tracked successfully.
2. according to the method described in claim 1, wherein the determination tracks and successfully includes:
Compare second human face region and the trace regions to determine the first offset；And
In response to first offset not higher than threshold value is examined, determination is tracked successfully.
3. according to the method described in claim 2, further comprising: it is tracked successfully in response to determination,
Based on the trace regions, new trace regions are calculated in the third video frame of the video sequence, described new chases after
The instruction of track region includes the candidate region of the target face.
4. according to the method described in claim 2, further include:
It is higher than the inspection threshold value in response to first offset, determines tracking failure；And
In response to determining tracking failure, it is based on second human face region, is calculated in the third video frame of the video sequence
New trace regions, the new trace regions instruction include the candidate region of the target face.
5. according to the method described in claim 1, further include:
Camera lens frame is set based on the trace regions, the camera lens frame is used to show that in the video sequence to include the target
The image-region of face.
6. according to the method described in claim 5, wherein the trace regions are defined as B 'track=(x, y, w, h), wherein x,
Y is the lower-left angular coordinate of the trace regions, and w is the width of the trace regions, and h is the height of the trace regions；
Being provided with the camera lens frame includes:
By the camera lens frame B 'camIt is set as B 'cam=(x-x0, y-y0, W, H), wherein x0、y0For predefined deviant, W is pre-
The width of the camera lens frame of definition, H are the height of predefined camera lens frame, wherein W > w and H > h.
7. method according to claim 5 or 6, further includes:
Compare the camera lens frame with the first front lens frame for the preceding video frame in the video sequence to determine the second offset；
And
It is not higher than smooth threshold value in response to second offset, the video sequence is shown based on the camera lens frame；And
It is higher than the smooth threshold value in response to second offset, the camera lens frame is smoothed.
8. according to the method described in claim 7, wherein the smoothing processing includes:
Obtain the interpolation between the camera lens frame and the first front lens frame；And
Based on the interpolation, the camera lens frame is set.
9. according to the method described in claim 1, wherein the trace regions are defined as B 'track=(x, y, w, h), x, y are
The coordinate in the lower left corner of the trace regions, w are the width of the trace regions, and h is the height of the trace regions；
Wherein determine that detection zone includes: by the detection zone in second video frame based on the trace regions
R′detectIt is set as R 'detect=(x-w, y-h, 3*w, 3*h).
10. according to the method described in claim 2, wherein the coordinate at the center of second human face region be (X, Y), it is described to chase after
The coordinate at the center in track region is (X ', Y '),
Wherein determine that the first offset includes: to set first offset d to
11. according to the method described in claim 8, wherein the lower-left angular coordinate of the camera lens frame is (X ', Y '), the previous mirror
The lower-left angular coordinate of head frame is (X, Y),
Wherein determine that second offset includes: to set the second offset s to
12. according to the method for claim 11, wherein obtaining the interpolation includes: by the interpolation B "camIt is set asWherein W is the width of predefined camera lens frame
Degree, H are the height of predefined camera lens frame.
13. a kind of device for face tracking, comprising:
Detection module, the detection module are configured as detecting target face in the first video frame of video sequence to determine the
One human face region；
Computing module, the computing module are configured as based on first human face region, and second in the video sequence
Trace regions are calculated in video frame, the trace regions instruction includes the candidate region of the target face；
Inspection module, the inspection module are configured as:
Determine that detection zone, the detection zone include at least described chase after in second video frame based on the trace regions
Track region；And
The target face is detected in the detection zone to determine the second human face region；
Determining module, determining module are configured to respond to second human face region and are matched with the trace regions, and determination chases after
Track success.
14. device according to claim 13, wherein the determination tracks and successfully includes:
Compare second human face region and the trace regions to determine the first offset；And
In response to first offset not higher than threshold value is examined, determination is tracked successfully.
15. device according to claim 14, further comprises: it is tracked successfully in response to determination,
Based on the trace regions, new trace regions are calculated in the third video frame of the video sequence, described new chases after
The instruction of track region includes the candidate region of the target face.
16. device according to claim 14, further includes:
It is higher than the inspection threshold value in response to first offset, determines tracking failure；And
In response to determining tracking failure, it is based on second human face region, is calculated in the third video frame of the video sequence
New trace regions, the new trace regions instruction include the candidate region of the target face.
17. device according to claim 13, further includes:
Display module, the display module are configured as that camera lens frame is arranged based on the trace regions, and the camera lens frame is for showing
Show the image-region including the target face in the video sequence.
18. device according to claim 17, wherein the trace regions are defined as B 'track=(x, y, w, h), wherein
X, y is the lower-left angular coordinate of the trace regions, and w is the width of the trace regions, and h is the height of the trace regions；
Being provided with the camera lens frame includes:
By the camera lens frame B 'camIt is set as B 'cam=(x-x0, y-y0, W, H), wherein x0、y0For predefined deviant, W is pre-
The width of the camera lens frame of definition, H are the height of predefined camera lens frame, wherein W > w and H > h.
19. device described in 7 or 18 according to claim 1, wherein the display module further includes smooth submodule,
The smooth submodule is configured as:
Compare the camera lens frame with the first front lens frame for the preceding video frame in the video sequence to determine the second offset；
And
It is not higher than smooth threshold value in response to second offset, the video sequence is shown based on the camera lens frame；And
It is higher than the smooth threshold value in response to second offset, the camera lens frame is smoothed.
20. device according to claim 19, wherein the smoothing processing includes:
Obtain the interpolation between the camera lens frame and the first front lens frame；And
Based on the interpolation, the camera lens frame is set.
21. device according to claim 13, wherein the trace regions are defined as B 'track=(x, y, w, h), x, y
For the coordinate in the lower left corner of the trace regions, w is the width of the trace regions, and h is the height of the trace regions；
Wherein determine that detection zone includes: by the detection zone in second video frame based on the trace regions
R′detectIt is set as R 'detect=(x-w, y-h, 3*w, 3*h).
22. device according to claim 14, wherein the coordinate at the center of second human face region is (X, Y), it is described
The coordinate at the center of trace regions is (X ', Y '),
Wherein determine that the first offset includes: to set first offset d to
23. device according to claim 20, wherein the lower-left angular coordinate of the camera lens frame is (X ', Y '), it is described previously
The lower-left angular coordinate of camera lens frame is (X, Y),
Wherein determine that second offset includes: to set the second offset s to
24. device according to claim 23, wherein obtaining the interpolation includes: by the interpolation B "camIt is set asWherein W is the width of predefined camera lens frame
Degree, H are the height of predefined camera lens frame.
25. a kind of equipment, the equipment include:
One or more processors；And
Storage device, for storing one or more programs, when one or more of programs are by one or more of processing
Device executes, so that one or more of processors are realized is used for face tracking according to claim 1 described in any one of -12
Method.
26. a kind of computer readable storage medium is stored thereon with computer program, realization when described program is executed by processor
The method of face tracking is used for described in any one of -12 according to claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810786024.2A CN109064489A (en) | 2018-07-17 | 2018-07-17 | Method, apparatus, equipment and medium for face tracking |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810786024.2A CN109064489A (en) | 2018-07-17 | 2018-07-17 | Method, apparatus, equipment and medium for face tracking |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109064489A true CN109064489A (en) | 2018-12-21 |
Family
ID=64817130
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810786024.2A Pending CN109064489A (en) | 2018-07-17 | 2018-07-17 | Method, apparatus, equipment and medium for face tracking |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109064489A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111401111A (en) * | 2019-01-03 | 2020-07-10 | 瑞昱半导体股份有限公司 | Object tracking system, object tracking method and non-transitory computer readable medium |
| CN111612813A (en) * | 2019-02-26 | 2020-09-01 | 北京海益同展信息科技有限公司 | Face tracking method and device |
| CN112102364A (en) * | 2020-09-22 | 2020-12-18 | 广州华多网络科技有限公司 | Target tracking method and device, electronic equipment and storage medium |
| CN113302911A (en) * | 2019-01-17 | 2021-08-24 | Mo-Sys工程有限公司 | Camera control |
| CN113689462A (en) * | 2020-05-19 | 2021-11-23 | 深圳绿米联创科技有限公司 | Target processing method and device and electronic equipment |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1924894A (en) * | 2006-09-27 | 2007-03-07 | 北京中星微电子有限公司 | Multiple attitude human face detection and track system and method |
| CN102509070A (en) * | 2011-10-12 | 2012-06-20 | 西安理工大学 | Video-based human face area tracking method for counting people paying close attention to advertisement |
| KR101381343B1 (en) * | 2013-09-14 | 2014-04-04 | 주식회사 코어셀 | The method for appling the masking for the protection of privacy to the face area detected by using face recognition technology |
| CN104915663A (en) * | 2015-07-03 | 2015-09-16 | 广东欧珀移动通信有限公司 | Method, system, and mobile terminal for improving face identification and display |
| CN104954664A (en) * | 2014-03-24 | 2015-09-30 | 东芝阿尔派·汽车技术有限公司 | Image processing apparatus and image processing method |
| CN105117022A (en) * | 2015-09-24 | 2015-12-02 | 北京零零无限科技有限公司 | Method and device for controlling unmanned aerial vehicle to rotate along with face |
| CN105264570A (en) * | 2013-06-14 | 2016-01-20 | 高通股份有限公司 | Tracker assisted image capture |
| CN105913028A (en) * | 2016-04-13 | 2016-08-31 | 华南师范大学 | Face tracking method and face tracking device based on face++ platform |
| CN106709932A (en) * | 2015-11-12 | 2017-05-24 | 阿里巴巴集团控股有限公司 | Face position tracking method and device and electronic equipment |
| CN106934818A (en) * | 2015-12-31 | 2017-07-07 | 芋头科技（杭州）有限公司 | A kind of hand exercise tracking and system |
| CN107959798A (en) * | 2017-12-18 | 2018-04-24 | 北京奇虎科技有限公司 | Video data real-time processing method and device, computing device |
-
2018
- 2018-07-17 CN CN201810786024.2A patent/CN109064489A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1924894A (en) * | 2006-09-27 | 2007-03-07 | 北京中星微电子有限公司 | Multiple attitude human face detection and track system and method |
| CN102509070A (en) * | 2011-10-12 | 2012-06-20 | 西安理工大学 | Video-based human face area tracking method for counting people paying close attention to advertisement |
| CN105264570A (en) * | 2013-06-14 | 2016-01-20 | 高通股份有限公司 | Tracker assisted image capture |
| KR101381343B1 (en) * | 2013-09-14 | 2014-04-04 | 주식회사 코어셀 | The method for appling the masking for the protection of privacy to the face area detected by using face recognition technology |
| CN104954664A (en) * | 2014-03-24 | 2015-09-30 | 东芝阿尔派·汽车技术有限公司 | Image processing apparatus and image processing method |
| CN104915663A (en) * | 2015-07-03 | 2015-09-16 | 广东欧珀移动通信有限公司 | Method, system, and mobile terminal for improving face identification and display |
| CN105117022A (en) * | 2015-09-24 | 2015-12-02 | 北京零零无限科技有限公司 | Method and device for controlling unmanned aerial vehicle to rotate along with face |
| CN106709932A (en) * | 2015-11-12 | 2017-05-24 | 阿里巴巴集团控股有限公司 | Face position tracking method and device and electronic equipment |
| CN106934818A (en) * | 2015-12-31 | 2017-07-07 | 芋头科技（杭州）有限公司 | A kind of hand exercise tracking and system |
| CN105913028A (en) * | 2016-04-13 | 2016-08-31 | 华南师范大学 | Face tracking method and face tracking device based on face++ platform |
| CN107959798A (en) * | 2017-12-18 | 2018-04-24 | 北京奇虎科技有限公司 | Video data real-time processing method and device, computing device |
Non-Patent Citations (1)
| Title |
|---|
| 杜军平等: "《跨尺度运动图像的目标检测与跟踪》", 30 June 2018, 北京邮电大学出版社 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111401111A (en) * | 2019-01-03 | 2020-07-10 | 瑞昱半导体股份有限公司 | Object tracking system, object tracking method and non-transitory computer readable medium |
| CN113302911A (en) * | 2019-01-17 | 2021-08-24 | Mo-Sys工程有限公司 | Camera control |
| US11949982B2 (en) | 2019-01-17 | 2024-04-02 | Mo-Sys Engineering Limited | Method of forming a processed video stream |
| CN111612813A (en) * | 2019-02-26 | 2020-09-01 | 北京海益同展信息科技有限公司 | Face tracking method and device |
| CN113689462A (en) * | 2020-05-19 | 2021-11-23 | 深圳绿米联创科技有限公司 | Target processing method and device and electronic equipment |
| CN112102364A (en) * | 2020-09-22 | 2020-12-18 | 广州华多网络科技有限公司 | Target tracking method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109064489A (en) | Method, apparatus, equipment and medium for face tracking | |
| US11727663B2 (en) | Method and apparatus for detecting face key point, computer device and storage medium | |
| CN106204660B (en) | A kind of Ground Target Tracking device based on characteristic matching | |
| CN108921782B (en) | Image processing method, device and storage medium | |
| US9361510B2 (en) | Efficient facial landmark tracking using online shape regression method | |
| WO2018102880A1 (en) | Systems and methods for replacing faces in videos | |
| CN110163211B (en) | Image recognition method, device and storage medium | |
| CN110147708B (en) | Image data processing method and related device | |
| CN107707899B (en) | Multi-view image processing method, device and electronic equipment comprising moving target | |
| CN112017212B (en) | Training and tracking method and system of face key point tracking model | |
| US20210056292A1 (en) | Image location identification | |
| CN111563490A (en) | Face key point tracking method and device and electronic equipment | |
| Firintepe et al. | The more, the merrier? A study on in-car IR-based head pose estimation | |
| Kantarcı et al. | Shuffled patch-wise supervision for presentation attack detection | |
| Hennings-Yeomans et al. | Recognition of low-resolution faces using multiple still images and multiple cameras | |
| CN109284722A (en) | Image processing method, device, face recognition device and storage medium | |
| US10032080B2 (en) | Evaluation of models generated from objects in video | |
| Martin et al. | An evaluation of different methods for 3d-driver-body-pose estimation | |
| Luo et al. | Alignment and tracking of facial features with component-based active appearance models and optical flow | |
| JP2004157778A (en) | Nose position extraction method, program for operating it on computer, and nose position extraction device | |
| CN111582654B (en) | Service quality evaluation method and device based on deep cycle neural network | |
| CN111401240B (en) | Classroom attention detection method, device, equipment and storage medium | |
| US20220067357A1 (en) | Full skeletal 3d pose recovery from monocular camera | |
| CN110298229B (en) | Video image processing method and device | |
| JP2023512359A (en) | Associated object detection method and apparatus |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20190122 Address after: Room 1702-03, Blue Sky and Sheng Building, 32 Zhongguancun Street, Haidian District, Beijing Applicant after: BEIJING CENTURY TAL EDUCATION TECHNOLOGY CO., LTD. Address before: Room 118, Building 3, 6-C, 8 High-tech Parks, Shijingshan District, Beijing, 100144 Applicant before: BEIJING XINTANG SICHUANG EDUCATIONAL TECHNOLOGY CO., LTD. |