CN106650682A - Method and device for face tracking - Google Patents
Method and device for face tracking Download PDFInfo
- Publication number
- CN106650682A CN106650682A CN201611248067.2A CN201611248067A CN106650682A CN 106650682 A CN106650682 A CN 106650682A CN 201611248067 A CN201611248067 A CN 201611248067A CN 106650682 A CN106650682 A CN 106650682A
- Authority
- CN
- China
- Prior art keywords
- face
- frame
- image
- region
- current image
- 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
- 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
Abstract
The invention is applied to the technical field of image detection and provides a method and a device for face tracking. The method comprises the steps of acquiring an image position of a face detection box in a previous image frame; locating a region which is the same as the image position in a current image frame; carrying out face alignment between an image in the region located in the current image frame and a front template of a face in the face detection box to acquire characteristic point positions of the face in a current image; determining the face detection box in the current image frame according to the acquired characteristic point positions. When face tracking is carried out on a video image, overall situation is not needed for each frame of image in a video, the calculation quantity for face tracking in the video is reduced, the processing efficiency is improved and the influence degree of video definition for processing is reduced.
Description
Technical field
The invention belongs to field of image detection, more particularly to a kind of method and device of face tracking.
Background technology
Face tracking technology is based on the face feature information of people, to face location and face in the video of camera acquisition
A kind of human face detection tech that feature is tracked.As shown in figure 1, generally there are three types position of the face on image：Lift
Head or the facial pitching for causing of bowing, the face driftage for causing of turning one's head to the left or to the right, and the face that to the left or to the right torticollis causes
Roll in portion.Its septum reset rolling refers to that face carries out the rotation of certain angle with the normal vector of face place plane as axle.
At present the object of face tracking research work is mainly front or Jie Jin positive facial image, i.e., by face pair
Together (Face Alignment) algorithm carries out face tracking and alignment to front face image as shown in Figure 2.Work as front face
Have and remain able to accurately be tracked and align compared with algorithm during primary deflector.For facial pitching and face driftage, common people
The face anglec of rotation is less, and the horizontal and vertical texture variations of face less, directly can carry out face using face alignment algorithm
Follow the trail of.But for face rolling, sometimes the face anglec of rotation is larger, can carry out 360 degree of rotations, such as the face when people stands upside down
The angle of rolling is 180 degree.Now, the horizontal and vertical texture variations of face are too big, it is impossible to directly entered using face alignment algorithm
Row is followed the trail of.In this case, the angle for detecting face rolling first is needed, then according to the angle of face rolling by entire image
Rotated accordingly, make the face in image become front face, finally using face alignment algorithm to the postrotational figure
As carrying out face tracking and alignment.
This face tracking methods, need to do Face datection to every two field picture overall situation of video, by face alignment algorithm
The tracking and alignment of face are realized, operand is big.When facial roll angle degree is larger, need first to carry out entire image rotation
Face alignment algorithm can be used, so the operand being tracked to face in the two field picture is significantly increased again.And with
The raising of definition in video, the face tracking operand per two field picture also can accordingly increase.For example, it is right on PC at present
1920 × 1080 HD video is processed, and more than ten milliseconds to more than 100 milliseconds are needed per two field picture.In mobile phone and camera
Etc. the requirement of real-time that cannot reach 30FPS in handheld device to the processing speed that video human face is followed the trail of, easily cause video and broadcast
Put interim card.
In sum, big to face tracking methods' operand of video at present, occupying system resources are more, and treatment effeciency is low,
Video playback interim card is easily caused, and processing speed is affected by face rolling angle and video definition.
The content of the invention
The purpose of the embodiment of the present invention is to provide a kind of method of face tracking, it is intended to solved at present to the face of video
Method for tracing operand is big, and occupying system resources are more, the low problem for the treatment of effeciency.
The embodiment of the present invention is achieved in that a kind of method of face tracking, including：
Obtain picture position of the Face datection frame in a upper picture frame；
In current image frame, the region of identical described image position is oriented；
The positive face mould of the image in region that will be oriented in the current image frame and face in the Face datection frame
Plate carries out face alignment, obtains the characteristic point position of face described in the present image；
According to the characteristic point position for getting, Face datection frame is determined in the current image frame.
The another object of the embodiment of the present invention is to provide a kind of device of face tracking, including：
Acquiring unit, for obtaining picture position of the Face datection frame in a upper picture frame；
Positioning unit, in current image frame, orienting the region of identical described image position；
Alignment unit, in the image in region that will orient in the current image frame and the Face datection frame
The front face template of face carries out face alignment, obtains the characteristic point position of face described in the present image；
Computing unit, for according to the characteristic point position for getting, in the current image frame face being determined
Detection block.
In the embodiment of the present invention, by the picture position using Face datection frame in video in a upper picture frame as current
Picture frame carries out the position of the image-region of face alignment, and to the image in the region feature that face positioned in alignment goes out face is carried out
Point position, and Face datection frame in current image frame is determined according to the position of the characteristic point of face.The embodiment of the present invention is to regarding
Frequency image need not be reduced and video human face is followed the trail of when carrying out face tracking to all carrying out global detection per two field picture in video
Operand, improve treatment effeciency, reduce video definition to process influence degree.
Description of the drawings
Fig. 1 is the schematic diagram of the face location in image provided in an embodiment of the present invention；
Fig. 2 is the schematic diagram of positive facial image provided in an embodiment of the present invention；
Fig. 3 is the flowchart of the method for face tracking provided in an embodiment of the present invention；
Fig. 4 be face tracking provided in an embodiment of the present invention method in carry out the flowchart of Face datection；
Fig. 5 be the face tracking that another embodiment of the present invention is provided method in carry out the flowchart of Face datection；
Fig. 6 is that method S303 of face tracking provided in an embodiment of the present invention implements flow chart；
Fig. 7 is the structured flowchart of the device of face tracking provided in an embodiment of the present invention.
Specific embodiment
In order that the objects, technical solutions and advantages of the present invention become more apparent, it is right below in conjunction with drawings and Examples
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only to explain the present invention, and
It is not used in the restriction present invention.
In the embodiment of the present invention, by the picture position using Face datection frame in video in a upper picture frame as current
Picture frame carries out the position of the image-region of face alignment, and to the image in the region feature that face positioned in alignment goes out face is carried out
Point position, and Face datection frame in current image frame is determined according to the characteristic point position of face.The embodiment of the present invention is to video
Image need not reduce what video human face was followed the trail of when carrying out face tracking to all carrying out global detection per two field picture in video
Operand, improves treatment effeciency, reduces influence degree of the video definition to process.
The present invention is applied to includes that mobile phone, flat board, notebook, intelligent television etc. can carry out the terminal of video playback and set
It is standby.Face datection frame is the frame that the characteristic point of face can be integrally incorporated in an interior rectangular area in image.Regarding
When frequency is played, Face datection frame can show, it is also possible to not shown.The characteristic point of face refers to the people in the middle of image
Face portion is easier the feature locations distinguished, such as lip outline, eye contour etc..The characteristic point position of face can be with image
Obtained from image information by face alignment algorithm, depending on the number of characteristic point is according to selected algorithm and actual demand.
Characteristic point can be used for face positioning in the picture, for example can be in eyes position according to the characteristic point of eyes in face in image
The control of the trims such as addition glasses is put, the modification to face in image is realized.Characteristic point can be also used for carrying out recognition of face
Different face in distinguish image.
Face tracking is on every two field picture in video the characteristic point information for finding out face, determines the position of face.Example
Eye position for face such as in video display process adds the control of glasses trim, when the position of the face in video is sent out
When raw mobile, the display location of glasses control can be adjusted according to the change of the characteristic point position of eyes, show control
Show that position is synchronous with the mobile holding of eyes all the time, realize more preferable modification effect.Face tracking can be with monitor video
All the time the characteristic point of a face is locked.Face tracking and human face characteristic point also have many other purposes, and here is not arranged one by one
Lift.Global detection is carried out to every two field picture to refer to, by the whole scan to whole two field picture, determine the region that face is located,
May then pass through the means such as image procossing to find out the position of the characteristic point of the face in the region.Because global detection needs
Whole two field picture is scanned, video definition is higher, processing speed is slower.
The conventional method positioned to the characteristic point of face is that image is processed by face alignment algorithm.People
Face alignment algorithm has a many kinds, including ASM algorithms, AAM algorithms, STASM algorithms etc., the characteristic point to face face can be realized
Positioning.Wherein, ASM algorithms are relatively easy, and processing speed is very fast, and STASM is relatively accurate to the positioning of characteristic point, but process speed
Degree is slower.Suitable face alignment algorithm can be according to the actual requirements selected to carry out the positioning feature point of face.
Fig. 3 is the flowchart of the method for face tracking provided in an embodiment of the present invention, and details are as follows：
In S301, picture position of the Face datection frame in a upper picture frame is obtained.
In the present embodiment, a upper picture frame is for relatively current picture frame.Face tracking is carried out to video image
When, the process to image is carried out one by one, and after a frame has been processed next frame is reprocessed.So described upper one in text
Comparatively picture frame, current image frame and next image frame are all.Current image frame is the two field picture being presently processing.
A upper picture frame refers to the two field picture for just completing to process before current image frame.After the completion of current image frame process, Xia Yitu
As frame just becomes for current image to be processed.
Face datection frame refers to the frame of a rectangular area for being included the characteristic point of face in image.The area
The size in domain can be set according to actual conditions and selected face alignment algorithm.Face datection frame is obtained in figure
Position as in can be the coordinate or rectangle frame two for obtaining the angular vertex of rectangle frame four pixel in the picture to angular vertex
The coordinate of pixel in the picture, is not limited thereto.
In S302, in current image frame, the region of identical described image position is oriented.
That is, the picture position according to the Face datection frame for getting in a upper picture frame, searches in current image frame
Identical position.Specifically can be positioned according to the pixel coordinate of the rectangular shaped rim of rectangular area.Due to the company of video playback
Continuous property, the usual two frame adjacent image spacing time is shorter, and image is more similar, and the position that face is located in image also differs not
It is many.Therefore picture position that can be by Face datection frame in a upper picture frame is come as the rectangular area in current image frame
Position, the possibility of characteristic point of the region comprising face is maximum.So equivalent to having predicted the general of face in the picture
Position, it is not necessary to entire image is scanned again.The operand of image procossing is reduced, treatment effeciency is improve.
In S303, the image in region that will be oriented in the current image frame and face in the Face datection frame
Front face template carry out face alignment, obtain the characteristic point position of face described in the present image.
In the present embodiment, the region oriented in current image frame be by Face datection frame in a upper picture frame
The region that picture position determines.Because the difference of Qian Hou two field pictures is little, can first predict that the region that face is located is not sent out
The larger change of life.The template selected in the present embodiment for face front face template because face alignment algorithm face in the picture
For front when search face characteristic point precision it is higher.In the region oriented in current image frame, using face pair
Neat algorithm, you can obtain the characteristic point position of face described in present image.
Consider that the factor in terms of processing speed selects ASM algorithms in the present embodiment.Using ASM algorithms, the front of face
Template is one and is based on statistical model, by the set expression put.Each local first in matching template and target image
Characteristic point calculating the new position of characteristic point, by iteration until convergence is making the characteristic point position and face in image of template
Characteristic point position match, the final characteristic point position for determining face in image.Other can also be selected according to actual conditions
The algorithm of face alignment, is not limited thereto.
In S304, according to the characteristic point position for getting, in the current image frame Face datection is determined
Frame.
Wherein, the characteristic point position is to carry out face pair by the image in the region to orienting in current image frame
Obtain together.By the position of these characteristic points, the rectangular area comprising these characteristic points is obtained in calculating.The side of the rectangular area
Frame is the Face datection frame determined in current image frame.The big I of Face datection frame according to the actual requirements depending on, but must
Whole characteristic point of face must be included.After the completion of current image frame process, start to process next image frame.When
Determine that Face datection frame carries out the foundation of rectangular area positioning of face alignment as next image frame in front picture frame.
It is also step according to this that other frame video images in video are carried out with process, by the way that face in a upper picture frame is examined
Survey the position of frame carries out face alignment as the rectangular area position in current image frame, it is not necessary to carry out global inspection to every two field picture
Survey, reduce operand.
As one embodiment of the present of invention, in the region that can be oriented in current image frame first, detect whether
There is facial image.If having detected facial image, then face alignment is carried out to the image in the region.As shown in figure 4,
After S302, before S303, methods described also includes：
In S401, in the region oriented in the current image frame, facial image has been detected whether.
In S402, if being not detected by facial image, the Face datection of the overall situation is carried out to the current image frame, with again
The region is positioned.
In the present embodiment, can detect in a designated area whether there is facial image by Face datection.Face is examined
Survey and the difference of face alignment is that Face datection is not positioned to the characteristic point position of face, only detects designated area
Inside there is facial image.And face alignment is to look for out the position of the characteristic point of face.The processing speed of Face datection is usual
It is faster than face alignment.To advanced row Face datection in the region oriented in current image frame, can avoid not having in the region
Face and the region is carried out face alignment the occurrence of.When face location difference is larger in adjacent two field pictures, carry
Before detect in rectangular area without face, invalid face alignment work is not carried out, so as to reduce unnecessary operand.
If without facial image in the region oriented in current image frame, then it represents that the people in two field pictures in front and back
Face position difference is larger.Now can be come again to current image frame by carrying out global Face datection to current image frame
The region that middle face is located is positioned.Then the image in rectangular area for finding to global detection again carries out face alignment,
Obtain the characteristic point position of face.
As an alternative embodiment of the invention, can be being not detected by the region oriented in current image frame
During facial image, current image frame is not processed, then process next image frame, region is carried out in next image frame
Positioning.As shown in figure 5, after S302, before S303, methods described also includes：
In S501, in the region oriented in the current image frame, facial image has been detected whether.
In S502, if being not detected by facial image, process next image frame as the current image frame.
In the present embodiment, video image causes some of which frame because of reasons such as shooting or data transmission faults
Image makes a mistake.The image for making a mistake differs larger with the image of other frames of video, it is impossible to mutually corresponding, in these figures
May be without facial image in the rectangular area oriented as in.In this case, orient in current image frame is detected
Region in when there is no facial image, the process work of Face datection and face alignment can not be carried out to current image frame.This
One two field picture is likely to be the image of mistake, this two field picture is not processed, and carries out the image procossing of next frame.According to face
Picture position of the detection block in a upper picture frame, searches the rectangular area for carrying out face alignment in the image of next frame.Such as
There is face in the rectangular area that fruit finds in next image frame, then the two field picture before showing is error image.So
Avoiding the image to making a mistake carries out invalid process, reduces unnecessary computing.
Preferably, if being all not detected by face in the picture frame of continuous default frame number, the current frame image is entered
The global Face datection of row, to position to the region again.
That is, when all not having facial image in the region oriented in continuous several two field pictures, reason is possible to not be image
Error of transmission, but the characteristic point position of face has deviation in the mistake in computation of rectangular area or before image.Now such as
Fruit goes to carry out face alignment with the rectangular area of mistake again, it is impossible to find the position of the characteristic point of face.Therefore, this feelings
Global Face datection can be carried out under condition to current frame image to reacquire the position of correct rectangular area, is made to regarding
The face tracking process of frequency restarts.This avoid because the positioning feature point of the face of certain two field picture in video is inaccurate,
Or the face tracking failure that causes of position calculation mistake of rectangular area that is located of face the occurrence of.
Wherein, the default frame number represents image is entered when the region in how many two field picture is not detected by facial image
Row global detection.Default frame number can be configured according to the speed of video playback, and the broadcasting speed of such as video is per second 30
Frame, can be set to 3 frames by default frame number.Depending on the occurrence of default frame number can be according to actual conditions, it is not limited thereto.
Used as one embodiment of the present of invention, the image in region positioned in current image frame is not positive face
Image, but face is when having the image of certain face rolling angle, first the image in the region can be passed through into affine transformation
It is positive image to be transformed to face, then carries out face alignment.As shown in fig. 6, in S303, it is described by the current image frame
The image in region oriented carries out face and aligns with the front face template of face in the Face datection frame, obtains described current
The characteristic point position of face described in image, including：
In S601, affine transformation is carried out to the image in the region oriented in the current image frame, so that the area
Facial image in domain is transformed to positive facial image.
Wherein, affine transformation refers to a kind of image procossing mode for being projected image by affine transformation matrix.It is imitative
Penetrate the not conllinear coordinate points of three couple that transformation matrix can be by image before conversion and after converting in image to obtain.When finding conversion
In front and back in image during the not conllinear coordinate points of more than three pairs, it is possible to use least square method is optimized, and obtains affine transformation
The optimal solution of matrix.
In the present embodiment, the image in the region oriented in current image frame is become to the front face template of face
Change.The location lookup of characteristic point obtained will be relatively easy in rectangular area by way of image procossing out, such as eyes
The characteristic points such as two canthus, two labial angles of lip.According to the correspondence in the coordinate of these characteristic points and the front face template of face
The coordinate of characteristic point is calculating the affine transformation matrix between two width images.Again by affine transformation matrix by rectangular area
Image is transformed to positive facial image.
In S602, the positive facial image is carried out into face with the front face template of face in the Face datection frame
Alignment, obtains the characteristic point position of face in the positive facial image.
That is, the image in region oriented in current image frame is converted to after positive facial image, positive to this
Facial image carries out face alignment, obtains the characteristic point position of face in the positive facial image.
In S603, affine transformation is carried out to the positive facial image, be transformed to be positioned in the current image frame
The image in region for going out, obtains the characteristic point position of face described in the current image frame.
Affine transformation matrix according to obtaining before obtains its inverse matrix.Using the inverse matrix by conversion after it is positive
Facial image switches back to the image in the region oriented originally.Obtain the characteristic point position of face in the positive facial image
Project to the position in the image in the region, you can obtain the characteristic point position of face described in current image frame.
Affine transformation is carried out by the image in the region to orienting and is converted to positive facial image, in the face of face
Portion's rolling angle can also carry out face alignment when larger, and entire image need not be rotated.Which reduce in face
Face rolling angle it is larger in the case of carry out the operand of face alignment, improve carries out the process of face tracking to video
Efficiency, and reduce influence degree of the video definition to face tracking.
As one embodiment of the present of invention, if current image frame is the first two field picture of video playback, to current
Picture frame carries out global detection, to obtain current image frame in Face datection frame position.
That is, process video the first two field picture when, by the global detection to image to obtain image in Face datection
The position of frame.Other two field pictures afterwards are then processed according to method said before.
In the embodiment of the present invention, by the picture position using Face datection frame in video in a upper picture frame as current
Picture frame carries out the position of the image-region of face alignment, and to the image in the region feature that face positioned in alignment goes out face is carried out
Point position, and Face datection frame in current image frame is determined according to the position of the characteristic point of face.The embodiment of the present invention is to regarding
Frequency image need not be reduced and video human face is followed the trail of when carrying out face tracking to all carrying out global detection per two field picture in video
Operand, improve treatment effeciency, reduce video definition to process influence degree.
It should be understood that in embodiments of the present invention, the size of the sequence number of above-mentioned each process is not meant to the elder generation of execution sequence
Afterwards, the execution sequence of each process should determine with its function and internal logic, and should not be to the implementation process structure of the embodiment of the present invention
Into any restriction.
The method of the face tracking provided corresponding to the embodiment of the present invention, Fig. 7 shows provided in an embodiment of the present invention
The structured flowchart of the device of face tracking.For convenience of description, illustrate only part related to the present embodiment.
With reference to Fig. 7, the device of the face tracking includes：
Acquiring unit 71, for obtaining picture position of the Face datection frame in a upper picture frame.
Positioning unit 72, in current image frame, orienting the region of identical described image position.
Alignment unit 73, for image and the Face datection frame in region that will orient in the current image frame
The front face template of middle face carries out face alignment, obtains the characteristic point position of face described in the present image.
Computing unit 74, for according to the characteristic point position for getting, in the current image frame people being determined
Face detection block.
Preferably, described device also includes the first detector unit, and first detector unit is used for：
In the region oriented in the current image frame, facial image has been detected whether.
If being not detected by facial image, the Face datection of the overall situation is carried out to the current image frame, with again to the area
Domain is positioned.
Preferably, described device also includes the second detector unit, and second detector unit is used for：
In the region oriented in the current image frame, facial image has been detected whether.
If being not detected by facial image, process next image frame as the current image frame.
Preferably, described device also includes the 3rd detector unit, and the 3rd detector unit is used for：
If being all not detected by face in the picture frame of continuous default frame number, the overall situation is carried out to the current frame image
Face datection, to position to the region again.
Preferably, the alignment unit 73 is used for：
Affine transformation is carried out to the image in the region oriented in the current image frame, so that the face in the region
Image is transformed to positive facial image.
The positive facial image is carried out into face with the front face template of face in the Face datection frame to align, is obtained
The characteristic point position of face in the positive facial image.
Affine transformation is carried out to the positive facial image, is transformed in the region oriented in the current image frame
Image, obtain the characteristic point position of face described in the current image frame.
In the embodiment of the present invention, by the picture position using Face datection frame in video in a upper picture frame as current
Picture frame carries out the position of the image-region of face alignment, and to the image in the region feature that face positioned in alignment goes out face is carried out
Point position, and Face datection frame in current image frame is determined according to the position of the characteristic point of face.The embodiment of the present invention is to regarding
Frequency image need not be reduced and video human face is followed the trail of when carrying out face tracking to all carrying out global detection per two field picture in video
Operand, improve treatment effeciency, reduce video definition to process influence degree.
Presently preferred embodiments of the present invention is the foregoing is only, not to limit the present invention, all essences in the present invention
Any modification, equivalent and improvement made within god and principle etc., should be included within the scope of the present invention.
Claims (10)
1. a kind of method of face tracking, it is characterised in that include：
Obtain picture position of the Face datection frame in a upper picture frame；
In current image frame, the region of identical described image position is oriented；
The image in region that will be oriented in the current image frame enters with the front face template of face in the Face datection frame
Pedestrian's face aligns, and obtains the characteristic point position of face described in the present image；
According to the characteristic point position for getting, Face datection frame is determined in the current image frame.
2. the method for claim 1, it is characterised in that in current image frame, orient identical described image position
Region after, the image in the region that will be oriented in the current image frame and face in the Face datection frame
Front face template is carried out before face alignment, and methods described also includes：
In the region oriented in the current image frame, facial image has been detected whether；
If being not detected by facial image, the Face datection of the overall situation is carried out to the current image frame, to enter to the region again
Row positioning.
3. the method for claim 1, it is characterised in that in current image frame, orient identical described image position
Region after, the image in the region that will be oriented in the current image frame and face in the Face datection frame
Front face template is carried out before face alignment, and methods described also includes：
In the region oriented in the current image frame, facial image has been detected whether；
If being not detected by facial image, process next image frame as the current image frame.
4. method as claimed in claim 3, it is characterised in that methods described also includes：
If being all not detected by face in the picture frame of continuous default frame number, the face of the overall situation is carried out to the current frame image
Detection, to position to the region again.
5. the method for claim 1, it is characterised in that in the region that will be oriented in the current image frame
Image carries out face and aligns with the front face template of face in the Face datection frame, obtains face described in the present image
Characteristic point position, including：
Affine transformation is carried out to the image in the region oriented in the current image frame, so that the facial image in the region
It is transformed to positive facial image；
The positive facial image is carried out into face with the front face template of face in the Face datection frame to align, obtains described
The characteristic point position of face in positive facial image；
Affine transformation is carried out to the positive facial image, the figure being transformed in the region oriented in the current image frame
Picture, obtains the characteristic point position of face described in the current image frame.
6. a kind of device of face tracking, it is characterised in that include：
Acquiring unit, for obtaining picture position of the Face datection frame in a upper picture frame；
Positioning unit, in current image frame, orienting the region of identical described image position；
Alignment unit, for image and face in the Face datection frame in region that will orient in the current image frame
Front face template carry out face alignment, obtain the characteristic point position of face described in the present image；
Computing unit, for according to the characteristic point position for getting, in the current image frame Face datection being determined
Frame.
7. device as claimed in claim 6, it is characterised in that described device also includes the first detector unit, first inspection
Surveying unit is used for：
In the region oriented in the current image frame, facial image has been detected whether；
If being not detected by facial image, the Face datection of the overall situation is carried out to the current image frame, to enter to the region again
Row positioning.
8. device as claimed in claim 6, it is characterised in that described device also includes the second detector unit, second inspection
Surveying unit is used for：
In the region oriented in the current image frame, facial image has been detected whether；
If being not detected by facial image, process next image frame as the current image frame.
9. device as claimed in claim 8, it is characterised in that described device also includes the 3rd detector unit, the 3rd inspection
Surveying unit is used for：
If being all not detected by face in the picture frame of continuous default frame number, the face of the overall situation is carried out to the current frame image
Detection, to position to the region again.
10. device as claimed in claim 6, it is characterised in that the alignment unit is used for：
Affine transformation is carried out to the image in the region oriented in the current image frame, so that the facial image in the region
It is transformed to positive facial image；
The positive facial image is carried out into face with the front face template of face in the Face datection frame to align, obtains described
The characteristic point position of face in positive facial image；
Affine transformation is carried out to the positive facial image, the figure being transformed in the region oriented in the current image frame
Picture, obtains the characteristic point position of face described in the current image frame.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611248067.2A CN106650682B (en) | 2016-12-29 | 2016-12-29 | Face tracking method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611248067.2A CN106650682B (en) | 2016-12-29 | 2016-12-29 | Face tracking method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106650682A true CN106650682A (en) | 2017-05-10 |
| CN106650682B CN106650682B (en) | 2020-05-01 |
Family
ID=58835921
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611248067.2A Active CN106650682B (en) | 2016-12-29 | 2016-12-29 | Face tracking method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106650682B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108108694A (en) * | 2017-12-21 | 2018-06-01 | 北京搜狐新媒体信息技术有限公司 | A kind of man face characteristic point positioning method and device |
| CN108197613A (en) * | 2018-02-12 | 2018-06-22 | 天津天地伟业信息系统集成有限公司 | A kind of Face datection optimization algorithm based on depth convolution cascade network |
| CN108875506A (en) * | 2017-11-17 | 2018-11-23 | 北京旷视科技有限公司 | Face shape point-tracking method, device and system and storage medium |
| CN109033924A (en) * | 2017-06-08 | 2018-12-18 | 北京君正集成电路股份有限公司 | The method and device of humanoid detection in a kind of video |
| CN109255337A (en) * | 2018-09-29 | 2019-01-22 | 北京字节跳动网络技术有限公司 | Face critical point detection method and apparatus |
| WO2019033571A1 (en) * | 2017-08-17 | 2019-02-21 | 平安科技（深圳）有限公司 | Facial feature point detection method, apparatus and storage medium |
| CN109614848A (en) * | 2018-10-24 | 2019-04-12 | 百度在线网络技术（北京）有限公司 | Human body recognition method, device, equipment and computer readable storage medium |
| CN110070479A (en) * | 2019-01-31 | 2019-07-30 | 北京字节跳动网络技术有限公司 | Anamorphose pulls independent positioning method and device |
| CN110874547A (en) * | 2018-08-30 | 2020-03-10 | 富士通株式会社 | Method and device for identifying object from video |
| WO2020133330A1 (en) * | 2018-12-29 | 2020-07-02 | Zhejiang Dahua Technology Co., Ltd. | Systems and methods for video surveillance |
| CN111368593A (en) * | 2018-12-25 | 2020-07-03 | 北京右划网络科技有限公司 | Mosaic processing method and device, electronic equipment and storage medium |
| CN111914757A (en) * | 2020-08-04 | 2020-11-10 | 深圳市英威诺科技有限公司 | Face detection method and device, electronic equipment and storage medium |
| CN113989689A (en) * | 2021-11-29 | 2022-01-28 | 沈阳农业大学 | Crop pest and disease identification method and system based on unmanned aerial vehicle |
| CN115661624A (en) * | 2022-11-07 | 2023-01-31 | 浙江莲荷科技有限公司 | Digital method and device for goods shelf and electronic equipment |
| WO2023142419A1 (en) * | 2022-01-29 | 2023-08-03 | 上海商汤智能科技有限公司 | Facial tracking recognition method and apparatus, and electronic device, medium, and program product |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1908962A (en) * | 2006-08-21 | 2007-02-07 | 北京中星微电子有限公司 | People face track display method and system for real-time robust |
| 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 |
| CN101169827A (en) * | 2007-12-03 | 2008-04-30 | 北京中星微电子有限公司 | Method and device for tracking characteristic point of image |
| US20080137917A1 (en) * | 2006-12-08 | 2008-06-12 | Atsushi Okubo | Information Processing Apparatus and Information Processing Method, Recognition Apparatus and Information Recognition Method, and Program |
| CN101216885A (en) * | 2008-01-04 | 2008-07-09 | 中山大学 | Passerby face detection and tracing algorithm based on video |
| CN101635028A (en) * | 2009-06-01 | 2010-01-27 | 北京中星微电子有限公司 | Image detecting method and image detecting device |
| CN102722698A (en) * | 2012-05-17 | 2012-10-10 | 上海中原电子技术工程有限公司 | Method and system for detecting and tracking multi-pose face |
| CN103377367A (en) * | 2012-04-28 | 2013-10-30 | 中兴通讯股份有限公司 | Facial image acquiring method and device |
| CN103679125A (en) * | 2012-09-24 | 2014-03-26 | 致伸科技股份有限公司 | Human face tracking method |
| US20140098988A1 (en) * | 2012-10-04 | 2014-04-10 | Adobe Systems Incorporated | Fitting Contours to Features |
| CN104036240A (en) * | 2014-05-29 | 2014-09-10 | 小米科技有限责任公司 | Face feature point positioning method and device |
| CN105760826A (en) * | 2016-02-03 | 2016-07-13 | 歌尔声学股份有限公司 | Face tracking method and device and intelligent terminal. |
| CN106250850A (en) * | 2016-07-29 | 2016-12-21 | 深圳市优必选科技有限公司 | Face datection tracking and device, robot head method for controlling rotation and system |
-
2016
- 2016-12-29 CN CN201611248067.2A patent/CN106650682B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1908962A (en) * | 2006-08-21 | 2007-02-07 | 北京中星微电子有限公司 | People face track display method and system for real-time robust |
| 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 |
| US20080137917A1 (en) * | 2006-12-08 | 2008-06-12 | Atsushi Okubo | Information Processing Apparatus and Information Processing Method, Recognition Apparatus and Information Recognition Method, and Program |
| CN101169827A (en) * | 2007-12-03 | 2008-04-30 | 北京中星微电子有限公司 | Method and device for tracking characteristic point of image |
| CN101216885A (en) * | 2008-01-04 | 2008-07-09 | 中山大学 | Passerby face detection and tracing algorithm based on video |
| CN101635028A (en) * | 2009-06-01 | 2010-01-27 | 北京中星微电子有限公司 | Image detecting method and image detecting device |
| CN103377367A (en) * | 2012-04-28 | 2013-10-30 | 中兴通讯股份有限公司 | Facial image acquiring method and device |
| CN102722698A (en) * | 2012-05-17 | 2012-10-10 | 上海中原电子技术工程有限公司 | Method and system for detecting and tracking multi-pose face |
| CN103679125A (en) * | 2012-09-24 | 2014-03-26 | 致伸科技股份有限公司 | Human face tracking method |
| US20140098988A1 (en) * | 2012-10-04 | 2014-04-10 | Adobe Systems Incorporated | Fitting Contours to Features |
| CN104036240A (en) * | 2014-05-29 | 2014-09-10 | 小米科技有限责任公司 | Face feature point positioning method and device |
| CN105760826A (en) * | 2016-02-03 | 2016-07-13 | 歌尔声学股份有限公司 | Face tracking method and device and intelligent terminal. |
| CN106250850A (en) * | 2016-07-29 | 2016-12-21 | 深圳市优必选科技有限公司 | Face datection tracking and device, robot head method for controlling rotation and system |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109033924A (en) * | 2017-06-08 | 2018-12-18 | 北京君正集成电路股份有限公司 | The method and device of humanoid detection in a kind of video |
| WO2019033571A1 (en) * | 2017-08-17 | 2019-02-21 | 平安科技（深圳）有限公司 | Facial feature point detection method, apparatus and storage medium |
| CN108875506A (en) * | 2017-11-17 | 2018-11-23 | 北京旷视科技有限公司 | Face shape point-tracking method, device and system and storage medium |
| CN108875506B (en) * | 2017-11-17 | 2022-01-07 | 北京旷视科技有限公司 | Face shape point tracking method, device and system and storage medium |
| CN108108694B (en) * | 2017-12-21 | 2020-09-29 | 北京搜狐新媒体信息技术有限公司 | Face feature point positioning method and device |
| CN108108694A (en) * | 2017-12-21 | 2018-06-01 | 北京搜狐新媒体信息技术有限公司 | A kind of man face characteristic point positioning method and device |
| CN108197613B (en) * | 2018-02-12 | 2022-02-08 | 天地伟业技术有限公司 | Face detection optimization method based on deep convolution cascade network |
| CN108197613A (en) * | 2018-02-12 | 2018-06-22 | 天津天地伟业信息系统集成有限公司 | A kind of Face datection optimization algorithm based on depth convolution cascade network |
| CN110874547A (en) * | 2018-08-30 | 2020-03-10 | 富士通株式会社 | Method and device for identifying object from video |
| CN110874547B (en) * | 2018-08-30 | 2023-09-12 | 富士通株式会社 | Method and apparatus for identifying objects from video |
| CN109255337B (en) * | 2018-09-29 | 2020-04-28 | 北京字节跳动网络技术有限公司 | Face key point detection method and device |
| CN109255337A (en) * | 2018-09-29 | 2019-01-22 | 北京字节跳动网络技术有限公司 | Face critical point detection method and apparatus |
| CN109614848B (en) * | 2018-10-24 | 2021-07-20 | 百度在线网络技术（北京）有限公司 | Human body recognition method, device, equipment and computer readable storage medium |
| US11790483B2 (en) | 2018-10-24 | 2023-10-17 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method, apparatus, and device for identifying human body and computer readable storage medium |
| CN109614848A (en) * | 2018-10-24 | 2019-04-12 | 百度在线网络技术（北京）有限公司 | Human body recognition method, device, equipment and computer readable storage medium |
| CN111368593A (en) * | 2018-12-25 | 2020-07-03 | 北京右划网络科技有限公司 | Mosaic processing method and device, electronic equipment and storage medium |
| CN111368593B (en) * | 2018-12-25 | 2023-11-28 | 北京右划网络科技有限公司 | Mosaic processing method and device, electronic equipment and storage medium |
| WO2020133330A1 (en) * | 2018-12-29 | 2020-07-02 | Zhejiang Dahua Technology Co., Ltd. | Systems and methods for video surveillance |
| CN110070479B (en) * | 2019-01-31 | 2020-12-04 | 北京字节跳动网络技术有限公司 | Method and device for positioning image deformation dragging point |
| CN110070479A (en) * | 2019-01-31 | 2019-07-30 | 北京字节跳动网络技术有限公司 | Anamorphose pulls independent positioning method and device |
| CN111914757A (en) * | 2020-08-04 | 2020-11-10 | 深圳市英威诺科技有限公司 | Face detection method and device, electronic equipment and storage medium |
| CN113989689A (en) * | 2021-11-29 | 2022-01-28 | 沈阳农业大学 | Crop pest and disease identification method and system based on unmanned aerial vehicle |
| CN113989689B (en) * | 2021-11-29 | 2023-09-22 | 沈阳农业大学 | Crop pest and disease damage identification method and system based on unmanned aerial vehicle |
| WO2023142419A1 (en) * | 2022-01-29 | 2023-08-03 | 上海商汤智能科技有限公司 | Facial tracking recognition method and apparatus, and electronic device, medium, and program product |
| CN115661624A (en) * | 2022-11-07 | 2023-01-31 | 浙江莲荷科技有限公司 | Digital method and device for goods shelf and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106650682B (en) | 2020-05-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106650682A (en) | Method and device for face tracking | |
| CN106652972B (en) | Processing circuit of display screen, display method and display device | |
| US10462362B2 (en) | Feature based high resolution motion estimation from low resolution images captured using an array source | |
| US10147192B2 (en) | Coordinate-conversion-parameter determination apparatus, coordinate-conversion-parameter determination method, and non-transitory computer readable recording medium having therein program for coordinate-conversion-parameter determination | |
| US9769443B2 (en) | Camera-assisted two dimensional keystone correction | |
| US10410089B2 (en) | Training assistance using synthetic images | |
| EP3576017A1 (en) | Method, apparatus, and device for determining pose of object in image, and storage medium | |
| AU2011202555B2 (en) | Multi-view alignment based on fixed-scale ground plane rectification | |
| US7554575B2 (en) | Fast imaging system calibration | |
| US8385595B2 (en) | Motion detection method, apparatus and system | |
| WO2019080229A1 (en) | Chess piece positioning method and system based on machine vision, storage medium, and robot | |
| US20100232643A1 (en) | Method, Apparatus, and Computer Program Product For Object Tracking | |
| CN106981078B (en) | Sight line correction method and device, intelligent conference terminal and storage medium | |
| CN102542282B (en) | Mosaic detection method and mosaic detection device for passive images | |
| WO2016032735A1 (en) | Systems and methods for determining a seam | |
| CN108470356A (en) | A kind of target object fast ranging method based on binocular vision | |
| US10346709B2 (en) | Object detecting method and object detecting apparatus | |
| US20230209204A1 (en) | Display apparatus and camera tracking method | |
| CN112307912A (en) | Method and system for determining personnel track based on camera | |
| CN107862713A (en) | Video camera deflection for poll meeting-place detects method for early warning and module in real time | |
| CN114742866A (en) | Image registration method and device, storage medium and electronic equipment | |
| JP2018182593A (en) | Image processing apparatus and image processing method | |
| US9413477B2 (en) | Screen detector | |
| CN112672057B (en) | Shooting method and device | |
| Zheng et al. | An accurate multi-row panorama generation using multi-point joint stitching |
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 |