CN105426827B - Living body verification method, device and system - Google Patents
Living body verification method, device and system Download PDFInfo
- Publication number
- CN105426827B CN105426827B CN201510756011.7A CN201510756011A CN105426827B CN 105426827 B CN105426827 B CN 105426827B CN 201510756011 A CN201510756011 A CN 201510756011A CN 105426827 B CN105426827 B CN 105426827B
- Authority
- CN
- China
- Prior art keywords
- image
- human face
- image information
- sight line
- line vector
- 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.)
- Active
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
- G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
-
- 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
- G06V40/40—Spoof detection, e.g. liveness detection
- G06V40/45—Detection of the body part being alive
Abstract
The invention discloses a kind of living body verification methods, device and system, the method comprise the steps that generating the optic centre point by desired guiding trajectory movement, and the facial image of multiframe measurand is acquired in the optic centre point motion process, in collection process, the sight of measurand follows the optic centre point to move always；Image information is extracted to every frame facial image collected；Sight line vector is estimated according to extracted image information；The projected footprint estimated according to the sight line vector estimated；The projected footprint of the estimation and the desired guiding trajectory of the optic centre point are compared, when similarity between the two is greater than or equal to preset threshold, judge that measurand for living body, when similarity between the two is less than preset threshold, judges that measurand is not living body.
Description
Technical field
The present invention relates to computer vision field, a kind of living body verification method, device and system are related in particular to.
Background technique
In recent years, face recognition technology has significant progress.But it is mobile in many applications, such as recognition of face
Payment, video witness are opened an account, and when verifying to facial image, while needing to judge that the facial image is living body
Facial image in facial image or photo or the video of recording.
Currently used face living body verification method mainly there are several types of:
1) it by the depth information of acquisition facial image, is modeled reconstruct and is matched with three-dimensional template.This method lacks
Point is limited by environmental condition, may be not easy to obtain complete depth information, and the accuracy of three-dimensional modeling is still to be improved.
2) according to face texture detail information, by facial image some characteristic points or statistical nature information and true people
Face template is compared.But when the resolution ratio of image to be detected is lower or sufficiently complete, it can not accurately obtain grain details
When information, this method is simultaneously not suitable for.
3) Chinese patent application CN201210331141.2 discloses a kind of biopsy method, acquires multiframe face figure
Picture positions face key point/block in each frame image, is sentenced by judging whether average difference values are greater than preset threshold
It is disconnected whether living body.Chinese patent application CN201510243778.X also discloses a kind of biopsy method, also acquires
Multiframe facial image is judged by judging whether the rule of attribute change value of key point meets the changing rule of real human face
Whether living body.However, the method for such detection face global motion is since the movement of face is single biological characteristic, and it is not easy
Change, when such method is in conjunction with the identity identifying method based on face, if a people is maliciously had collected greatly by other people
My face moving image is measured for detecting, then the reliability of such method will reduce.
4) in method disclosed in Chinese patent application CN201310363154.2, pass through blinking for the face in detection image
The movement such as eye is to determine whether be living body.However, being easier to be forged, movement simple in this way of blinking so that such
The anti-fake reliability of method reduces.
Summary of the invention
It is complicated that technical problem to be solved by the present invention lies in existing living body proof scheme devices, and accuracy and reliability
It is not high.
For this purpose, the embodiment of the present invention proposes a kind of living body verification method, comprising: generate the vision by desired guiding trajectory movement
Central point, and in the optic centre point motion process acquire multiframe measurand facial image, in collection process, quilt
The sight for surveying object follows the optic centre point to move always；Image information is extracted to every frame facial image collected,
Include: that Face datection is carried out to every frame facial image, obtains human face region；To obtained human face region, face spy is determined
Sign point；According to the human face characteristic point, the position of eye is obtained, cuts out the eyes image of left and right two；Wherein described image
Information includes eyes image；Sight line vector is estimated according to extracted image information；It is obtained according to the sight line vector estimated
The projected footprint of estimation；The projected footprint of the estimation and the desired guiding trajectory of the optic centre point are compared, the two is worked as
Between similarity be greater than or equal to preset threshold when, judge measurand for living body, when similarity between the two be less than preset threshold
When value, judge that measurand is not living body.
Preferably, it is described according to extracted image information estimate sight line vector be by described image information input to mind
The sight line vector estimated through network model, the neural network model are obtained by following steps: acquisition magnanimity is different
Facial image under people, different sight；Image information and sight line vector are extracted from collected facial image；According to acquired
Image information and sight line vector obtain the neural network model.
Preferably, it is described from extracted in collected facial image in image information and sight line vector from collected people
The step of image information is extracted in face image includes: to carry out Face datection to collected every width facial image, obtains face area
Domain；To obtained human face region, human face characteristic point is determined；According to the human face characteristic point, the position of eye is obtained, is cut
The eyes image of left and right two out.
Preferably, it is described from extracted in collected facial image in image information and sight line vector from collected people
The step of image information is extracted in face image further include: the eyes image cut out is unified to same pixel size.
Preferably, the step of neural network model is obtained according to obtained image information and sight line vector packet
It includes: using obtained eyes image as input, building multilayer depth convolutional neural networks, the multilayer depth convolutional Neural net
Network is sequentially connected by convolutional layer, down-sampled layer, non-linear layer, and the last layer is the full articulamentum of f dimension, obtained view
Line vector is as output layer；Using obtained eyes image and sight line vector, to the depth convolutional neural networks built into
Row training, the training are based on back-propagation algorithm, update model parameter using stochastic gradient descent on the training data.
Preferably, it is described from extracted in collected facial image in image information and sight line vector from collected people
The step of image information is extracted in face image includes: to carry out Face datection to collected every width facial image, obtains face area
Domain；To obtained human face region, human face characteristic point is determined；The human face characteristic point is normalized；According to normalization
Human face characteristic point, obtain the position of eye, cut out two eyes images in left and right.
Preferably, it is described from extracted in collected facial image in image information and sight line vector from collected people
The step of image information is extracted in face image further include: the eyes image cut out is unified to same pixel size.
Preferably, the step of neural network model is obtained according to obtained image information and sight line vector packet
It includes: using obtained eyes image as input, building multilayer depth convolutional neural networks, the multilayer depth convolutional Neural net
Network is sequentially connected by convolutional layer, down-sampled layer, non-linear layer, and the last layer is the full articulamentum of f dimension, and will be acquired
Normalization after human face characteristic point as human face posture feature and this f dimension full articulamentum be stitched together, as expansion
Full articulamentum, obtained sight line vector is as output layer；Utilize obtained human face posture feature, eyes image and sight
Vector is trained the depth convolutional neural networks built, and the training is based on back-propagation algorithm, on the training data
Model parameter is updated using stochastic gradient descent.
Preferably, it is described from extracted in collected facial image in image information and sight line vector from collected people
The step of sight line vector is extracted in face image includes: to obtain head threedimensional model；The human face characteristic point is snapped into the head
On portion's threedimensional model；According to alignment result and optic centre the point position of obtained human face characteristic point and head threedimensional model,
Calculate sight line vector.
Preferably, described the step of image information is extracted to every frame facial image collected further include: will be cut out
Eyes image it is unified to same pixel size.
Preferably, described image information includes human face posture feature, described to extract figure to every frame facial image collected
As information further include: the human face characteristic point is normalized.
Preferably, described that image information is extracted to every frame facial image collected further include: the eye that will be cut out
Image is unified to same pixel size.
The embodiment of the invention also provides a kind of living bodies to verify device, comprising: track generates and image acquisition units, is used for
The optic centre point by desired guiding trajectory movement is generated, and acquires the multiframe of measurand in the optic centre point motion process
Facial image, in collection process, the sight of measurand follows the optic centre point to move always；Image information is extracted single
Member, for extracting image information to every frame facial image collected comprising: Face datection subelement, for every frame people
Face image carries out Face datection, obtains human face region；Human face characteristic point marks subelement, is used for obtained human face region,
Determine human face characteristic point；Eyes image cuts subelement, for obtaining the position of eye according to the human face characteristic point, cuts out
Cut the eyes image of left and right two；Wherein described image information includes eyes image；Sight line vector estimation unit is used for basis
Extracted image information estimates sight line vector；Projected footprint generation unit is estimated according to the sight line vector estimated
Projected footprint；Comparison unit compares the projected footprint of the estimation and the desired guiding trajectory of the optic centre point, when
When similarity between the two is greater than or equal to preset threshold, measurand is judged for living body, when similarity between the two is less than in advance
If when threshold value, judging that measurand is not living body.
Preferably, the sight line vector estimation unit is estimated described image information input to neural network model
Sight line vector, the neural network model obtained by following subelement: acquisition subelement, for acquire magnanimity different people,
Facial image under different sight；Extract subelement, for extracted from collected facial image image information and sight to
Amount；Neural network model generates subelement, for obtaining the neural network according to obtained image information and sight line vector
Model.
Preferably, the eyes image cuts subelement and is also used to the eyes image cut out unification to same pixel
Size.
Preferably, described image information further includes human face posture feature, described image information extraction unit further include: face
Characteristic point normalizes subelement, for the human face characteristic point to be normalized.
Preferably, the eyes image cuts subelement and is also used to the eyes image cut out unification to same pixel
Size.
The embodiment of the present invention further additionally provides a kind of living body verifying system, comprising: display device, it is default for showing
The optic centre point of track movement；Image collecting device, for acquiring measurand in the optic centre point motion process
Multiframe facial image, in collection process, the sight of measurand follows the optic centre point to move always；Processor,
For generating the optic centre point for pressing desired guiding trajectory movement；Image letter is extracted to every frame facial image collected
Breath comprising: Face datection is carried out to every frame facial image, obtains human face region；To obtained human face region, people is determined
Face characteristic point；According to the human face characteristic point, the position of eye is obtained, cuts out the eyes image of left and right two；It is wherein described
Image information includes eyes image；Sight line vector is estimated according to extracted image information；According to the sight line vector estimated
The projected footprint estimated；The projected footprint of estimation and the desired guiding trajectory of optic centre point compared, when between the two
When similarity is greater than or equal to preset threshold, measurand is judged for living body, when similarity between the two is less than preset threshold,
Judge that measurand is not living body.
Living body verification method according to an embodiment of the present invention, device and system, by acquiring quilt in real time in verification process
The facial image for surveying object, estimates the sight track of measurand according to facial image, and by the sight track that will estimate and
The actual motion track comparison of optic centre point come judge measurand whether living body, it is only necessary to setting with camera and screen
It is standby to complete to judge, do not need complicated peripheral apparatus；Allowed by the way of Eye-controlling focus the sight of measurand with
The desired guiding trajectory generated at random is mobile, it is difficult to be forged, greatly improve the accuracy and reliability of living body verifying.
Living body verification method according to an embodiment of the present invention, device and system, using obtained image information as input,
Multilayer depth convolutional neural networks are built, the multilayer depth convolutional neural networks pass through convolutional layer, down-sampled layer, non-linear layer
Be sequentially connected, the last layer be a f dimension full articulamentum, obtained sight line vector be used as output layer, and utilization obtained by
Image information and sight line vector, the depth convolutional neural networks built are trained to obtain neural network model, from
And sight line vector rapidly and accurately can be estimated from the image information extracted, and then improve the accurate of living body judgement
Degree.
Living body verification method according to an embodiment of the present invention, device and system choose eyes image as image information
It is calculated so as to simplify, detection is rapidly completed；In further preferred embodiment, human face posture feature and eye are chosen
Both images further contemplate the head of measurand there is a situation where moving, in measurand as image information
When head moves, In vivo detection still can be accurately carried out.
Detailed description of the invention
The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, and attached drawing is schematically without that should manage
Solution is carries out any restrictions to the present invention, in the accompanying drawings:
Fig. 1 shows the flow chart of living body verification method according to an embodiment of the present invention；
Fig. 2 shows the step of extracting image information to every frame facial image collected according to an embodiment of the present invention
Flow chart；
Fig. 3 shows the schematic diagram of 21 human face characteristic points；
Fig. 4 shows the flow chart of the acquisition methods of neural network model according to an embodiment of the present invention；
Fig. 5 shows the schematic diagram of neural network model according to an embodiment of the present invention；
Fig. 6 shows the flow chart of living body verification method according to another embodiment of the present invention；
Fig. 7 shows the step according to another embodiment of the present invention that image information is extracted to every frame facial image collected
Rapid flow chart；
Fig. 8 shows the schematic diagram of living body verifying device according to an embodiment of the present invention；
Fig. 9 shows the schematic diagram of living body verifying system according to an embodiment of the present invention.
Specific embodiment
Below in conjunction with attached drawing, embodiments of the present invention is described in detail.
Embodiment 1
As shown in Figure 1, living body verification method provided in this embodiment, which only needs one to have camera and screen
The equipment of curtain can be completed, and include the following steps:
S11. the optic centre point by desired guiding trajectory movement is generated, and acquires multiframe in the optic centre point motion process
Facial image.In collection process, the sight of measurand needs that optic centre point is followed to move always, watches attentively in optic centre
On point, which moves, running track can be screen on the screen according to preset running track and the method for operation
The straight line of curtain from left to right, straight line from top to bottom, a circular trace etc., but it is not limited to cited these types operation rail
Mark.In order to promote the difficulty of forgery, running track is also possible to verify the random walk track generated at random every time.Optic centre
The point method of operation can be uniform speed, be also possible to variable Rate.During the test, facial image can be also acquired in real time,
The mode of Image Acquisition can be continuous acquisition, can also be spaced acquisition, and interval time can be the same or different.
S12. image information is extracted to every frame facial image collected, in the present embodiment, chooses eyes image to make
For image information, is calculated so as to simplify, detection is rapidly completed.It is of course also possible to be extracted from facial image collected
More information, to promote the accuracy of verifying.
S13. sight line vector is estimated according to extracted eyes image.In general, the step can use neural network mould
Type is realized, i.e., extracted eyes image is input to the neural network of trained association eyes image and sight line vector
In model, so that it may the sight line vector estimated, so as to rapidly and accurately estimate sight line vector.
S14. the projected footprint estimated according to the sight line vector estimated.In general, can be first according to each frame image institute
The sight line vector estimated obtains the motion profile of sight line vector, which is projected to the plane where screen, then
The projected footprint of the motion profile for the sight line vector that can be estimated on the screen.
S15. the projected footprint of estimation is compared with desired guiding trajectory, is preset when similarity between the two is greater than or equal to
When threshold value, it can be determined that measurand is living body, when similarity between the two is less than preset threshold, it can be determined that measurand
It is not living body.
Living body verification method according to the present embodiment is the method based on Eye-controlling focus to realize that living body is verified, however existing
Eye-controlling focus scheme generally require the cooperations of the peripheral apparatus such as complicated and expensive eye tracker and light source, platform, and can not
It can apply it in In vivo detection, unlikely apply it in the In vivo detection of mobile terminal, and existing not use eye
Then often accuracy is inadequate again for the Eye-controlling focus scheme of the peripheral apparatus such as dynamic instrument, can not be applied in In vivo detection.
Living body verification method disclosed in the embodiment of the present invention, by the people for acquiring measurand in real time in verification process
Face image estimates the sight track of measurand, and the sight track by that will estimate and optic centre point according to facial image
Actual motion track comparison come judge measurand whether living body, it is only necessary to the equipment with camera and screen can be complete
At judgement, complicated peripheral apparatus is not needed；Allow the sight of measurand with generating at random by the way of Eye-controlling focus
Desired guiding trajectory is mobile, it is difficult to be forged, greatly improve the accuracy and reliability of living body verifying.In addition, implementing in the present invention
In living body verification process disclosed in example, verify between device and user without going through auditory tone cues, user need not also speak.
Preferably, as shown in Fig. 2, above-mentioned steps S12 may include:
S121. Face datection is carried out to every frame facial image, obtains human face region.
S122. to obtained human face region, human face characteristic point is marked.It is, for example, possible to use facial modelings
Method mark human face characteristic point, referring to document Supervised Descent Method and its Applications
To Face Alignment, Computer Vision and Pattern Recognition (CVPR), 2013IEEE
Conference, the 532-539 pages.The schematic diagram of 21 human face characteristic points, i.e. each 6 characteristic points of right and left eyes are shown in Fig. 3,
4 characteristic points of nose, 5 characteristic points of mouth, it will be appreciated by those skilled in the art that using more or fewer human face characteristic points
It is also feasible.
S123. according to the human face characteristic point marked, the position of eye is obtained, cuts out the eyes image of left and right two.
S124., the eyes image cut out is normalized to unified resolution ratio m × n-pixel size.
Through the above steps, it can easily, eyes image is rapidly extracted from facial image.Hereinafter will
The acquisition methods of neural network model are discussed in detail, as shown in figure 4, this method may include steps of:
S21. the facial image under magnanimity different people, different sight is acquired, can specifically include following steps:
S21a) using the method for mature camera calibration, the inner parameter of camera is obtained, meanwhile, using based on mirror
The calibration method in face estimates the three-dimensional position of screen.
S21b) using the equipment for having camera and screen, several optic centre points are generated one by one at random on the screen.
S21c collected people) is required to watch sight attentively optic centre on the screen point, when acquisition people confirms that sight has been infused
When depending on to optic centre point, facial image at this time is acquired.
S21d) repeat the above steps S21b and S21c, acquire the facial image under the different people of magnanimity, different sight.Its
In, acquisition equipment includes the camera and screen of different model, such as laptop, tablet computer, smart phone, is collected
Everybody count it is numerous, acquisition environment it is changeable.
S22. eyes image and sight line vector are extracted from collected facial image, can specifically include following steps:
S22a method identical with step S12) is utilized, eyes image is extracted from collected facial image, herein not
It repeats again.
S22b head threedimensional model) is obtained, which can be the head for modeling the obtained specific collected people in advance
Threedimensional model can also use an average head threedimensional model of predefined to different people.
S22c such as existing algorithm of EPnP algorithm) is utilized, the human face characteristic point marked is snapped into head threedimensional model
On, and successive optimization, obtain the optimum results that human face characteristic point is snapped to head threedimensional model.
S22d) according to alignment result and optic centre the point position of obtained human face characteristic point and head threedimensional model,
Calculate sight line vector, i.e., eyes (position of the eye intermediate features point as eyes) to optic centre point vector, and by this to
Amount normalizes to unit length.
S23. neural network model is obtained according to obtained eyes image and sight line vector, can specifically include following step
It is rapid:
S23a deep neural network model) is built, using the eyes image of obtained m × n resolution ratio as input, is built
Multilayer depth convolutional neural networks, depth convolutional neural networks are sequentially connected by convolutional layer, down-sampled layer, non-linear layer, most
Later layer is the full articulamentum of f dimension, and obtained sight line vector v is as output layer, as shown in Figure 5.
S23b) utilize obtained eyes image and sight line vector data, to the depth convolutional neural networks built into
Row training, obtains the deep neural network model of Eye-controlling focus.Training be based on back-propagation algorithm, utilize on the training data with
The decline of machine gradient updates model parameter.
Sight line vector is estimated by using the deep neural network model, the peripheral hardwares such as expensive eye tracker without complexity
Equipment can rapidly and accurately carry out sight line vector estimation, to improve the accuracy of living body judgement.
Embodiment 2
When the head of measurand moves, the accuracy of sight line vector estimation is influenced whether, and then influence whether
The accuracy of living body judgement.Therefore, unlike the first embodiment, when carrying out In vivo detection, it is also necessary to consider measurand
The image that moves of head.For this purpose, as shown in fig. 6, living body verification method provided in this embodiment equally only needs a band
There is the equipment of camera and screen that can complete, includes the following steps:
S31. the optic centre point by desired guiding trajectory movement is generated, and acquisition is more in the optic centre point motion process
Frame facial image.
S32. image information is extracted to every frame facial image.It is transported to eliminate the head for the measurand that may occur
The dynamic influence generated to living body verifying accuracy chooses human face posture feature and eyes image in the present embodiment as figure
As information.
S33. sight line vector is estimated according to extracted human face posture feature and eyes image.In general, can will be mentioned
The human face posture feature and eyes image taken is input to trained association face posture feature, eyes image and sight line vector
Neural network model in, so that it may the sight line vector estimated.
S34. the projected footprint estimated according to the sight line vector estimated.
S35. projected footprint and desired guiding trajectory are compared, when similarity between the two is greater than or equal to preset threshold
When, it can be determined that measurand is living body, when similarity between the two is less than preset threshold, it can be determined that measurand is not
Living body.
According to the living body verification method of the present embodiment, by the face figure for acquiring measurand in real time in verification process
Picture estimates the sight track of measurand according to facial image, and passes through the reality of the sight track and optic centre point that will estimate
Border motion profile comparison come judge measurand whether living body, it is only necessary to the equipment with camera and screen can be completed to sentence
It is disconnected, do not need complicated peripheral apparatus；Allow the sight of measurand default with what is generated at random by the way of Eye-controlling focus
Track is mobile, it is difficult to be forged, greatly improve the accuracy and reliability of living body verifying, and also further contemplate tested
The head of object is there is a situation where moving, and when measurand head moves, still can accurately carry out In vivo detection.
Preferably, as shown in fig. 7, above-mentioned steps S32 may include:
S321. every frame facial image carries out Face datection, obtains human face region.
S322. to obtained human face region, human face characteristic point is marked.
S323. the human face characteristic point marked is normalized, is located at coordinate in the coordinate range of [0,1] × [0,1],
Using the human face characteristic point after normalization as human face posture feature.
S324. according to the human face characteristic point marked, the position of eye is obtained, cuts out the eyes image of left and right two.
S325., the eyes image cut out is normalized to unified resolution ratio m × n-pixel size.
Through the above steps, it can easily, human face posture feature and eye are rapidly extracted from facial image
Image.It hereinafter will be described in detail the acquisition methods of neural network model, as shown in fig. 7, this method may include walking as follows
It is rapid:
S41. the facial image under magnanimity different people, different sight is acquired, can specifically include following steps:
S41a) using the method for mature camera calibration, the inner parameter of camera is obtained, meanwhile, using based on mirror
The calibration method in face estimates the three-dimensional position of screen.
S41b) using the equipment for having camera and screen, several optic centre points are generated one by one at random on the screen.
S41c collected people) is required to watch sight attentively optic centre on the screen point, when acquisition people confirms that sight has been infused
When depending on to optic centre point, facial image at this time is acquired.
S41d) repeat the above steps S41b and S41c, acquire the facial image under the different people of magnanimity, different sight.Together
Sample, acquisition equipment includes the camera and screen of different model, such as laptop, tablet computer, smart phone, is adopted
Collect everybody count it is numerous, acquisition environment it is changeable.
S42. human face posture feature, eyes image and sight line vector are extracted from collected facial image, it specifically can be with
Include the following steps:
S42a Face datection) is carried out to collected facial image, obtains human face region.
S42b) to obtained human face region, human face characteristic point is marked.
S42c) human face characteristic point marked is normalized, is located at coordinate in the coordinate range of [0,1] × [0,1],
Using the human face characteristic point after normalization as human face posture feature.
S42d) according to the human face characteristic point marked, the position of eye is obtained, cuts out the eyes image of left and right two,
And normalize to unified resolution ratio m × n-pixel size.
S42e head threedimensional model) is obtained.
S42f) obtained human face characteristic point is corresponded on the threedimensional model of head.
S42g) according to alignment result and optic centre the point position of obtained human face characteristic point and head threedimensional model,
Calculate sight line vector, i.e., eyes (position of the eye intermediate features point as eyes) to optic centre point vector, and by this to
Amount normalizes to unit length.
S43. neural network model is obtained according to obtained human face posture feature, eyes image and sight line vector, specifically
It may include steps of:
S23a deep neural network model) is built, using the eyes image of obtained m × n resolution ratio as input, is built
Multilayer depth convolutional neural networks, the multilayer depth convolutional neural networks are successively connected by convolutional layer, down-sampled layer, non-linear layer
It connects, the last layer is the full articulamentum of f dimension, and the full articulamentum of obtained human face posture feature and this f dimension is spelled
It is connected together, as the full articulamentum of expansion, obtained sight line vector v is also shown in FIG. 5 as output layer.
S23b) using obtained human face posture feature, eyes image and sight line vector data, the depth built is neural
Network is trained, and obtains the deep neural network model of Eye-controlling focus.Training is based on back-propagation algorithm, on the training data
Model parameter is updated using stochastic gradient descent.
Sight line vector is estimated by using the deep neural network model, the peripheral hardwares such as expensive eye tracker without complexity
Equipment can rapidly and accurately carry out sight line vector estimation, to improve the accuracy of living body judgement.
Embodiment 3
Present embodiment discloses a kind of living bodies to verify device, as shown in Figure 8, comprising:
Track generates and image acquisition units 11, for generating the optic centre point for pressing desired guiding trajectory movement, and in vision
Multiframe facial image is acquired in central point motion process, in collection process, the sight of measurand follows the vision always
Central point movement；
Image information extraction unit 12, for extracting image information to every frame facial image collected；
Sight line vector estimation unit 13, for estimating sight line vector according to extracted image information；
Projected footprint generation unit 14, the projected footprint estimated according to the sight line vector estimated；
Comparison unit 15 compares the projected footprint of estimation and the desired guiding trajectory of optic centre point, when between the two
When similarity is greater than or equal to preset threshold, measurand is judged for living body, when similarity between the two is less than preset threshold,
Judge that measurand is not living body.
Device is verified according to the living body of the present embodiment, by the face figure for acquiring measurand in real time in verification process
Picture estimates the sight track of measurand according to facial image, and passes through the reality of the sight track and optic centre point that will estimate
Border motion profile comparison come judge measurand whether living body, it is only necessary to the equipment with camera and screen can be completed to sentence
It is disconnected, compared with living body verification method in the prior art, do not need complicated platform and light source cooperation, and the accuracy judged
And high reliablity.
When image information is eyes image, which may include:
Face datection subelement obtains human face region for carrying out Face datection to every frame facial image；
Human face characteristic point marks subelement, for marking human face characteristic point to obtained human face region；
Eyes image cuts subelement, for obtaining the position of eye, cutting out a left side according to the human face characteristic point marked
Right two eyes images, and the eyes image cut out is unified to same pixel size.
As a preferred implementation manner, when image information is human face posture feature and eyes image, image information is mentioned
The unit is taken to may include:
Face datection subelement obtains human face region for carrying out Face datection to every frame facial image；
Human face characteristic point marks subelement, for marking human face characteristic point to obtained human face region；
Human face characteristic point normalizes subelement, for the human face characteristic point marked to be normalized, after normalization
Human face characteristic point as human face posture feature；
Eyes image cuts subelement, for obtaining the position of eye, cutting out a left side according to the human face characteristic point marked
Right two eyes images, and the eyes image cut out is unified to same pixel size.
The head of measurand is further contemplated as a result, there is a situation where moving, and is moved on measurand head
When, it still can accurately carry out In vivo detection.
Preferably, sight line vector estimation unit 13 is estimated described image information input to neural network model
Sight line vector, the neural network model can be obtained by following subelement:
Subelement is acquired, for acquiring the facial image under magnanimity different people, different sight；
Subelement is extracted, for extracting image information and sight line vector from collected facial image；
Neural network model generates subelement, for obtaining the nerve according to obtained image information and sight line vector
Network model.
The generation method of neural network model is identical as embodiment 1 and embodiment 2, and details are not described herein.By using this
Deep neural network model estimates sight line vector, and without complexity, the peripheral apparatus such as expensive eye tracker can be quick and precisely
Ground carries out sight line vector estimation, to improve the accuracy of living body judgement.
Embodiment 4
Present embodiment discloses a kind of living body verify system, the living body verifying system can be applied to mobile phone, tablet computer,
Laptop, PC machine and other all equipment with camera and screen, as shown in figure 9, the system includes:
Display device 21, the optic centre point 24 moved for showing desired guiding trajectory 25, which for example can be with
It is display screen；
Image collecting device 22, for acquiring the multiframe face figure of measurand in 24 motion process of optic centre point
Picture, in collection process, the sight of measurand follows optic centre point 24 to move always, which for example may be used
To be camera；
Processor 23, for generating the optic centre point 24 moved by desired guiding trajectory 25；To every frame face collected
Image zooming-out image information；Sight line vector is estimated according to extracted image information；It is obtained according to the sight line vector estimated
The projected footprint 26 of estimation；The projected footprint 26 of estimation and the desired guiding trajectory 25 of optic centre point are compared, when between the two
Similarity when being greater than or equal to preset threshold, judge measurand for living body, when similarity between the two is less than preset threshold
When, judge that measurand is not living body.
System is verified according to the living body of the present embodiment, by the face figure for acquiring measurand in real time in verification process
Picture estimates the sight track of measurand according to facial image, and passes through the reality of the sight track and optic centre point that will estimate
Border motion profile comparison come judge measurand whether living body, it is only necessary to the equipment with camera and screen can be completed to sentence
It is disconnected, do not need complicated peripheral apparatus；Allow the sight of measurand default with what is generated at random by the way of Eye-controlling focus
Track is mobile, it is difficult to be forged, greatly improve the accuracy and reliability of living body verifying.
When image information is eyes image, the step of this extracts image information to every frame facial image collected, can be with
Include:
Face datection is carried out to every frame facial image, obtains human face region；
To obtained human face region, human face characteristic point is marked；
According to the human face characteristic point marked, the position of eye is obtained, cuts out two eyes images in left and right, and by institute
The eyes image cut out is unified to same pixel size.
As a preferred implementation manner, when image information is human face posture feature and eyes image, this is to being acquired
Every frame facial image extract image information the step of may include:
Face datection is carried out to every frame facial image, obtains human face region；
To obtained human face region, human face characteristic point is marked；
The human face characteristic point marked is normalized, using the human face characteristic point after normalization as human face posture spy
Sign；
According to the human face characteristic point marked, the position of eye is obtained, cuts out two eyes images in left and right, and by institute
The eyes image cut out is unified to same pixel size.
The head of measurand is further contemplated as a result, there is a situation where moving, and is moved on measurand head
When, it still can accurately carry out In vivo detection.
As a preferred implementation manner, by by extracted eyes image or human face posture feature and eyes image two
Person is input to trained neural network model, with the sight line vector estimated, the generation method of neural network model with
Embodiment 1 and embodiment 2 are identical, and details are not described herein.Sight line vector, nothing are estimated by using the deep neural network model
Need the peripheral apparatus such as complicated and expensive eye tracker that can rapidly and accurately carry out sight line vector estimation, to improve living body
The accuracy of judgement.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the present invention
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the present invention, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Although the embodiments of the invention are described in conjunction with the attached drawings, but those skilled in the art can not depart from this hair
Various modifications and variations can be made in the case where bright spirit and scope, and such modifications and variations are each fallen within by appended claims
Within limited range.
Claims (18)
1. a kind of living body verification method characterized by comprising
The optic centre point by desired guiding trajectory movement is generated, and acquires tested pair of multiframe in the optic centre point motion process
The facial image of elephant, in collection process, the sight of measurand follows the optic centre point to move always；
Image information is extracted to every frame facial image collected comprising: Face datection is carried out to every frame facial image, is obtained
Human face region；To obtained human face region, human face characteristic point is determined；According to the human face characteristic point, the position of eye is obtained
It sets, cuts out the eyes image of left and right two；Wherein described image information includes eyes image；
Sight line vector is estimated according to extracted image information；
The projected footprint estimated according to the sight line vector estimated；
The projected footprint of the estimation and the desired guiding trajectory of the optic centre point are compared, when similarity between the two is big
When preset threshold, measurand is judged for living body, when similarity between the two is less than preset threshold, judgement is tested
Object is not living body.
2. the method according to claim 1, wherein it is described according to extracted image information estimate sight to
Amount is the sight line vector for being estimated described image information input to neural network model, and the neural network model passes through
Following steps obtain:
Acquire the facial image under magnanimity different people, different sight；
Image information and sight line vector are extracted from collected facial image；
The neural network model is obtained according to obtained image information and sight line vector.
3. according to the method described in claim 2, it is characterized in that, described extract image information from collected facial image
Include: with the step of extracting image information in sight line vector from collected facial image
Face datection is carried out to collected every width facial image, obtains human face region；
To obtained human face region, human face characteristic point is determined；
According to the human face characteristic point, the position of eye is obtained, cuts out the eyes image of left and right two.
4. according to the method described in claim 3, it is characterized in that, described extract image information from collected facial image
With the step of image information is extracted in sight line vector from collected facial image further include:
The eyes image cut out is unified to same pixel size.
5. according to the method described in claim 3, it is characterized in that, described obtain according to obtained image information and sight line vector
Include: to the step of neural network model
Using obtained eyes image as input, multilayer depth convolutional neural networks, the multilayer depth convolutional Neural are built
Network is sequentially connected by convolutional layer, down-sampled layer, non-linear layer, and the last layer is the full articulamentum of f dimension, obtained
Sight line vector is as output layer；
Using obtained eyes image and sight line vector, the depth convolutional neural networks built are trained, the instruction
Practice and be based on back-propagation algorithm, updates model parameter using stochastic gradient descent on the training data.
6. according to the method described in claim 2, it is characterized in that, described extract image information from collected facial image
Include: with the step of extracting image information in sight line vector from collected facial image
Face datection is carried out to collected every width facial image, obtains human face region；
To obtained human face region, human face characteristic point is determined；
The human face characteristic point is normalized；
According to normalized human face characteristic point, the position of eye is obtained, cuts out the eyes image of left and right two.
7. according to the method described in claim 6, it is characterized in that, described extract image information from collected facial image
With the step of image information is extracted in sight line vector from collected facial image further include:
The eyes image cut out is unified to same pixel size.
8. according to the method described in claim 6, it is characterized in that, described obtain according to obtained image information and sight line vector
Include: to the step of neural network model
Using obtained eyes image as input, multilayer depth convolutional neural networks, the multilayer depth convolutional Neural are built
Network is sequentially connected by convolutional layer, down-sampled layer, non-linear layer, and the last layer is the full articulamentum of f dimension, and by gained
To normalization after human face characteristic point be stitched together as the full articulamentum of human face posture feature and this f dimension, as opening up
The full articulamentum of exhibition, obtained sight line vector is as output layer；
Using obtained human face posture feature, eyes image and sight line vector, to the depth convolutional neural networks built into
Row training, the training are based on back-propagation algorithm, update model parameter using stochastic gradient descent on the training data.
9. according to the method described in claim 2, it is characterized in that, described extract image information from collected facial image
Include: with the step of extracting sight line vector in sight line vector from collected facial image
Obtain head threedimensional model；
The human face characteristic point is snapped on the head threedimensional model；
According to alignment result and optic centre the point position of obtained human face characteristic point and head threedimensional model, calculate sight to
Amount.
10. method according to claim 1 to 9, which is characterized in that described to every frame face figure collected
As the step of extracting image information further include:
The eyes image cut out is unified to same pixel size.
11. method according to claim 1 to 9, which is characterized in that described image information further includes face appearance
State feature, it is described that image information is extracted to every frame facial image collected further include:
The human face characteristic point is normalized.
12. according to the method for claim 11, which is characterized in that described to extract image to every frame facial image collected
Information further include:
The eyes image cut out is unified to same pixel size.
13. a kind of living body verifies device characterized by comprising
Track generates and image acquisition units, for generating the optic centre point for pressing desired guiding trajectory movement, and in the vision
The multiframe facial image of measurand is acquired in heart point motion process, in collection process, the sight of measurand follows always
The optic centre point movement；
Image information extraction unit, for extracting image information to every frame facial image collected comprising: Face datection
Unit obtains human face region for carrying out Face datection to every frame facial image；Human face characteristic point mark subelement, for pair
Obtained human face region, determines human face characteristic point；Eyes image cuts subelement, is used for according to the human face characteristic point,
The position of eye is obtained, the eyes image of left and right two is cut out；Wherein described image information includes eyes image；
Sight line vector estimation unit, for estimating sight line vector according to extracted image information；
Projected footprint generation unit, the projected footprint estimated according to the sight line vector estimated；
Comparison unit compares the projected footprint of the estimation and the desired guiding trajectory of the optic centre point, when between the two
Similarity when being greater than or equal to preset threshold, judge measurand for living body, when similarity between the two is less than preset threshold
When, judge that measurand is not living body.
14. device according to claim 13, which is characterized in that the sight line vector estimation unit is by described image information
The sight line vector for being input to neural network model to be estimated, the neural network model are obtained by following subelement:
Subelement is acquired, for acquiring the facial image under magnanimity different people, different sight；
Subelement is extracted, for extracting image information and sight line vector from collected facial image；
Neural network model generates subelement, for obtaining the neural network according to obtained image information and sight line vector
Model.
15. device described in 3 or 14 according to claim 1, which is characterized in that the eyes image cut subelement be also used to by
The eyes image cut out is unified to same pixel size.
16. device described in 3 or 14 according to claim 1, which is characterized in that described image information further includes human face posture spy
Sign, described image information extraction unit further include:
Human face characteristic point normalizes subelement, for the human face characteristic point to be normalized.
17. device according to claim 16, which is characterized in that the eyes image cuts subelement and is also used to be cut out
The eyes image cut is unified to same pixel size.
18. a kind of living body verifies system characterized by comprising
Display device, for showing the optic centre point of desired guiding trajectory movement；
Image collecting device, for acquiring the multiframe facial image of measurand in the optic centre point motion process,
In collection process, the sight of measurand follows the optic centre point to move always；
Processor, for generating the optic centre point for pressing desired guiding trajectory movement；To every frame facial image collected
Extract image information comprising: Face datection is carried out to every frame facial image, obtains human face region；To obtained face area
Human face characteristic point is determined in domain；According to the human face characteristic point, the position of eye is obtained, cuts out the eye figure of left and right two
Picture；Wherein described image information includes eyes image；Sight line vector is estimated according to extracted image information；According to estimating
The projected footprint estimated of sight line vector；The projected footprint of estimation and the desired guiding trajectory of optic centre point are compared,
When similarity between the two is greater than or equal to preset threshold, measurand is judged for living body, the similarity when between the two is less than
When preset threshold, judge that measurand is not living body.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510756011.7A CN105426827B (en) | 2015-11-09 | 2015-11-09 | Living body verification method, device and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510756011.7A CN105426827B (en) | 2015-11-09 | 2015-11-09 | Living body verification method, device and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105426827A CN105426827A (en) | 2016-03-23 |
| CN105426827B true CN105426827B (en) | 2019-03-08 |
Family
ID=55505027
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510756011.7A Active CN105426827B (en) | 2015-11-09 | 2015-11-09 | Living body verification method, device and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105426827B (en) |
Families Citing this family (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105956572A (en) * | 2016-05-15 | 2016-09-21 | 北京工业大学 | In vivo face detection method based on convolutional neural network |
| CN106599883B (en) * | 2017-03-08 | 2020-03-17 | 王华锋 | CNN-based multilayer image semantic face recognition method |
| CN108734057A (en) * | 2017-04-18 | 2018-11-02 | 北京旷视科技有限公司 | The method, apparatus and computer storage media of In vivo detection |
| CN107066983B (en) * | 2017-04-20 | 2022-08-09 | 腾讯科技（上海）有限公司 | Identity verification method and device |
| CN108229284B (en) * | 2017-05-26 | 2021-04-09 | 北京市商汤科技开发有限公司 | Sight tracking and training method and device, system, electronic equipment and storage medium |
| CN108875469A (en) * | 2017-06-14 | 2018-11-23 | 北京旷视科技有限公司 | In vivo detection and identity authentication method, device and computer storage medium |
| CN107590429A (en) * | 2017-07-20 | 2018-01-16 | 阿里巴巴集团控股有限公司 | The method and device verified based on eyeprint feature |
| CN107545248B (en) * | 2017-08-24 | 2021-04-02 | 北京小米移动软件有限公司 | Biological characteristic living body detection method, device, equipment and storage medium |
| CN109726613B (en) | 2017-10-27 | 2021-09-10 | 虹软科技股份有限公司 | Method and device for detection |
| CN107992842B (en) * | 2017-12-13 | 2020-08-11 | 深圳励飞科技有限公司 | Living body detection method, computer device, and computer-readable storage medium |
| CN108829247B (en) * | 2018-06-01 | 2022-11-15 | 北京市商汤科技开发有限公司 | Interaction method and device based on sight tracking and computer equipment |
| CN108921209A (en) * | 2018-06-21 | 2018-11-30 | 杭州骑轻尘信息技术有限公司 | Image identification method, device and electronic equipment |
| CN110853073A (en) * | 2018-07-25 | 2020-02-28 | 北京三星通信技术研究有限公司 | Method, device, equipment and system for determining attention point and information processing method |
| CN109376595B (en) * | 2018-09-14 | 2023-06-23 | 杭州宇泛智能科技有限公司 | Monocular RGB camera living body detection method and system based on human eye attention |
| CN110969061A (en) * | 2018-09-29 | 2020-04-07 | 北京市商汤科技开发有限公司 | Neural network training method, neural network training device, visual line detection method, visual line detection device and electronic equipment |
| CN109635554A (en) * | 2018-11-30 | 2019-04-16 | 努比亚技术有限公司 | A kind of red packet verification method, terminal and computer storage medium |
| CN111291607B (en) * | 2018-12-06 | 2021-01-22 | 广州汽车集团股份有限公司 | Driver distraction detection method, driver distraction detection device, computer equipment and storage medium |
| CN109711309B (en) * | 2018-12-20 | 2020-11-27 | 北京邮电大学 | Method for automatically identifying whether portrait picture is eye-closed |
| CN109886080A (en) * | 2018-12-29 | 2019-06-14 | 深圳云天励飞技术有限公司 | Human face in-vivo detection method, device, electronic equipment and readable storage medium storing program for executing |
| CN109977764A (en) * | 2019-02-12 | 2019-07-05 | 平安科技（深圳）有限公司 | Vivo identification method, device, terminal and storage medium based on plane monitoring-network |
| CN111967293A (en) * | 2020-06-22 | 2020-11-20 | 云知声智能科技股份有限公司 | Face authentication method and system combining voiceprint recognition and attention detection |
| CN111881431B (en) * | 2020-06-28 | 2023-08-22 | 百度在线网络技术（北京）有限公司 | Man-machine verification method, device, equipment and storage medium |
| CN112287909B (en) * | 2020-12-24 | 2021-09-07 | 四川新网银行股份有限公司 | Double-random in-vivo detection method for randomly generating detection points and interactive elements |
| CN112633217A (en) * | 2020-12-30 | 2021-04-09 | 苏州金瑞阳信息科技有限责任公司 | Human face recognition living body detection method for calculating sight direction based on three-dimensional eyeball model |
| CN113505756A (en) * | 2021-08-23 | 2021-10-15 | 支付宝(杭州)信息技术有限公司 | Face living body detection method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103400122A (en) * | 2013-08-20 | 2013-11-20 | 江苏慧视软件科技有限公司 | Method for recognizing faces of living bodies rapidly |
| CN103440479A (en) * | 2013-08-29 | 2013-12-11 | 湖北微模式科技发展有限公司 | Method and system for detecting living body human face |
| CN103593598A (en) * | 2013-11-25 | 2014-02-19 | 上海骏聿数码科技有限公司 | User online authentication method and system based on living body detection and face recognition |
| CN104966070A (en) * | 2015-06-30 | 2015-10-07 | 北京汉王智远科技有限公司 | Face recognition based living body detection method and apparatus |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8260008B2 (en) * | 2005-11-11 | 2012-09-04 | Eyelock, Inc. | Methods for performing biometric recognition of a human eye and corroboration of same |
-
2015
- 2015-11-09 CN CN201510756011.7A patent/CN105426827B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103400122A (en) * | 2013-08-20 | 2013-11-20 | 江苏慧视软件科技有限公司 | Method for recognizing faces of living bodies rapidly |
| CN103440479A (en) * | 2013-08-29 | 2013-12-11 | 湖北微模式科技发展有限公司 | Method and system for detecting living body human face |
| CN103593598A (en) * | 2013-11-25 | 2014-02-19 | 上海骏聿数码科技有限公司 | User online authentication method and system based on living body detection and face recognition |
| CN104966070A (en) * | 2015-06-30 | 2015-10-07 | 北京汉王智远科技有限公司 | Face recognition based living body detection method and apparatus |
Non-Patent Citations (1)
| Title |
|---|
| 人脸识别中的活体检测技术研究;孙霖;《中国博士学位论文全文数据库 信息科技辑》;20110815(第8期);I138-69 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105426827A (en) | 2016-03-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105426827B (en) | Living body verification method, device and system | |
| CN108256433B (en) | Motion attitude assessment method and system | |
| Dikovski et al. | Evaluation of different feature sets for gait recognition using skeletal data from Kinect | |
| CN112184705B (en) | Human body acupuncture point identification, positioning and application system based on computer vision technology | |
| CN104966070B (en) | Biopsy method and device based on recognition of face | |
| CN103177269B (en) | For estimating the apparatus and method of object gesture | |
| CN108345869A (en) | Driver's gesture recognition method based on depth image and virtual data | |
| CN108369785A (en) | Activity determination | |
| CN105518708A (en) | Method and equipment for verifying living human face, and computer program product | |
| CN104123543B (en) | A kind of eye movement recognition methods based on recognition of face | |
| Viraktamath et al. | Face detection and tracking using OpenCV | |
| CN107895160A (en) | Human face detection and tracing device and method | |
| CN109472198A (en) | A kind of video smiling face's recognition methods of attitude robust | |
| CN105740780A (en) | Method and device for human face in-vivo detection | |
| CN106650619A (en) | Human action recognition method | |
| CN110211222B (en) | AR immersion type tour guide method and device, storage medium and terminal equipment | |
| CN109598242A (en) | A kind of novel biopsy method | |
| KR20220028654A (en) | Apparatus and method for providing taekwondo movement coaching service using mirror dispaly | |
| Zhu | Computer vision-driven evaluation system for assisted decision-making in sports training | |
| CN110796101A (en) | Face recognition method and system of embedded platform | |
| CN107480586A (en) | Bio-identification photo bogus attack detection method based on human face characteristic point displacement | |
| CN108765014A (en) | A kind of intelligent advertisement put-on method based on access control system | |
| CN109543629A (en) | A kind of blink recognition methods, device, equipment and readable storage medium storing program for executing | |
| Yang et al. | Research on face recognition sports intelligence training platform based on artificial intelligence | |
| RU2005100267A (en) | METHOD AND SYSTEM OF AUTOMATIC VERIFICATION OF THE PRESENCE OF A LIVING FACE OF A HUMAN IN BIOMETRIC SECURITY SYSTEMS |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |