CN103956028B - The polynary driving safety means of defence of a kind of automobile - Google Patents

Abstract

The invention provides the polynary driving safety means of defence of a kind of automobile and specialized equipment thereof.The method is by gathering facial information and the pulse information of driver, use the empirical mode decomposition scheduling algorithm based on information entropy, extract in facial information and pulse information the characteristic quantity that effectively can reflect human body indignation mood, fatigue state and identity information, and by intelligent algorithm, all kinds of characteristic parameter is merged, to realize the accurate identification to driver's indignation mood, fatigue state and identity information.This specialized equipment realizes based on the technology such as DSP and Bluetooth wireless communication, when device detects that the states such as abnormal feeling, driving fatigue or identity change appear in driver, adopt the driving behavior of mode to driver such as alarm and backstage record to intervene respectively, and then protect from the driving safety of multiple angle to driver.

Description

The polynary driving safety means of defence of a kind of automobile

Technical field

The present invention relates to the polynary driving safety means of defence of a kind of automobile, particularly relate to a kind of simultaneously can monitoring the angry mood of driver, fatigue state and identity information, and then implement the polynary driving safety means of defence of automobile of prompting and safeguard procedures.

Background technology

Along with the fast development with Modern Traffic forwarding business that improves constantly of people's living standard, transport need increases day by day.The vehicles, particularly on-road vehicle are increasing, cause frequent accidents to occur.Statistical data shows, in China, toll on traffic accounts for more than 3/4 of all kinds of particularly serious security incident death toll, and traffic hazard has become " the first killer " in various accident.

In whole traffic system, transport driver, as taxi driver, bus driver, long-distance passenger transportation driver and teamster etc., plays very important role, has important impact to the safety of whole traffic system.The specification of driver drives the important leverage of the transportation system safety being its place of guarantee, but in real operation, but there are following three kinds of nonstandard driving phenomenons:

The first, " road anger " phenomenon ubiquity.In vehicle travel process, the mood of driver can be subject to the impact of driving environment.Road anger disease refers to the indignation that driver causes because of dysthymia in driving procedure, even aggressive behavior.Likely there is road anger disease in all kinds of driver, particularly in urban highway traffic, there is the bad steering customs such as random intertrack crosstalk because traffic congestion, new hand or Apprentice 0ffice drive in unskilled, part driver startup procedure, other drivers are caused to occur unhealthy emotion, even bring out road anger disease, so cause fighting for signal, scramble for roads joyride or the dangerous traffic behavior such as to blow a whistle loudly.Road anger disease had both been unfavorable for the social environment of creating policy, also can become the potential risk of traffic safety.

The second, fatigue driving phenomenon often has generation.For coach, the report of coach generation traffic hazard is caused again and again to occur because of fatigue driving.Once there is traffic hazard, not only driver in coach, in car, the safety of life and property of each passenger all can be subject to serious threat, brings danger also can to other vehicles simultaneously, cause great casualties and property loss.In taxi and long-haul over-the-highway truck, driver is in order to pursue larger economic interests, and frequent long-time continuous is driven, and finally causes, because of driving fatigue, traffic hazard occurs.Fatigue driving is as one of the main hidden danger of traffic safety, and it identifies and warning problems demand solves.

Three, Professional drivers is arbitrarily taken over other's shift phenomenon and is often occurred.Random relief phenomenon is the most common in cab driving.Each taxi is interior at a fixed time should be driven by the driver specified, but in reality, the selection of relief person but has very large randomness, even there will be the people not possessing this car driving qualification and takes over other's shift.This adds the possibility that traffic accidents occurs undoubtedly, and after there is traffic hazard, is also not easy to the identification of responsibility, makes troubles to traffic administration and law enforcement agency.

In order to reduce traffic hazard, ensure traffic safety, people wish the traffic environment by laws and regulations or technological means creating policy.For this reason, country has put into effect some corresponding laws and regulations, and the development of modern science and technology is also for traffic safety problem have found effective solution route simultaneously.Such as, utilize advanced video technique and sensor technology, facial information and the physiologic information of human body can be obtained easily.Can identify the identity of driver easily based on human body face expression and mode identification technology, and the part that can obtain driver is tired and angry information.Human-body fatigue can be reflected by physiological parameter better with the emotional change of indignation, and various Wearable sensor also provides possibility for the physiologic information of driver in duty detects.Study the Fusion Features based on human body face information and physiological signal and recognition technology, and in the middle of the driving environment being applied to driver, must relieving fatigue driving and road anger disease problem effectively.Meanwhile, facial information itself comprises identity information, also can be realized the identification of driver by face-image easily.If can identify the angry mood of driver and fatigue state and remind in time, the identity of driver be confirmed simultaneously, good protective action must be played to the driving safety of driver.

Summary of the invention

In order to overcome existing driving safety safeguard measure function singleness and the problem such as detecting and assessing means are too simple; the invention provides the polynary driving safety means of defence of a kind of automobile and specialized equipment thereof; the method is based on the fusion to facial information and pulse information; the comprehensive identification of the fatigue state to driver, angry mood and identity information can be realized simultaneously, thus the safe driving of multiple angle protection driver.

The present invention is the polynary driving safety means of defence of a kind of automobile, realizes mainly through following steps:

Step (1): the facial information and the pulse information that gather driver respectively by two CMOS camera and Wrist belt-type wireless pulse sensor, and carry out filtering operation respectively with the interference of stress release treatment;

Step (2): in facial information and pulse information, by the empirical mode decomposition algorithm based on information entropy, extracts the characteristic parameter that can reflect human body indignation mood, fatigue state and identity information respectively;

Step (3): by Fisher linear classification scheduling algorithm, fusion treatment is carried out to the characteristic parameter corresponding to the various states extracted in step (2), and realizes simultaneously to the identification of the angry mood of driver, fatigue state and identity information;

Step (4): according to the result of step (3) state recognition, takes the mode of alarm or backstage record to respond driver, protects from the driving safety of multiple angle to driver.

As the further improvement of invention, in described step (1), the facial information of driver is gathered by two CMOS camera, the wide-angle camera being wherein arranged on A post position is responsible for locating the head position of driver, the high-resolution camera being arranged on windshield upper limb is responsible for gathering driver's face-image, and the position of this camera and shooting angle can be finely tuned; Gathered the pulse information of driver by Wrist belt-type pulse transducer, adopt lithium battery as power supply, the pulse information collected outwards transmits in blue teeth wireless mode, thus avoids line, farthest avoids the interference to driver's normal driving operations.To the facial information collected and pulse information, carry out filtering operation respectively with the interference of stress release treatment.

As the further improvement of invention, in described step (2), needed for different state recognitions, from facial image information and pulse information, extract one or more in angry emotional characteristics parameter, fatigue state characteristic parameter and identity characteristic parameter respectively, be specially:

1) angry emotional characteristics parameter,

For facial information, first by the position of mixed integrating method sciagraphy determination human eye, mixed integrating method projection formula is:

$H_v\left(x\right)=\frac{1}{2}{{\mathrm{\σ}}_v}^2\left(x\right)+\frac{1}{2}{M}_v\left(x\right),H_h\left(y\right)=\frac{1}{2}{{\mathrm{\σ}}_h}^2\left(y\right)+\frac{1}{2}{M}_h\left(y\right)$

H in formula _v(x) and H _hy () represents image hybrid projection in the vertical and horizontal direction respectively, M _v(x) and M _hy () represents average integral projection in the vertical and horizontal direction, σ _v(x) and σ _hy () represents variance integral projection in the vertical and horizontal direction; Feature according to the statistical law relation between position of human eye and face contour and image itself determines accurate human face region, and then uses Principal Component Analysis Algorithm (PCA) to extract principal component vector, major component Z to facial image _icomputing formula be:

Z _i＝(X ₁,X ₂,X ₃,…,X _p)*(L ₁,L ₂,L ₃,…,L _p) ^T

X in formula _i(i=1,2 ..., p) represent each column vector of image array X, L _i(i=1,2 ..., p) representing each column vector of major component loading matrix L, is that characteristic quantity carries out angry Emotion identification with Zi;

For pulse information, when acquisition one section of continuous print pulse signal, use method of characteristic point to determine main ripple position, and then obtain pulse frequency and pulse frequency average, the computing formula of pulse frequency is

$h=\frac{60}{\▿t}$

In formula, ▽ t represents the mistiming of adjacent two main ripples, to h carry out calculus of differences HRV HRV; Use FFT that pulse signal is converted to frequency field, and calculate pulse power spectral energies than r, formula is

$r=\frac{\underset{k=k_1}{\overset{k_2}{\mathrm{\Σ}}}{\left|X\left(k\right)\right|}^2}{\underset{k=k_3}{\overset{k_4}{\mathrm{\Σ}}}{\left|X\left(k\right)\right|}^2}$

Observed reading when X (k) represents that pulse signal frequency is k in formula, k1, k2, k3, k4 represent specific frequency values.Finally, carry out mode energy decomposition to pulse information, calculate pulse mode energy entropy, formula is:

$T_{(i-j)/(h-k)}=\frac{\underset{x=i}{\overset{j}{\mathrm{\Σ}}}T\left(x\right)}{\underset{x=h}{\overset{k}{\mathrm{\Σ}}}T\left(x\right)}$

In formula, T (x) is the mode energy after the natural mode of vibration component normalization of pulse signal kth, and i, j, k, h represent the sequence number of natural mode of vibration component;

2) fatigue state characteristic parameter,

For facial information, when acquisition piece image, by the position of mixed integrating method sciagraphy determination human eye, use mean-shift algorithm segmentation pupil portion, the computing formula of mean-shift is:

$M_h\left(x\right)\≡\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}G\left(\frac{x_i-x}{h}\right)w\left(x_i\right)(x_i-x)}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}G\left(\frac{x_i-x}{h}\right)w\left(x_i\right)}$

M in formula _hx () represents the mean-shift value of sample, x _ifor sample point, point centered by x, G (y) represents kernel function, and h represents the window size of kernel function, w (x _i) be sample point x _iweight; Continue to use mean-shift algorithm to follow the trail of pupil position in ensuing image, thus obtain the continuous pupil image information of some width in a period of time, and then PERCLOS, frequency of wink and average pupil size three parameters can be calculated, in order to carry out fatigue state identification; Wherein PERCLOS refers to the ratio within a certain period of time shared by eyes closed, and computing formula is:

$f=\frac{t_3-t_2}{t_4-t_1}\×100\%$

T in formula ₁, t ₂, t ₃, t ₄being illustrated respectively in an eye moves in the cycle, and eyes aperture is the time point of 80%, 20%, 20% and 80%; Frequency of wink s refers to the number of times of blinking within the unit interval, and formula is

$s=\frac{n}{\mathrm{\Δt}}$

In formula, n is the picture number that eyes aperture is less than 10%, and Δ t is the time interval; The computing formula of average pupil size q is:

$q=\frac{1}{n}\mathrm{\Σ}{l}_i$

L in formula _irepresent the spacing of pupil lower edges, n is pupil image number interior during this period of time; For pulse information, first replace heart rate so that pulse frequency is approximate, and then determine HRV parameter, and fatigue state identification can be carried out than with mode energy entropy in conjunction with pulse spectrum;

3) identity information characteristic parameter, for facial information, first by the position of mixed integrating method sciagraphy determination human eye, accurate human face region is determined according to position of human eye and facial statistical law and characteristics of image, and then PCA algorithm classification principal component vector is used, in order to carry out identity information identification to facial image; Because the identity information comprised in the middle of pulse information is less, therefore pulse information does not participate in the identification of identity information.

As the further improvement of invention, in described step (3), information fusion and pattern-recognition refer to by intelligent algorithm, and each category feature parameter is carried out fusion treatment, and realize the accurate measurements to corresponding state, are specially:

1) angry mood monitoring, using 5 seconds as the time interval, by the facial information gathered in every 5 seconds and pulse information in order to perform step (2), extract angry emotional characteristics parameter, using the angry emotional characteristics parameter of extraction in step (2) as input quantity, use Fisher linear classifier by its fusion treatment, namely the process of classification achieves the monitoring to angry mood, judge that if double driver is as angry mood, then assert that driver's current emotional is indignation; Wherein Fisher linear classifier is method common in pattern classification, its thought makes linear projection to input feature vector, after making projection, the inter _ class relationship of data reaches maximum and within-cluster variance reaches minimum, linear projection axle used is called discriminant vectors, then by discriminant vectors, sample is projected, thus determine the classification belonging to sample, obtain classification results;

2) fatigue state monitoring, using 1 minute as the time interval, by the facial information gathered in each minute and pulse information in order to perform step (2), extract fatigue state characteristic parameter, using the fatigue state characteristic parameter of extraction in step (2) as input quantity, use SVM algorithm by its fusion treatment, namely the process of classification achieves the monitoring to fatigue state, judge that if double driver is as fatigue state, then assert that driver is current and enter fatigue state; Wherein SVM algorithm is a kind of two conventional sorting algorithms, its step is, by interior Product function by input feature vector linear transformation to higher dimensional space, and by training mode in higher dimensional space, find optimal classification surface, then according to sample in this space with the position relationship determination classification results of optimal classification surface;

3) identity information monitoring, using 5 minutes as the time interval, by the facial information that gathers in every 5 minutes and pulse information in order to perform step (2), extract identity characteristic parameter, using the identity information characteristic parameter of extraction in step (2) as input quantity, use Fisher linear classifier by its fusion treatment, namely the process of classification achieves the monitoring to identity information, if the identity information of double judgement driver is not inconsistent, then assert that current driver is not the driver specified;

As the further improvement of invention, in described step (4), the specific implementation of antifeedback measures is: when monitoring driver and occurring fatigue state or angry mood, fed back to driver by the mode playing alarm sound, thus correct the bad steering state of driver, ensure driving safety; When the identity information monitoring driver is wrong, is there is by one memory storage preservation driver's face-image now of access rights, wait until supvr and make regular check on process, thus stop arbitrarily to take over other's shift phenomenon, ensure driving safety.

The invention has the beneficial effects as follows: by the accurate acquisition to facial information and pulse information, intelligent algorithm is used to carry out effective fusion treatment to two kinds of information, achieve simultaneously to the comprehensive identification of the angry mood of driver, fatigue state and identity information, there is the feature of multifunctionality, accuracy and simplification, can protect from the driving safety of multiple angle to driver, thus the generation avoided traffic accident, reduce casualties and property loss.

Accompanying drawing explanation

Fig. 1 is hardware structure diagram of the present invention.

Fig. 2 is process flow diagram of the present invention.

Fig. 3 is that Wrist belt type sensor wears schematic diagram.

Fig. 4 is that real vehicle installs front view.

Fig. 5 is that real vehicle installs vertical view.

In Fig. 4-5, the representative of circle frame is decided to be camera, and oval frame represents acquisition camera, and rectangle frame represents main computer unit.

Embodiment

Below in conjunction with accompanying drawing and embodiment, the invention will be further described.

The invention provides the car assisted polynary driving safety means of defence of a kind of automobile multifunctional and specialized equipment thereof, its specific implementation step is as follows:

Step 1: the collection of facial information and pulse information;

(1) facial information of driver is gathered by two CMOS camera, the head position of wide-angle camera to driver being wherein arranged on A post position positions, the high-resolution camera being arranged on windshield upper limb is responsible for gathering driver's face-image details, and the position of this camera and shooting angle can be finely tuned;

(2) pulse information of driver is gathered by Wrist belt-type pulse transducer, adopt lithium battery as power supply, the pulse information collected outwards transmits in blue teeth wireless mode, thus avoids line, farthest reduces the interference to driver's normal driving operations;

Step 2: information processing;

Facial information and pulse information are respectively in wired and mode that is blue teeth wireless, be sent to and be arranged on copilot glove box place, using DSP as the central processing module of MCU, processing module is by realizing the state recognition to driver to the fusion treatment of two kinds of signals, processing procedure comprises:

(1) pre-service, carries out filtering, with the impact of stress release treatment to facial image information and pulse information respectively;

(2) feature extraction, needed for different state recognitions, from facial image information and pulse information, extract angry emotional characteristics parameter, fatigue state characteristic parameter and identity characteristic parameter respectively, be specially:

A) angry emotional characteristics parameter, for facial information, first in conjunction with the position of template matches and mixed integrating method sciagraphy determination human eye, feature according to the statistical law relation between position of human eye and face contour and image itself determines accurate human face region, and then use PCA algorithm to extract principal component vector, in order to carry out angry Emotion identification to facial image; For pulse information, when acquisition one section of continuous print pulse signal, use method of characteristic point to determine main ripple position respectively, and then obtain pulse frequency and pulse frequency average, the computing formula of pulse frequency is

$h=\frac{60}{\▿t}$

In formula, ▽ t represents the mistiming of adjacent two main ripples, to h carry out calculus of differences HRV HRV; Use FFT that pulse signal is converted to frequency field, and calculate pulse power spectral energies than r, formula is

$r=\frac{\underset{k=k_1}{\overset{k_2}{\mathrm{\Σ}}}{\left|X\left(k\right)\right|}^2}{\underset{k=k_3}{\overset{k_4}{\mathrm{\Σ}}}{\left|X\left(k\right)\right|}^2}$

Observed reading when X (k) represents that pulse signal frequency is k in formula, k1, k2, k3, k4 represent specific frequency values.Finally, carry out mode energy decomposition to pulse information, calculate pulse mode energy entropy, formula is:

$T_{(i-j)/(h-k)}=\frac{\underset{x=i}{\overset{j}{\mathrm{\Σ}}}T\left(x\right)}{\underset{x=h}{\overset{k}{\mathrm{\Σ}}}T\left(x\right)}$

In formula, T (x) is the mode energy after the natural mode of vibration component normalization of pulse signal kth, and i, j, k, h represent the sequence number of natural mode of vibration component;

B) fatigue state characteristic parameter, for facial information, when acquisition piece image, in conjunction with the position of template matches and mixed integrating method sciagraphy determination human eye, use mean-shift algorithm segmentation pupil portion, continue to use mean-shift algorithm to follow the trail of pupil position in ensuing image, thus obtain the continuous pupil image information of some width in a period of time, and then PERCLOS, frequency of wink and average pupil size three parameters can be calculated, in order to carry out fatigue state identification; Wherein PERCLOS refers to the ratio within a certain period of time shared by eyes closed, and computing formula is:

$f=\frac{t_3-t_2}{t_4-t_1}\×100\%$

T in formula ₁, t ₂, t ₃, t ₄being illustrated respectively in an eye moves in the cycle, and eyes aperture is the time point of 80%, 20%, 20% and 80%; Frequency of wink refers within the unit interval, and eyes aperture is less than the picture number n of 10% and the ratio of time span t.Average pupil size refers in a period of time, the mean value of pupil lower edges spacing in each image; For pulse information, first replace heart rate so that pulse frequency is approximate, and then determine HRV parameter, and fatigue state identification can be carried out than with mode energy entropy in conjunction with pulse spectrum;

C) identity information characteristic parameter, for facial information, first in conjunction with the position of template matches and mixed integral projection method determination human eye, accurate human face region is determined according to position of human eye and facial statistical law and characteristics of image, and then PCA algorithm classification principal component vector is used, in order to carry out identity information identification to facial image; Because the identity information comprised in the middle of pulse information is less, therefore pulse information does not participate in the identification of identity information.

(3) information fusion and pattern-recognition, uses intelligent algorithm, each category feature parameter is carried out fusion treatment, and realizes the accurate measurements to corresponding state, is specially:

A) angry mood monitoring, using the angry emotional characteristics parameter of extraction in step (2) as input quantity, use Fisher linear classifier by its fusion treatment, namely the process of classification achieves the monitoring to angry mood;

B) fatigue state monitoring, using the fatigue state characteristic parameter of extraction in step (2) as input quantity, use SVM algorithm by its fusion treatment, namely the process of classification achieves the monitoring to fatigue state;

C) identity information monitoring, using the identity information characteristic parameter of extraction in step (2) as input quantity, use Fisher linear classifier by its fusion treatment, namely the process of classification achieves the monitoring to identity information;

Step 3: condition responsive;

(1) when monitoring driver and occurring angry mood or fatigue state, fed back to driver by the mode playing alarm, thus correct the bad steering state of driver, ensure driving safety;

(2) when the identity information monitoring driver is wrong, be there is by one memory storage preservation driver's face-image now of access rights, wait until supvr and make regular check on process, thus stop arbitrarily to take over other's shift phenomenon, ensure driving safety.

By reference to the accompanying drawings the specific embodiment of the present invention is described although above-mentioned; but not limiting the scope of the invention; one of ordinary skill in the art should be understood that; on the basis of technical scheme of the present invention, those skilled in the art do not need to pay various amendment or distortion that creative work can make still within protection scope of the present invention.

Claims (4)

1. the polynary driving safety means of defence of automobile, is characterized in that, realize mainly through following steps:

Step (1): the facial information and the pulse information that gather driver respectively by two CMOS camera and Wrist belt-type wireless pulse sensor, and carry out filtering operation respectively with the interference of stress release treatment;

Step (2): in facial information and pulse information, by the empirical mode decomposition algorithm based on information entropy, extracts the characteristic parameter that can reflect human body indignation mood, fatigue state and identity information respectively;

Step (3): by Fisher linear classify algorithm, fusion treatment is carried out to the characteristic parameter corresponding to the various states extracted in step (2), and realizes simultaneously to the identification of the angry mood of driver, fatigue state and identity information;

Step (4): according to the result of step (3) state recognition, takes the mode of alarm or backstage record to respond driver;

In described step (1), the facial information of driver is gathered by two CMOS camera, the camera being wherein arranged on A post position is responsible for locating the head position of driver, the camera being arranged on windshield upper limb is responsible for gathering driver's face-image, and the position of this camera and shooting angle can be finely tuned; Gathered the pulse information of driver by Wrist belt-type pulse transducer, the pulse information collected outwards transmits in blue teeth wireless mode.

2. the polynary driving safety means of defence of automobile according to claim 1, it is characterized in that, in described step (2), from facial image information and pulse information, extract one or more in angry emotional characteristics parameter, fatigue state characteristic parameter and identity characteristic parameter respectively, be specially:

1) angry emotional characteristics parameter,

For facial information, first by the position of mixed integrating method sciagraphy determination human eye, mixed integrating method projection formula is:

$H_v\left(x\right)=\frac{1}{2}{{\mathrm{\σ}}_v}^2\left(x\right)+\frac{1}{2}{M}_v\left(x\right),H_h\left(y\right)=\frac{1}{2}{{\mathrm{\σ}}_h}^2\left(y\right)+\frac{1}{2}{M}_h\left(y\right)$

H in formula _v(x) and H _hy () represents image hybrid projection in the vertical and horizontal direction respectively, M _v(x) and M _hy () represents average integral projection in the vertical and horizontal direction, σ _v(x) and σ _vy () represents variance integral projection in the vertical and horizontal direction; Feature according to the statistical law relation between position of human eye and face contour and image itself determines accurate human face region, and then uses Principal Component Analysis Algorithm (PCA) to extract principal component vector, major component Z to facial image _icomputing formula be:

Z _i＝(X ₁，X ₂，X ₃，…，X _p)*(L ₁，L ₂，L ₃，…，L _p) ^T

X in formula _i(i=1,2 ..., p) represent each column vector of image array X, L _i(i=1,2 ..., p) represent each column vector of major component loading matrix L, with Z _ifor characteristic quantity carries out angry Emotion identification;

For pulse information, when acquisition one section of continuous print pulse signal, use method of characteristic point to determine main ripple position, and then obtain pulse frequency and pulse frequency average, the computing formula of pulse frequency is

$h=\frac{60}{\▿t}$

In formula represent the mistiming of adjacent two main ripples, to h carry out calculus of differences HRV HRV; Use FFT that pulse signal is converted to frequency field, and calculate pulse power spectral energies than r, formula is

$r=\frac{\underset{k=k_1}{\overset{k_2}{\Σ}}|X\left(k\right){|}^2}{\underset{k=k_3}{\overset{k_4}{\Σ}}|X\left(k\right){|}^2}$

Observed reading when X (k) represents that pulse signal frequency is k in formula, k1, k2, k3, k4 represent specific frequency values; Finally, carry out mode energy decomposition to pulse information, calculate pulse mode energy entropy, formula is:

$T_{\left(i-j\right)/\left(h-k\right)}=\frac{\underset{x=i}{\overset{j}{\Σ}}T\left(x\right)}{\underset{x=h}{\overset{k}{\Σ}}T\left(x\right)}$

In formula, T (x) is the mode energy after the natural mode of vibration component normalization of pulse signal kth, and i, j, k, h represent the sequence number of natural mode of vibration component;

2) fatigue state characteristic parameter,

For facial information, when acquisition piece image, by the position of mixed integrating method sciagraphy determination human eye, use mean-shift algorithm segmentation pupil portion, the computing formula of mean-shift is:

$M_h\left(x\right)\≡\frac{\underset{i=1}{\overset{n}{\Σ}}G\left(\frac{x_i-x}{h}\right)w\left(x_i\right)(x_i-x)}{\underset{i=1}{\overset{n}{\Σ}}G\left(\frac{x_i-x}{h}\right)w\left(x_i\right)}$

M in formula _hx () represents the mean-shift value of sample, x _ifor sample point, point centered by x, G (y) represents kernel function, and h represents the window size of kernel function, w (x _i) be sample point x _iweight; Continue to use mean-shift algorithm to follow the trail of pupil position in ensuing image, thus obtain the continuous pupil image information of some width in a period of time, and then PERCLOS, frequency of wink and average pupil size three parameters can be calculated, in order to carry out fatigue state identification; Wherein PERCLOS refers to the ratio within a certain period of time shared by eyes closed, and computing formula is:

$f=\frac{t_3-t_2}{t_4-t_1}\×100\%$

T in formula ₁, t ₂, t ₃, t ₄being illustrated respectively in an eye moves in the cycle, and eyes aperture is the time point of 80%, 20%, 20% and 80%; Frequency of wink s refers to the number of times of blinking within the unit interval, and formula is

$s=\frac{n}{\mathrm{\Δ}t}$

In formula, n is the picture number that eyes aperture is less than 10%, is the time interval; The computing formula of average pupil size q is:

$q=\frac{1}{n}{\mathrm{\Σl}}_i$

L in formula _irepresent the spacing of pupil lower edges, n is pupil image number interior during this period of time; For pulse information, first replace heart rate so that pulse frequency is approximate, and then determine HRV parameter, and fatigue state identification can be carried out than with mode energy entropy in conjunction with pulse spectrum;

3) identity information characteristic parameter, for facial information, first by the position of mixed integrating method sciagraphy determination human eye, accurate human face region is determined according to position of human eye and facial statistical law and characteristics of image, and then PCA algorithm classification principal component vector is used, in order to carry out identity information identification to facial image.

3. the polynary driving safety means of defence of automobile according to claim 1, is characterized in that, in described step (3), information fusion and pattern-recognition refer to passes through intelligent algorithm, each category feature parameter is carried out fusion treatment, and realizes the accurate measurements to corresponding state, be specially:

1) angry mood monitoring, using 5 seconds as the time interval, by the facial information gathered in every 5 seconds and pulse information in order to perform step (2), extract angry emotional characteristics parameter, using the angry emotional characteristics parameter of extraction in step (2) as input quantity, use Fisher linear classifier by its fusion treatment, namely the process of classification achieves the monitoring to angry mood, judge that if double driver is as angry mood, then assert that driver's current emotional is indignation;

2) fatigue state monitoring, using 1 minute as the time interval, by the facial information gathered in each minute and pulse information in order to perform step (2), extract fatigue state characteristic parameter, using the fatigue state characteristic parameter of extraction in step (2) as input quantity, use SVM algorithm by its fusion treatment, namely the process of classification achieves the monitoring to fatigue state, judge that if double driver is as fatigue state, then assert that driver is current and enter fatigue state;

3) identity information monitoring, using 5 minutes as the time interval, by the facial information that gathers in every 5 minutes and pulse information in order to perform step (2), extract identity characteristic parameter, using the identity information characteristic parameter of extraction in step (2) as input quantity, use Fisher linear classifier by its fusion treatment, namely the process of classification achieves the monitoring to identity information, if the identity information of double judgement driver is not inconsistent, then assert that current driver is not the driver specified.

4. the polynary driving safety means of defence of automobile according to claim 1, it is characterized in that, in described step (4), the specific implementation of antifeedback measures is: when monitoring driver and occurring fatigue state or angry mood, fed back to driver by the mode playing alarm sound, thus correct the bad steering state of driver, ensure driving safety; When the identity information monitoring driver is wrong, is there is by one memory storage preservation driver's face-image now of access rights, wait until supvr and make regular check on process, thus stop arbitrarily to take over other's shift phenomenon, ensure driving safety.