CN109934207A - A kind of characteristic distance modification method of driver face based on facial expression fatigue driving detection algorithm - Google Patents
A kind of characteristic distance modification method of driver face based on facial expression fatigue driving detection algorithm Download PDFInfo
- Publication number
- CN109934207A CN109934207A CN201910297915.6A CN201910297915A CN109934207A CN 109934207 A CN109934207 A CN 109934207A CN 201910297915 A CN201910297915 A CN 201910297915A CN 109934207 A CN109934207 A CN 109934207A
- Authority
- CN
- China
- Prior art keywords
- mouth
- head
- distance
- driver
- angle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of characteristic distance modification methods of driver face based on facial expression fatigue driving detection algorithm, fluctuated using driver, rock from side to side head or torticollis when, laterally or longitudinally the variable angle of Eigenvector and camera positive direction releases the actual value of mouth and eyes laterally or longitudinally characteristic distance for mouth and eyes；Right hand rectangular coordinate system in space is established centered on the head of driver, the direction of x-axis is that left ear is directed toward auris dextra, and the direction of y-axis is to be directed toward the crown, and the direction of z-axis is that forehead is directed toward the back side of head.Obtain rotation angle of the head in three reference axis, wherein pitch angle yaw indicates the angle that driver head rotates around y-axis, yaw angle pitch indicates the distance that driver head rotates around x-axis, roll angle roll indicates the angle that driver head rotates around z-axis, the i.e. Eulerian angles of driver head in three dimensions.Characteristic distance amendment is carried out using obtained Eulerian angles, promotes driver's facial expression fatigue driving accuracy in detection and precision.
Description
Technical field
The present invention relates to fatigue drivings to detect septum reset human facial feature extraction technical field, specifically based on facial table
A kind of characteristic distance modification method of driver face of feelings fatigue driving detection algorithm.
Background technique
A variety of typical sleepy performances can be presented in the rear face for entering fatigue state in people, such as upper palpebra inferior be can't help
Ground occurs that the time that closure trend, frequency of wink in a period of time are significantly raised, blink every time is progressively longer, mouth is opened slightly
It opens, yawn.As the most active two parts of face, the movement of eyes and mouth best embody out it is above-mentioned it is sleepy under table
End of love.With the development of computer vision technique, human face detection tech has been more and more maturation, is based on computer at this stage
The Face datection of vision either detection accuracy still detects speed and is substantially accomplished commercial degree, this is also based on computer
The fatigue driving detection of vision is laid a good foundation.The fatigue-driving detection technology based on computer vision of mainstream at present
First is that in real time acquisition driver facial expression picture, the fatigue characteristic of facial expression is extracted, by dividing fatigue characteristic
Analysis is to differentiate whether driver belongs to fatigue driving.Wherein the fatigue characteristic of facial expression is primarily referred to as the closure of mouth and eyes
Degree, the lateral distance between fore-and-aft distance and upper and lower lip feature point between upper lip and lower lip are known as the spy of mouth
Distance is levied, the closure degree of mouth is measured with this；Fore-and-aft distance and upper lower eyelid feature between upper eyelid and lower eyelid
Lateral distance between point is known as the characteristic distance of eyes, and the closure degree of eyes is measured with this.
The characteristic distance of eyes and mouth be reflect mouth and eyes closed degree apart from variable, it is practical to be closed degree
On be the distance between palpebra inferior on eyes, or for being the distance between upper lower lip for mouth, best embody out this
Plant range information is exactly the spacing between characteristic point, and this spacing is claimed to be characterized distance (Feature Length, FL), such as attached
Shown in Fig. 1.
In view of directly using characteristic distance as fatigue characteristic and infeasible, because in a specific facial expression figure
In piece, the shadow for the factors such as the characteristic distance of eyes and mouth can be not allowed by facial differences, camera lens distance, Partial Feature point location
It rings, so that it is inaccurate to cause fatigue characteristic to extract.Therefore consider to use the combination of characteristic distance as fatigue characteristic.Combined distance
Feature extraction it is as shown in Fig. 2, extract formula it is as follows.
Wherein e_ll be left eye on the left of upper lower eyelid between characteristic distance, e_lr be left eye on the right side of upper lower eyelid it
Between characteristic distance；E_lu is the transverse features distance of left eye upper eyelid, and e_ld is the transverse features distance of left eye lower eyelid；e_
Rl is the characteristic distance between the upper lower eyelid on the left of right eye, and e_rr is the characteristic distance between the upper lower eyelid on the right side of right eye；
E_ru is the transverse features distance of right eye upper eyelid, and e_rd is the transverse features distance of right eye lower eyelid；M_l is on mouth left side
Characteristic distance between lower lip, m_r are the characteristic distance on the right side of mouth between upper lower lip；M_u is the cross of mouth upper lip
To characteristic distance, m_d is the transverse features distance of mouth lower lip.
Use the combination of characteristic distance as fatigue characteristic can be well solved because of facial differences, camera lens is far and near, part is special
Influence of the factors such as sign point location is inaccurate to characteristic distance and caused by fatigue characteristic extract inaccurate problem.But in reality also
There are another merely with the above-mentioned insoluble problem of fatigue characteristic extracting mode.Consider following several Driving Scenes:
1) driver observes vehicle left side or right side；2) driver, which bows, checks instrument board；3) road conditions are poor, and vehicle jolts in driving
Seriously；4) camera actual installation angle.In above several situations, although can be solved using the width-to-height ratio of eyes and mouth
The problem of certainly facial dimension camera distance causes, it is contemplated that actual conditions, since the installation site of camera can not
Complete face driver face, in addition drive the complexity of road conditions, driving conditions jolt or the needs of actual conditions (are such as checked
Instrument board perhaps rearview mirror) driver head is likely to occur left and right or upper and lower shaking by a small margin, so camera captures
To face-image can not be always the same angle.In view of 2D camera used in practical application, such as 3 institute of attached drawing
Show, when driver's face camera, the characteristic distance size that the length of line segment exactly needs to extract, and when the non-face in head is taken the photograph
When as head, since face occurs certain deflection angle relative to camera, the characteristic distance that camera captures at this time occurs
Deviation, thus it is according to the calculated fatigue characteristic of fatigue characteristic extracting method set forth above not accurate enough.
In order to overcome this defect, it is necessary to use a kind of reasonable modification method, to the characteristic distances of eyes and mouth into
Row amendment has the fatigue driving detection based on facial expression feature higher to guarantee the accuracy that fatigue characteristic extracts
Precision.
Summary of the invention
The purpose of the invention is to overcome in above-mentioned fatigue characteristic extracting method to draw because of driver head's attitudes vibration
The characteristic distance error risen, provides a kind of characteristic distance modification method of driver's face, corrects feature using head pose angle
Distance.To promote the accuracy extracted based on fatigue characteristic in facial expression fatigue driving detection algorithm, driver's head is reduced
Influence of portion's posture to fatigue driving detection accuracy promotes the accuracy and precision of fatigue driving detection.
Realizing the specific technical solution of the object of the invention is:
A kind of characteristic distance modification method of driver's face based on facial expression fatigue driving detection algorithm, this method
Comprising the following specific steps
Step 1: the facial expression feature of driver being refined as mouth feature and eye feature, respectively to eyes and mouth
Characteristic point mark is carried out, demarcates the transverse features distance and longitudinal feature of eyes and mouth respectively using the characteristic point marked
Distance, solves longitudinal characteristic distance of left eye, right eye and mouth and the ratio of transverse features distance respectively, and building facial expression is special
Levy vector Ft, calculate the opening and closing degree of eyes and mouth；Facial expression feature vector solution formula is as follows:
Wherein e_ll be left eye on the left of upper lower eyelid between characteristic distance, e_lr be left eye on the right side of upper lower eyelid it
Between characteristic distance；E_lu is the transverse features distance of left eye upper eyelid, and e_ld is the transverse features distance of left eye lower eyelid；e_
Rl is the characteristic distance between the upper lower eyelid on the left of right eye, and e_rr is the characteristic distance between the upper lower eyelid on the right side of right eye；
E_ru is the transverse features distance of right eye upper eyelid, and e_rd is the transverse features distance of right eye lower eyelid；M_l is on mouth left side
Characteristic distance between lower lip, m_r are the characteristic distance on the right side of mouth between upper lower lip；M_u is the cross of mouth upper lip
To characteristic distance, m_d is the transverse features distance of mouth lower lip；
Step 2: establishing right hand rectangular coordinate system in space centered on driver head, obtain head in three dimensions three
Rotational angle yaw, pitch, roll in a reference axis, the i.e. Eulerian angles of head in three dimensions；
Wherein pitch angle yaw indicates the angle that driver head rotates around y-axis, and yaw angle pitch indicates driver head
Around the distance of x-axis rotation, roll angle roll indicates the angle that driver head rotates around z-axis；Wherein:
Described that right hand rectangular coordinate system in space is established centered on driver head, wherein the direction of x-axis is left ear direction
Auris dextra, the direction of y-axis are to be directed toward the crown, and the direction of z-axis is that forehead is directed toward the back side of head, obtain driver head in three dimensions
Three Eulerian angles yaw, pitch and roll, characterize the posture of driver head in three dimensions with this three Eulerian angles；
Step 3: when driver fluctuate head, rock from side to side head or torticollis when, extract at this time head in three-dimensional space
Between in three Eulerian angles, while extract eyes and mouth transverse features distance and longitudinal characteristic distance, with mouth and eye
Lateral, longitudinal characteristic distance of eyeball is the right-angle side of right angled triangle；Using the Eulerian angles of head in three dimensions as right angle
The angle of right-angle side and bevel edge in triangle；Construct flat square triangle；It specifically includes:
(1) when driver fluctuates head, i.e., head is rotated around x-axis, and pitch yaw angle changes；For
The transverse features distance of eyes and mouth, when facial face camera, lateral Eigenvector at this time perpendicular to camera just
Pair direction；Driver makes the movement bowed or come back and fluctuates head, at this time these transverse features line segments with take the photograph
As the angle of head positive direction does not change, therefore the transverse features that camera captures are constant apart from size；For eye
Eyeball and mouth longitudinal direction characteristic distance, when facial face camera, longitudinal Eigenvector is also perpendicularly to camera face at this time
Direction；Driver makes the movement bowed or come back and fluctuates head, longitudinal Eigenvector and camera positive direction
Angle become smaller, the longitudinal characteristic distance captured by camera shortens, therefore vertical line segment is (as indicated eyes and mouth
Hold the line segment of degree) it then can shorter；Longitudinal characteristic distance of the eyes and mouth that are observed at this time using camera is right angle
Right side of the triangle, using longitudinal characteristic distance of revised eyes and mouth as the bevel edge of right angled triangle, with the angle pitch
For the angle between the right-angle side and bevel edge that are constructed, flat square triangle is constructed；
(2) when driver or so shakes the head, i.e., head is rotated around y-axis, and yaw pitch angle changes, for eyes and
Longitudinal characteristic distance of mouth, when driver turn one's head observe left and right vehicle wheel two sides rock from side to side head when, longitudinal feature at this time
Line segment and the angle of camera positive direction do not change, therefore longitudinal characteristic distance size that camera captures is not
Become；For the transverse features distance of eyes and mouth, when driver turn one's head observe left and right vehicle wheel two sides rock from side to side head when,
Transverse features line segment becomes smaller with camera positive direction angle, the transverse features Distance Shortened captured by camera, therefore horizontal
It can then shorten to line segment；The transverse features distance of the eyes and mouth that are observed at this time using camera is the right angle of right angled triangle
Side, using the transverse features distance of revised eyes and mouth as the bevel edge of right angled triangle, with the angle yaw for the right angle that is constructed
Angle between side and bevel edge constructs flat square triangle；
(3) when driver is to the left or when torticollis to the right, i.e., head is moved around z-axis, and roll roll angle changes, right
In the transverse features distance of eyes and mouth, when driver is to the left or when torticollis to the right, transverse features line segment is perpendicular to taking the photograph
As head positive direction plane on rotate, the angle relative to camera positive direction do not change, therefore camera captures
The transverse features of eyes and mouth will not change apart from size；For longitudinal characteristic distance of eyes and mouth, work as driving
Member to the left or to the right torticollis when, longitudinal Eigenvector rotates in the plane perpendicular to camera positive direction, relative to taking the photograph
As the angle of the positive direction of head does not change, longitudinal characteristic distance size of eyes and mouth that camera captures will not occur
Variation；The characteristic distance of all eyes and mouth rotates in the plane perpendicular to z-axis at this time, the eyes that camera captures
It will not change with the characteristic distance length of mouth；Therefore, lateral, longitudinal characteristic distance of eyes and mouth is without being repaired
Just, i.e., at this time without establishing flat square triangle.
Step 4: solving the length of bevel edge, required bevel edge using Pythagorean theorem in the flat square triangle of building
The true value of lateral, longitudinal characteristic distance of eyes and mouth when length is driver's face non-face camera, i.e. base
The characteristic distance of eyes and mouth is corrected in the head pose of driver；Its modified detailed process are as follows:
(1) when driver fluctuates head, i.e., head is rotated around x-axis, and pitch yaw angle changes；If eye
Length of the longitudinal characteristic distance of eyeball and mouth before pitch rotation occurs is l_r, the length that cameras view arrives after rotation
Degree is l_s, and according to the flat square triangle that (1) of step 3 constructs, longitudinal feature of eyes and mouth can be obtained by Pythagorean theorem
The modified computing formulae of distance is as follows:
L_r=l_s/cos (pitch)
Wherein length of the l_r for longitudinal characteristic distance of eyes and mouth before pitch rotation occurs, l_s is eyes
With longitudinal characteristic distance of mouth length that cameras view arrives after pitch rotation occurs, pitch be driver bow or
Person come back (fluctuate head) when angle i.e. driver head around x-axis rotation angle；
(2) when driver or so shakes the head, i.e., head is rotated around y-axis, and yaw pitch angle changes；If eyes and mouth
Bar transverse features distance occur yaw rotation before length be l_r, the length that cameras view arrives after rotation be l_s,
According to the flat square triangle that (2) of step 3 construct, repairing for the transverse features distance of eyes and mouth can be obtained by Pythagorean theorem
Positive calculation formula is as follows:
L_r=l_s/cos (yaw)
Wherein l_r is length of the transverse features distance of eyes and mouth before yaw rotation occurs, l_s be eyes and
The transverse features distance of the mouth length that cameras view arrives after yaw rotation occurs, yaw are when driver or so shakes the head
The rotation angle of angle, i.e. driver head around y-axis；
(3) when driver is to the left or when torticollis to the right, i.e., head is moved around z-axis, and roll roll angle changes；This
When eyes and lateral, longitudinal characteristic distance of mouth be not necessarily to be modified, therefore without correction formula；
Step 5: solving facial expression feature vector again using lateral, longitudinal characteristic distance of eyes after amendment and mouth
F, solution formula are as follows:
Wherein e_ll be left eye on the left of upper lower eyelid between characteristic distance, e_lr be left eye on the right side of upper lower eyelid it
Between characteristic distance；E_lu is the transverse features distance of left eye upper eyelid, and e_ld is the transverse features distance of left eye lower eyelid；e_
Rl is the characteristic distance between the upper lower eyelid on the left of right eye, and e_rr is the characteristic distance between the upper lower eyelid on the right side of right eye；
E_ru is the transverse features distance of right eye upper eyelid, and e_rd is the transverse features distance of right eye lower eyelid；M_l is on mouth left side
Characteristic distance between lower lip, m_r are the characteristic distance on the right side of mouth between upper lower lip；M_u is the cross of mouth upper lip
To characteristic distance, m_d is the transverse features distance of mouth lower lip；Angle when pitch bows or comes back for driver is i.e.
Rotation angle of the driver head around x-axis；Yaw is rotation of angle i.e. driver head when driver or so shakes the head around y-axis
Angle；FtFor the facial expression feature vector before amendment；
Step 6: using the facial expression feature vector F solved again as driver fatigue feature extraction formula, extraction is driven
The fatigue characteristic for the person of sailing carries out fatigue driving judgement.
The present invention is made to promote the facial fatigue characteristic extraction accuracy based on facial expression fatigue driving detection algorithm
The facial fatigue characteristic of driver is unrelated with the head pose of driver.And it proposes to be based on facial expression fatigue driving detection algorithm
A kind of driver face characteristic distance modification method.
It can be used for using the present invention as the new fatigue characteristic extraction extracted driver's face fatigue characteristic of formula tired
Please the fatigue driving state judgement for sailing detection system, improves the precision and real-time of fatigue driving detection.It simultaneously can be used for
Other any need carry out the scene of fatigue state detection, as the input of fatigue detecting system, to realize in specific application
The high-precision fatigue state detection unrelated with head pose.
The present invention calculates head Eulerian angles in three dimensions using head pose algorithm, gradually analyzes and how to utilize Europe
Angle is drawn to solve the problems, such as to extract characteristic distance in facial non-face camera that final export considers that characteristic distance is repaired there are deviation
New fatigue characteristic extracts formula after just, is proposed the fatigue characteristic extracting method unrelated with head pose.It improves and is based on
The accuracy of facial expression feature method for detecting fatigue driving, method proposed by the present invention, which can be used for other, at the same time needs
Facial expression is carried out in the scene of characteristic distance extraction, to promote the accuracy of facial expression feature extraction.Therefore, originally
Invention is considered a kind of general facial expression feature distance correction methods, has very strong practicability.
Detailed description of the invention
Fig. 1 is the schematic diagram of eyes and mouth characteristic distance；
Fig. 2 is the extraction schematic diagram of assemblage characteristic；
Fig. 3 is the influence schematic diagram that head pose extracts fatigue characteristic；
Fig. 4 is rectangular coordinate system in space schematic diagram；
Fig. 5 is the Eulerian angles schematic diagram of head in a space rectangular coordinate system；
Fig. 6 is characterized distance correction schematic diagram；
Fig. 7 is the partial data schematic diagram of YawDD public data collection.
Specific embodiment
Present invention will now be described in detail with reference to the accompanying drawings..
The present invention comprising the following specific steps
Step 1: the facial video data of driver is acquired in real time using the video camera being located at immediately ahead of driver, to being adopted
Driver's face picture in every one-frame video data of collection carries out the characteristic point mark of eyes and mouth, and with lower lip and
Lateral distance between fore-and-aft distance and upper lower lip and upper lower eyelid characteristic point between upper lower eyelid characteristic point, which is used as, drives
Lateral, longitudinal characteristic distance of the person's of sailing eyes and mouth, specific implementation are as shown in Fig. 1.
Step 2: shown in attached drawing 2, describing the characteristic distance between the upper lower eyelid on the left of left eye with e_ll, e_lr description is left
Characteristic distance between the upper lower eyelid on eye right side；E_lu describes the transverse features distance of left eye upper eyelid, and e_ld describes left eye
The transverse features distance of lower eyelid；E_rl describes the characteristic distance between the upper lower eyelid on the left of right eye, and e_rr describes the right eye right side
Characteristic distance between the upper lower eyelid of side；E_ru describes the transverse features distance of right eye upper eyelid, and e_rd describes eye under right eye
The transverse features distance of skin；M_l describes the characteristic distance on the left of mouth between upper lower lip, and m_r describes mouth above and below on the right side of mouth
Characteristic distance between lip；M_u describes the transverse features distance of mouth upper lip, m_ describe the transverse features of mouth lower lip away from
From.To construct facial expression assemblage characteristic vector Ft, specific formula is as follows:
Step 3, as shown in Fig. 5, right hand rectangular coordinate system in space is established centered on driver head, head is obtained and exists
Rotational angle yaw, pitch, roll in three-dimensional space in three reference axis, the i.e. Eulerian angles of head in three dimensions.Its
The direction of middle x-axis is that left ear is directed toward auris dextra, and the direction of y-axis is to be directed toward the crown, and the direction of z-axis is that forehead is directed toward the back side of head；Pitching
Angle yaw indicates the angle that driver head rotates around y-axis, and yaw angle pitch indicates the distance that driver head rotates around x-axis,
Roll angle roll indicates the angle that driver head rotates around z-axis
Step 4: as shown in Fig. 6, for driver fluctuate head, rock from side to side the three kinds of situations in head or torticollis,
Lateral, longitudinal characteristic distance of corresponding driver head's posture Eulerian angles and corresponding eyes and mouth is extracted respectively, and
Building is used for the modified flat square triangle of lateral, longitudinal characteristic distance of eyes and mouth.
(1) when driver fluctuates head, i.e., head is rotated around x-axis, and pitch yaw angle changes.For
The transverse features distance of eyes and mouth, when facial face camera, lateral Eigenvector at this time perpendicular to camera just
Pair direction.Driver makes the movement bowed or come back and fluctuates head, at this time these transverse features line segments with take the photograph
As the angle of head positive direction does not change, therefore the transverse features that camera captures are constant apart from size.For eye
Eyeball and mouth longitudinal direction characteristic distance, when facial face camera, longitudinal Eigenvector is also perpendicularly to camera face at this time
Direction.Driver makes the movement bowed or come back and fluctuates head, longitudinal Eigenvector and camera positive direction
Angle become smaller, the longitudinal characteristic distance captured by camera shortens, therefore vertical line segment is (as indicated eyes and mouth
Hold the line segment of degree) it then can shorter.Longitudinal characteristic distance of the eyes and mouth that are arrived at this time using cameras view is right angle
Right side of the triangle, using longitudinal characteristic distance of revised eyes and mouth as the bevel edge of right angled triangle, with the angle pitch
For the angle between the right-angle side and bevel edge that are constructed, formation level right angled triangle；
(2) when driver or so shakes the head, i.e., head is rotated around y-axis, and yaw pitch angle changes, for eyes and
Longitudinal characteristic distance of mouth, when driver turn one's head observe left and right vehicle wheel two sides rock from side to side head when, longitudinal feature at this time
Line segment and the angle of camera positive direction do not change, therefore longitudinal characteristic distance size that camera captures is not
Become.For the transverse features distance of eyes and mouth, when driver turn one's head observe left and right vehicle wheel two sides rock from side to side head when,
Transverse features line segment becomes smaller with camera positive direction angle, the transverse features Distance Shortened captured by camera, therefore horizontal
It can then shorten to line segment.The transverse features distance of the eyes and mouth that are arrived at this time using cameras view is the right angle of right angled triangle
Side, using the transverse features distance of revised eyes and mouth as the bevel edge of right angled triangle, with the angle yaw for the right angle that is constructed
Angle between side and bevel edge, formation level right angled triangle；
(3) when driver is to the left or when torticollis to the right, i.e., head is moved around z-axis, and roll roll angle changes, right
In the transverse features distance of eyes and mouth, when driver is to the left or when torticollis to the right, transverse features line segment is perpendicular to taking the photograph
As head positive direction plane on rotate, the angle relative to camera positive direction do not change, therefore camera captures
The transverse features of eyes and mouth will not change apart from size.For longitudinal characteristic distance of eyes and mouth, work as driving
Member to the left or to the right torticollis when, longitudinal Eigenvector rotates in the plane perpendicular to camera positive direction, relative to taking the photograph
As the angle of the positive direction of head does not change, therefore longitudinal characteristic distance size of eyes and mouth that camera captures will not
It changes.So when all eyes and the characteristic distance of mouth rotated in the plane perpendicular to z-axis, camera captures
Eyes and the characteristic distance length of mouth will not change.Therefore, lateral, longitudinal characteristic distance of eyes and mouth is not necessarily to
It is modified, i.e., at this time without establishing flat square triangle.
Step 5: solving the length of bevel edge, required bevel edge using Pythagorean theorem in the flat square triangle of building
The true value of lateral, longitudinal characteristic distance of eyes and mouth when length is driver's face non-face camera, i.e. base
The characteristic distance of eyes and mouth is corrected in the head pose of driver.
(1) when driver fluctuates head, i.e., head is rotated around x-axis, and pitch yaw angle changes.If eye
Length of the longitudinal characteristic distance of eyeball and mouth before pitch rotation occurs is l_r, the length that cameras view arrives after rotation
Degree is that l_s can be obtained longitudinal characteristic distance of eyes and mouth by Pythagorean theorem after constructing flat square triangle according to step 3
Modified computing formulae it is as follows:
L_r=l_s/cos (pitch)
Wherein length of the l_r for longitudinal characteristic distance of eyes and mouth before pitch rotation occurs, l_s is eyes
With longitudinal characteristic distance of mouth length that cameras view arrives after pitch rotation occurs, pitch be driver bow or
Person come back (fluctuate head) when angle i.e. driver head around x-axis rotation angle.
(2) when driver or so shakes the head, i.e., head is rotated around y-axis, and yaw pitch angle changes.If eyes and mouth
Bar transverse features distance occur yaw rotation before length be l_r, the length that cameras view arrives after rotation be l_s,
After constructing flat square triangle according to step 3, the amendment meter of the transverse features distance of eyes and mouth can be obtained by Pythagorean theorem
It is as follows to calculate formula:
L_r=l_s/cos (yaw)
Wherein l_r is length of the transverse features distance of eyes and mouth before yaw rotation occurs, l_s be eyes and
The transverse features distance of the mouth length that cameras view arrives after yaw rotation occurs, yaw are when driver or so shakes the head
The rotation angle of angle, i.e. driver head around y-axis.
(3) when driver is to the left or when torticollis to the right, i.e., head is moved around z-axis, and roll roll angle changes.Root
According to according to step 3 be not necessarily to formation level right angled triangle, so when eyes and mouth lateral, longitudinal characteristic distance it is equal
Without being modified, therefore without correction formula.
Step 6: by corresponding situation lateral, longitudinal characteristic distance of revised eyes and mouth (such as step 5 institute
State) substitute into step 1 in facial expression assemblage characteristic vector Ft solution formula in must can be modified after facial expression combine
Feature vector F, specific solution formula are as follows:
Wherein e_ll be left eye on the left of upper lower eyelid between characteristic distance, e_lr be left eye on the right side of upper lower eyelid it
Between characteristic distance；E_lu is the transverse features distance of left eye upper eyelid, and e_ld is the transverse features distance of left eye lower eyelid；e_
Rl is the characteristic distance between the upper lower eyelid on the left of right eye, and e_rr is the characteristic distance between the upper lower eyelid on the right side of right eye；
E_ru is the transverse features distance of right eye upper eyelid, and e_rd is the transverse features distance of right eye lower eyelid；M_l is on mouth left side
Characteristic distance between lower lip, m_r are the characteristic distance on the right side of mouth between upper lower lip；M_u is the cross of mouth upper lip
To characteristic distance, m_d is the transverse features distance of mouth lower lip；Angle when pitch bows or comes back for driver is i.e.
Rotation angle of the driver head around x-axis；Yaw is rotation of angle i.e. driver head when driver or so shakes the head around y-axis
Angle；FtFor the facial expression feature vector before amendment.
Step 7: using revised facial expression assemblage characteristic vector F as new driver fatigue feature extraction formula,
The fatigue characteristic for extracting driver, to carry out fatigue driving judgement.
The present invention uses the method that is modified to eyes and mouth characteristic distance of head pose angle, as shown in Fig. 4 with
Right hand rectangular coordinate system in space is established centered on camera, analyzes the established available head of rectangular coordinate system in space three
Eulerian angles in dimension space.With this corresponding, as shown in Fig. 5, rotational angle of the available head in three reference axis
yaw,pitch,roll.Using head pose angle to based on facial expression feature fatigue driving detection in eyes and mouth into
It has gone characteristic distance amendment (Feature Length Correction, FLC).
Firstly, that is, head is rotated around x-axis when driver bows or comes back, pitch angle changes.For cross
To characteristic distance, when facial face camera, lateral Eigenvector is at this time perpendicular to the direction of camera face.Driver
It makes the movement bowed or come back to fluctuate head, at this time the angle of these transverse features line segments and camera positive direction
It does not change, therefore the transverse features that camera captures are constant apart from size, the transverse direction that camera captures at this time
Line segment (line segment as indicated the width of eyes and mouth) length variation can be ignored substantially.For longitudinal characteristic distance, face to face
When portion's face camera, longitudinal Eigenvector is also perpendicularly to the direction of camera face at this time.Driver make bow or
The movement of new line is the head that fluctuates, and longitudinal Eigenvector and the angle of camera positive direction become smaller, captured by camera
The longitudinal characteristic distance arrived shortens, therefore can mutually strain if vertical line segment (line segment as indicated eyes and mouth opening degree)
It is short.As shown in Fig. 6, it is assumed that length of the vertical line segment before pitch rotation occurs is l_r, cameras view after rotation
The length arrived is l_s.By attached drawing 6 according to the cosine of pitch, available following formula:
L_r=l_s/cos (pitch)
With should driver or so shake the head when, i.e., head is rotated around y-axis, and yaw angle changes, for longitudinal feature away from
From, when driver turn one's head observe left and right vehicle wheel two sides rock from side to side head when, longitudinal Eigenvector and camera are square at this time
To angle do not change, therefore longitudinal characteristic distance size that camera captures is constant, and camera is observed at this time
To vertical line segment length variation can ignore substantially.For transverse features distance, left and right vehicle wheel two is observed when driver turns one's head
Side is when rocking from side to side head, and transverse features line segment becomes smaller with camera positive direction angle, the transverse direction captured by camera
Characteristic distance shortens, therefore lateral line segment can then shorten.Assuming that length of the lateral line segment before yaw rotation occurs is l_r, rotation
The length that cameras view arrives after turning is l_s, at this time according to the available following formula of the cosine of yaw:
L_r=l_s/cos (yaw)
And when driver is to the left or when torticollis to the right, i.e. head is moved around z-axis, and roll angle changes, for cross
To characteristic distance, when driver is to the left or when torticollis to the right, transverse features line segment is in the plane perpendicular to camera positive direction
Upper rotation, the angle relative to camera positive direction do not change still, therefore the transverse features distance that camera captures
Size will not change.For longitudinal characteristic distance, when driver is to the left or when torticollis to the right, longitudinal Eigenvector is hanging down
It is directly rotated in the plane of camera positive direction, the angle of the positive direction relative to camera does not change still, therefore takes the photograph
The longitudinal characteristic distance size captured as head will not change.So when all line segment in the plane perpendicular to z-axis
Upper rotation, the line segment length that camera captures can't change.
In conclusion fatigue new according to the feature extraction formula of combined distance after considering characteristic distance amendment is special
Sign extracts formula are as follows:
Being 90 degree there is no consideration pitch and yaw angle in the revised feature extraction formula of above formula makes denominator zero
Situation, this is because Face datection algorithm can't detect face in this case, so be not in denominator being zero
Situation.So the modified method of the characteristic distance herein proposed is just for the adjustment in accuracy in special angle variation range, and
It is not fatigue detecting failure problem caused by thorough solution head deflection angle is excessive.
Fatigue detecting precision of the present invention in standard data set (YawDD) public data of fatigue detecting and detection speed
To carry out recruitment evaluation.The partial data of YawDD public data collection is as shown in Fig. 7.
It is tested and has been counted experimental result for the data set present invention in YawDD, uses this number due to many
It is not identical according to experimental evaluation standard final in the scheme of collection, in order to it is as much as possible with other schemes in method carry out pair
Than the present invention calculates recall rate (Recall), accurate rate (Precision), accuracy rate using following three formula respectively
(Accuracy), application effect of the invention is assessed with this.
Wherein occur the number of samples yawned and correctly detected in TP representative sample, does not go out in TN representative sample
Testing result of now yawning is that the number of samples yawned is not detected, and the testing result that do not occur yawning in FP representative sample is
It detects the number of samples yawned, occurs yawning in FN representative sample but testing result is that the sample yawned is not detected
Number.
The present invention is as shown in table 1 with the comparing result of other schemes in fatigue detecting precision.
1 fatigue detecting accuracy comparison result of table
The present invention is as shown in table 2 with the comparing result of other schemes in fatigue detecting speed.
2 fatigue detecting velocity contrast's result of table
Claims (4)
1. a kind of characteristic distance modification method of driver face based on facial expression fatigue driving detection algorithm, feature exist
In, this method comprising the following specific steps
Step 1: the facial expression feature of driver being refined as mouth feature and eye feature, eyes and mouth are carried out respectively
Characteristic point mark, using the characteristic point marked demarcate respectively eyes and mouth transverse features distance and longitudinal feature away from
From the ratio of longitudinal characteristic distance of solution left eye, right eye and mouth and transverse features distance, constructs facial expression feature respectively
Vector Ft indicates the opening and closing degree of eyes and mouth；Facial expression feature vector solution formula is as follows:
Wherein e_ll is the characteristic distance between the upper lower eyelid on the left of left eye, and e_lr is between the upper lower eyelid on the right side of left eye
Characteristic distance；E_lu is the transverse features distance of left eye upper eyelid, and e_ld is the transverse features distance of left eye lower eyelid；E_rl is
The characteristic distance between upper lower eyelid on the left of right eye, e_rr are the characteristic distance between the upper lower eyelid on the right side of right eye；e_ru
For the transverse features distance of right eye upper eyelid, e_rd is the transverse features distance of right eye lower eyelid；M_l is mouth left side mouth up and down
Characteristic distance between lip, m_r are the characteristic distance on the right side of mouth between upper lower lip；M_u is the laterally special of mouth upper lip
Distance is levied, m_d is the transverse features distance of mouth lower lip；
Step 2: establishing right hand rectangular coordinate system in space centered on driver head, obtain head three seats in three dimensions
Rotational angle yaw, pitch, roll on parameter, the i.e. Eulerian angles of head in three dimensions；
Wherein pitch angle yaw indicates the angle that driver head rotates around y-axis, and yaw angle pitch indicates driver head around x-axis
The distance of rotation, roll angle roll indicate the angle that driver head rotates around z-axis；
Step 3: when driver fluctuate head, rock from side to side head or torticollis when, extract at this time that head is in three dimensions
Three Eulerian angles, while extract eyes and mouth transverse features distance and longitudinal characteristic distance, with mouth and eyes
Laterally, longitudinal characteristic distance is the right-angle side of right angled triangle；Using the Eulerian angles of head in three dimensions as right angle trigonometry
The angle of right-angle side and bevel edge in shape；Construct flat square triangle；
Step 4: solving the length of bevel edge, required the length of the hypotenuse using Pythagorean theorem in the flat square triangle of building
The true value of lateral, longitudinal characteristic distance of eyes and mouth when being driver's non-face camera of face, that is, be based on driving
The head pose for the person of sailing is corrected the characteristic distance of eyes and mouth；
Step 5: solving facial expression feature vector F again using lateral, longitudinal characteristic distance of eyes after amendment and mouth, ask
It is as follows to solve formula:
Wherein e_ll is the characteristic distance between the upper lower eyelid on the left of left eye, and e_lr is between the upper lower eyelid on the right side of left eye
Characteristic distance；E_lu is the transverse features distance of left eye upper eyelid, and e_ld is the transverse features distance of left eye lower eyelid；E_rl is
The characteristic distance between upper lower eyelid on the left of right eye, e_rr are the characteristic distance between the upper lower eyelid on the right side of right eye；e_ru
For the transverse features distance of right eye upper eyelid, e_rd is the transverse features distance of right eye lower eyelid；M_l is mouth left side mouth up and down
Characteristic distance between lip, m_r are the characteristic distance on the right side of mouth between upper lower lip；M_u is the laterally special of mouth upper lip
Distance is levied, m_d is the transverse features distance of mouth lower lip；Pitch is that angle when driver bows or comes back drives
Rotation angle of the member head around x-axis；Yaw is rotation angle of angle i.e. driver head when driver or so shakes the head around y-axis；
Ft is the facial expression feature vector before amendment；
Step 6: using the facial expression feature vector F solved again as driver fatigue feature extraction formula, extracting driver
Fatigue characteristic, carry out fatigue driving judgement.
2. the characteristic distance of a kind of driver face based on facial expression fatigue driving detection algorithm according to claim 1
Modification method, it is characterised in that: it is described that right hand rectangular coordinate system in space is established centered on driver head in step 2, wherein
The direction of x-axis is that left ear is directed toward auris dextra, and the direction of y-axis is to be directed toward the crown, and the direction of z-axis is that forehead is directed toward the back side of head, is driven
Three Eulerian angles yaw, pitch and the roll of the person of sailing head in three dimensions characterize driver's head with this three Eulerian angles
The posture of portion in three dimensions.
3. the characteristic distance of a kind of driver face based on facial expression fatigue driving detection algorithm according to claim 1
Modification method, it is characterised in that: in step 3, the building flat square triangle is specifically included:
(1) when driver fluctuates head, i.e., head is rotated around x-axis, and pitch yaw angle changes；For eyes
With the transverse features distance of mouth, when facial face camera, lateral Eigenvector is at this time perpendicular to camera face
Direction；Driver makes the movement bowed or come back and fluctuates head, at this time these transverse features line segments and camera
The angle of positive direction does not change, therefore the transverse features that camera captures are constant apart from size；For eyes and
Mouth longitudinal direction characteristic distance, when facial face camera, longitudinal Eigenvector is also perpendicularly to the side of camera face at this time
To；Driver makes the movement bowed or come back and fluctuates head, the folder of longitudinal Eigenvector and camera positive direction
Angle becomes smaller, and is shortened by longitudinal characteristic distance that camera captures, therefore vertical line segment then can shorter；At this time with camera shooting
Longitudinal characteristic distance of eyes and mouth that head observes is the right-angle side of right angled triangle, with revised eyes and mouth
Longitudinal characteristic distance is the bevel edge of right angled triangle, is the angle between the right-angle side and bevel edge that are constructed with the angle pitch, constructs
Flat square triangle；
(2) when driver or so shakes the head, i.e., head is rotated around y-axis, and yaw pitch angle changes, for eyes and mouth
Longitudinal characteristic distance, when driver turn one's head observe left and right vehicle wheel two sides rock from side to side head when, longitudinal Eigenvector at this time
It does not change with the angle of camera positive direction, therefore longitudinal characteristic distance size that camera captures is constant；It is right
In the transverse features distance of eyes and mouth, when driver turn one's head observe left and right vehicle wheel two sides rock from side to side head when, laterally
Eigenvector becomes smaller with camera positive direction angle, the transverse features Distance Shortened captured by camera, therefore x wire
Duan Zehui shortens；The transverse features distance of the eyes and mouth that are observed at this time using camera as the right-angle side of right angled triangle,
Using the transverse features distance of revised eyes and mouth as the bevel edge of right angled triangle, with the angle yaw for the right-angle side that is constructed
Angle between bevel edge constructs flat square triangle；
(3) when driver is to the left or when torticollis to the right, i.e., head is moved around z-axis, and roll roll angle changes, for eye
The transverse features distance of eyeball and mouth, when driver is to the left or when torticollis to the right, transverse features line segment is perpendicular to camera
It is rotated in the plane of positive direction, the angle relative to camera positive direction does not change, therefore the eyes that camera captures
It will not change with the transverse features of mouth apart from size；For longitudinal characteristic distance of eyes and mouth, when driver to
When a left side or to the right torticollis, longitudinal Eigenvector rotates in the plane perpendicular to camera positive direction, relative to camera
The angle of positive direction do not change, longitudinal characteristic distance size of eyes and mouth that camera captures will not become
Change；The characteristic distance of all eyes and mouth rotates in the plane perpendicular to z-axis at this time, the eyes that camera captures and
The characteristic distance length of mouth will not change；Therefore, lateral, longitudinal characteristic distance of eyes and mouth is without being repaired
Just, i.e., at this time without establishing flat square triangle.
4. the characteristic distance of a kind of driver face based on facial expression fatigue driving detection algorithm according to claim 1
Modification method, it is characterised in that: step 4 specifically includes:
(1) when driver fluctuates head, i.e., head is rotated around x-axis, and pitch yaw angle changes；If eyes and
Longitudinal characteristic distance of mouth occur pitch rotation before length be l_r, after rotation cameras view to length be
L_s, according to the flat square triangle of building, the corrected Calculation for obtaining longitudinal characteristic distance of eyes and mouth by Pythagorean theorem is public
Formula is as follows:
L_r=l_s/cos (pitch)
Wherein length of the l_r for longitudinal characteristic distance of eyes and mouth before pitch rotation occurs, l_s are eyes and mouth
Bar longitudinal characteristic distance after pitch rotation occurs the length that arrives of cameras view, pitch be that driver bows or lifts
Head when angle, that is, driver head around x-axis rotation angle；
(2) when driver or so shakes the head, i.e., head is rotated around y-axis, and yaw pitch angle changes；If eyes and mouth
Length of the transverse features distance before yaw rotation occurs is l_r, and the length that cameras view arrives after rotation is l_s, according to
The flat square triangle of building, the modified computing formulae for obtaining the transverse features distance of eyes and mouth by Pythagorean theorem are as follows:
L_r=l_s/cos (yaw)
Wherein length of the l_r for the transverse features distance of eyes and mouth before yaw rotation occurs, l_s are eyes and mouth
Transverse features distance after yaw rotation occurs the length that arrives of cameras view, yaw be angle when driver or so shakes the head
Degree, i.e. rotation angle of the driver head around y-axis；
(3) when driver is to the left or when torticollis to the right, i.e., head is moved around z-axis, and roll roll angle changes；Eye at this time
Lateral, longitudinal characteristic distance of eyeball and mouth is not necessarily to be modified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910297915.6A CN109934207A (en) | 2019-04-15 | 2019-04-15 | A kind of characteristic distance modification method of driver face based on facial expression fatigue driving detection algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910297915.6A CN109934207A (en) | 2019-04-15 | 2019-04-15 | A kind of characteristic distance modification method of driver face based on facial expression fatigue driving detection algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109934207A true CN109934207A (en) | 2019-06-25 |
Family
ID=66990007
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910297915.6A Pending CN109934207A (en) | 2019-04-15 | 2019-04-15 | A kind of characteristic distance modification method of driver face based on facial expression fatigue driving detection algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109934207A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110796838A (en) * | 2019-12-03 | 2020-02-14 | 吉林大学 | Automatic positioning and recognition system for facial expressions of driver |
| CN112396805A (en) * | 2019-07-31 | 2021-02-23 | Oppo广东移动通信有限公司 | Fatigue driving reminding method, device, terminal and storage medium |
| CN113208591A (en) * | 2020-01-21 | 2021-08-06 | 初速度（苏州）科技有限公司 | Method and device for determining eye opening and closing distance |
| WO2021218117A1 (en) * | 2020-04-28 | 2021-11-04 | 佛山市顺德区美的电热电器制造有限公司 | Distance measurement method, apparatus and device, and storage medium |
| CN115019377A (en) * | 2022-08-08 | 2022-09-06 | 京东方艺云(杭州)科技有限公司 | Eye state judgment method, eye state judgment device and electronic equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH03167698A (en) * | 1989-11-28 | 1991-07-19 | Nissan Motor Co Ltd | Driving state detecting device for driver |
| US20120010533A1 (en) * | 2005-10-17 | 2012-01-12 | Arnett G William | Method for determining the correct natural head position location of references planes relative to a three-dimensional computerized image of a patient's head |
| US20160224852A1 (en) * | 2015-01-30 | 2016-08-04 | GM Global Technology Operations LLC | Vehicle operator monitoring system and method |
| DE102016113374B3 (en) * | 2016-07-20 | 2017-10-26 | Carl Zeiss Vision International Gmbh | Remote viewpoint determination for a spectacle lens |
| CN107679468A (en) * | 2017-09-19 | 2018-02-09 | 浙江师范大学 | A kind of embedded computer vision detects fatigue driving method and device |
| CN108256427A (en) * | 2017-12-18 | 2018-07-06 | 佛山正能光电有限公司 | Face recognition module |
| CN108446600A (en) * | 2018-02-27 | 2018-08-24 | 上海汽车集团股份有限公司 | A kind of vehicle driver's fatigue monitoring early warning system and method |
| CN109191791A (en) * | 2018-10-30 | 2019-01-11 | 罗普特（厦门）科技集团有限公司 | A kind of fatigue detection method and device merging multiple features |
-
2019
- 2019-04-15 CN CN201910297915.6A patent/CN109934207A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH03167698A (en) * | 1989-11-28 | 1991-07-19 | Nissan Motor Co Ltd | Driving state detecting device for driver |
| US20120010533A1 (en) * | 2005-10-17 | 2012-01-12 | Arnett G William | Method for determining the correct natural head position location of references planes relative to a three-dimensional computerized image of a patient's head |
| US20160224852A1 (en) * | 2015-01-30 | 2016-08-04 | GM Global Technology Operations LLC | Vehicle operator monitoring system and method |
| DE102016113374B3 (en) * | 2016-07-20 | 2017-10-26 | Carl Zeiss Vision International Gmbh | Remote viewpoint determination for a spectacle lens |
| CN107679468A (en) * | 2017-09-19 | 2018-02-09 | 浙江师范大学 | A kind of embedded computer vision detects fatigue driving method and device |
| CN108256427A (en) * | 2017-12-18 | 2018-07-06 | 佛山正能光电有限公司 | Face recognition module |
| CN108446600A (en) * | 2018-02-27 | 2018-08-24 | 上海汽车集团股份有限公司 | A kind of vehicle driver's fatigue monitoring early warning system and method |
| CN109191791A (en) * | 2018-10-30 | 2019-01-11 | 罗普特（厦门）科技集团有限公司 | A kind of fatigue detection method and device merging multiple features |
Non-Patent Citations (8)
| Title |
|---|
| 冯春等: "追踪与目标航天器相对位姿参数四元数的解析算法", 《中国机械工程》 * |
| 张波等: "基于人脸3D模型的驾驶人头部姿态检测", 《汽车工程》 * |
| 张腾飞等: "基于特征区域自动分割的人脸表情识别", 《计算机工程》 * |
| 李响等: "基于Zernike矩的人眼定位与状态识别", 《电子测量与仪器学报》 * |
| 杨欢: "列车驾驶员多视角实时疲劳检测方法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
| 杨欢等: "基于逆投影修正和眼睛凝视修正的列车驾驶员疲劳检测方法", 《铁道学报》 * |
| 王琼等: "基于眼睛状态识别的驾驶员疲劳监测", 《南京理工大学学报(自然科学版)》 * |
| 赵磊等: "基于ASM局部定位和特征三角形的列车驾驶员头部姿态估计", 《铁道学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112396805A (en) * | 2019-07-31 | 2021-02-23 | Oppo广东移动通信有限公司 | Fatigue driving reminding method, device, terminal and storage medium |
| CN112396805B (en) * | 2019-07-31 | 2022-10-25 | Oppo广东移动通信有限公司 | Fatigue driving reminding method, device, terminal and storage medium |
| CN110796838A (en) * | 2019-12-03 | 2020-02-14 | 吉林大学 | Automatic positioning and recognition system for facial expressions of driver |
| CN113208591A (en) * | 2020-01-21 | 2021-08-06 | 初速度（苏州）科技有限公司 | Method and device for determining eye opening and closing distance |
| CN113208591B (en) * | 2020-01-21 | 2023-01-06 | 魔门塔(苏州)科技有限公司 | Method and device for determining eye opening and closing distance |
| WO2021218117A1 (en) * | 2020-04-28 | 2021-11-04 | 佛山市顺德区美的电热电器制造有限公司 | Distance measurement method, apparatus and device, and storage medium |
| CN115019377A (en) * | 2022-08-08 | 2022-09-06 | 京东方艺云(杭州)科技有限公司 | Eye state judgment method, eye state judgment device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109934207A (en) | A kind of characteristic distance modification method of driver face based on facial expression fatigue driving detection algorithm | |
| WO2022088254A1 (en) | Vehicle-mounted display screen adjustment device and vehicle | |
| EP2338416B1 (en) | Line-of-sight direction determination device and line-of-sight direction determination method | |
| US9373156B2 (en) | Method for controlling rotation of screen picture of terminal, and terminal | |
| CN101489467B (en) | Visual axis direction detection device and visual line direction detection method | |
| Tsukada et al. | Illumination-free gaze estimation method for first-person vision wearable device | |
| CN107492123B (en) | Road monitoring camera self-calibration method using road surface information | |
| CN105574518A (en) | Method and device for human face living detection | |
| WO2019137065A1 (en) | Image processing method and apparatus, vehicle-mounted head up display system, and vehicle | |
| CN101840509B (en) | Measuring method for eye-observation visual angle and device thereof | |
| CN107958479A (en) | A kind of mobile terminal 3D faces augmented reality implementation method | |
| US20140204193A1 (en) | Driver gaze detection system | |
| WO2009091029A1 (en) | Face posture estimating device, face posture estimating method, and face posture estimating program | |
| CN103605965A (en) | Multi-pose face recognition method and device | |
| CN101419664A (en) | Sight direction measurement method and sight direction measurement device | |
| CN107103309A (en) | A kind of sitting posture of student detection and correcting system based on image recognition | |
| CN105740688A (en) | Unlocking method and device | |
| CN111814556A (en) | Teaching assistance method and system based on computer vision | |
| CN105844227A (en) | Driver identity authentication method for school bus safety | |
| CN108921148A (en) | Determine the method and device of positive face tilt angle | |
| CN104122983A (en) | Adjusting method and device for screen displaying direction | |
| CN112101247B (en) | Face pose estimation method, device, equipment and storage medium | |
| CN103544478A (en) | All-dimensional face detection method and system | |
| CN109886173A (en) | The autonomous service robot of side face attitude algorithm method and mood sensing of view-based access control model | |
| CN114022514A (en) | Real-time sight line inference method integrating head posture and eyeball tracking |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190625 |
|
| WD01 | Invention patent application deemed withdrawn after publication |