CN104269028B - Fatigue driving detection method and system - Google Patents
Fatigue driving detection method and system Download PDFInfo
- Publication number
- CN104269028B CN104269028B CN201410568886.XA CN201410568886A CN104269028B CN 104269028 B CN104269028 B CN 104269028B CN 201410568886 A CN201410568886 A CN 201410568886A CN 104269028 B CN104269028 B CN 104269028B
- Authority
- CN
- China
- Prior art keywords
- face
- state
- value
- eyes
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
- G08B21/06—Alarms for ensuring the safety of persons indicating a condition of sleep, e.g. anti-dozing alarms
Landscapes
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a fatigue driving detection method and system. The fatigue driving detection method comprises the steps of S1, image acquisition and preprocessing; S2, face localization and detection; S3, face tracking; S4, eye detection and state recognition; S5, calculation of the fatigue PERCLOS value; S6, comparison between the obtained PERCLOS value with a preset threshold value and judgment of fatigue driving of drivers. By adopting the fatigue driving detection method and system, the microcosmic change of the faces of the drivers can be automatically captured, the state of the drivers can be judged through scientific probability calculation, the drivers who are in the fatigue driving state can be warned in time, and therefore driving safety can be guaranteed.
Description
Technical field
Design vehicle security technology area of the present invention, is specifically designed a kind of method for detecting fatigue driving and system.
Background technology
With the development of economic society, motor vehicles sharply increase, and the vehicle accident that therefore fatigue driving leads to is also more next
More.For this problem, create many fatigue-driving detection technologies, be summed up three kinds.
(1) detection technique based on physiological signal, according to the blood pressure of driver, the physiological signal such as brain wave judges whether tired
Labor, it is desirable to have high-precision detecting instrument, and contact driver, driver is had a certain impact and high cost.
(2) fatigue detecting technology based on running information, when steering wheel does not revise direction for a long time, or left and right vehicle wheel shakes
Shake frequency or amplitude is excessive, speed wobble is judged as fatigue driving, but the driver for different driving habits, and
, it is difficult to driving is sought unity of standard, degree of accuracy is not high for driving conditions on different roads.
(3) fatigue detecting technology based on physiological feature, when driver fatigue, can show frequency of bowing and close one's eyes
The physiological features such as increase.By the method for machine vision, detect the above-mentioned physiological feature of driver it can be determined that driver whether
Fatigue.Based on the method for Machine Vision Detection physiological driver's feature, there is noncontact, cost is low, the advantages of degree of accuracy is high,
It is widely adopted in current fatigue detecting system.
In above-mentioned three kinds of methods, the third method is substantially better than first two method.But it also has shortcoming, the third side at present
Method passes through image processing techniquess, locating human face mostly, then analyzes the state of eyes in the range of face, judges whether fatigue.
Wherein locating human face is the hunting zone in order to reduce eyes, improves treatment effeciency, is the crucial step of fatigue detecting.Wherein fixed
The method that position face and eyes are commonly used has two kinds, and a kind of is method based on morphological image, and one kind is to be based on machine learning
The method of grader.The former is computationally intensive, and speed is slow, big by illumination effect.The latter's calculating speed is fast, quilt little by illumination effect
Widely used with Face detection technology in.But in the current technology of grader locating human face adopting, when driver head moves
Make excessive it is impossible to collect during face image it is impossible to be properly positioned face location, there is fatigue detecting algorithm Problem of Failure.
Content of the invention
In order to solve above-mentioned technical problem, the present invention discloses a kind of method for detecting fatigue driving and system, and the present invention adopts
Following technical scheme is solving above-mentioned technical problem:
A kind of method for detecting fatigue driving, comprises the steps:
S1, collection image and pretreatment, gather driver's image information by image capture interface and transmit to maincenter
Reason device, hub processor carries out pretreatment to collection image；
S2, Face detection and detection, using the face classification device based on haar feature, detect driver's human face region；
S3, face tracking, using the Face tracking algorithm of Kalman filter algorithm, track human faces；
S4, eye detection and state recognition, position driver's eyes using eye opening grader and identify eye state, record
Recognition result；
S5, the tired perclos value of calculating, after obtaining the state recognition result of eyes, by eyes in the unit of account time
Percentage ratio shared by closing time is perclos (percentage of eyelid closure over time):
S6, the perclos obtaining value is compared with default threshold value, judges driver's whether fatigue driving.
Preferably, in a kind of above-mentioned method for detecting fatigue driving, described step s2 specifically includes step a, using base
In Kalman filter algorithm parameter tracking face:
A1, set up system state equation；
x(k+1)=a(k)x(k)+w(k)
A2, set up systematic observation equation；
z(k)=h(k)x(k)+v(k)
A3, prediction face location, according to the predictive equation group of Kalman filter, wave filter is using state on last stage
Estimate, make the estimation to current state；
Predictive equation group:
(1) predicted state
x(k|k-1)=a(k)x(k|k-1)
(2) predicted estimate covariance matrix
p(k|k-1)=a(k)p(k-1|k-1)at (k)+q(k)
A4, the actual face location of detection, behind the position predicting current face place, call haar grader in prediction
Detect face in region, obtain current state information；When can't detect face in estimation range, just detect face in full figure；When
When full figure also can't detect face, illustrate driver head's action excessive it is impossible to face is detected；
A5, Kalman filter is updated using current actual face location by renewal equation group, allow Kalman to filter
The predictive value of ripple device and actual value become closer to, and reach the effect of tracking；
Renewal equation group:
(3) optimum Kalman gain
k(k)=p(k|k-1)ht k[h(k)p(k|k-1)ht (k)+r(k)]-1
(4) state estimation updating
x(k|k)=x(k|k-1)+k(k)[z(k+1)-h(k)x](k|k-1)
(5) covariance updating is estimated
p(k|k)=[e-k(k)h(k)]p(k|k-1)
A6, mark human face region in the input image.
Wherein xkIt is the state estimation in face in the k moment, a(k)It is to act on x(k)On state-transition matrix, h(k)It is
The observing matrix of system, z(k)It is the observational variable of system, w(k)It is process noise, and assume that it meets average for zero covariance square
Battle array is q(k)Gauss distribution, i.e. wk～n (0, q(k))；v(k)It is observation noise, also assuming that it meets average is null covariance matrix
For r(k)Gauss distribution, i.e. vk～n (0, r(k))；pkFor error correlation matrix, measure the levels of precision of estimated value；
Preferably, in a kind of above-mentioned method for detecting fatigue driving, the eye opening grader in described step s3 is through as follows
Step is trained:
B1, collect the sample source of face, intercept the human eye area of face sample source and non-human eye area, as positive sample and
Negative sample；
B2, using machine learning algorithm train eye opening grader.
Preferably, in a kind of above-mentioned method for detecting fatigue driving, in step s5, perclos value adopts equation below
Calculate:
Perclos value=(frame number of eyes closed/fixation frame number) * 100%
Preferably, in a kind of above-mentioned method for detecting fatigue driving, described step b2 includes:
B21, the positive and negative pattern representation file of generation；
B22, the vector description file of generation positive sample；
B23, it is trained using the executable file that opencv carries.
Compared with prior art, the present invention has the following technical effect that
Using the design of the present invention, can automatically catch face's micro-variations of driver, by the probability calculation of science
Differentiate driver status, and warning in time is in the driver of fatigue driving state it is ensured that traffic safety.The present invention uses simultaneously
Kalman filter and related algorithm are predicted through continuous, update, predictive value and actual value will be made to become closer to, and are holding
In the range of error permitted, predictive value can be considered as actual value.When driver head's action excessive it is impossible to people is positioned by grader
When face position, followed the tracks of using Kalman filter, replace real face location with the face location of prediction, continue fatigue
Drive detection, it is to avoid cannot locating human face, the problem that fatigue detecting algorithm lost efficacy.
Brief description
Fig. 1 is principle of the invention schematic diagram；
Fig. 2 is face tracking principle schematic in the present invention；
Fig. 3 is Kalman filter recursive process in the present invention；
Fig. 4 is the training method flow process of eye opening grader in the present invention.
Specific embodiment
Below in conjunction with the accompanying drawings the present invention is described in further detail, but the present invention is not limited solely to following enforcement
Example, as shown in Figure 1, Figure 2, Figure 3 shows:
A kind of method for detecting fatigue driving of the present invention, comprises the steps:
S1, collection image and pretreatment, gather driver's image information by image capture interface and transmit to maincenter
Reason device, hub processor carries out the pretreatment such as gray processing, histogram equalization to collection image；
S2, Face detection and detection, using the face classification device based on haar feature, detect driver's human face region；
S3, face tracking, using the Face tracking algorithm of Kalman filter algorithm, track human faces, specifically include as follows
Step:
A1, set up system state equation；
x(k+1)=a(k)x(k)+w(k)
A2, set up systematic observation equation；
z(k)=h(k)x(k)+v(k)
A3, prediction face location, according to the predictive equation group of Kalman filter, wave filter is using state on last stage
Estimate, make the estimation to current state；
Predictive equation group:
(1) predicted state
x(k|k-1)=a(k)x(k|k-1)
(2) predicted estimate covariance matrix
p(k|k-1)=a(k)p(k-1|k-1)at (k)+q(k)
A4, the actual face location of detection, behind the position predicting current face place, call haar grader in prediction
Detect face in region, obtain current state information；When can't detect face in estimation range, just detect face in full figure；When
When full figure also can't detect face, illustrate driver head's action excessive it is impossible to face is detected；
A5, Kalman filter is updated using current actual face location by renewal equation group, allow Kalman to filter
The predictive value of ripple device and actual value become closer to, and reach the effect of tracking；
Renewal equation group:
(3) optimum Kalman gain
k(k)=p(k|k-1)ht k[h(k)p(k|k-1)ht (k)+r(k)]-1
(4) state estimation updating
x(k|k)=x(k|k-1)+k(k)[z(k+1)-h(k)x](k|k-1)
(5) covariance updating is estimated
p(k|k)=[e-k(k)h(k)]p(k|k-1)
A6, mark human face region in the input image.
Wherein xkIt is the state estimation in face in the k moment, a(k)It is to act on x(k)On state-transition matrix, h(k)It is
The observing matrix of system, z(k)It is the observational variable of system, w(k)It is process noise, and assume that it meets average for zero covariance square
Battle array is q(k)Gauss distribution, i.e. wk～n (0, q(k))；v(k)It is observation noise, also assuming that it meets average is null covariance matrix
For r(k)Gauss distribution, i.e. vk～n (0, r(k))；pkFor error correlation matrix, measure the levels of precision of estimated value；
Directly carry out Face detection with grader, exist computationally intensive, take long, the problems such as error detection is high.In order to improve
The speed of Face detection and accuracy, the present invention is using the track algorithm of the face based on Kalman filter.
But Kalman filter algorithm is applied to linear system, almost all of system is all nonlinear in practice,
Only real system is approximately linear system, Kalman filter algorithm could be used.Between two frames every in video,
Time interval is short, and the speed of target and position are basically identical between two frames, so the motion of face can be approximately at the uniform velocity straight
Line moves, and meets Kalman filter and system linearity is assumed.
Using Kalman filter algorithm follow the tracks of target it is necessary first to set up system model, then by algorithm filter with
Track target.
1.1 set up Linear system model:
The state of Kalman filter is represented by following two variables: x(k)Estimation for the state in k moment；pkFor error
Correlation matrix, for measuring the levels of precision of estimated value；
The general mathematics model of system is: x(k+1)=a(k)x(k)+w(k)(system state equation), z(k)=h(k)x(k)+v(k)
(systematic observation equation).x(k)It is the system state variables sequence in k moment, x(k)={ x, y, w, h, vx, vy, vw, vh)t, the present invention
The status switch chosen is coordinate (x, y) and the speed (v of face area geometric centerx, vy), the width of face's boundary rectangle and height
(w, h), and wide and high pace of change (vw, vh).a(k)It is to act on x(k)On state-transition matrix；h(k)It is the sight of system
Survey matrix；z(k)It is the observational variable of system；w(k)It is process noise, and to assume that it meets average be zero, covariance matrix is q(k)
Gauss distribution, i.e. wk～n (0, q(k)).v(k)It is observation noise, also assuming that it meets average is zero, covariance matrix is r(k)
Gauss distribution, i.e. vk～n (0, r(k)).
Set up system model, need to set up system state equation, and observational equation.
(1) set up system state equation:
Because between two frames every in video, time interval is short, the position (x of target in next frame(k+1), y(k+1)) and speed
(vx(k+1), vy(k+1)), wide and high (w(k+1), hk+1)), and wide and high pace of change (vw(k+1), vh(k+1)) too big all without having
Change, then can approximately obtain following equations:
Because of the state variable above making the k moment it is again:
x(k)={ x(k)y(k)w(k)h(k)vx(k)vy(k)vw(k)vh(k)}t
The state variable of subsequent time is x(k+1), then equation group (1) can be converted into:
Then equation group (2) can be converted to:
x(k+1)=a(k)·x(k)
Because this is approximate linear movement, so not being completely credible, there is certain error, this error be with
Machine value, also referred to as process noise.Kalman filter assumes that the random error of system meets the Gauss distribution that average is 0.If crossing
Journey noise w(k)Meeting average is zero, and covariance matrix is q(k)Gauss distribution, i.e. wk～n (0, q(k)), taken by test of many times
Empirical value q (k)=0.01 × e, e is the unit matrix of 8*8.Then consider that the system state equation after noise is changed into:
x(k+1)=a(k)x(k)+w(k)
(2) set up systematic observation equation:
So in multisystem state variable, we are finally only concerned the value of the wide height (w, h) in position (x, y) of target, that is, see
Direction finding amount z(k)={ x y w h }t.So obtaining equation group:
Formula (3) is represented by: z(k)=h(k)·x(k)
The same with system state equation, the observational variable of system is also to there is random error, is called observation noise, karr
Graceful wave filter assumes that observation noise also obeys the Gauss distribution that average is 0.If observation noise v(k)Meeting average is zero, covariance
Matrix is q(k)Gauss distribution, i.e. vk～n (0, r(k)), empirical value r is taken by test of many times(k)=0.1 × e, e are the list of 8*8
Bit matrix.Then consider that the system state equation after noise is changed into:
z(k)=h(k)x(k)+v(k)
The tracking process of 1.2 Kalman filter:
The operation of Kalman filter includes prediction and updates two stages.In forecast period, wave filter uses a upper shape
The estimation of state, makes the estimation to current state.In the more new stage, wave filter utilizes the observation optimization to current state pre-
The predictive value that the survey stage obtains, to obtain a more accurate new estimation value.
Predictive equation group: (a) predicted state
x(k|k-1)=a(k)x(k|k-1)
(b) predicted estimate covariance matrix
p(k|k-1)=a(k)p(k-1|k-1)at (k)+q(k)
Renewal equation group: (c) optimum Kalman gain
k(k)=p(k|k-1)ht k[h(k)p(k|k-1)ht (k)+r(k)]-1
D state estimation that () updates
x(k|k)=x(k|k-1)+k(k)[z(k+1)-h(k)x](k|k-1)
E covariance that () updates is estimated
p(k|k)=[e-k(k)h(k)]p(k|k-1)
By formula (a), state x obtaining target in subsequent time can be predicted(1|0).Instantly after being carved into for the moment, pre-
Near location is put, local detection is carried out to image, detect virtual condition x of target(1), it is right to be completed by formula (b) to formula (e)
The renewal of Kalman filter, the dbjective state for predicting again subsequent time is ready.So continuous recursion, can be continuous
The state of prediction subsequent time moving target, as shown in Figure 3.
Be can be seen that by above-mentioned steps, Kalman filter is predicted through continuous, updates, will make predictive value and actual value
Become closer to, in the range of tolerance, predictive value can be considered as actual value.When driver head's action excessive it is impossible to
When by grader locating human face position, can be followed the tracks of using Kalman filter, be replaced truly with the face location of prediction
Face location, continue fatigue driving detection.So avoid cannot locating human face, fatigue detecting algorithm lost efficacy problem
S4, eye detection and state recognition, position driver's eyes using eye opening grader and identify eye state, record
Recognition result, the eye opening grader of use is trained through following method, as shown in Figure 4:
B1, collecting sample as much as possible, including positive negative sample, wherein positive sample is the human eye of state of opening, negative sample
Human eye for closure state and other non-eye opening regions；
B21, the positive and negative pattern representation file of generation；
B22, the vector description file of generation positive sample；
B23, it is trained using the executable file that opencv carries.
The training of grader can be carried out according to three steps of Fig. 4.
The description file of the generation sample of the first step, generally refers to describe the path of positive sample and target image-open
Eye pattern picture number in the picture and position.Negative sample only needs to describe the path of its sample.The form of therefore positive sample
It is as follows,
<file path of sample><number of target image><position in the picture of target image>.
Remaining positive sample form is by that analogy.
Negative sample only needs to describe sample path in the picture, and its form is as follows:
<position in the image of sample>
The descriptor format of remaining negative sample is by that analogy
The vector description file of the generation positive sample of second step, calls the file of correlation.
Being trained using the executable file that opencv carries of 3rd step.
The present invention trains grader according to the method for adaboost it is stipulated that the verification and measurement ratio of each layer of strong classifier of training is
0.998, false drop rate is 0.5.It is initially using the sample collected most, including positive negative sample totally 25000 sample sets, instructed
Practice, during training, every layer of strong classifier must is fulfilled for 0.998 verification and measurement ratio and 0.5 false drop rate, so that training
Obtain the requirement that grader effect can meet system.
Using adaboost algorithm training in advance, eye opening grader carries out target detection well, compares other detection methods,
The speed processing is fast, and the impact receiving illumination is little, and accuracy rate is high.And directly train eye opening grader, while positioning eyes
Just can judge that eyes are in the state of opening, when be not detected by eyes all when, be all considered as closed-eye state, so by eyes
Positioning and state-detection combine, and improve detection efficiency.
S5, the tired perclos value of calculating, after obtaining the state recognition result of eyes, by eyes in the unit of account time
Percentage ratio shared by closing time is perclos (percentage of eyelid closure over time)；
S6, the perclos obtaining value is compared with default threshold value, judges driver's whether fatigue driving.
In step s4, perclos value is the percentage ratio in the unit interval shared by the eyes closed time, using perclos value
The fatigue state of differentiation driver is the most efficient method generally acknowledged at present, and its computing formula is as follows:
Perclos value=(eyes closed time/set time) * 100%
For the ease of programming, the molecule denominator on above formula left side average time divided by each detection simultaneously, obtain following formula:
Perclos value=(frame number of eyes closed/fixation frame number) * 100%
In step s5, default threshold value is rule of thumb preset generally according to substantial amounts of experimental data and environment, will
To perclos value be compared with default threshold value, when perclos value be more than or equal to predetermined threshold value when, system judge drive
For fatigue driving and the system of ordering about sends alarm to member, reminds driver's rest, to ensure traffic safety.
Using the design of the present invention, can automatically catch face's micro-variations of driver, by the probability calculation of science
Differentiate driver status, and warning in time is in the driver of fatigue driving state it is ensured that traffic safety.The present invention uses simultaneously
Kalman filter and related algorithm are predicted through continuous, update, predictive value and actual value will be made to become closer to, and are holding
In the range of error permitted, predictive value can be considered as actual value.When driver head's action excessive it is impossible to people is positioned by grader
When face position, followed the tracks of using Kalman filter, replace real face location with the face location of prediction, continue fatigue
Drive detection, it is to avoid cannot locating human face, the problem that fatigue detecting algorithm lost efficacy.
The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all essences in the present invention
Any modification, equivalent and improvement made within god and principle etc., should be included within the scope of the present invention.
Claims (3)
1. a kind of method for detecting fatigue driving is it is characterised in that comprise the steps:
S1, collection image and pretreatment, gather driver's image information by image capture interface and transmit to hub processor,
Hub processor carries out pretreatment to collection image；
S2, Face detection and detection, using the face classification device based on haar feature, detect driver's human face region；
S3, face tracking, using the Face tracking algorithm of Kalman filter algorithm, track human faces；
S4, eye detection and state recognition, position driver's eyes using eye opening grader and identify eye state, record identification
Result；
S5, the tired perclos value of calculating, after obtaining the state recognition result of eyes, by eyes closed in the unit of account time
Percentage ratio shared by time is perclos (percentage of eyelid closure over time)；
S6, the perclos obtaining value is compared with default threshold value, judges driver's whether fatigue driving；
Described step s3 specifically includes step a, using based on Kalman filter algorithm parameter tracking face:
A1, set up system state equation；
x(k+1)=a(k)x(k)+w(k)
x(k)={ x, y, w, h, vx, vy, vw, vh)t, the status switch of selection is coordinate (x, y) and the speed of face area geometric center
Degree (vx, vy), the width of face's boundary rectangle and high (w, h), and wide and high pace of change (vw, vh)；
A2, set up systematic observation equation；
z(k)=h(k)x(k)+v(k)
A3, prediction face location, according to the predictive equation group of Kalman filter, wave filter is estimated using state on last stage
Meter, makes the estimation to current state；
Predictive equation group:
(1) predicted state
x(k|k-1)=a(k)x(k|k-1)
(2) predicted estimate covariance matrix
p(k|k-1)=a(k)p(k-1|k-1)at (k)+q(k)
A4, the actual face location of detection, behind the position predicting current face place, call haar grader in estimation range
Interior detection face, obtains current state information；When can't detect face in estimation range, just detect face in full figure；When complete
Figure is when also can't detect face, illustrate driver head's action excessive it is impossible to face is detected；
A5, Kalman filter is updated using current actual face location by renewal equation group, allow Kalman filter
Predictive value become closer to actual value, reach the effect of tracking；
Renewal equation group:
(3) optimum Kalman gain
k(k)=p(k|k-1)ht k[h(k)p(k|k-1)ht (k)+r(k)]-1
(4) state estimation updating
x(k|k)=x(k|k-1)+k(k)[z(k+1)-h(k)x(k|k-1)]
(5) covariance updating is estimated
p(k|k)=[e-k(k)h(k)]p(k|k-1)
A6, mark human face region in the input image；
Wherein x(k)It is the state estimation in face in the k moment, a(k)It is to act on x(k)On state-transition matrix, h(k)It is system
Observing matrix, z(k)It is the observational variable of system, w(k)It is process noise, and assume that it meets average and for null covariance matrix is
q(k)Gauss distribution, i.e. w(k)～n (0, q(k))；v(k)It is observation noise, also assume that it meets average and for null covariance matrix is
r(k)Gauss distribution, i.e. v(k)～n (0, r(k))；p(k)For error correlation matrix, measure the levels of precision of estimated value；
State-transition matrix
Systematic observation matrix
Eye opening grader in described step s4 is trained through following steps:
B1, the sample source of collection face, intercept the human eye area of face sample source and non-human eye area, as positive sample and negative sample
This；
B2, using machine learning algorithm train eye opening grader；Judge that eyes are in the state of opening while positioning eyes,
When being not detected by eyes, it is considered as closed-eye state.
2. a kind of method for detecting fatigue driving according to claim 1 it is characterised in that: in step s5, perclos value is adopted
Calculated with equation below:
Perclos value=(frame number of eyes closed/fixation frame number) * 100%.
3. a kind of method for detecting fatigue driving according to claim 1 is it is characterised in that described step b2 includes:
B21, the positive and negative pattern representation file of generation；
B22, the vector description file of generation positive sample；
B23, it is trained using the executable file that opencv carries.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410568886.XA CN104269028B (en) | 2014-10-23 | 2014-10-23 | Fatigue driving detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410568886.XA CN104269028B (en) | 2014-10-23 | 2014-10-23 | Fatigue driving detection method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104269028A CN104269028A (en) | 2015-01-07 |
| CN104269028B true CN104269028B (en) | 2017-02-01 |
Family
ID=52160545
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410568886.XA Expired - Fee Related CN104269028B (en) | 2014-10-23 | 2014-10-23 | Fatigue driving detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104269028B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573657B (en) * | 2015-01-09 | 2018-04-27 | 安徽清新互联信息科技有限公司 | It is a kind of that detection method is driven based on the blind of feature of bowing |
| CN106295474B (en) * | 2015-05-28 | 2019-03-22 | 交通运输部水运科学研究院 | Fatigue detection method, system and the server of deck officer |
| CN105160913B (en) * | 2015-08-17 | 2018-02-06 | 上海斐讯数据通信技术有限公司 | A kind of method and device of specification driver driving behavior |
| CN106203394B (en) * | 2016-07-26 | 2019-04-26 | 浙江捷尚视觉科技股份有限公司 | Fatigue driving safety monitoring method based on human eye state detection |
| CN106447828A (en) * | 2016-10-10 | 2017-02-22 | 深圳市驰勇科技有限公司 | Automobile data recorder with iris recognition function |
| CN108229359A (en) * | 2017-12-26 | 2018-06-29 | 大唐软件技术股份有限公司 | A kind of face image processing process and device |
| 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 |
| KR102529919B1 (en) * | 2018-11-28 | 2023-05-08 | 현대자동차주식회사 | Apparatus, system and method for managing drowsy driving |
| CN110852247A (en) * | 2019-11-07 | 2020-02-28 | 北京云迹科技有限公司 | Abnormality detection method, abnormality detection device, electronic apparatus, and computer-readable storage medium |
| CN111035096A (en) * | 2020-01-09 | 2020-04-21 | 郑州铁路职业技术学院 | Engineering constructor fatigue detection system based on safety helmet |
| CN112668393A (en) * | 2020-11-30 | 2021-04-16 | 海纳致远数字科技(上海)有限公司 | Fatigue degree detection device and method based on face recognition and key point detection |
| CN112950900A (en) * | 2021-01-27 | 2021-06-11 | 吉林云帆智能工程有限公司 | Driver behavior monitoring and detecting method |
| DE102021111465A1 (en) | 2021-05-04 | 2022-11-10 | Bayerische Motoren Werke Aktiengesellschaft | DETERMINING A PERSON'S FATIGUE BASED ON AN ANALYSIS OF AN EYE BLINK |
| CN113536900A (en) * | 2021-05-31 | 2021-10-22 | 浙江大华技术股份有限公司 | Method and device for evaluating quality of face image and computer readable storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1680779A (en) * | 2005-02-04 | 2005-10-12 | 江苏大学 | Fatigue monitoring method and device for driver |
| CN101090482A (en) * | 2006-06-13 | 2007-12-19 | 唐琎 | Driver fatigue monitoring system and method based on image process and information mixing technology |
| CN101540090A (en) * | 2009-04-14 | 2009-09-23 | 华南理工大学 | Driver fatigue monitoring device based on multivariate information fusion and monitoring method thereof |
| CN102622600A (en) * | 2012-02-02 | 2012-08-01 | 西南交通大学 | High-speed train driver alertness detecting method based on face image and eye movement analysis |
| CN104013414A (en) * | 2014-04-30 | 2014-09-03 | 南京车锐信息科技有限公司 | Driver fatigue detecting system based on smart mobile phone |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0318351A (en) * | 1989-06-16 | 1991-01-25 | Omron Corp | Driving fatigue calculating device |
-
2014
- 2014-10-23 CN CN201410568886.XA patent/CN104269028B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1680779A (en) * | 2005-02-04 | 2005-10-12 | 江苏大学 | Fatigue monitoring method and device for driver |
| CN101090482A (en) * | 2006-06-13 | 2007-12-19 | 唐琎 | Driver fatigue monitoring system and method based on image process and information mixing technology |
| CN101540090A (en) * | 2009-04-14 | 2009-09-23 | 华南理工大学 | Driver fatigue monitoring device based on multivariate information fusion and monitoring method thereof |
| CN102622600A (en) * | 2012-02-02 | 2012-08-01 | 西南交通大学 | High-speed train driver alertness detecting method based on face image and eye movement analysis |
| CN104013414A (en) * | 2014-04-30 | 2014-09-03 | 南京车锐信息科技有限公司 | Driver fatigue detecting system based on smart mobile phone |
Non-Patent Citations (1)
| Title |
|---|
| 基于卡尔曼滤波的人脸跟踪算法;李晶等;《西安邮电学院学报》;20100331;第15卷(第3期);第101-104页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104269028A (en) | 2015-01-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104269028B (en) | Fatigue driving detection method and system | |
| Ramzan et al. | A survey on state-of-the-art drowsiness detection techniques | |
| Ohn-Bar et al. | Head, eye, and hand patterns for driver activity recognition | |
| Bergasa et al. | Real-time system for monitoring driver vigilance | |
| CN104013414B (en) | A kind of Study in Driver Fatigue State Surveillance System based on intelligent movable mobile phone | |
| CN105286802A (en) | Driver fatigue detection method based on video information | |
| CN105564436A (en) | Advanced driver assistance system | |
| Celona et al. | A multi-task CNN framework for driver face monitoring | |
| Sun et al. | An integrated solution for lane level irregular driving detection on highways | |
| Wu et al. | Reasoning-based framework for driving safety monitoring using driving event recognition | |
| CN106184220B (en) | Abnormal driving detection method in a kind of track based on vehicle location track | |
| CN104794855A (en) | Driver's attention comprehensive assessment system | |
| CN104021370A (en) | Driver state monitoring method based on vision information fusion and driver state monitoring system based on vision information fusion | |
| CN105404862A (en) | Hand tracking based safe driving detection method | |
| Rezaei et al. | Simultaneous analysis of driver behaviour and road condition for driver distraction detection | |
| Pech et al. | Head tracking based glance area estimation for driver behaviour modelling during lane change execution | |
| Ma et al. | Real time drowsiness detection based on lateral distance using wavelet transform and neural network | |
| CN101872419A (en) | Method for detecting fatigue of automobile driver | |
| Bergasa et al. | Visual monitoring of driver inattention | |
| Joseph et al. | Real time drowsiness detection using Viola jones & KLT | |
| Coetzer et al. | Driver fatigue detection: A survey | |
| Ananthi et al. | Drivers Drowsiness Detection using Image Processing and I-Ear Techniques | |
| Bergasa et al. | Real-time system for monitoring driver vigilance | |
| Billah et al. | Tracking-based detection of driving distraction from vehicular interior video | |
| Guo et al. | Monitoring and detection of driver fatigue from monocular cameras based on Yolo v5 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170201 Termination date: 20171023 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |