CN110020632A - A method of the recognition of face based on deep learning is for detecting fatigue driving - Google Patents
A method of the recognition of face based on deep learning is for detecting fatigue driving Download PDFInfo
- Publication number
- CN110020632A CN110020632A CN201910294584.0A CN201910294584A CN110020632A CN 110020632 A CN110020632 A CN 110020632A CN 201910294584 A CN201910294584 A CN 201910294584A CN 110020632 A CN110020632 A CN 110020632A
- Authority
- CN
- China
- Prior art keywords
- face
- deep learning
- recognition
- mouth
- fatigue driving
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/59—Context or environment of the image inside of a vehicle, e.g. relating to seat occupancy, driver state or inner lighting conditions
- G06V20/597—Recognising the driver's state or behaviour, e.g. attention or drowsiness
-
- 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
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V40/174—Facial expression recognition
Abstract
The invention belongs to fatigue-driving detection technology fields, a kind of method for being used to detect fatigue driving the invention discloses recognition of face based on deep learning, comprising steps of S1, using deep learning neural network, training human face recognition model and facial expression disaggregated model；S2, in detection cycle, by frequency acquisition acquisition person under test face picture, face picture is identified using human face recognition model and facial expression disaggregated model；S3, according to recognition result, judge whether person under test is in fatigue driving state.Present invention application deep learning neural network acquires data based on detection cycle sample mode, judgment models is added in tired human facial expression recognition, fusion eye dynamic, headwork and mouth action are to determine whether in a state of fatigue；To which the present invention has the characteristics that Detection accuracy is high, detection is fireballing.
Description
Technical field
The invention belongs to fatigue-driving detection technology fields, and in particular to a kind of recognition of face based on deep learning is used for
The method for detecting fatigue driving.
Background technique
There are mainly two types of detection methods at present for anti-fatigue-driving: one, being based on sensor: head position sensor, steering wheel
Sensor etc., the data based on physical trait sensor make a decision whether fatigue driving；Two, based on eye dynamic: frequency of wink,
Direction of gaze, closed-eye time, pupil detection etc..Objectively both detection modes all achieve good results.
The first sensor-based detection mode proposes that the time is more early, and scheme is also more mature, but does not have in recent years
Technical too many progress.Sensor detection data is generally reliable, and the accuracy rate obtained is also relatively high.But due to sensor
At high cost, multipair driver has invasive greatly, is unfavorable for the driving experience of driver.And a variety of models layout is different, generally needs
It customizes.Since the problems such as its is at high cost, construction is complicated, is difficult to promote.
Second be based on the dynamic identification method of eye, mainly use traditional machine learning algorithm.Knowledge based on pupil
The monitoring device of more high definition is not needed, and higher cost, domestic only less scientific research institution grasps core technology at present.It is based on
Simple closed-eye time, frequency of wink, there is no very high accuracys rate for industry at present.Practical application is mostly that eye motion adds pupil
It identifies mixed mode, but since equipment equipment is higher, pupil identification technology level requirement is higher, is not also promoted effectively.
The concept of deep learning is derived from the research of artificial neural network.Multilayer perceptron containing more hidden layers is exactly a kind of depth
Learning structure.Deep learning, which forms more abstract high level by combination low-level feature, indicates attribute classification or feature, with discovery
The distributed nature of data indicates.
The concept of deep learning was proposed by Hinton et al. in 2006.Non- prison is proposed based on depth confidence network (DBN)
The layer-by-layer training algorithm of greed is superintended and directed, hope is brought to solve the relevant optimization problem of deep structure, then proposes multilayer autocoding
Device deep structure.Furthermore the convolutional neural networks that Lecun et al. is proposed are first real multilayered structure learning algorithms, it is utilized
Spatial correlation reduces number of parameters to improve training performance.
Deep learning is a kind of based on the method for carrying out representative learning to data in machine learning.Observation (such as a width
Image) various ways can be used to indicate, such as vector of each pixel intensity value, or be more abstractively expressed as a series of
Side, region of specific shape etc..And use certain specific representation methods be easier from example learning tasks (for example, face
Identification or human facial expression recognition).The benefit of deep learning is feature learning and the layered characteristic with non-supervisory formula or Semi-supervised
It extracts highly effective algorithm and obtains feature by hand to substitute.
Deep learning is a new field in machine learning research, and motivation is that foundation, simulation human brain are divided
The neural network of study is analysed, it imitates the mechanism of human brain to explain data, such as image, sound and text.
It is the same with machine learning method, point of depth machine learning method also supervised learning and unsupervised learning.It is different
Learning framework under the learning model established it is very different.For example, convolutional neural networks (Convolutional neural
Networks, abbreviation CNNs) be exactly a kind of depth supervised learning under machine learning model, and depth confidence net (Deep
Belief Nets, abbreviation DBNs) it is exactly a kind of machine learning model under unsupervised learning.
Deep learning fatigue driving detection field is not applied to also in currently available technology, the prior art is badly in need of a kind of logical
The method for crossing deep learning to realize fatigue driving detection.
Summary of the invention
In order to solve the above problems existing in the present technology, it is an object of that present invention to provide a kind of people based on deep learning
Face identifies the method for detecting fatigue driving.
The technical scheme adopted by the invention is as follows:
A method of the recognition of face based on deep learning is used to detect fatigue driving, comprising steps of
S1, using deep learning neural network, training human face recognition model and facial expression disaggregated model；
S2, in detection cycle, by frequency acquisition acquisition person under test face picture, using human face recognition model and face
Expression classification model identifies face picture；
S3, according to recognition result, judge whether person under test is in fatigue driving state.
Further, the human face recognition model in the step S1 detects face five by label face features point
The size position of official；Human face recognition model identifies face skew direction by label face posture.
Further, the facial expression disaggregated model in the step S1 identifies face picture type by classification data.
Further, the step S2 obtains the frequency of wink and blink space-number of face picture using human face recognition model
According to the specific method is as follows:
1) the upper pixel in upper eyelid that human face recognition model obtains same eyes, that upper and lower position is symmetrical is utilized to sit
The lower pixel coordinate of mark and palpebra inferior, calculates blink width in real time, and blink width is the=symmetrical upper eyelid in upper and lower position
Upper pixel coordinate and palpebra inferior lower pixel coordinate distance；The upper limit value of blink width is M, and the lower limit value for width of blinking is
N；
2) in detection cycle, record blink width >=0.8M number, as eye opening number；Record blink width≤
The number of 0.2N, as eye closing number；If the adjacent blink width twice in front and back meets blink width >=0.8M and blink width
≤ 0.2N is then blink, is denoted as number of winks；
3) the adjacent longest interval time blinked twice is recorded；By formula: based on frequency of wink=number of winks/detection cycle
Calculate frequency of wink.
Further, when the step S2 opens one's mouth size with opening one's mouth using the mouth that human face recognition model obtains face picture
Long data, the specific method is as follows:
1) the upper pixel coordinate of upper lip and lower lip upper and lower position symmetrically is obtained under using human face recognition model
Pixel coordinate, real-time calculating are opened one's mouth highly, and height of opening one's mouth is=upper the pixel coordinate and lower pixel seat of upper and lower position symmetrically
Target distance；2) in detection cycle, the numerical value for height of opening one's mouth is recorded；
3) duration opened one's mouth is recorded.
Further, the step S2 obtains the facial expression of face picture by facial expression disaggregated model, will own
Face picture according to excited, normal or tired classification.
Further, the step S3 judges whether person under test is in fatigue driving according to recognition result as follows
State；If person under test is in fatigue driving state, at least meet two in the following conditions:
1) frequency of wink is lower than default frequency of wink value；
2) the adjacent longest interval time blinked twice is lower than default wink time spacing value；
3) duration opened one's mouth is lower than the default duration value opened one's mouth；
4) numerical value for height of opening one's mouth is greater than the numerical value of default height of opening one's mouth；
5) face picture that total 1/2 or more quantity is accounted in detection cycle belongs to fatigue state.
Further, further include step S4: whether being fatigue driving in the predicted detection period based on the analysis results.
Further, each detection cycle of the step S2 is 30 seconds；Frequency acquisition is 30 frame picture per second.
The invention has the benefit that a kind of recognition of face based on deep learning of the invention is for detecting fatigue driving
Method data are acquired based on detection cycle sample mode, tired human facial expression recognition is added using deep learning neural networks
Enter judgment models, fusion eye dynamic, headwork and mouth action are to determine whether in a state of fatigue；To present invention tool
There are Detection accuracy height, the fireballing feature of detection；Improvement and installation of the invention is at low cost, it is only necessary to which driver's seat is installed common
Camera is connect with the system of operation deep learning neural network.To which the present invention is readily produced popularization and use.
Specific embodiment
Further explaination is done to the present invention combined with specific embodiments below.
Embodiment 1
A method of the recognition of face based on deep learning is for detecting fatigue driving, comprising steps of S1, utilizing depth
Learning neural network GoogleNet trains human face recognition model, face features point is on the one hand marked, for detecting face
Size position etc..On the other hand the posture of face is marked, face is to the left to the right for identification and nods and faces upward head；Utilize deep learning
Neural network ResNet trains facial expression disaggregated model, classification data are as follows: excited, normal, fatigue.Require early period prepare compared with
Big data set is cut, artificial mark, and then training pattern, is not that the emphasis of the present embodiment is not described in detail, other people can be with
Reach effect by similar methods.Human face recognition model recognition accuracy based on GoogleNet is fine, based on ResNet's
Facial expression disaggregated model classification speed and generalization ability are stablized.
S2,1) frequency of wink is calculated.The present embodiment is using 30 seconds a cycles detection per second 30 frame picture collections blink
Frequency, frequency of nodding and the maximum duration closed one's eyes in 30 seconds, the amplitude peak bowed obtain eyes using FaceModel
The picture pixels coordinate of top and low calculates the pixel wide of top-low (blinkLen), as eyes blink width in real time.
Since blink speed is exceedingly fast, it is likely that when capturing most fast and most narrow less than eyes blink；So eyes are blinked, width has a
Upper and lower bound, blink width maximum value take camera to start the maximum value (eyeHigh) of rear eye-level, the upper limit it is smaller this
Value, the present embodiment take its 0.8 times；Lower limit bigger 0, the present embodiment takes 0.2 times.When blinkLen is greater than eyeHigh*0.8, note
It is primary to open eyes；When blinkLen is less than eyeHigh*0.2, it is primary to be denoted as eye closing.The two conditions are blinked when meeting simultaneously
Eye is primary.
Normal frequency of wink is different from according to the frequency of wink of medical judgment fatigue state.
2) calculate in a short time every.Calculate blink one time while, we in addition record this blink used how long, protect
Deposit blink used time maximum duration in 30 seconds.
It can be longer according to the blink time-consuming of medical judgment fatigue state.
3) mouth is calculated to open one's mouth largest amount and duration.Similarly calculate in mouth 30 seconds intervals the amplitude peaks opened one's mouth and when
It is long.Yawn can be longer than the duration of opening one's mouth to open one's mouth more greatly of normally speaking.
4) facial expression is identified.The sampling prediction of 30 seconds pictures is belonged to using human face recognition model excited, normal, tired
Labor.
5) predict 30 second period in whether fatigue driving.
Actual use is empirically, as a result more accurate when the result accounting weight that (1)-(4) calculate is bigger.(1)-(4) meter
It calculates result and is judged as that fatigue driving by sound early warning, can accurately judge that fatigue is driven when having and meeting more than two conditions
It sails.
To the method that a kind of recognition of face based on deep learning of the present invention is used to detect fatigue driving, using depth
Learning neural network acquires data based on detection cycle sample mode, and judgment models, fusion is added in tired human facial expression recognition
Eye dynamic, headwork and mouth action are to determine whether in a state of fatigue；To the present invention have Detection accuracy it is high,
Detect fireballing feature；Improvement and installation of the invention is at low cost, it is only necessary to which it is deep that driver's seat installs common camera and operation
Spend the system connection of learning neural network.To which the present invention is readily produced popularization and use.
Embodiment 2
A method of the recognition of face based on deep learning is used to detect fatigue driving, comprising steps of
S1, using deep learning neural network, training human face recognition model and facial expression disaggregated model；
S2, in detection cycle, by frequency acquisition acquisition person under test face picture, using human face recognition model and face
Expression classification model identifies face picture；
S3, according to recognition result, judge whether person under test is in fatigue driving state.
Further, the human face recognition model in the step S1 detects face five by label face features point
The size position of official；Human face recognition model identifies face skew direction by label face posture.
The method that a kind of recognition of face based on deep learning of the present embodiment is used to detect fatigue driving, step S1, is adopted
With deep learning neural network, training human face recognition model and facial expression disaggregated model, it is preferable that training human face recognition model
Deep learning neural network be deep learning neural network GoogleNet；The deep learning of training facial expression disaggregated model
Neural network is deep learning neural network ResNet.
On the one hand the human face recognition model of the present embodiment marks face features point, for detecting the size position of face
Deng.On the other hand the posture of face is marked, face is to the left to the right for identification and nods and faces upward head；To which the face of the present embodiment is known
Other model still can mark face features point, accurately lock five when face is in off-normal position or has deflection
Official position.
The training facial expression disaggregated model of the present embodiment is using deep learning neural network ResNet training, number of classifying
According to being excited, normal or tired.
Step S2, in detection cycle, by the face picture of frequency acquisition acquisition person under test, using human face recognition model and
Facial expression disaggregated model identifies face picture；
By in certain detection cycle, such as 30s, 1min, 2min；By the face figure of frequency acquisition acquisition person under test
Piece；The frequency acquisition can be 20/s, 30/s or 40/s etc.；Using human face recognition model and facial expression classification mould
Type identifies face picture, to judge whether that fatigue driving provides foundation.
Further, the facial expression disaggregated model in the step S1 identifies face picture type by classification data.
S3, according to recognition result, judge whether person under test is in fatigue driving state.
Specifically, known using the human face recognition model of deep learning neural metwork training especially by by the way of following
Not.
The step S2 obtains the frequency of wink and blink interval data of face picture, specific side using human face recognition model
Method is as follows:
1) the upper pixel in upper eyelid that human face recognition model obtains same eyes, that upper and lower position is symmetrical is utilized to sit
The lower pixel coordinate of mark and palpebra inferior, calculates blink width in real time, and blink width is the=symmetrical upper eyelid in upper and lower position
Upper pixel coordinate and palpebra inferior lower pixel coordinate distance；The upper limit value of blink width is M, and the lower limit value for width of blinking is
N；
When i.e. blink width is understood that stand for people, the vertical range between upper eyelid and palpebra inferior；When eye opening, upper eye
The critical distance of vertical range between eyelid and palpebra inferior is then the upper limit value of blink width；It is then blink width when eye closing
Lower limit value, theoretic lower limit value are zero.
2) in detection cycle, record blink width >=0.8M number, as eye opening number；Record blink width≤
The number of 0.2N, as eye closing number；If the adjacent blink width twice in front and back meets blink width >=0.8M and blink width
≤ 0.2N is then blink, is denoted as number of winks.
3) the adjacent longest interval time blinked twice is recorded；By formula: based on frequency of wink=number of winks/detection cycle
Calculate frequency of wink.
Further, when the step S2 opens one's mouth size with opening one's mouth using the mouth that human face recognition model obtains face picture
Long data, the specific method is as follows:
1) the upper pixel coordinate of upper lip and lower lip upper and lower position symmetrically is obtained under using human face recognition model
Pixel coordinate, real-time calculating are opened one's mouth highly, and height of opening one's mouth is=upper the pixel coordinate and lower pixel seat of upper and lower position symmetrically
Target distance；2) in detection cycle, the numerical value for height of opening one's mouth is recorded；
Vertical range of the height opened one's mouth between upper lip and lower lip.
3) duration opened one's mouth is recorded.
Further, the step S2 obtains the facial expression of face picture by facial expression disaggregated model, will own
Face picture according to excited, normal or tired classification.
To, the present embodiment by frequency of wink, in a short time every, open one's mouth height, duration, the facial expression opened one's mouth etc.
Several aspects go identification judges data required for fatigue driving, thus the present embodiment can by frequency of wink, in a short time every,
Height, duration for opening one's mouth and the facial expression of opening one's mouth come whether comprehensive descision is fatigue driving state.
Embodiment 3
A method of the recognition of face based on deep learning is used to detect fatigue driving, comprising steps of
S1, using deep learning neural network, training human face recognition model and facial expression disaggregated model；
S2, in detection cycle, by frequency acquisition acquisition person under test face picture, using human face recognition model and face
Expression classification model identifies face picture；
S3, according to recognition result, judge whether person under test is in fatigue driving state.
Further, the human face recognition model in the step S1 detects face five by label face features point
The size position of official；Human face recognition model identifies face skew direction by label face posture.
The method that a kind of recognition of face based on deep learning of the present embodiment is used to detect fatigue driving, step S1, is adopted
With deep learning neural network, training human face recognition model and facial expression disaggregated model, it is preferable that training human face recognition model
Deep learning neural network be deep learning neural network GoogleNet；The deep learning of training facial expression disaggregated model
Neural network is deep learning neural network ResNet.
On the one hand the human face recognition model of the present embodiment marks face features point, for detecting the size position of face
Deng.On the other hand the posture of face is marked, face is to the left to the right for identification and nods and faces upward head；To which the face of the present embodiment is known
Other model still can mark face features point, accurately lock five when face is in off-normal position or has deflection
Official position.
The training facial expression disaggregated model of the present embodiment is using deep learning neural network ResNet training, number of classifying
According to being excited, normal or tired.
Step S2, in detection cycle, by the face picture of frequency acquisition acquisition person under test, using human face recognition model and
Facial expression disaggregated model identifies face picture；
By in certain detection cycle, such as 30s, 1min, 2min；By the face figure of frequency acquisition acquisition person under test
Piece；The frequency acquisition can be 20/s, 30/s or 40/s etc.；Using human face recognition model and facial expression classification mould
Type identifies face picture, to judge whether that fatigue driving provides foundation.
Further, the facial expression disaggregated model in the step S1 identifies face picture type by classification data.
S3, according to recognition result, judge whether person under test is in fatigue driving state.
Specifically, known using the human face recognition model of deep learning neural metwork training especially by by the way of following
Not.
The step S2 obtains the frequency of wink and blink interval data of face picture, specific side using human face recognition model
Method is as follows:
3) the upper pixel in upper eyelid that human face recognition model obtains same eyes, that upper and lower position is symmetrical is utilized to sit
The lower pixel coordinate of mark and palpebra inferior, calculates blink width in real time, and blink width is the=symmetrical upper eyelid in upper and lower position
Upper pixel coordinate and palpebra inferior lower pixel coordinate distance；The upper limit value of blink width is M, and the lower limit value for width of blinking is
N；
When i.e. blink width is understood that stand for people, the vertical range between upper eyelid and palpebra inferior；When eye opening, upper eye
The critical distance of vertical range between eyelid and palpebra inferior is then the upper limit value of blink width；It is then blink width when eye closing
Lower limit value, theoretic lower limit value are zero.
4) in detection cycle, record blink width >=0.8M number, as eye opening number；Record blink width≤
The number of 0.2N, as eye closing number；If the adjacent blink width twice in front and back meets blink width >=0.8M and blink width
≤ 0.2N is then blink, is denoted as number of winks.
3) the adjacent longest interval time blinked twice is recorded；By formula: based on frequency of wink=number of winks/detection cycle
Calculate frequency of wink.
Further, when the step S2 opens one's mouth size with opening one's mouth using the mouth that human face recognition model obtains face picture
Long data, the specific method is as follows:
2) the upper pixel coordinate of upper lip and lower lip upper and lower position symmetrically is obtained under using human face recognition model
Pixel coordinate, real-time calculating are opened one's mouth highly, and height of opening one's mouth is=upper the pixel coordinate and lower pixel seat of upper and lower position symmetrically
Target distance；2) in detection cycle, the numerical value for height of opening one's mouth is recorded；
Vertical range of the height opened one's mouth between upper lip and lower lip.
3) duration opened one's mouth is recorded.
Further, the step S2 obtains the facial expression of face picture by facial expression disaggregated model, will own
Face picture according to excited, normal or tired classification.
Further, the step S3 judges whether person under test is in fatigue driving according to recognition result as follows
State；If person under test is in fatigue driving state, at least meet two in the following conditions:
1) frequency of wink is lower than default frequency of wink value；
2) the adjacent longest interval time blinked twice is lower than default wink time spacing value；
3) duration opened one's mouth is lower than the default duration value opened one's mouth；
4) numerical value for height of opening one's mouth is greater than the numerical value of default height of opening one's mouth；
5) face picture that total 1/2 or more quantity is accounted in detection cycle belongs to fatigue state.
Further, further include step S4: whether being fatigue driving in the predicted detection period based on the analysis results.
Further, each detection cycle of the step S2 is 30 seconds；Frequency acquisition is 30 frame picture per second.
The present embodiment optimizes each index, optimizes through frequency of wink, in a short time every, height of opening one's mouth, holding of opening one's mouth
The method of the continuous index comprehensives such as time and facial expression judgement detection fatigue driving, so that the present embodiment be made to sentence fatigue driving
Break or detects more acurrate.
The present invention is not limited to above-mentioned optional embodiment, anyone can show that other are each under the inspiration of the present invention
The product of kind form.Above-mentioned specific embodiment should not be understood the limitation of pairs of protection scope of the present invention, protection of the invention
Range should be subject to be defined in claims, and specification can be used for interpreting the claims.
Claims (9)
1. a kind of method that recognition of face based on deep learning is used to detect fatigue driving, it is characterised in that: comprising steps of
S1, using deep learning neural network, training human face recognition model and facial expression disaggregated model；
S2, in detection cycle, by frequency acquisition acquisition person under test face picture, using human face recognition model and facial expression
Disaggregated model identifies face picture；
S3, according to recognition result, judge whether person under test is in fatigue driving state.
2. the method that a kind of recognition of face based on deep learning according to claim 1 is used to detect fatigue driving,
Be characterized in that: the human face recognition model in the step S1 detects the size of face face by label face features point
Position；Human face recognition model identifies face skew direction by label face posture.
3. the method that a kind of recognition of face based on deep learning according to claim 2 is used to detect fatigue driving,
Be characterized in that: the facial expression disaggregated model in the step S1 identifies face picture type by classification data.
4. the method that a kind of recognition of face based on deep learning according to claim 3 is used to detect fatigue driving,
Be characterized in that: the step S2 obtains the frequency of wink and blink interval data of face picture using human face recognition model, specifically
Method is as follows:
1) using human face recognition model obtains same eyes, the upper pixel coordinate in upper eyelid that upper and lower position is symmetrical and
The lower pixel coordinate of palpebra inferior, in real time calculate blink width, blink width be=the symmetrical upper eyelid in upper and lower position it is upper
The distance of the lower pixel coordinate of pixel coordinate and palpebra inferior；The upper limit value of blink width is M, and the lower limit value for width of blinking is N；
2) in detection cycle, record blink width >=0.8M number, as eye opening number；Record blink width≤0.2N
Number, as eye closing number；If the adjacent blink width twice in front and back meets blink width >=0.8M and blink width≤0.2N
It is then blink, is denoted as number of winks；
3) the adjacent longest interval time blinked twice is recorded；By formula: frequency of wink=number of winks/detection cycle calculating is blinked
Eye frequency.
5. the method that a kind of recognition of face based on deep learning according to claim 4 is used to detect fatigue driving,
Be characterized in that: the step S2 opens one's mouth size and to open one's mouth duration data using the mouth that human face recognition model obtains face picture,
The specific method is as follows:
1) the symmetrical upper pixel coordinate of upper lip and lower lip upper and lower position and lower pixel are obtained using human face recognition model
Coordinate, real-time to calculate height of opening one's mouth, height of opening one's mouth is=upper the pixel coordinate and lower pixel coordinate of upper and lower position symmetrically
Distance；2) in detection cycle, the numerical value for height of opening one's mouth is recorded；
3) duration opened one's mouth is recorded.
6. the method that a kind of recognition of face based on deep learning according to claim 5 is used to detect fatigue driving,
Be characterized in that: the step S2 obtains the facial expression of face picture by facial expression disaggregated model, by all people's face figure
Piece is according to excited, normal or tired classification.
7. the method that a kind of recognition of face based on deep learning according to claim 6 is used to detect fatigue driving,
Be characterized in that: the step S3 judges whether person under test is in fatigue driving state according to recognition result as follows；If to
Survey person is in fatigue driving state, then at least meets two in the following conditions:
1) frequency of wink is lower than default frequency of wink value；
2) the adjacent longest interval time blinked twice is lower than default wink time spacing value；
3) duration opened one's mouth is lower than the default duration value opened one's mouth；
4) numerical value for height of opening one's mouth is greater than the numerical value of default height of opening one's mouth；
5) face picture that total 1/2 or more quantity is accounted in detection cycle belongs to fatigue state.
8. the method that a kind of recognition of face based on deep learning according to claim 7 is used to detect fatigue driving,
It is characterized in that: further including step S4: whether being fatigue driving in the predicted detection period based on the analysis results.
9. the method that a kind of recognition of face based on deep learning according to claim 8 is used to detect fatigue driving,
Be characterized in that: each detection cycle of the step S2 is 30 seconds；Frequency acquisition is 30 frame picture per second.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910294584.0A CN110020632A (en) | 2019-04-12 | 2019-04-12 | A method of the recognition of face based on deep learning is for detecting fatigue driving |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910294584.0A CN110020632A (en) | 2019-04-12 | 2019-04-12 | A method of the recognition of face based on deep learning is for detecting fatigue driving |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110020632A true CN110020632A (en) | 2019-07-16 |
Family
ID=67191253
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910294584.0A Pending CN110020632A (en) | 2019-04-12 | 2019-04-12 | A method of the recognition of face based on deep learning is for detecting fatigue driving |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110020632A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111310730A (en) * | 2020-03-17 | 2020-06-19 | 扬州航盛科技有限公司 | Driving behavior early warning system based on facial expressions |
| CN111553190A (en) * | 2020-03-30 | 2020-08-18 | 浙江工业大学 | Image-based driver attention detection method |
| CN111611939A (en) * | 2020-05-22 | 2020-09-01 | 云知声智能科技股份有限公司 | Eye fatigue state detection method and device |
| CN111695510A (en) * | 2020-06-12 | 2020-09-22 | 浙江工业大学 | Image-based fatigue detection method for computer operator |
| CN111708437A (en) * | 2020-06-11 | 2020-09-25 | 湖北美和易思教育科技有限公司 | Artificial intelligence thing networking e-book |
| CN111931748A (en) * | 2020-10-12 | 2020-11-13 | 天能电池集团股份有限公司 | Worker fatigue detection method suitable for storage battery production workshop |
| CN112270215A (en) * | 2020-10-13 | 2021-01-26 | 杭州电子科技大学 | Face recognition method based on sequence feature gradient vector structure |
| CN112989978A (en) * | 2021-03-04 | 2021-06-18 | 扬州微地图地理信息科技有限公司 | Driving assistance recognition method based on high-precision map |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107697069A (en) * | 2017-10-31 | 2018-02-16 | 上海汽车集团股份有限公司 | Fatigue of automobile driver driving intelligent control method |
| CN108460345A (en) * | 2018-02-08 | 2018-08-28 | 电子科技大学 | A kind of facial fatigue detection method based on face key point location |
| CN108609019A (en) * | 2018-05-09 | 2018-10-02 | 浙江金之路信息科技有限公司 | A kind of electric vehicle automatic Pilot method based on artificial intelligence platform |
-
2019
- 2019-04-12 CN CN201910294584.0A patent/CN110020632A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107697069A (en) * | 2017-10-31 | 2018-02-16 | 上海汽车集团股份有限公司 | Fatigue of automobile driver driving intelligent control method |
| CN108460345A (en) * | 2018-02-08 | 2018-08-28 | 电子科技大学 | A kind of facial fatigue detection method based on face key point location |
| CN108609019A (en) * | 2018-05-09 | 2018-10-02 | 浙江金之路信息科技有限公司 | A kind of electric vehicle automatic Pilot method based on artificial intelligence platform |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111310730A (en) * | 2020-03-17 | 2020-06-19 | 扬州航盛科技有限公司 | Driving behavior early warning system based on facial expressions |
| CN111553190A (en) * | 2020-03-30 | 2020-08-18 | 浙江工业大学 | Image-based driver attention detection method |
| CN111611939A (en) * | 2020-05-22 | 2020-09-01 | 云知声智能科技股份有限公司 | Eye fatigue state detection method and device |
| CN111708437A (en) * | 2020-06-11 | 2020-09-25 | 湖北美和易思教育科技有限公司 | Artificial intelligence thing networking e-book |
| CN111695510A (en) * | 2020-06-12 | 2020-09-22 | 浙江工业大学 | Image-based fatigue detection method for computer operator |
| CN111931748A (en) * | 2020-10-12 | 2020-11-13 | 天能电池集团股份有限公司 | Worker fatigue detection method suitable for storage battery production workshop |
| CN111931748B (en) * | 2020-10-12 | 2021-01-26 | 天能电池集团股份有限公司 | Worker fatigue detection method suitable for storage battery production workshop |
| CN112270215A (en) * | 2020-10-13 | 2021-01-26 | 杭州电子科技大学 | Face recognition method based on sequence feature gradient vector structure |
| CN112270215B (en) * | 2020-10-13 | 2024-03-26 | 杭州电子科技大学 | Face recognition method based on sequence feature gradient vector structure |
| CN112989978A (en) * | 2021-03-04 | 2021-06-18 | 扬州微地图地理信息科技有限公司 | Driving assistance recognition method based on high-precision map |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110020632A (en) | A method of the recognition of face based on deep learning is for detecting fatigue driving | |
| Zhuang et al. | Driver fatigue detection method based on eye states with pupil and iris segmentation | |
| CN107491740A (en) | A kind of neonatal pain recognition methods based on facial expression analysis | |
| CN107358206A (en) | Micro- expression detection method that a kind of Optical-flow Feature vector modulus value and angle based on area-of-interest combine | |
| CN107007257B (en) | The automatic measure grading method and apparatus of the unnatural degree of face | |
| CN107085715A (en) | A kind of television set intelligently detects the dormant system and method for user | |
| CN108664947A (en) | A kind of fatigue driving method for early warning based on Expression Recognition | |
| CN106446811A (en) | Deep-learning-based driver's fatigue detection method and apparatus | |
| CN105286802B (en) | Driver Fatigue Detection based on video information | |
| CN109389806A (en) | Fatigue driving detection method for early warning, system and medium based on multi-information fusion | |
| CN110334600A (en) | A kind of multiple features fusion driver exception expression recognition method | |
| CN110119672A (en) | A kind of embedded fatigue state detection system and method | |
| CN110147738B (en) | Driver fatigue monitoring and early warning method and system | |
| CN106909220A (en) | A kind of sight line exchange method suitable for touch-control | |
| CN104123549B (en) | Eye positioning method for real-time monitoring of fatigue driving | |
| CN110472512A (en) | A kind of face state identification method and its device based on deep learning | |
| CN107403142A (en) | A kind of detection method of micro- expression | |
| CN110472511A (en) | A kind of driver status monitoring device based on computer vision | |
| CN107844780A (en) | A kind of the human health characteristic big data wisdom computational methods and device of fusion ZED visions | |
| CN106599821A (en) | Controller fatigue detection method and system based on BP neural network | |
| CN110428908A (en) | A kind of eyelid movement functional assessment system based on artificial intelligence | |
| CN106691440A (en) | Controller fatigue detection method and system based on BP neural network | |
| CN107480586A (en) | Bio-identification photo bogus attack detection method based on human face characteristic point displacement | |
| CN109543577A (en) | A kind of fatigue driving detection method for early warning based on facial expression feature | |
| CN107967944A (en) | A kind of outdoor environment big data measuring of human health method and platform based on Hadoop |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190716 |
|
| RJ01 | Rejection of invention patent application after publication |