CN106778538A - Intelligent driving behavior evaluation method based on analytic hierarchy process (AHP) - Google Patents

Intelligent driving behavior evaluation method based on analytic hierarchy process (AHP) Download PDF

Info

Publication number
CN106778538A
CN106778538A CN201611067704.6A CN201611067704A CN106778538A CN 106778538 A CN106778538 A CN 106778538A CN 201611067704 A CN201611067704 A CN 201611067704A CN 106778538 A CN106778538 A CN 106778538A
Authority
CN
China
Prior art keywords
normal
driving
lane change
turn
face
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201611067704.6A
Other languages
Chinese (zh)
Inventor
方志军
姚兴华
刘翔
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai University of Engineering Science
Original Assignee
Shanghai University of Engineering Science
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai University of Engineering Science filed Critical Shanghai University of Engineering Science
Priority to CN201611067704.6A priority Critical patent/CN106778538A/en
Publication of CN106778538A publication Critical patent/CN106778538A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/59—Context or environment of the image inside of a vehicle, e.g. relating to seat occupancy, driver state or inner lighting conditions
    • G06V20/597—Recognising the driver's state or behaviour, e.g. attention or drowsiness
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

The present invention relates to a kind of intelligent driving behavior evaluation method based on analytic hierarchy process (AHP), including：Step 1) eyes of face and people are followed the trail of by using Open CV, obtain to the fatigue detecting of driver and note force detection；Step 2) comprehensive marking is carried out come the driving behavior to driver in terms of six based on analytic hierarchy process (AHP) and BP neural network training, six of which aspect is respectively normal driving, fatigue driving, normal turn, break turn, normal lane change and frequent lane change.Compared with prior art, the present invention is accurate with scoring for scoring the driving behavior of driver custom, the advantages of reference value is high.

Description

Intelligent driving behavior evaluation method based on analytic hierarchy process (AHP)
Technical field
The present invention relates to a kind of driving behavior evaluation method., driven more particularly, to a kind of intelligence based on analytic hierarchy process (AHP) Sail behavior evaluation method.
Background technology
According to statistics, the domestic car owner that is not in danger for always having 82.5% checks in the client that is often in danger for 17.5%.Take in vehicle insurance When the rate marketization is put into effect, with driving habit offset premium, the car owner that allowing has good driving habit enjoys minimum living expense, allows and always like The car owner of " dangerous driving " is the car networking product of the driving behavior " checking " of oneself, into the topic that nearest in the market is burning hot. Such as：The products such as the road treasured that Tencent releases, can be by data for the driving behavior of car owner is given a mark.Driving behavior custom is commented Sentence system, be possible to carry out quantization judge to the driving behavior of driver.The scoring that evaluation system is given is used as driver's car One Primary Reference foundation of dangerous premium, can effectively strengthen the safe driving consciousness of driver, the driving of specification driver Behavior, reduces the loss of lives and properties.
The content of the invention
The purpose of the present invention is exactly that a kind of scoring is provided for the defect for overcoming above-mentioned prior art to exist accurately, reference The costly intelligent driving behavior evaluation method based on analytic hierarchy process (AHP).
The purpose of the present invention can be achieved through the following technical solutions：
A kind of intelligent driving behavior evaluation method based on analytic hierarchy process (AHP), it is characterised in that including：
Step 1) eyes of face and people are followed the trail of by using Open CV, obtain to the fatigue detecting of driver and note Meaning force detection；
Step 2) entered come the driving behavior to driver in terms of six based on analytic hierarchy process (AHP) and BP neural network training The comprehensive marking of row, six of which aspect is respectively normal driving, fatigue driving, normal turn, break turn, normal lane change and frequency Numerous lane change；Marking grade includes：100~90,90~80,80~60,60~40,40~20,20~0.
Described step 1) it is specially：
Face and the eyes of driver are caught by the built-in camera in driver's cabin and Open CV technologies, eyes are judged State, calculate and driver fatigue and judge whether driver attention concentrates, the state of described eyes includes eye opening, closes The frequency of eye and eyes blink.
Described driver fatigue detecting is specific as follows：
(101) human eye Primary Location, the priori of the position distribution according to human eye in face, in the base of recognition of face On plinth, it is assumed that the face width w for collecting, is highly H, if the starting point coordinate of face rectangle frame is (X, Y), according to the several of human eye The width of face, is divided into 6 parts of equal lengths by what position relationship, is highly divided into 4 parts of equal lengths, the position of human eye that frame is selected Starting point coordinate isThe coordinate in final position isThe position for then being determined by the rectangle two ends It is exactly position where eye to put；
(102) people's ocular pursuit, with reference to vision built-in function interface and external camera that OpenCV is provided, first to scene Video acquisition, to the picture of normal acquisition, by loading face classification device, detection identifies face, in the face base for identifying On plinth, based on experience with the geometric position feature of human eye, select region Primary Location interested and go out human eye substantially Position, then gathers next two field picture, does same treatment, realizes the tracking to human eye；
(103) judge the state of eyes, calculate the fatigue strength of driver.
Described driver attention's detecting is specific as follows：
(201) first it is face's detecting after obtaining in-vehicle image, Open CV, face in interception image is used in detecting Part, tracking face is taken followed by the mode of COLOR COMPOSITION THROUGH DISTRIBUTION and statistics；
(202) region more black during face influences, i.e. eye areas are searched using critical value method, with k-means groups after allowing Many algorithms find out the position of eyes；
(203) whether the eyes closed time is judged more than 0.3 second, if exceeding, judges that the behavior of driver constitutes danger Behavior.
Described step 2) it is specially：
31) the various signals collected for driving behavior identification, by according to each signal parameter meaning, to whole factors The contrast judgement square A=(a of structural factor relative importance are contrasted two-by-twoij)k×k, wherein aij=f (xi, xj) it is contrast letter Number, contrast function value such as table, so that weight vectors W can be determined according to the criterion of setting
Classification proportion quotiety reference table is as shown in table 1：
Table 1
Assignment (xi/xj) Implication
1 Represent index xiWith xjCompare, with same importance
3 Represent index xiWith xjCompare, index xiCompare xjIt is somewhat important
5 Represent index xiWith xjCompare, index xiCompare xjIt is substantially important
7 Represent index xiWith xjCompare, index xiCompare xjIt is strong important
9 Represent index xiWith xjCompare, index xiCompare xjIt is extremely important
2、4、6、8 The more than correspondence intermediate state of two adjacent judgements
It is reciprocal Index xiWith xjCompare to obtain aij, then index xjWith xiComparing has aji=1/aij
(1) the analytic hierarchy process (AHP) classification proportion quotiety table according to table 1, six kinds of information pair that construction system acquisition treatment is obtained The comparison sheet two-by-two of 100~90 marking, it is as shown in table 2 below.
Table 2
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change
Normal driving 1 9 2 9 2 9
Fatigue driving 1/9 1 1/9 1/2 1/9 1/2
Normal turn 1/2 9 1 9 2 9
Break turn 1/9 2 1/9 1 1/9 1/2
Normal lane change 1/2 9 1/2 9 1 9
Frequent lane change 1/9 2 1/9 2 1/9 1
Then, the synthetical information evaluating table two-by-two for passing ratio scale reference table being built, is converted into and judges square accordingly Battle array A1,
Can be calculated, the Maximum characteristic root of matrix A 1 is λmax=6.27, corresponding characteristic vector is：
For matrix A 1, its consistency check index：
RI=1.24；
From CR ＜ 0.1, judgment matrix A1 meets coherence request；
(2) to such as table 3 below of comparison sheet two-by-two of 90~80 marking：
Table 3
Corresponding judgment matrix A2 can be obtained by table 3,
Its Maximum characteristic root λmax=6.27, corresponding characteristic vector is
(3) to such as table 4 below of comparison sheet two-by-two of 80~60 marking：
Table 4
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change
Normal driving 1 6 2 6 2 6
Fatigue driving 1/6 1 1/6 1/2 1/6 1/2
Normal turn 1/2 6 1 6 2 6
Break turn 1/6 2 1/6 1 1/6 1/2
Normal lane change 1/2 6 1/2 6 1 6
Frequent lane change 1/6 2 1/6 2 1/6 1
Corresponding judgment matrix A3 can be obtained by table 4,
Its Maximum characteristic root is λmax=6.27, individual features vector is
(4) to such as table 5 below of comparison sheet two-by-two of 60~40 marking：
Table 5
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change
Normal driving 1 1/2 1/2 1/3 1/2 1/3
Fatigue driving 2 1 3 2 3 2
Normal turn 2 1/3 1 1/3 1/2 1/3
Break turn 3 1/2 3 1 3 2
Normal lane change 2 1/3 2 1/3 1 1/3
Frequent lane change 3 1/2 3 1/2 3 1
Corresponding judgment matrix A4 can be obtained by table 5,
Its Maximum characteristic root is λmax=6.36, individual features vector is
(5) to such as table 6 below of comparison sheet two-by-two of 40~20 marking：
Table 6
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change
Normal driving 1 1/4 1/2 1/4 1/2 1/4
Fatigue driving 4 1 4 2 4 2
Normal turn 2 1/4 1 1/4 1/2 1/4
Break turn 4 1/2 4 1 4 2
Normal lane change 2 1/4 2 1/4 1 1/4
Frequent lane change 4 1/2 4 1/2 4 1
Corresponding judgment matrix A5 can be obtained by table 6,
Its Maximum characteristic root is λmax=6.27, individual features vector is：
(6) to such as table 7 below of comparison sheet two-by-two of 20~0 marking：
Table 7
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change
Normal driving 1 1/6 1/2 1/6 1/2 1/6
Fatigue driving 6 1 6 2 6 2
Normal turn 2 1/6 1 1/6 1/2 1/6
Break turn 6 1/2 6 1 6 2
Normal lane change 2 1/6 2 1/6 1 1/6
Frequent lane change 6 1/2 6 1/2 6 1
Corresponding judgment matrix A6 can be obtained by table 7,
Its Maximum characteristic root is λmax=6.27, individual features vector is
(7) eigenvectors matrix that driving behavior is judged
According to (1) above~(6), can obtain on normal driving, fatigue driving, normal turn, break turn, normal Lane change, six matrixes of driving behavior evaluation rank of frequent lane change factor, driving behavior judge corresponding vector be：
T={ normal driving, fatigue driving, normal turn, break turn, normal lane change, frequent lane change }T,
Character pair value constitute vector be：
λ=[6.27 6.27 6.27 6.36 6.27 6.27]T,
Character pair vector constitute matrix be：
32) BP neural network training
Each information obtained using the chromatographic assays of driving behavior grade is on different driving behavior evaluation ranks Decision weights coefficient, with reference to expert's assessment, the related vectorial Preliminary design BP neural network training sample of driving behavior identification of design.
Input information vector is expressed as：
S=[the frequent lane change of the normal driving normal lane change of fatigue driving normal turn break turn]
Output vector is expressed as：
T=[100~90 90~80 80~60 60~40 40~20 20~0]
Correspondence position is 1, then it represents that recognition result is exactly the driving behavior evaluation rank that this position represents；
The training sample of BP neural network is as shown in table 8：
Table 8
Realize the training of BP neural network using MATLAB, and by training result by means of MATLAB figure shows work( Dynamic shows.
Compared with prior art, the present invention is for scoring the driving behavior of driver custom, and insurance company can be with What this evaluation method was given scores as a foundation for calculating driver's vehicle insurance premium, so that, promote the strong of vehicle insurance market Kang Fazhan.
Brief description of the drawings
Fig. 1 is human eye position tracking flow chart；
Fig. 2 is driver attention detecting flow process figure；
Fig. 3 is driving behavior discriminant analysis hierarchy chart；
Fig. 4 is BP neural network error curve diagram.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.
Embodiment
A kind of intelligent driving behavior evaluation method based on analytic hierarchy process (AHP), including：Step 1) come by using Open CV The eyes of face and people are followed the trail of, the fatigue detecting to driver and attention force detection is obtained；Step 2) based on analytic hierarchy process (AHP) and BP neural network is trained carries out comprehensive marking in terms of six come the driving behavior to driver, and six of which aspect is respectively just Normal driving, fatigue driving, normal turn, break turn, normal lane change and frequent lane change；Marking grade includes：100~90,90 ~80,80~60,60~40,40~20,20~0.
Described step 1) it is specially：Catch driver's by the built-in camera in driver's cabin and Open CV technologies Face and eyes, judge the state of eyes, calculate driver fatigue and judge whether driver attention concentrates, described eye The state of eyeball includes the frequency opened eyes, eye closing and eyes are blinked.
1st, driver fatigue detecting：
Data of literatures shows that generally, human eye closing time is between 0.2~0.3s；If driver exists On the run the eyes closed time reach 0.5~3s, then easily there is traffic accident.Accordingly, it would be desirable to complete within this time The analysis and judgement of paired driver fatigue.This patent evaluation method, is caught by built-in camera and Open CV methods and driven The face of the person of sailing and eyes, judge the state of eyes, and then such as eye opening, eye closing, the frequency of eyes blink calculate driver's Fatigue strength.
1.1 human eye Primary Locations
Human eye Primary Location, is the priori of the position distribution according to human eye in face, on the basis of recognition of face On, it is assumed that the face width W for collecting, is highly H, if the starting point coordinate of face rectangle frame is (X, Y)
According to the geometry site of human eye, the width of face is divided into 6 parts of equal lengths herein, is highly divided into 4 parts Equal length, the starting point coordinate of the position of human eye that frame is selected isThe coordinate in final position is The position for then being determined by the rectangle two ends is exactly the position where eye.
1.2 people's ocular pursuits
With reference to vision built-in function interface and external camera that OpenCV is provided, live video is gathered first, aligned Often the picture of collection, by loading face classification device, detects and identifies face, on the basis of the face for identifying, rule of thumb The geometric position feature of knowledge and human eye, selects the approximate location that region Primary Location interested goes out human eye, then gathers Next two field picture, does same treatment, realizes the tracking to human eye.
First, cvCaptureFromCAM (0) video flowing for being provided using OpenCV gathers function, and live video is adopted Collection, then loads face classification device and calls a function to detect face, after Face datection goes out, to the positioning of human face target, Ran Hougen The Heuristicses such as the geometric position distribution according to human eye in face are extracted to eye position, are realized to the preliminary fixed of human eye Position, finally shows the position of human eye, is so achieved that the position tracking to present frame human eye.Its design flow diagram such as Fig. 1 It is shown.
2. driver attention's detecting
To the method and step of driver attention's detecting：
Step1：First it is face's detecting after obtaining in-vehicle image, Open CV, people in interception image is used in detecting Face part, tracking face is taken followed by the mode of COLOR COMPOSITION THROUGH DISTRIBUTION and statistics
Step2：More black region in being influenceed using critical value method face, that is, eye areas, are drilled with the k-means masses Algorithm finds out the position of eyes
Step3：By face's detecting, eye detecting, three nodes, then design one and judge quasi- survey to judge driver Behavior and behavior whether constitute hazardous act.
Driver attention detecting flow process figure is as shown in Figure 2.
3rd, principal factor analysis (PFA) is judged in the driving behavior based on analytic hierarchy process (AHP)
For the much informations such as people-car-road are collected, introduce analytic hierarchy process (AHP) is carried out to the relative importance of multi information Compare, both important sexual intercourse are determined according to expertise, using 1~9 ratio scale, multilevel iudge, quantification, composition Judgment matrix.For the driving behavior various signals that are collected of identification, by expert according to each signal parameter meaning, to it is whole because Element is contrasted the contrast judgement square A=(a of structural factor relative importance two-by-twoij)k×k, wherein aij=f (xi, xj) it is contrast Function, contrast function value such as table, so as to weight vectors W can be determined according to certain criterion.
The various information of people-car that project is obtained to acquisition process, analyze much information and driving behavior are differentiated using AHP Recognition decision weight, the information for relating generally to has normal driving, fatigue driving, normal turn, break turn, normal lane change, frequency Numerous lane change etc., shown in level of analysis Fig. 3 as shown below.
(1) the analytic hierarchy process (AHP) classification proportion quotiety table according to table 1, six kinds of information pair that construction system acquisition treatment is obtained The comparison sheet two-by-two of 100~90 marking, as shown in table 2.
(2) comparison sheet two-by-two to 90~80 marking is as shown in table 3.
(3) comparison sheet two-by-two to 80~60 marking is as shown in table 4.
(4) comparison sheet two-by-two to 60~40 marking is as shown in table 5 below.
(5) comparison sheet two-by-two to 40~20 marking is as shown in table 6 below.
(6) comparison sheet two-by-two to 20~0 marking is as shown in table 7 below.
(7) eigenvectors matrix that driving behavior is judged
According to (1) above~(6), can obtain on normal driving, fatigue driving, normal turn, break turn, just Six matrixes of driving behavior evaluation rank of Chang Biandao, frequent lane change factor.Driving behavior judge corresponding vector be：
T={ normal driving, fatigue driving, normal turn, break turn, normal lane change, frequent lane change }T,
Character pair value constitute vector be：
λ=[6.27 6.27 6.27 6.36 6.27 6.27]T,
Character pair vector constitute matrix be：
4th, BP neural network training
Each information obtained using the chromatographic assays of driving behavior grade is on different driving behavior evaluation ranks Decision weights coefficient, with reference to expert's assessment, the related vectorial Preliminary design BP neural network training sample of driving behavior identification of design.
Input information vector is expressed as：
S=[the frequent lane change of the normal driving normal lane change of fatigue driving normal turn break turn]
Output vector is expressed as：
T=[100~90 90~80 80~60 60~40 40~20 20~0].
Correspondence position is 1, then it represents that recognition result is exactly the driving behavior evaluation rank that this position represents.BP neural network Training sample it is as shown in table 8.The training of BP neural network is realized using MATLAB, and by training result by means of MATLAB Powerful graphical display function Dynamic Announce is out.The overall error for setting nerve network system is 0.01, show that neutral net is selected Even error performance curve is as shown in Figure 4：
Figure 4, it is seen that by the training of 14 steps or so, network system reaches stabilization, and error reaches 0.0077984<0.01, meet expected training requirement.

Claims (5)

1. a kind of intelligent driving behavior evaluation method based on analytic hierarchy process (AHP), it is characterised in that including：
Step 1) eyes of face and people are followed the trail of by using Open CV, obtain to the fatigue detecting of driver and notice Detecting；
Step 2) carried out come the driving behavior to driver in terms of six based on analytic hierarchy process (AHP) and BP neural network training it is comprehensive Marking is closed, six of which aspect is respectively normal driving, fatigue driving, normal turn, break turn, normal lane change and frequently becomes Road.
2. a kind of intelligent driving behavior evaluation method based on analytic hierarchy process (AHP) according to claim 1, it is characterised in that Described step 1) it is specially：
Face and the eyes of driver are caught by the built-in camera in driver's cabin and Open CV technologies, the shape of eyes is judged State, calculates and driver fatigue and judges whether driver attention concentrates, the state of described eyes includes opening eyes, eye closing with And the frequency of eyes blink.
3. a kind of intelligent driving behavior evaluation method based on analytic hierarchy process (AHP) according to claim 2, it is characterised in that Described driver fatigue detecting is specific as follows：
(101) human eye Primary Location, the priori of the position distribution according to human eye in face, on the basis of recognition of face On, it is assumed that the face width w for collecting, is highly H, if the starting point coordinate of face rectangle frame is (X, Y), according to the geometry of human eye The width of face, is divided into 6 parts of equal lengths by position relationship, is highly divided into 4 parts of equal lengths, and the position of human eye that frame is selected rises Point coordinates isThe coordinate in final position isThe position for then being determined by the rectangle two ends It is exactly position where eye to put；
(102) people's ocular pursuit, with reference to vision built-in function interface and external camera that OpenCV is provided, first to live video Collection, to the picture of normal acquisition, by loading face classification device, detection identifies face, on the basis of the face for identifying, Based on experience with the geometric position feature of human eye, the approximate location that region Primary Location interested goes out human eye is selected, Then next two field picture is gathered, same treatment is done, the tracking to human eye is realized；
(103) judge the state of eyes, calculate the fatigue strength of driver.
4. a kind of intelligent driving behavior evaluation method based on analytic hierarchy process (AHP) according to claim 2, it is characterised in that Described driver attention's detecting is specific as follows：
(201) first it is face's detecting after obtaining in-vehicle image, Open CV, face in interception image is used in detecting Point, take tracking face followed by the mode of COLOR COMPOSITION THROUGH DISTRIBUTION and statistics；
(202) region more black during face influences, i.e. eye areas are searched using critical value method, is drilled with the k-means masses after allowing Algorithm finds out the position of eyes；
(203) whether the eyes closed time is judged more than 0.3 second, if exceeding, judges that the behavior of driver constitutes hazardous act.
5. a kind of intelligent driving behavior evaluation method based on analytic hierarchy process (AHP) according to claim 2, it is characterised in that Described step 2) it is specially：
31) the various signals collected for driving behavior identification, by according to each signal parameter meaning, being carried out to whole factors The contrast judgement square A=(a of structural factor relative importance are contrasted two-by-twoij)k×k, wherein aij=f (xi, xj) it is contrast function, it is right Than function value such as table, so that weight vectors W can be determined according to the criterion of setting
Classification proportion quotiety reference table is as shown in table 1：
Table 1
Assignment (xi/xj ) Implication 1 Represent index xiWith xjCompare, with same importance 3 Represent index xiWith xjCompare, index xiCompare xjIt is somewhat important 5 Represent index xiWith xjCompare, index xiCompare xjIt is substantially important 7 Represent index xiWith xjCompare, index xiCompare xjIt is strong important 9 Represent index xiWith xjCompare, index xiCompare xjIt is extremely important 2、4、6、8 The more than correspondence intermediate state of two adjacent judgements It is reciprocal Index xiWith xjCompare to obtain aij, then index xjWith xiComparing has aji=1/aij
(1) the analytic hierarchy process (AHP) classification proportion quotiety table according to table 1, six kinds of information that construction system acquisition treatment is obtained are to 100 The comparison sheet two-by-two of~90 marking, it is as shown in table 2 below.
Table 2
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change Normal driving 1 9 2 9 2 9 Fatigue driving 1/9 1 1/9 1/2 1/9 1/2 Normal turn 1/2 9 1 9 2 9 Break turn 1/9 2 1/9 1 1/9 1/2 Normal lane change 1/2 9 1/2 9 1 9 Frequent lane change 1/9 2 1/9 2 1/9 1
Then, the synthetical information evaluating table two-by-two for passing ratio scale reference table being built, is converted into corresponding judgment matrix A1,
A 1 = 1 9 2 9 2 9 0.11 1 0.11 0.5 0.11 0.5 0.5 9 1 9 2 9 0.11 2 0.11 1 0.11 0.5 0.5 9 0.5 9 1 9 0.11 2 0.11 2 0.11 1
Can be calculated, the Maximum characteristic root of matrix A 1 is λmax=6.27, corresponding characteristic vector is：
W &RightArrow; = 0.37 0.03 0.29 0.03 0.24 0.04 T
For matrix A 1, its consistency check index：
RI=1.24；
C R = 0.054 1.24 = 0.044 < 0.1 ;
From CR ＜ 0.1, judgment matrix A1 meets coherence request；
(2) to such as table 3 below of comparison sheet two-by-two of 90~80 marking：
Table 3
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change Normal driving 1 8 2 8 2 8 Fatigue driving 1/8 1 1/8 1/2 1/8 1/2 Normal turn 1/2 8 1 8 2 8 Break turn 1/8 2 1/8 1 1/8 1/2 Normal lane change 1/2 8 1/2 8 1 8 Frequent lane change 1/8 2 1/8 2 1/8 1
Corresponding judgment matrix A2 can be obtained by table 3,
A 2 = 1 8 2 8 2 8 0.13 1 0.13 0.5 0.13 0.5 0.5 8 1 8 2 8 0.13 2 0.13 1 0.13 0.5 0.5 8 0.5 8 1 8 0.13 2 0.13 2 0.13 1
Its Maximum characteristic root λmax=6.27, corresponding characteristic vector is
(3) to such as table 4 below of comparison sheet two-by-two of 80~60 marking：
Table 4
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change Normal driving 1 6 2 6 2 6 Fatigue driving 1/6 1 1/6 1/2 1/6 1/2 Normal turn 1/2 6 1 6 2 6 Break turn 1/6 2 1/6 1 1/6 1/2 Normal lane change 1/2 6 1/2 6 1 6 Frequent lane change 1/6 2 1/6 2 1/6 1
Corresponding judgment matrix A3 can be obtained by table 4,
A 3 = 1 6 2 6 2 6 0.17 1 0.17 0.5 0.17 0.5 0.5 6 1 6 2 6 0.17 2 0.17 1 0.17 0.5 0.5 6 0.5 6 1 6 0.17 2 0.17 2 0.17 1
Its Maximum characteristic root is λmax=6.27, individual features vector is
(4) to such as table 5 below of comparison sheet two-by-two of 60~40 marking：
Table 5
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change Normal driving 1 1/2 1/2 1/3 1/2 1/3 Fatigue driving 2 1 3 2 3 2 Normal turn 2 1/3 1 1/3 1/2 1/3 Break turn 3 1/2 3 1 3 2 Normal lane change 2 1/3 2 1/3 1 1/3 Frequent lane change 3 1/2 3 1/2 3 1
Corresponding judgment matrix A4 can be obtained by table 5,
A 4 = 1 0.5 0.5 0.33 0.5 0.33 2 1 3 2 3 2 2 0.33 1 0.33 0.5 0.33 3 0.5 3 1 3 2 2 0.33 2 0.33 1 0.33 3 0.5 3 0.5 3 1
Its Maximum characteristic root is λmax=6.36, individual features vector is
(5) to such as table 6 below of comparison sheet two-by-two of 40~20 marking：
Table 6
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change Normal driving 1 1/4 1/2 1/4 1/2 1/4 Fatigue driving 4 1 4 2 4 2 Normal turn 2 1/4 1 1/4 1/2 1/4 Break turn 4 1/2 4 1 4 2 Normal lane change 2 1/4 2 1/4 1 1/4 Frequent lane change 4 1/2 4 1/2 4 1
Corresponding judgment matrix A5 can be obtained by table 6,
A 5 = 1 0.25 0.5 0.25 0.5 0.25 4 1 4 2 4 2 2 0.25 1 0.25 0.5 0.25 4 0.5 4 1 4 2 2 0.25 2 0.25 1 0.25 4 0.5 4 0.5 4 1
Its Maximum characteristic root is λmax=6.27, individual features vector is：
(6) to such as table 7 below of comparison sheet two-by-two of 20~0 marking：
Table 7
Normal driving Fatigue driving Normal turn Break turn Normal lane change Frequent lane change Normal driving 1 1/6 1/2 1/6 1/2 1/6 Fatigue driving 6 1 6 2 6 2 Normal turn 2 1/6 1 1/6 1/2 1/6 Break turn 6 1/2 6 1 6 2 Normal lane change 2 1/6 2 1/6 1 1/6 Frequent lane change 6 1/2 6 1/2 6 1
Corresponding judgment matrix A6 can be obtained by table 7,
A 6 = 1 0.17 0.5 0.17 0.5 0.17 6 1 6 2 6 2 2 0.17 1 0.17 0.5 0.17 6 0.5 6 1 6 2 2 0.17 2 0.17 1 0.17 6 0.5 6 0.5 6 1
Its Maximum characteristic root is λmax=6.27, individual features vector is
(7) eigenvectors matrix that driving behavior is judged
According to (1) above~(6), can obtain on normal driving, fatigue driving, normal turn, break turn, normal change Six matrixes of driving behavior evaluation rank in road, frequent lane change factor, driving behavior judge corresponding vector be：
T={ normal driving, fatigue driving, normal turn, break turn, normal lane change, frequent lane change }T,
Character pair value constitute vector be：
λ=[6.27 6.27 6.27 6.36 6.27 6.27]T,
Character pair vector constitute matrix be：
W = 0.37 0.37 0.35 0.07 0.05 0.05 0.03 0.03 0.04 0.30 0.33 0.35 0.29 0.29 0.28 0.08 0.07 0.04 0.03 0.04 0.05 0.25 0.26 0.28 0.24 0.22 0.22 0.10 0.08 0.06 0.04 0.05 0.06 0.20 0.21 0.22 ;
32) BP neural network training
Decision-making of each information on different driving behavior evaluation ranks obtained using the chromatographic assays of driving behavior grade Weight coefficient, with reference to expert's assessment, the related vectorial Preliminary design BP neural network training sample of driving behavior identification of design.
Input information vector is expressed as：
S=[the frequent lane change of the normal driving normal lane change of fatigue driving normal turn break turn]
Output vector is expressed as：
T=[100~90 90~80 80~60 60~40 40~20 20~0]
Correspondence position is 1, then it represents that recognition result is exactly the driving behavior evaluation rank that this position represents；
The training sample of BP neural network is as shown in table 8：
Table 8
Driving behavior is judged Input vector X Preferable output vector Y 100~90 [0.37 0.03 0.29 0.03 0.24 0.04] [1 0 0 0 0 0] 100~90 [0.38 0.04 0.28 0.04 0.23 0.03] [1 0 0 0 0 0] 100~90 [0.36 0.03 0.31 0.03 0.23 0.04] [1 0 0 0 0 0] 100~90 [0.40 0.02 0.30 0.02 0.21 0.05] [1 0 0 0 0 0] 90~80 [0.37 0.03 0.29 0.04 0.22 0.05] [0 1 0 0 0 0] 90~80 [0.36 0.04 0.28 0.05 0.24 0.03] [0 1 0 0 0 0] 90~80 [0.38 0.05 0.27 0.05 0.21 0.06] [0 1 0 0 0 0] 90~80 [0.36 0.04 0.30 0.04 0.23 0.03] [0 1 0 0 0 0] 80~60 [0.35 0.04 0.28 0.05 0.22 0.06] [0 0 1 0 0 0] 80~60 [0.36 0.05 0.26 0.06 0.23 0.04] [0 0 1 0 0 0] 80~60 [0.34 0.03 0.30 0.04 0.21 0.08] [0 0 1 0 0 0] 80~60 [0.35 0.05 0.29 0.06 0.21 0.04] [0 0 1 0 0 0] 60~40 [0.07 0.30 0.08 0.25 0.10 0.20] [0 0 0 1 0 0] 60~40 [0.08 0.31 0.09 0.23 0.09 0.20] [0 0 0 1 0 0] 60~40 [0.06 0.29 0.07 0.27 0.10 0.21] [0 0 0 1 0 0] 60~40 [0.07 0.29 0.09 0.24 0.11 0.20] [0 0 0 1 0 0] 40~20 [0.05 0.33 0.07 0.26 0.08 0.21] [0 0 0 0 1 0] 40~20 [0.06 0.34 0.08 0.24 0.08 0.20] [0 0 0 0 1 0] 40~20 [0.04 0.32 0.06 0.27 0.10 0.21] [0 0 0 0 1 0] 40~20 [0.05 0.34 0.08 0.25 0.09 0.19] [0 0 0 0 1 0] 20~0 [0.05 0.35 0.04 0.28 0.06 0.22] [0 0 0 0 0 1] 20~0 [0.06 0.36 0.05 0.26 0.07 0.20] [0 0 0 0 0 1] 20~0 [0.05 0.34 0.05 0.27 0.05 0.24] [0 0 0 0 0 1] 20~0 [0.04 0.36 0.05 0.26 0.07 0.22] [0 0 0 0 0 1]
Realize the training of BP neural network using MATLAB, and graphical display function by training result by means of MATLAB is moved State shows.
CN201611067704.6A 2016-11-28 2016-11-28 Intelligent driving behavior evaluation method based on analytic hierarchy process (AHP) Pending CN106778538A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611067704.6A CN106778538A (en) 2016-11-28 2016-11-28 Intelligent driving behavior evaluation method based on analytic hierarchy process (AHP)

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611067704.6A CN106778538A (en) 2016-11-28 2016-11-28 Intelligent driving behavior evaluation method based on analytic hierarchy process (AHP)

Publications (1)

Publication Number Publication Date
CN106778538A true CN106778538A (en) 2017-05-31

Family

ID=58904198

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611067704.6A Pending CN106778538A (en) 2016-11-28 2016-11-28 Intelligent driving behavior evaluation method based on analytic hierarchy process (AHP)

Country Status (1)

Country Link
CN (1) CN106778538A (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107491764A (en) * 2017-08-25 2017-12-19 电子科技大学 A kind of violation based on depth convolutional neural networks drives detection method
CN109118040A (en) * 2018-06-27 2019-01-01 上海经达信息科技股份有限公司 A kind of car steering behavior evaluation method based on Fuzzy AHP
CN109523652A (en) * 2018-09-29 2019-03-26 百度在线网络技术（北京）有限公司 Processing method, device, equipment and the storage medium of insurance based on driving behavior
CN109649396A (en) * 2019-01-18 2019-04-19 长安大学 A kind of commercial vehicle drivers safety detecting method
CN110386146A (en) * 2018-04-17 2019-10-29 通用汽车环球科技运作有限责任公司 For handling the method and system of driver attention's data
CN110619482A (en) * 2019-09-27 2019-12-27 深圳前海车米云图科技有限公司 Driving behavior scoring method based on logistic regression and single-level analysis weighting method
CN111598453A (en) * 2020-05-15 2020-08-28 中国兵器工业计算机应用技术研究所 Control work efficiency analysis method, device and system based on execution force in virtual scene
CN112131972A (en) * 2020-09-07 2020-12-25 重庆邮电大学 Method for recognizing human body behaviors by using WiFi data based on attention mechanism
CN113658088A (en) * 2021-08-27 2021-11-16 诺华视创电影科技(江苏)有限公司 Face synthesis method and device based on multiple discriminators

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101615292A (en) * 2009-07-24 2009-12-30 云南大学 Human eye accurate positioning method based on half-tone information
CN101941425A (en) * 2010-09-17 2011-01-12 上海交通大学 Intelligent recognition device and method for fatigue state of driver
CN103646508A (en) * 2013-11-25 2014-03-19 虞静丽 Device and operation method for preventing fatigue driving

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101615292A (en) * 2009-07-24 2009-12-30 云南大学 Human eye accurate positioning method based on half-tone information
CN101941425A (en) * 2010-09-17 2011-01-12 上海交通大学 Intelligent recognition device and method for fatigue state of driver
CN103646508A (en) * 2013-11-25 2014-03-19 虞静丽 Device and operation method for preventing fatigue driving

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
肖献强: "基于信息融合的驾驶行为识别关键技术研究", 《中国优秀博士学位论文全文数据库》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107491764A (en) * 2017-08-25 2017-12-19 电子科技大学 A kind of violation based on depth convolutional neural networks drives detection method
CN110386146A (en) * 2018-04-17 2019-10-29 通用汽车环球科技运作有限责任公司 For handling the method and system of driver attention's data
CN109118040A (en) * 2018-06-27 2019-01-01 上海经达信息科技股份有限公司 A kind of car steering behavior evaluation method based on Fuzzy AHP
CN109523652A (en) * 2018-09-29 2019-03-26 百度在线网络技术（北京）有限公司 Processing method, device, equipment and the storage medium of insurance based on driving behavior
CN109649396A (en) * 2019-01-18 2019-04-19 长安大学 A kind of commercial vehicle drivers safety detecting method
CN110619482A (en) * 2019-09-27 2019-12-27 深圳前海车米云图科技有限公司 Driving behavior scoring method based on logistic regression and single-level analysis weighting method
CN111598453A (en) * 2020-05-15 2020-08-28 中国兵器工业计算机应用技术研究所 Control work efficiency analysis method, device and system based on execution force in virtual scene
CN111598453B (en) * 2020-05-15 2021-08-24 中国兵器工业计算机应用技术研究所 Control work efficiency analysis method, device and system based on execution force in virtual scene
CN112131972A (en) * 2020-09-07 2020-12-25 重庆邮电大学 Method for recognizing human body behaviors by using WiFi data based on attention mechanism
CN113658088A (en) * 2021-08-27 2021-11-16 诺华视创电影科技(江苏)有限公司 Face synthesis method and device based on multiple discriminators

Similar Documents

Publication Publication Date Title
CN106778538A (en) Intelligent driving behavior evaluation method based on analytic hierarchy process (AHP)
CN101923645B (en) Iris splitting method suitable for low-quality iris image in complex application context
CN104778453B (en) A kind of night pedestrian detection method based on infrared pedestrian&#39;s brightness statistics feature
CN103886308B (en) A kind of pedestrian detection method of use converging channels feature and soft cascade grader
CN102122357B (en) Fatigue detection method based on human eye opening and closure state
CN103049740B (en) Fatigue state detection method based on video image and device
CN109147279B (en) Driver fatigue driving monitoring and early warning method and system based on Internet of vehicles
CN110070078B (en) Drunk driving detection method and system based on sensor and machine vision
CN109902560A (en) A kind of fatigue driving method for early warning based on deep learning
CN107038422A (en) The fatigue state recognition method of deep learning is constrained based on space geometry
EP3392806A1 (en) Neural network systems
CN104637246A (en) Driver multi-behavior early warning system and danger evaluation method
CN102567744B (en) Method for determining quality of iris image based on machine learning
CN105608446A (en) Video stream abnormal event detection method and apparatus
CN108921372A (en) Based on step analysis and the freeway tunnel operation security risk evaluating method that matter-element can be opened up
CN105354988A (en) Driver fatigue driving detection system based on machine vision and detection method
CN103136516A (en) Face recognition method and system fusing visible light and near-infrared information
CN104123549A (en) Eye positioning method for real-time monitoring of fatigue driving
CN103839033A (en) Face identification method based on fuzzy rule
CN106780266B (en) Principal component contribution degree parameter-based accident hotspot internal characteristic analysis and driving guidance method
CN105117705A (en) Iris image quality cascade type evaluation method
CN105303771A (en) Fatigue judging system and method
CN107145778A (en) A kind of intrusion detection method and device
CN104156717A (en) Method for recognizing rule breaking of phoning of driver during driving based on image processing technology
CN105184823A (en) Moving target detection algorithm performance evaluating method based on visual perception

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20170531