CN106384096A - Fatigue driving monitoring method based on blink detection - Google Patents
Fatigue driving monitoring method based on blink detection Download PDFInfo
- Publication number
- CN106384096A CN106384096A CN201610835397.5A CN201610835397A CN106384096A CN 106384096 A CN106384096 A CN 106384096A CN 201610835397 A CN201610835397 A CN 201610835397A CN 106384096 A CN106384096 A CN 106384096A
- Authority
- CN
- China
- Prior art keywords
- image
- face image
- face
- ocular
- driver
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a fatigue driving monitoring method based on blink detection, and the method comprises the steps: 1, image collection: uploading a face image collected at each sampling moment to an image processor synchronously; 2, image processing: enabling the image processor to carry out the analysis processing of the face images received in each analysis processing period, wherein the analysis processing process comprises the following steps: 201, carrying out the analysis processing of the face images collected in the first analysis processing period; 202, carrying out the analysis processing of the face images collected in the next analysis processing period; 203, repeatedly carrying out the step 202 for M-3 times; 204, carrying out the analysis processing of the face images collected in the next analysis processing period and carrying out the fatigue driving judgment; 205, returning to step 204. The method is simple in step, is reasonable in design, is convenient to implement, is good in use effect, can carry out the accurate monitoring of the fatigue driving state of a driver conveniently and quickly based on the blink detection of a driver face image, and is high in practical value.
Description
Technical field
The invention belongs to fatigue driving monitoring technical field, especially relate to a kind of prison of the fatigue driving based on blink detection
Survey method.
Background technology
According to statistical data, fatigue driving occupies very big ratio in vehicle accident, especially tired in highway
Please the vehicle accident ratio sailing initiation is far above common road, mainly includes long-duration driving, does not have enough sleep or of poor quality, raw
The reasons such as the reason rhythm and pace of moving things, driver's factor.In recent years, driver fatigue monitoring or even Physiological Psychology state analysiss have attracted the whole world
The broad interest of researcher, the research with regard to fatigue driving monitoring has been achieved for certain achievement.
At present, method for detecting fatigue driving mainly has following 4 kinds：Firstth, the fatigue driving based on physiological driver's signal
Detection, in corresponding site installation detecting device or the electrode of driver, the heart to driver specifically in biological information field
Rate, brain wave or electromyographic signal are monitored judging the state of driver, and this method achieves relatively in the research of early stage
Good effect；Wu Shaobin of Beijing Institute of Technology et al. utilizes physiological detector to gather the brain wave of driver, relative analyses brain
Relation between the Subjective fatigue evaluation and test of the different frequency bands signal power spectrum of the signal of telecommunication and driver, result shows：Subjective tired
Labor evaluation and test is corresponding with the change of power spectral value in EEG signals, and ratio (α+the θ)/β of EEG power spectrum is bigger, and level of fatigue is got over
Height, but because equipment installation is loaded down with trivial details, wears inconvenience, cost is very high, cause to be widely popularized in actual applications；Secondth,
Based on operator behavior fatigue driving detection, Wang Fei of Northeastern University et al. will be special with trailer reversing for electroencephalogram identification
Property the fatigue state to detect driver that combines provide theoretical and experimental basis it is anticipated that building fatigue driving detecting system,
And devise drive simulating experiment, electroencephalogram (EEG) signal of collection subjectss and corresponding steering wheel operating data；According to car
Handling characteristic assessment driver's fatigue degree is determining the criteria for classification of EEG signal, and selects support vector machine to EEG signal
Classified to complete the qualitative analyses to driver's mental fatigue state；Due to the different operating habit of different drivers or
Nonstandard operational motion causes the Detection results of this method undesirable；3rd, the detection method based on car status information,
Most representational is the AutoVue system that Iteris company of the U.S. develops, and it is taken the photograph by a CCD towards road ahead
As head detects the wheelpath of driver, when driver leads to unconscious run-off-road because of fatigue (as steering indicating light is not opened),
Can give a warning to driver in time；Zhong et al. utilizes energy spectrometer and wavelet analysis technology, by vehicle driving trace
With the monitoring of steering wheel rotation angle case, also achieve the detection to driver's whether fatigue driving；But due to real road conditions
Complexity, road conditions change is complicated, and in actual environment, Detection results are poor；4th, based on physiological driver's response feature
Detection method, the method can accurately identify the state of driver, but the detection difficulty of physiological driver's response feature is larger.
Content of the invention
The technical problem to be solved is for above-mentioned deficiency of the prior art, provides a kind of being based on to blink
The fatigue driving monitoring method of detection, its method and step is simple, reasonable in design and realization convenience, using effect are good, based on driving
The blink detection of member's face image can carry out accurate measurements by easy, the quick fatigue driving state to driver, and practical value is high.
For solving above-mentioned technical problem, the technical solution used in the present invention is：A kind of fatigue driving based on blink detection
Monitoring method is it is characterised in that the method comprises the following steps：
Step one, image acquisition：Using image capture device and according to sample frequency f set in advances, drive to monitored
The face image of the person of sailing is acquired, and the face image synchronous driving that each sampling instant is gathered is to image procossing
Device；Described image collecting device is connected with image processor；
Wherein, fs=FsHz, FsFor positive integer and Fs=25～35；
Step 2, image procossing：Described image processor is according to analyzing and processing frequency f set in advance and first according to the time
Order afterwards, is analyzed processing to the described face image receiving in each analyzing and processing cycle respectively；Wherein,n
For positive integer and n=5,6,10,12 or 15；
When the described face image receiving in each analyzing and processing cycle is analyzed processing respectively, process is as follows：
Step 201, in first analyzing and processing cycle, received face image analyzes and processes, and comprises the following steps：
Step 2011, face image synchronously storage and blink detection：Described image processor is suitable according to sampling time priority
Sequence, synchronizes storage respectively to the face image described in N width being received in this analysis process cycle, and to each width being received
Described face image carries out blink detection respectively；Wherein, N is an analyzing and processing cycle interior described face image being received
Quantity and N=n × Fs；
To being received in this analysis process cycle when described in arbitrary width, face image carries out blink detection, using simple eye inspection
Survey method or eyes detection method carry out blink detection to this face image；
Wherein, when carrying out blink detection using simple eye detection method to this face image, process is as follows：
Step A1, ocular Image Acquisition：Call eye location module, to monitored driver in this face image
Left eye or right eye present position are positioned, and obtain the ocular image of monitored driver；
Described ocular image is monitored driver's left eye or the image in region residing for right eye；
Step A2, image normalization are processed：Calling figure as normalized module, to ocular figure described in step A1
As being normalized；
In this step, the size of the described ocular image after normalized is M1×N1Individual pixel, wherein M1For
Positive integer and M1=75～85, N1For positive integer and N1=28～32；
Step A3, image binaryzation are processed：Calling figure as binary conversion treatment module, after normalized in step A2
Described ocular image carries out binary conversion treatment；
Step A4, pupil image remove：Calling figure picture removes module, the described eye after binary conversion treatment in removal step A3
The image of pupil in portion's area image, obtains eyes image；
Step A5, circle matching：Call round fitting module, to the circle residing for upper eyelid line in eyes image described in step A4
It is fitted, and to simulating round central point pixel coordinate (X0,Y0) recorded；
Step A6, nictation judge：Central point pixel coordinate (X according to step A50,Y0), this face image is gathered
When testee whether blink and judged：Work as Y0＜ ymWhen, it is judged as that during the collection of this face image, testee is in nictation shape
State, and the blink detection result of this face image is nictation；Otherwise, be judged as this face image collection when testee be in non-
Nictation state, and the blink detection result of this face image is non-nictation；Wherein, ymFor nictation condition adjudgement threshold set in advance
Value and ym=11～14；
When carrying out blink detection using eyes detection method to this face image, process is as follows：
Step B1, ocular Image Acquisition：Call eye location module, to monitored driver in this face image
Left eye and right eye present position are positioned respectively, obtain the ocular image described in two width of monitored driver；
Ocular image described in two width is respectively the image in monitored driver's left eye and region residing for right eye；
Step B2, image normalization are processed：According to the method described in step A2, to eye area described in two width in step B1
Area image is normalized respectively；
Step B3, image binaryzation are processed：According to the method described in step A3, after normalized in step B2
Ocular image described in two width carries out binary conversion treatment respectively, obtains the ocular figure described in two width after binary conversion treatment
Picture；
Step B4, pupil image remove：According to the method described in step A4, eye described in two width in removal step B4 respectively
The image of pupil in portion's image, obtains eyes image described in two width；
Step B5, circle matching：Call round fitting module, to residing for upper eyelid line in eyes image described in two width in step B4
Circle be fitted respectively, and the central point pixel coordinate simulating two circles is recorded respectively, the central point of two circles
Pixel coordinate is denoted as (X respectively1,Y1) and (X2,Y2)；
Step B6, nictation judge：According to the described central point pixel coordinate simulating two circles in step B5, to this face
During image acquisition, whether monitored driver blinks and is judged：Work as Y1＜ ymAnd Y2＜ ymWhen, it is judged as that this face image gathers
When monitored driver be in nictation state, and the blink detection result of this face image be nictation；Otherwise, it is judged as this face
During image acquisition, monitored driver is in non-nictation state, and the blink detection result of this face image is non-nictation；
Step 2012, number of winks statistics：According in step 2011 to this analysis process cycle in described in the N width that received
The blink detection result of face image, counts to the number of winks of monitored driver in this analysis process cycle；
The number of winks of monitored driver and the interior N width institute being received of this analysis process cycle in this analysis process cycle
The total quantity stating the face image that blink detection result in face image is nictation is identical；
Step 202, interior received face image analyzing and processing of next analyzing and processing cycle：According to step 2011 to step
Method described in 2012, is analyzed to received face image in the next analyzing and processing cycle processing, obtains this analysis
The number of winks of monitored driver in process cycle；
Step 203, M-3 repeat step 202, until complete received face image in the front M-1 analyzing and processing cycle
Analyzing and processing process, the number of winks of monitored driver in the M-1 analyzing and processing cycle before acquisition；
Wherein,
Step 204, interior received face image analyzing and processing of next analyzing and processing cycle and fatigue driving judge, process
As follows：
Step 2041, interior received face image analyzing and processing of next analyzing and processing cycle：According to step 2011 to step
Method described in rapid 2012, is analyzed to received face image in the next analyzing and processing cycle processing, obtains one's duty
The number of winks of monitored driver in analysis process cycle；
Step 2042, fatigue driving judge：According to monitored driving in this analysis process cycle drawing in step 2041
In the M-1 analyzing and processing cycle before number of winks and this analysis process cycle of member the number of winks of monitored driver it
And Nz, judge to whether now monitored driver is in fatigue driving state：Work as Nz＞ N0When, it is judged as now being supervised
Survey driver and be in fatigue driving state；Otherwise, it is judged as that monitored driver is in abnormal driving state；
Wherein NzFor number of winks judgment threshold set in advance and Nz=25～30；
Step 205, return to step 204, according to the method described in step 2041 to step 2042, carry out next analysis
In process cycle, received face image analyzing and processing and fatigue driving judge.
A kind of above-mentioned fatigue driving monitoring method based on blink detection, is characterized in that：M described in step A21=80,
N1=30.
A kind of above-mentioned fatigue driving monitoring method based on blink detection, is characterized in that：Y described in step A6m=12.
A kind of above-mentioned fatigue driving monitoring method based on blink detection, is characterized in that：Image acquisition described in step one
Equipment includes obtaining the photographic head of the face image of monitored driver in real time and according to sample frequency f set in advancesTo quilt
The image pick-up card that the face image of monitoring driver is acquired, described photographic head is connected with image pick-up card, described image
Capture card is connected with image processor.
A kind of above-mentioned fatigue driving monitoring method based on blink detection, is characterized in that：Face image described in step one
Resolution be 1024 × 768, the video stream synchronization being gathered is sent to image processor by described image collecting device, described
The frame rate of video flowing is Fsfps.
A kind of above-mentioned fatigue driving monitoring method based on blink detection, is characterized in that：To this analysis place in step 2011
Before the face image described in arbitrary width being received in the reason cycle synchronizes storage, described image processor first adopts face to examine
Survey grader and judge that whether this face image is the face image of monitored driver：When judgement show that this face image is to be supervised
During the face image of survey driver, then this face image is synchronized with storage and blink detection；Otherwise, complete this face image
Image processing process.
A kind of above-mentioned fatigue driving monitoring method based on blink detection, is characterized in that：Judge in step A6 and step B6
When monitored driver is in nictation state when gathering for this face image, monitored driver during the collection of this face image is described
It is in closed-eye state；When being judged as that during the collection of this face image, monitored driver is in non-nictation state, illustrate that this face schemes
As during collection, monitored driver is in eyes-open state；
Described Face datection grader be based on Haar feature and using Adaboost algorithm build grader；To described
Before Face datection grader is built, first obtain training sample set；Described training sample is concentrated and is included two class training samples,
When one class training sample is to be in eyes-open state using the monitored driver that image capture device described in step one collects
Face image, another kind of training sample is at the monitored driver being collected using image capture device described in step one
Face image when closed-eye state.
A kind of above-mentioned fatigue driving monitoring method based on blink detection, is characterized in that：Carry out pupil image in step A4
During removal, by the described ocular image after binary conversion treatment in step A3 and face image or described eye described in step A1
Portion's area image is contrasted, and calls described image to remove module, removes the described ocular image after binary conversion treatment
The image of middle pupil, obtains eyes image；
Circle fitting module described in step A5 justifies fitting module for method of least square.
A kind of above-mentioned fatigue driving monitoring method based on blink detection, is characterized in that：In step A3, calling figure is as two-value
Change processing module, when binary conversion treatment is carried out to the described ocular image after normalized, by described ocular figure
As in each pixel pixel value of background area be all revised as 255, and by driver's eye monitored in described ocular image
Each pixel pixel value in portion is all revised as 0；
When carrying out pupil image removal in step A4, by residing for pupil in the described ocular image after binary conversion treatment
Each pixel pixel value in region is all revised as 255.
A kind of above-mentioned fatigue driving monitoring method based on blink detection, is characterized in that：Carry out pupil image in step A4
During removal, the central point of pupil in the described ocular image after first calling nine grids algorithm to obtain binary conversion treatment, then obtain
Take region residing for pupil in the described ocular image after binary conversion treatment, region residing for acquired pupil is with pupil
Central point is the border circular areas of round dot, and region residing for acquired pupil includes all pixels of monitored driver's pupil
Point；Afterwards, then by the pixel value of pixel each in region residing for acquired pupil all it is revised as 255.
The present invention compared with prior art has advantages below：
1st, method and step is simple, reasonable in design and realizes conveniently, and input cost is relatively low.
2nd, fatigue driving monitoring velocity is fast, Synchronization Analysis can process the fatigue driving state drawing driver.
3rd, the hardware configuration that adopted is simple, only includes at image capture device and the image that is connected with image capture device
Reason device, wherein image capture device are conventional image collecting device, and image processor is conventional image processing equipment, hardware
Input cost is low and installs laying simplicity, only need to be laid in image capture device and drive indoor, easy-to-connect.
4th, adopt eyes image as the decision-making foundation of physiological fatigue, with traditional behavior characteristicss analysis, image procossing skill
Art etc. has very big inherent advantage, proposes a kind of new fatigue driving monitoring side based on physiological driver's response feature
Method.
6th, adopt computer vision technique, by in-car facial information (the i.e. face installing photographic head real-time capture driver
Portion's image) and accordingly judge to draw driver status, the method is non-contact detection, and easy to use, driver need not wear
Equipment, monitoring process is as good as with normal driving situation.
7th, the grader of Adaboost (The Adaptive Boosting Algorithm) Algorithm for Training is adopted to enter pedestrian
Face identifies, recognition speed is fast, and detection accuracy is high.And, carry out again after monitored driver's eyes are positioned justifying matching,
And judge whether to be in nictation state, it is to avoid other regions of face such as nostril, mouth and forehead area above are to eye
The impact of region detection, can significantly enhance accuracy and the detection efficiency of blink detection.
8th, propose the blink detection method based on circle matching, different driver's human eyes can be tracked positioning and detect,
Widely applicable and detection method is easy, reasonable in design, detection speed is fast, detection accuracy is high it is only necessary to Y to fitting circle central point
Axle pixel coordinate carries out threshold value and compares, and just can accurately judge nictation state, thus judging the fatigue state of monitored driver.
9th, the eye pupil image minimizing technology being adopted is simple, reasonable in design and using effect is good, by removing pupil
Image can reduce or eliminate to the noise of ocular and interference, improves the accuracy rate of blink detection.
10th, propose a kind of new blink detection method, being detected, robustness is good to the nictation state of people, to real ring
Border strong adaptability, and blink detection can be carried out to this face image using simple eye detection method or eyes detection method, detection
Mode flexibly, and adapts to the side of face and turns and inclination conditions, improves human eye and blinks the accuracy rate of state-detection and detection
Efficiency.
11st, safe and reliable, can effectively reduce fatigue driving contingency occurrence probability, be in fatigue driving when detecting driver
During state, image processor can be reported to the police to remind driver in time.
12nd, using effect is good and practical value is high, and economic benefit and social benefit are notable, can the easy fatigue to driver
Driving condition carries out real-time monitoring, can effectively prevent the generation of fatigue, reduces the generation of fatigue driving accident.First obtain monitored
The ocular image of driver, then remove pupil image acquisition eyes image, then using method of least square, ocular is entered
Row circle matching, enables the track and localization to different measured's human eyes and detection, can effectively improve nictation state-detection simultaneously
Accuracy rate and efficiency, are reduced or eliminated to the noise of ocular and interference by removing pupil image, improve further
The accuracy rate of blink detection.And, the present invention is easy and simple to handle, be easy to popularization, with low cost, realize easy.
In sum, the inventive method step is simple, reasonable in design and realization convenience, using effect are good, based on driver
The blink detection of face image can carry out accurate measurements by easy, the quick fatigue driving state to driver, and practical value is high.
Below by drawings and Examples, technical scheme is described in further detail.
Brief description
Fig. 1 is method of the present invention FB(flow block).
Fig. 2 carries out method flow block diagram during blink detection for the present invention to face image.
Fig. 3 is the schematic block circuit diagram of the present invention adopted fatigue driving monitoring device.
Fig. 4 is in the ocular image schematic diagram under eyes-open state for monitored driver acquired in the present invention.
Fig. 5 be using the present invention ocular image in Fig. 4 is carried out image normalization process, image binaryzation process,
The structural representation of obtained fitting circle after pupil image removal and circle matching.
Fig. 6 is in the ocular image schematic diagram under closed-eye state for monitored driver acquired in the present invention.
Fig. 7 be using the present invention ocular image in Fig. 6 is carried out image normalization process, image binaryzation process,
The structural representation of obtained fitting circle after pupil image removal and circle matching.
Description of reference numerals:
1 image capture device；1-1 photographic head；2 image pick-up cards；
3 image processors.
Specific embodiment
A kind of fatigue driving monitoring method based on blink detection as shown in Figure 1, comprises the following steps：
Step one, image acquisition：Using image capture device 1 and according to sample frequency f set in advances, to monitored
The face image of driver is acquired, and the face image synchronous driving that each sampling instant is gathered is to image procossing
Device 3；Described image collecting device 1 is connected with image processor 3；
Wherein, fs=FsHz, FsFor positive integer and Fs=25～35；
Step 2, image procossing：Described image processor 3 is according to analyzing and processing frequency f set in advance and according to the time
Sequencing, is analyzed processing to the described face image receiving in each analyzing and processing cycle respectively；Wherein,
N is positive integer and n=5,6,10,12 or 15；
When the described face image receiving in each analyzing and processing cycle is analyzed processing respectively, process is as follows：
Step 201, in first analyzing and processing cycle, received face image analyzes and processes, and comprises the following steps：
Step 2011, face image synchronously storage and blink detection：Described image processor 3 is suitable according to sampling time priority
Sequence, synchronizes storage respectively to the face image described in N width being received in this analysis process cycle, and to each width being received
Described face image carries out blink detection respectively；Wherein, N is an analyzing and processing cycle interior described face image being received
Quantity and N=n × Fs；
To being received in this analysis process cycle when described in arbitrary width, face image carries out blink detection, using simple eye inspection
Survey method or eyes detection method carry out blink detection to this face image；
Wherein, as shown in Fig. 2 when carrying out blink detection using simple eye detection method to this face image, process is as follows：
Step A1, ocular Image Acquisition：Call eye location module, to monitored driver in this face image
Left eye or right eye present position are positioned, and obtain the ocular image of monitored driver；
Described ocular image is monitored driver's left eye or the image in region residing for right eye；
Step A2, image normalization are processed：Calling figure as normalized module, to ocular figure described in step A1
As being normalized；
In this step, the size of the described ocular image after normalized is M1×N1Individual pixel, wherein M1For
Positive integer and M1=75～85, N1For positive integer and N1=28～32；
Step A3, image binaryzation are processed：Calling figure as binary conversion treatment module, after normalized in step A2
Described ocular image carries out binary conversion treatment；
Step A4, pupil image remove：Calling figure picture removes module, the described eye after binary conversion treatment in removal step A3
The image of pupil in portion's area image, obtains eyes image；
Step A5, circle matching：Call round fitting module, to the circle residing for upper eyelid line in eyes image described in step A4
It is fitted, and to simulating round central point pixel coordinate (X0,Y0) recorded；
Step A6, nictation judge：Central point pixel coordinate (X according to step A50,Y0), this face image is gathered
When testee whether blink and judged：Work as Y0＜ ymWhen, it is judged as that during the collection of this face image, testee is in nictation shape
State, and the blink detection result of this face image is nictation；Otherwise, be judged as this face image collection when testee be in non-
Nictation state, and the blink detection result of this face image is non-nictation；Wherein, ymFor nictation condition adjudgement threshold set in advance
Value and ym=11～14；
When carrying out blink detection using eyes detection method to this face image, process is as follows：
Step B1, ocular Image Acquisition：Call eye location module, to monitored driver in this face image
Left eye and right eye present position are positioned respectively, obtain the ocular image described in two width of monitored driver；
Ocular image described in two width is respectively the image in monitored driver's left eye and region residing for right eye；
Step B2, image normalization are processed：According to the method described in step A2, to eye area described in two width in step B1
Area image is normalized respectively；
Step B3, image binaryzation are processed：According to the method described in step A3, after normalized in step B2
Ocular image described in two width carries out binary conversion treatment respectively, obtains the ocular figure described in two width after binary conversion treatment
Picture；
Step B4, pupil image remove：According to the method described in step A4, eye described in two width in removal step B4 respectively
The image of pupil in portion's image, obtains eyes image described in two width；
Step B5, circle matching：Call round fitting module, to residing for upper eyelid line in eyes image described in two width in step B4
Circle be fitted respectively, and the central point pixel coordinate simulating two circles is recorded respectively, the central point of two circles
Pixel coordinate is denoted as (X respectively1,Y1) and (X2,Y2)；
Step B6, nictation judge：According to the described central point pixel coordinate simulating two circles in step B5, to this face
During image acquisition, whether monitored driver blinks and is judged：Work as Y1＜ ymAnd Y2＜ ymWhen, it is judged as that this face image gathers
When monitored driver be in nictation state, and the blink detection result of this face image be nictation；Otherwise, it is judged as this face
During image acquisition, monitored driver is in non-nictation state, and the blink detection result of this face image is non-nictation；
Step 2012, number of winks statistics：According in step 2011 to this analysis process cycle in described in the N width that received
The blink detection result of face image, counts to the number of winks of monitored driver in this analysis process cycle；
The number of winks of monitored driver and the interior N width institute being received of this analysis process cycle in this analysis process cycle
The total quantity stating the face image that blink detection result in face image is nictation is identical；
Step 202, interior received face image analyzing and processing of next analyzing and processing cycle：According to step 2011 to step
Method described in 2012, is analyzed to received face image in the next analyzing and processing cycle processing, obtains this analysis
The number of winks of monitored driver in process cycle；
Step 203, M-3 repeat step 202, until complete received face image in the front M-1 analyzing and processing cycle
Analyzing and processing process, the number of winks of monitored driver in the M-1 analyzing and processing cycle before acquisition；
Wherein,
Step 204, interior received face image analyzing and processing of next analyzing and processing cycle and fatigue driving judge, process
As follows：
Step 2041, interior received face image analyzing and processing of next analyzing and processing cycle：According to step 2011 to step
Method described in rapid 2012, is analyzed to received face image in the next analyzing and processing cycle processing, obtains one's duty
The number of winks of monitored driver in analysis process cycle；
Step 2042, fatigue driving judge：According to monitored driving in this analysis process cycle drawing in step 2041
In the M-1 analyzing and processing cycle before number of winks and this analysis process cycle of member the number of winks of monitored driver it
And Nz, judge to whether now monitored driver is in fatigue driving state：Work as Nz＞ N0When, it is judged as now being supervised
Survey driver and be in fatigue driving state；Otherwise, it is judged as that monitored driver is in abnormal driving state；
Wherein NzFor number of winks judgment threshold set in advance and Nz=25～30；
Step 205, return to step 204, according to the method described in step 2041 to step 2042, carry out next analysis
In process cycle, received face image analyzing and processing and fatigue driving judge.
In the present embodiment, the F described in step ones=30.
When actually used, can according to specific needs, to F described in step onesValue size adjust accordingly.
In step 2, a described analyzing and processing cycle is the n second.
N described in step 2042zNumber of winks sum for monitored driver in the continuous N analyzing and processing cycle.
Wherein,
In the present embodiment, the n=6 described in step 2, that is, a described analyzing and processing cycle is 6s.Institute in step 2011
N=n × the F statings=6 × 30=180.When actually used, can according to specific needs, value size to n described in step 2
Adjust accordingly.
Described in step 203Thus, after M-3 repeat step 202 in step 203, obtain first 9 points
The number of winks of monitored driver in analysis process cycle.
Number of winks according to monitored driver in this analysis process cycle and this analysis process cycle in step 2042
Number of winks sum N of monitored driver in 9 analyzing and processing cycles beforez, whether now monitored driver is located
Judged in fatigue driving state, " now " described herein is this analysis process cycle finish time.And, described Nz
Number of winks sum for monitored driver in continuous 10 analyzing and processing cycles.
In the present embodiment, the N described in step 204z=28.When actually used, can according to specific needs, to NzValue
Size adjusts accordingly.
In the present embodiment, the resolution of face image described in step one is 1024 × 768.
When actually used, the resolution of face image described in step one is carried out according to the resolution of image capture device 1
Determine.
In the present embodiment, the M described in step A21=80, N1=30.
When actually used, can according to specific needs, to the M described in step A21And N1Value size carry out respectively accordingly
Adjustment.
In the present embodiment, the y described in step A6m=12.
When actually used, can according to specific needs, to the y described in step A6mValue size accordingly adjusted respectively
Whole.
In the present embodiment, the video stream synchronization being gathered is sent to image processor 3, institute by described image collecting device 1
The frame rate stating video flowing is Fsfps.
And, described video flowing is the video flowing being made up of face image several described.In the present embodiment, described video flowing
Frame rate be 30fps.
In the present embodiment, image capture device 1 described in step one includes the face's figure obtaining monitored driver in real time
The photographic head 1-1 of picture and according to sample frequency f set in advancesThe image that the face image of monitored driver is acquired
Capture card 2, described photographic head 1-1 is connected with image pick-up card 2, and described image capture card 2 is connected with image processor 3.
When actually used, described image collecting device 1 forms fatigue driving monitoring device with image processor 3, refers to figure
3.
In the present embodiment, described photographic head 1-1 is USB camera.
And, described USB camera is conventional USB camera.When actually used, described photographic head 1-1 can also be
Other kinds of regular camera.
In the present embodiment, described image processor 3 is PC.
When actually used, described image processor 3 can also be at palm PC, smart mobile phone or other types of data
Reason device.
In the present embodiment, when carrying out image normalization process in step A2, the normalization processing method being adopted is routine
Image normalization method.
And, when carrying out eye location in step B1, the localization method of employing in the localization method being adopted and step A1
Identical.
In the present embodiment, in step A6 and step B6, it is judged as that during the collection of this face image, monitored driver is in nictation
During state, described image processor 3 also needs calling figure, as mark module, this face image is labeled as image of blinking；And, sentence
When breaking that when gathering for this face image, monitored driver is in nictation state, described image processor 3 also needs calling figure picture mark
This face image is labeled as non-nictation image by note module.
In the present embodiment, in step 2011, the face image described in arbitrary width being received in this analysis process cycle is carried out
Before synchronous storage, using Face datection grader, described image processor 3 first judges whether this face image is monitored driving
The face image of the person of sailing：When judging to draw the face image that this face image is monitored driver, then to this face image
Synchronize storage and blink detection；Otherwise, complete the image processing process of this face image.
And, the Face datection grader being adopted is the grader of conventional Face datection or recognition of face.
In the present embodiment, in step A6 and step B6, it is judged as that during the collection of this face image, monitored driver is in nictation
During state, illustrate that during the collection of this face image, monitored driver is in closed-eye state；It is judged as quilt during the collection of this face image
When monitoring driver is in non-nictation state, illustrate that during the collection of this face image, monitored driver is in eyes-open state.
In the present embodiment, described Face datection grader is based on Haar feature and using dividing that Adaboost algorithm builds
Class device；Before described Face datection grader is built, first obtain training sample set；Described training sample is concentrated and is included two
Class training sample, a class training sample is at the monitored driver being collected using image capture device 1 described in step one
Face image when eyes-open state, another kind of training sample is collected using image capture device 1 described in step one
Monitored driver is in face image during closed-eye state.
Wherein, described Adaboost algorithm is conventional Adaboost algorithm.The extracting method of Haar feature is routine
Haar feature extracting method, Haar feature be also referred to as Harr feature or Haar-like feature.
Described Adaboost algorithm is a kind of self adaptation boosting (ballot weight) iterative algorithm.Its core concept is
Train the general grader (i.e. Weak Classifier) of different classifications ability for same training sample set, then by these weak typings
Device superposition gets up, and constitutes a higher strong classifier.In training and detection, each strong classifier is to Haar to be detected
Feature (also referred to as rectangular characteristic) makes decisions, and these strong classifiers is cascaded up and just can generate one accurately, quickly
Grader.Its feature is exactly that detection speed is fast, because each strong classifier can veto rectangular characteristic to be detected, institute
Just the feature discharge of most of mistake can be removed with strong classifier, finally give the higher classification results of accuracy rate.
In the present embodiment, described training sample is concentrated and is included n0Individual training sample, n0I-th training sample in individual training sample
Originally it is denoted as (xi,yi), wherein i is positive integer and i=1,2 ..., n0, xiFor extracting face image in i-th training sample
Haar feature, yiOutput for Face datection grader and yi=-1 or 1, wherein yi=-1 represents that i-th training sample is negative
In sample and i-th training sample, face image is not real face image, yi=1 represents that i-th training sample is negative sample
In basis and i-th training sample, face image is real face image.
During the grader being built using Adaboost algorithm, under initial situation, described training sample concentrates each training sample
Weight be all identical, and the weight of i-th training sampleIt is the initial probability distribution of training sample.First
During secondary iteration, using the training sample training Weak Classifier of current probability distribution, and calculate error rates of weak classifiers, select simultaneously
Appropriate threshold is taken to make error minimum, then the weight updating training sample；So, after multiple circulation, multiple Weak Classifiers are obtained,
By the weighted superposition updating, the strong classifier finally giving.
It is demonstrated experimentally that can effectively be detected based on the Face datection grader that the Adaboost algorithm of Haar feature trains
Face and ocular are simultaneously effectively positioned to eye pupil position.When actually used, detection speed and accuracy rate are much high
In other algorithms, take between several milliseconds to a few tens of milliseconds, reach real-time detection and require.
In the present embodiment, Application No. 200810030010.4 and denomination of invention disclosed in 14 days January in 2009 can be adopted
For《Sight tracing and the disabled assisting system of application the method》Application for a patent for invention file disclosed in sight line with
Track method carries out Face datection and eye location, and the Face datection grader accordingly being adopted is in this application for a patent for invention file
Disclosed grader.And, also can adopt Application No. 201110200287.9 disclosed in 23 days November in 2011 and invention name
Referred to as《A kind of method for detecting human face based on AdaBoost algorithm》Application for a patent for invention file disclosed in Face datection side
Method carries out Face datection (also referred to as recognition of face or Face datection classification), and the Face datection grader accordingly being adopted is this
Grader disclosed in bright patent application document.
Meanwhile, can be using Application No. 201310244614.X disclosed in September in 2013 4 days and invention entitled《A kind of
Based on the eye locating method improving Adaboost algorithm and Face geometric eigenvector》Application for a patent for invention file disclosed in people
Face classification carries out Face datection and eye location with eye locating method, and the Face datection grader accordingly being adopted is this invention
Face classification device disclosed in patent application document.Thus, call eye location module in step A1 to quilt in this face image
When the left eye of monitoring driver or right eye present position are positioned, fixed using the eyes disclosed in this application for a patent for invention file
Method for position is positioned.
When actually used, call eye location module to the left eye of monitored driver in this face image in step A1 or
When right eye present position is positioned, it would however also be possible to employ other conventional eye locating methods.
In the present embodiment, in step A3, binary conversion treatment mistake is carried out to the described ocular image after normalized
Journey, also referred to as gray processing processing procedure.
In the present embodiment, in step A3 calling figure as binary conversion treatment module, to the described eye area after normalized
When area image carries out binary conversion treatment, each pixel pixel value of background area in described ocular image is all revised as
255, and each pixel pixel value of monitored driver eye in described ocular image is all revised as 0.
In the present embodiment, the planar rectangular coordinate system on eyes image described in step A4 is with described eyes image
The upper left corner (i.e. the summit in the upper left corner) is zero, length direction (the i.e. quilt in described eyes image with described eyes image
The lateral length direction of monitoring driver eye) it is X-axis, the width with described eyes image is (i.e. in described eyes image
The vertical height direction of monitored driver eye) it is Y-axis.
Wherein, described eyes image is X-axis positive direction from left to right, on described eyes image from top to bottom for Y-axis just
Direction.
In the present embodiment, circle fitting module described in step A5 justifies fitting module for method of least square,
And, when carrying out in step A5 justifying matching, the method for least square using routine carries out justifying matching.
In the present embodiment, when the circle residing for upper eyelid line in described eyes image being fitted in step A5, first call
Edge detection module carries out rim detection to described eyes image, specifically carries out side to the upper eyelid line in described eyes image
Edge detects, and detects multiple marginal points, the pixel coordinate of detected each marginal point is recorded simultaneously.Detect is many
Individual marginal point forms marginal point set, and edge point set is denoted as { (x'1,y'1),(x'2,y'2),...,(x'i',y'i'),...,
(x'm',y'm'), wherein (x'i',y'i') for the i-th ' individual marginal point in edge point set pixel coordinate, i' is positive integer and i'
=1,2 ..., m', m' be edge point set in included marginal point total quantity.
(the X of central point pixel coordinate described in step A50,Y0) in, X0For simulating the X-axis pixel coordinate of round central point
(simulating the X-axis coordinate of round central point), Y0Y-axis pixel coordinate for simulating round central point (simulates round
The Y-axis coordinate of central point).Wherein, simulating round central point is to simulate the round center of circle, and simulates round central point and be
One of described eyes image pixel.
Correspondingly, (the X of central point pixel coordinate described in step B51,Y1) in, X1The X of one for simulating round central point
Axle pixel coordinate (i.e. the X-axis coordinate of this circle central point), Y1For simulating the Y-axis pixel coordinate of this circle central point (i.e. in this circle
The Y-axis coordinate of heart point).Described central point pixel coordinate (X2,Y2) in, X2The X-axis pixel of another circle central point for simulating
Coordinate (i.e. the X-axis coordinate of this circle central point), Y2For simulating Y-axis pixel coordinate (the i.e. Y of this circle central point of this circle central point
Axial coordinate).
In the present embodiment, in step A5, call round fitting module that the circle residing for upper eyelid line in described eyes image is carried out
During matching, using conventional method of least square, the multiple marginal points in described edge point set are carried out justifying matching.And, to described
When multiple marginal points in edge point set carry out justifying matching, error of fitting
Try to achieve the minima of S, find out best-fit-circle, wherein r is the radius of fitting circle.
In the present embodiment, when carrying out pupil image removal in step A4, by the described eye after binary conversion treatment in step A3
Portion's area image is contrasted with face image described in step A1 or described ocular image, and calls described image to remove
Module, removes the image of pupil in the described ocular image after binary conversion treatment, obtains eyes image.
In the present embodiment, when carrying out pupil image removal in step A4, by the described ocular figure after binary conversion treatment
In picture, each pixel pixel value in region residing for pupil is all revised as 255.
For easy and simple to handle, when carrying out pupil image in step A4 and removing, nine grids algorithm is first called to obtain binary conversion treatment
The central point of pupil in described ocular image afterwards, then obtain pupil in the described ocular image after binary conversion treatment
Residing region, region residing for acquired pupil is the border circular areas with the central point of pupil as round dot, and acquired pupil
Residing region includes all pixels point of monitored driver's pupil；Afterwards, then by picture each in region residing for acquired pupil
The pixel value of vegetarian refreshments is all revised as 255.
In the present embodiment, the present invention is in the ocular figure under eyes-open state to monitored driver as shown in Figure 4
As carrying out circle (the i.e. matching after image normalization process, image binaryzation process, pupil image removal and circle matching, simulating
Circle) refer to Fig. 5；The ocular image monitored driver as shown in Figure 6 being under closed-eye state carries out image normalizing
After change process, image binaryzation process, pupil image removal and circle matching, the circle (i.e. fitting circle) simulating refers to Fig. 7.
For verifying simplicity and the accuracy of blink detection of the present invention, monitored driver is under eyes-open state and closes one's eyes
Several ocular images under state carry out respectively image normalization process, image binaryzation process, pupil image remove and
After circle matching, draw the central point pixel coordinate of multiple fitting circles, refer to table 1：
In table 1, the fitting circle of serial number 0 to 7 is the ocular image monitored driver being under eyes-open state
The fitting circle obtaining after carrying out image normalization process, image binaryzation process, pupil image removal and circle matching, serial number 8
Fitting circle to 16 is that the ocular image that monitored driver is under closed-eye state carries out image normalization process, figure
As the fitting circle obtaining after binary conversion treatment, pupil image removal and circle matching, can directly, clearly be drawn by table 1：In fitting circle
The Y-axis pixel coordinate of heart point also carries out bright to the ocular image under the ocular image under eyes-open state and closed-eye state
Really distinguish, thus using the present invention can easy to face image, fast and accurately carry out blink detection.
The above, be only presently preferred embodiments of the present invention, not the present invention imposed any restrictions, every according to the present invention
Any simple modification, change and equivalent structure change that technical spirit is made to above example, all still fall within skill of the present invention
In the protection domain of art scheme.
Claims (10)
1. a kind of fatigue driving monitoring method based on blink detection is it is characterised in that the method comprises the following steps：
Step one, image acquisition：Using image capture device (1) and according to sample frequency f set in advances, to monitored driving
The face image of member is acquired, and the face image synchronous driving that each sampling instant is gathered is to image processor
(3)；Described image collecting device (1) is connected with image processor (3)；
Wherein, fs=FsHz, FsFor positive integer and Fs=25～35；
Step 2, image procossing：Described image processor (3) is according to analyzing and processing frequency f set in advance and first according to the time
Order afterwards, is analyzed processing to the described face image receiving in each analyzing and processing cycle respectively；Wherein,n
For positive integer and n=5,6,10,12 or 15；
When the described face image receiving in each analyzing and processing cycle is analyzed processing respectively, process is as follows：
Step 201, in first analyzing and processing cycle, received face image analyzes and processes, and comprises the following steps：
Step 2011, face image synchronously storage and blink detection：Described image processor (3) is suitable according to sampling time priority
Sequence, synchronizes storage respectively to the face image described in N width being received in this analysis process cycle, and to each width being received
Described face image carries out blink detection respectively；Wherein, N is an analyzing and processing cycle interior described face image being received
Quantity and N=n × Fs；
To being received in this analysis process cycle when described in arbitrary width, face image carries out blink detection, using simple eye detection side
Method or eyes detection method carry out blink detection to this face image；
Wherein, when carrying out blink detection using simple eye detection method to this face image, process is as follows：
Step A1, ocular Image Acquisition：Call eye location module, the left eye to monitored driver in this face image
Or right eye present position is positioned, obtain the ocular image of monitored driver；
Described ocular image is monitored driver's left eye or the image in region residing for right eye；
Step A2, image normalization are processed：Calling figure, as normalized module, is entered to ocular image described in step A1
Row normalized；
In this step, the size of the described ocular image after normalized is M1×N1Individual pixel, wherein M1For just whole
Number and M1=75～85, N1For positive integer and N1=28～32；
Step A3, image binaryzation are processed：Calling figure as binary conversion treatment module, described in after normalized in step A2
Ocular image carries out binary conversion treatment；
Step A4, pupil image remove：Calling figure picture removes module, the described eye area after binary conversion treatment in removal step A3
The image of pupil in area image, obtains eyes image；
Step A5, circle matching：Call round fitting module, the circle residing for upper eyelid line in eyes image described in step A4 is carried out
Matching, and to simulating round central point pixel coordinate (X0,Y0) recorded；
Step A6, nictation judge：Central point pixel coordinate (X according to step A50,Y0), quilt when this face image is gathered
Whether tester blinks is judged：Work as Y0＜ ymWhen, it is judged as that during the collection of this face image, testee is in nictation state,
And the blink detection result of this face image is nictation；Otherwise, it is judged as that during the collection of this face image, testee is in non-blinking
Eye state, and the blink detection result of this face image is non-nictation；Wherein, ymFor nictation condition adjudgement threshold value set in advance
And ym=11～14；
When carrying out blink detection using eyes detection method to this face image, process is as follows：
Step B1, ocular Image Acquisition：Call eye location module, the left eye to monitored driver in this face image
Positioned respectively with right eye present position, obtained the ocular image described in two width of monitored driver；
Ocular image described in two width is respectively the image in monitored driver's left eye and region residing for right eye；
Step B2, image normalization are processed：According to the method described in step A2, to ocular figure described in two width in step B1
As being normalized respectively；
Step B3, image binaryzation are processed：According to the method described in step A3, to two width after normalized in step B2
Described ocular image carries out binary conversion treatment respectively, obtains the ocular image described in two width after binary conversion treatment；
Step B4, pupil image remove：According to the method described in step A4, eye figure described in two width in removal step B4 respectively
The image of pupil in picture, obtains eyes image described in two width；
Step B5, circle matching：Call round fitting module, to the circle residing for upper eyelid line in eyes image described in two width in step B4
It is fitted respectively, and the central point pixel coordinate simulating two circles is recorded respectively, the central point pixel of two circles
Coordinate is denoted as (X respectively1,Y1) and (X2,Y2)；
Step B6, nictation judge：According to the described central point pixel coordinate simulating two circles in step B5, to this face image
During collection, whether monitored driver blinks and is judged：Work as Y1＜ ymAnd Y2＜ ymWhen, it is judged as quilt during the collection of this face image
Monitoring driver is in nictation state, and the blink detection result of this face image is nictation；Otherwise, it is judged as this face image
During collection, monitored driver is in non-nictation state, and the blink detection result of this face image is non-nictation；
Step 2012, number of winks statistics：According in step 2011 to this analysis process cycle in the face described in N width that received
The blink detection result of image, counts to the number of winks of monitored driver in this analysis process cycle；
The number of winks of monitored driver and the interior face described in N width being received of this analysis process cycle in this analysis process cycle
In portion's image blink detection result be nictation face image total quantity identical；
Step 202, interior received face image analyzing and processing of next analyzing and processing cycle：According to step 2011 to step 2012
Described in method, received face image in the next analyzing and processing cycle is analyzed process, obtain this analyzing and processing
The number of winks of monitored driver in cycle；
Step 203, M-3 repeat step 202, until complete dividing of interior received face image of front M-1 analyzing and processing cycle
Analysis processing procedure, the number of winks of monitored driver in the M-1 analyzing and processing cycle before acquisition；
Wherein,
Step 204, received face image analyzing and processing and fatigue driving judgement in the next analyzing and processing cycle, process is such as
Under：
Step 2041, interior received face image analyzing and processing of next analyzing and processing cycle：According to step 2011 to step
Method described in 2012, is analyzed to received face image in the next analyzing and processing cycle processing, obtains this analysis
The number of winks of monitored driver in process cycle；
Step 2042, fatigue driving judge：According to monitored driver in this analysis process cycle drawing in step 2041
The number of winks sum of monitored driver in the M-1 analyzing and processing cycle before number of winks and this analysis process cycle
Nz, judge to whether now monitored driver is in fatigue driving state：Work as Nz＞ N0When, it is judged as now monitored
Driver is in fatigue driving state；Otherwise, it is judged as that monitored driver is in abnormal driving state；
Wherein NzFor number of winks judgment threshold set in advance and Nz=25～30；
Step 205, return to step 204, according to the method described in step 2041 to step 2042, carry out next analyzing and processing
In cycle, received face image analyzing and processing and fatigue driving judge.
2. according to a kind of fatigue driving monitoring method based on blink detection described in claim 1 it is characterised in that：Step A2
Described in M1=80, N1=30.
3. according to a kind of fatigue driving monitoring method based on blink detection described in claim 1 or 2 it is characterised in that：Step
Y described in rapid A6m=12.
4. according to a kind of fatigue driving monitoring method based on blink detection described in claim 1 or 2 it is characterised in that：Step
Image capture device (1) described in rapid one includes obtaining the photographic head (1-1) of face image of monitored driver in real time and presses
According to sample frequency f set in advancesThe image pick-up card (2) that the face image of monitored driver is acquired, described takes the photograph
As head (1-1) is connected with image pick-up card (2), described image capture card (2) is connected with image processor (3).
5. according to a kind of fatigue driving monitoring method based on blink detection described in claim 1 or 2 it is characterised in that：Step
The resolution of face image described in rapid one is 1024 × 768, and described image collecting device (1) is by the video stream synchronization being gathered
It is sent to image processor (3), the frame rate of described video flowing is Fsfps.
6. according to a kind of fatigue driving monitoring method based on blink detection described in claim 1 or 2 it is characterised in that：Step
Before storage being synchronized to the face image described in arbitrary width being received in this analysis process cycle in rapid 2011, described image
Using Face datection grader, processor (3) first judges that whether this face image is the face image of monitored driver：When sentencing
Disconnected when drawing the face image that this face image is monitored driver, then this face image is synchronized with storage and nictation inspection
Survey；Otherwise, complete the image processing process of this face image.
7. according to a kind of fatigue driving monitoring method based on blink detection described in claim 6 it is characterised in that：Step A6
During with being judged as in step B6 that during the collection of this face image, monitored driver is in nictation state, illustrate that this face image gathers
When monitored driver be in closed-eye state；It is judged as that during the collection of this face image, monitored driver is in non-nictation state
When, illustrate that during the collection of this face image, monitored driver is in eyes-open state；
Described Face datection grader be based on Haar feature and using Adaboost algorithm build grader；To described face
Before detection grader is built, first obtain training sample set；Described training sample is concentrated and is included two class training samples, a class
Training sample is to be in during eyes-open state using the monitored driver that image capture device described in step one (1) collects
Face image, another kind of training sample is the monitored driver being collected using image capture device described in step one (1)
It is in face image during closed-eye state.
8. according to a kind of fatigue driving monitoring method based on blink detection described in claim 1 or 2 it is characterised in that：Step
When carrying out pupil image removal in rapid A4, by the described ocular image after binary conversion treatment in step A3 and institute in step A1
State face image or described ocular image is contrasted, and call described image to remove module, after removing binary conversion treatment
Described ocular image in pupil image, obtain eyes image；
Circle fitting module described in step A5 justifies fitting module for method of least square.
9. according to a kind of fatigue driving monitoring method based on blink detection described in claim 1 or 2 it is characterised in that：Step
In rapid A3, calling figure, as binary conversion treatment module, carries out binary conversion treatment to the described ocular image after normalized
When, each pixel pixel value of background area in described ocular image is all revised as 255, and by described ocular figure
In picture, each pixel pixel value of monitored driver eye is all revised as 0；
When carrying out pupil image removal in step A4, by region residing for pupil in the described ocular image after binary conversion treatment
Each pixel pixel value be all revised as 255.
10. according to a kind of fatigue driving monitoring method based on blink detection described in claim 9 it is characterised in that：Step
When carrying out pupil image removal in A4, pupil in the described ocular image after first calling nine grids algorithm to obtain binary conversion treatment
The central point in hole, then obtain region residing for pupil in the described ocular image after binary conversion treatment, acquired pupil institute
Place region is the border circular areas with the central point of pupil as round dot, and region residing for acquired pupil includes monitored driving
The all pixels point of member's pupil；Afterwards, then by the pixel value of pixel each in region residing for acquired pupil all it is revised as 255.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610835397.5A CN106384096B (en) | 2016-09-20 | 2016-09-20 | A kind of fatigue driving monitoring method based on blink detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610835397.5A CN106384096B (en) | 2016-09-20 | 2016-09-20 | A kind of fatigue driving monitoring method based on blink detection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106384096A true CN106384096A (en) | 2017-02-08 |
| CN106384096B CN106384096B (en) | 2018-07-10 |
Family
ID=57936722
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610835397.5A Expired - Fee Related CN106384096B (en) | 2016-09-20 | 2016-09-20 | A kind of fatigue driving monitoring method based on blink detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106384096B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108245171A (en) * | 2017-12-28 | 2018-07-06 | 东软集团股份有限公司 | Get parms method, fatigue detection method and device, medium and the equipment of model |
| CN109446999A (en) * | 2018-10-31 | 2019-03-08 | 中电科新型智慧城市研究院有限公司 | Quick sensing system and method based on the dynamic human body movement that statistics calculates |
Citations (7)
| 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 |
| CN1830389A (en) * | 2006-04-21 | 2006-09-13 | 太原理工大学 | Device for monitoring fatigue driving state and its method |
| CN101059837A (en) * | 2007-06-07 | 2007-10-24 | 上海交通大学 | Colorful contact lens false-proof detection method |
| CN101344919A (en) * | 2008-08-05 | 2009-01-14 | 华南理工大学 | Sight tracing method and disabled assisting system using the same |
| CN101732055A (en) * | 2009-02-11 | 2010-06-16 | 北京智安邦科技有限公司 | Method and system for testing fatigue of driver |
| CN103279752A (en) * | 2013-06-19 | 2013-09-04 | 山东大学 | Eye locating method based on improved Adaboost algorithm and human face geometrical characteristics |
| CN103400471A (en) * | 2013-08-12 | 2013-11-20 | 电子科技大学 | Detecting system and detecting method for fatigue driving of driver |
-
2016
- 2016-09-20 CN CN201610835397.5A patent/CN106384096B/en not_active Expired - Fee Related
Patent Citations (7)
| 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 |
| CN1830389A (en) * | 2006-04-21 | 2006-09-13 | 太原理工大学 | Device for monitoring fatigue driving state and its method |
| CN101059837A (en) * | 2007-06-07 | 2007-10-24 | 上海交通大学 | Colorful contact lens false-proof detection method |
| CN101344919A (en) * | 2008-08-05 | 2009-01-14 | 华南理工大学 | Sight tracing method and disabled assisting system using the same |
| CN101732055A (en) * | 2009-02-11 | 2010-06-16 | 北京智安邦科技有限公司 | Method and system for testing fatigue of driver |
| CN103279752A (en) * | 2013-06-19 | 2013-09-04 | 山东大学 | Eye locating method based on improved Adaboost algorithm and human face geometrical characteristics |
| CN103400471A (en) * | 2013-08-12 | 2013-11-20 | 电子科技大学 | Detecting system and detecting method for fatigue driving of driver |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108245171A (en) * | 2017-12-28 | 2018-07-06 | 东软集团股份有限公司 | Get parms method, fatigue detection method and device, medium and the equipment of model |
| CN109446999A (en) * | 2018-10-31 | 2019-03-08 | 中电科新型智慧城市研究院有限公司 | Quick sensing system and method based on the dynamic human body movement that statistics calculates |
| CN109446999B (en) * | 2018-10-31 | 2021-08-31 | 中电科新型智慧城市研究院有限公司 | Rapid sensing system and method for dynamic human body movement based on statistical calculation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106384096B (en) | 2018-07-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109993093B (en) | Road rage monitoring method, system, equipment and medium based on facial and respiratory characteristics | |
| CN103824420B (en) | Fatigue driving identification system based on heart rate variability non-contact measurement | |
| CN101593425B (en) | Machine vision based fatigue driving monitoring method and system | |
| CN104200192B (en) | Driver's gaze detection system | |
| CN109389806B (en) | Fatigue driving detection early warning method, system and medium based on multi-information fusion | |
| CN106846734B (en) | A kind of fatigue driving detection device and method | |
| CN106446811A (en) | Deep-learning-based driver's fatigue detection method and apparatus | |
| CN104013414A (en) | Driver fatigue detecting system based on smart mobile phone | |
| CN109460703B (en) | Non-invasive fatigue driving identification method based on heart rate and facial features | |
| CN104461007B (en) | A kind of driver assistance people's car mutual system based on EEG signals | |
| CN110811649A (en) | Fatigue driving detection method based on bioelectricity and behavior characteristic fusion | |
| CN105286802B (en) | Driver Fatigue Detection based on video information | |
| Li et al. | Yawning detection for monitoring driver fatigue based on two cameras | |
| CN112434611B (en) | Early fatigue detection method and system based on eye movement subtle features | |
| CN107595307A (en) | Fatigue driving detection device and detection method based on machine vision eye recognition | |
| CN106250801A (en) | Based on Face datection and the fatigue detection method of human eye state identification | |
| CN110751051A (en) | Abnormal driving behavior detection method based on machine vision | |
| CN102054163A (en) | Method for testing driver fatigue based on monocular vision | |
| CN109953763A (en) | A kind of vehicle carried driving behavioral value early warning system and method based on deep learning | |
| CN108596087B (en) | Driving fatigue degree detection regression model based on double-network result | |
| CN107669283A (en) | Spectacle driving fatigue monitoring warning device based on multi-feature fusion and method | |
| CN111753674A (en) | Fatigue driving detection and identification method based on deep learning | |
| CN106919913A (en) | Method for detecting fatigue driving and device based on computer vision | |
| CN109977930A (en) | Method for detecting fatigue driving and device | |
| CN104068868A (en) | Method and device for monitoring driver fatigue on basis of machine vision |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180710 Termination date: 20190920 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |